The Siphon Effect of Consumption End on Production End in the Value Chain under the Data Factor Flow: Evidence from the Regional Comprehensive Economic Partnership Region ^†

Abstract

The data factor strengthens the power of the consumption end, and its high liquidity alters the relationship between the consumption end of the value chain (CEVC) and the production end of the value chain (PEVC). In this paper, we used an interaction term model to empirically analyze the role of the data factor flow in the impact of the CEVC on the PEVC at the Regional Comprehensive Economic Partnership (RCEP) functional integration stage. The results indicate that the data factor flow facilitates the effect of CEVC on PEVC, which brings about the siphon effect (SE). We conducted heterogeneity tests at the national and industrial levels, which revealed the following: (1) The SEs of data capital flow in larger economies and data researcher flow in smaller economies are significant. (2) In most instances, technology-intensive industries suppress the SE because they control the flow of core technologies. (3) Due to the integration and penetration of the data factor with conventional factors, the SE is most prevalent in capital-intensive industries. (4) In labor-intensive industries, the SE is not evident and is even suppressed due to capital substitution for labor. This study provides policy recommendations that would help to reduce the RCEP region’s reliance on external demand and foster regional sustainable development.

1. Introduction

Global economic expansion is decelerating, and the external demand market remains sluggish, meaning that, in the Regional Comprehensive Economic Partnership (RCEP) region, the conventional export-oriented development model is untenable. Driven by the technological revolution, the production and consumption ends of the value chain (VC) production system have changed. Regarding the production end of the value chain (PEVC), the RCEP region has increased the proportion of high-tech manufacturing and accelerated the digital transformation of traditional industries [1]. Digital technology offers a technological avenue through which the VC in the RCEP region can be reshaped [2]. Regarding the consumption end of the value chain (CEVC), the consumption level in Asia is increasing; Asia’s share of global consumption increased from 7% to 18% between 1995 and 2017 [3,4]. Meanwhile, the CEVC and PEVC are highly concentrated; the PEVC continues to congregate in the consumption area, and the proportion of complex VCs in the region is constantly growing [5]. At the CEVC, the proportion of final goods from RCEP countries in the RCEP region exceeded 30% in 2000, 2010, and 2017; at the PEVC, the proportion of intermediate goods from RCEP countries in the region also surpassed 30% in the same years [2]. The RCEP region VC has evolved into a system with close ties and vast market potential.

Faced with this phenomenon, we must consider the following questions: What is the relationship between the CEVC and PEVC? What role does digital technology play in this relationship? Has the traditional VC division of RCEP been significantly reshaped and developed on the basis of the SE of the VC driven by the digital technology revolution after years of functional integration development?

Therefore, in this paper, we used the RCEP region as the study area to examine the role of the data factor flow in the relationship between the CEVC and PEVC. The most important findings indicate that the data factor flow facilitates the effect of CEVC on PEVC, which brings about the siphon effect (SE). Furthermore, the data factor flow drives the effect of downstream and upstream CEVCs on the PEVC. In addition, the SE varies at the national and industrial levels.

In this paper, we aimed to investigate whether the VC production system in the RCEP region has been reshaped under the conditions of digital technology, forming the SE characteristic of the CEVC attracting the PEVC, so as to provide a new development pathway for the RCEP region, which is overly dependent on external demand, as well as to promote sustainable RCEP regional economic development.

Consequently, this paper makes substantial contributions to the field.

(i)

In terms of research, in this paper, we incorporated the data factor flow and the VC production system into a unified analysis framework, focusing on the role of the data factor flow on the internal structure of the VC production system. This study is a useful supplement to the existing literature, and it helps us to understand and grasp the relationship between digital technology and VC reshaping more thoroughly. This means that we can provide useful policy recommendations to improve the development model of RCEP, which is overly dependent on external demand, and ultimately, we can embark on a pathway of sustainable development.

(ii)

In terms of theoretical analysis, in this paper, we systematically expounded the mechanism of the effect of the data factor flow on the production and consumption ends of the VC, meaning that we could develop a deeper understanding of the SE of the CEVC attracting the PEVC and provide a theoretical basis for empirical research.

(iii)

In terms of the construction of indicators, in this paper, we innovatively constructed the indicator of the CEVC according to the meanings of VC indicators, so as to carry out a relevant quantitative analysis.

(iv)

Regarding the empirical research, in this paper, we conducted detailed research, divided the data factor into data capital and data researcher, and divided the CEVC into upstream and downstream. To a certain extent, this reduced the aggregation bias in the overall research to a certain extent, meaning that the conclusions provided were more specific and reliable.

2. Theoretical Background

2.1. Related Background

The literature associated with this study mainly falls into three categories.

To begin, we discuss the term data factor. In a narrow sense, the data factor tends to be informational. The information generated by social and economic activities [6,7,8] can be encoded into a binary sequence of zeros and ones to predict unknown future states, resolve uncertainties [9], and reduce forecasting errors [10]. To some extent, data and information are compatible [11]. It is widely acknowledged that the data factor possesses the following economic attributes:

(i)

The high liquidity of the data factor. The information in data is easily refined and forms a circulating language in the digital realm, thereby reducing the complexity of the information flow and enhancing the efficiency of its diffusion [12]. Data are intermediate factors of production [13]. The integration of the data factor and other production factors fosters subsequent innovation, creates a mechanism for innovation to be spontaneously strengthened, and provides the impetus for sustainable economic development [14]. To some extent, the innovation process of data is the process of its continuous flow to generate informational knowledge.

(ii)

The low marginal cost of the data factor. The marginal cost of data is close to zero [9,15,16]. The value of data rises with increased usage [17]. This attribute is the basis for scale and scope economies. It is less costly to transmit and convert data, which diminishes the significance of boundaries and enables more businesses to participate in economic production activities.

(iii)

The non-competitiveness of the data factor. The integration of non-competitive factors with production factors generates the law of increasing returns to scale, thereby promoting sustainable economic growth [18]. This indicates that the extensive use of the data factor can make the production point close to the production possibility curve, which is crucial for fostering economic growth.

With the expansion of data factor applications, the classification of the data factor has become increasingly specific. In these studies, the emphasis has shifted to data capital. Xu and Zhao [19] defined data capital as digitalized and production factor-oriented information and data. This definition divides an initial portion of the data factor category and provides a starting point for the classification of the data factor. Li and Zhao [20] elaborated on the transformation of data from a resource to a production factor (data capital) through the evolution of data forms and value forms and summarized the conditions under which data can become production factors. Su and Chen [21] argued that data capital was the sum of social relations with a lossless value-creating effect on social production. The uniqueness of data capital is derived from the contrast between general capital and data capital. According to Tang [22], data capital is the sum of the value added, which is derived from the inflow or outflow of data assets. This study sought to reclassify capital using a scientific measurement method.

Theoretically, the data factor flow is inseparable from the “materialized” carrier, allowing for the deconstruction and classification of the data factor, thereby advancing quantitative research. Using the classification of the research and development (R&D) factor by Bai et al. [23], in this study, we deconstructed the data factor into data capital and data researcher.

In this study, we also highlighted the impact of digital technology on the SE in the VC. The data factor flow cannot be separated from the application of digital technology; to a large extent, the digital technology application is the data factor flow.

Typically, traditional international trade in the pursuit of a scale economy will reduce the diversity of needs. In contrast, the application of digital technology establishes the law of increasing returns to scale, reconciles the contradiction between scale economy and product differentiation, and more effectively satisfies diverse needs [24]. Thus, the consumption end is important in attracting the production end. Qi et al. [12] examined the economic theory of the evolution of consumer value. They concluded that digital technology enhanced consumers’ value-demand proposition and increased consumer demand’s supply-side efficiency.

Consumption-wise, digital technology promotes scaled-down, customized production by enhancing worker skills, expanding consumer rights, enhancing digital capabilities, and mining front-end and back-end personalized demand [25]. Production wise, digital information and traditional industries are highly integrated and permeated, thereby enhancing the VC performance of traditional industries. As an enterprise’s external transaction costs are lower than its internal transaction costs, an enterprise’s organizational scale tends to be miniaturized and specialized to meet massive customization demands [26]. The new digital economy model drives the diversification of consumer demands and the long-tail effect of consumption, thereby compelling businesses to transform and upgrade [27]. Consumption growth influences market fluctuations. The market should implement technological innovation, not only in the PEVC but also in CEVC’s marketing channels and brands, thus fostering the innovation of the PEVC with the CEVC [2]. A larger market creates a scale economy that compensates for the high cost of innovation and attracts innovation agglomeration, which becomes a vital force in sustainable development. During the transformation at the VC level, one sees the SE of CEVC attract the PEVC as a consequence of the changes mentioned above. Although the studies mentioned above provide a theoretical basis for the impact of digital technology on the SE in the VC, few empirical studies have been carried out, and there is no convincing empirical evidence.

Regarding the metrics associated with the SE, Ma et al. [28] explored the regionalization trend in the VC by using the VC database and adopting the difference between the internal and external proportion of intermediate input. Liu and Zhao [29] established upstream and downstream dependence indicators for the VC in the East Asian region on the basis of the theory of VC decomposition. However, due to the limitations of the research methods and objectives, these indicators do not accurately reflect the characteristics of the relationship between the CEVC and PEVC.

Finally, some of the literature addresses to the issues surrounding the VC and the economic model in East Asian nations.

In terms of comparative advantage, there are the essential comparative advantages of traditional trade are labor, land, and capital, among others. These comparative advantages allow East Asian nations to participate in the division [30]. As digitization develops, the proportion of labor allocated to productivity falls [31]. As a result, the comparative advantage of labor diminishes [32], data factors that rely on new infrastructure construction and human resources become the new comparative advantage, and the advantage of the scale economy becomes more prominent. Individual demands emphasize the advantages of local production and bring production closer to the regional market [33].

Agglomeration affects the VC. Enterprises are in both the division of GVC and localized clusters [34]. The agglomeration of exporters promotes the added value of exports [35], thus enhancing the position of the VC [36]. A more detailed dimensional study has shown that the industry concentration inhibits the position of the VC, while the specialization level promotes it [37], and synergistic industrial agglomeration increases the domestic value-added export rate [38]. The VC also influences agglomeration. According to research, deepening the domestic VC would tap market potential and promote agglomeration [39]. Currently, platform transactions enabled by digital technology successfully match supply and demand, aggregate upstream and downstream enterprises in the VC, create a virtual network agglomeration effect, and exert scale and network effects [36].

In terms of fragmentation and the VC, fragmentation production refers to the de-centralization of international production processes [40]. This is more evident in the context of globalization [41]. Under the influence of digital technology, production fragmentation modifies the equilibrium world market price and factor prices [42], thereby widening the income gap [43]. In East Asia, fragmentation production increases the intraregional economic interdependence and maintains the dependence on the global economy [44].

Regarding the SE, in essence, the current study demonstrated that digital technology has altered the production system’s power structure, and the GVC has amplified the shift [45]. The GVC’s mode of production has shifted from series to parallel. The organizational structure is typically simplified, market-oriented, and modularized, and it responds quickly to market changes [46]. The production mode shifts its emphasis from production to products, emphasizing the diversification and refinement of products [47]. This indicates the significance of the CEVC within the production system. This type of research begins with the shift in the production system caused by digital technology and, to some extent, identifies the essence of the SE.

Regarding the economic development model in East Asia, the existing studies generally conclude that the East Asian economy has followed a geese-shaped development model since the 1960s. In the late 1980s, as the economic bubble burst, East Asia established an industrial cluster chain division [48]. In the 1990s, the degree of integration strengthened [49], and a relatively stable triangular trade was created [50,51]. With the slowdown in global trade and the development of the VC, this export-oriented economic development model, which is excessively reliant on external markets, has become increasingly unsustainable. Due to the impact of digital technology, in East Asia, a node driven by domestic demand within the region and symmetrical development inside and outside the region are being reshaped [1]. The RCEP region provides the necessary conditions for the VC to be reshaped: consensus, an internal and external environment, VC introversion, and economic dependence [2]. The Asian VC is evolving toward diversification and regionalization [4]. As a result, the proportions of East Asian VC production and consumption have simultaneously increased [52].

The literature mentioned above examines the VC from multiple perspectives, ranging from customary conditions to the digital background. Understanding the formation of the SE is helpful, but we still lack a comprehensive and systematic framework for analysis.

2.2. Theoretical Analysis

With the widespread adoption of digital technology, the data factor is reshaping national factor endowment structures. This challenges the traditional scale economy theory and modifies the trend of supply and demand curves. (Krugman pointed out that the supply and demand curve with network externality diverged from the traditional theory. That is, expanding market scale would produce more value, and thus the demand curve would tilt to the upper right; meanwhile, as learning and experience make it easier to produce goods, as the quantity of goods increases, the price falls, and thus the supply curve slopes to the lower right.) Moreover, it causes an adjustment to the VC system by inverting the status of production and consumption.

From the perspective of PEVC, data factor flow transforms the traditional trade theory law of diminishing returns to scale into the law of increasing returns to scale. First, the noncompetitive nature of the data factor enables sharing to maximize the value. Consequently, the scale economy has been a significant source of enterprise revenue, and the market size has become an essential factor in stimulating the PEVC to benefit the VC. The use of data-carried information and information-generated knowledge expands traditional trade content and facilitates the formation of a scale economy. Second, the technology level is endogenous. As the production scale has increased, digital technology has become an endogenous variable. Its widespread application reduces the costs associated with data collection, storage, categorization, and extraction. Finally, the return to scale is becoming more prevalent because management efficiency is increasing faster than the enterprise scale [24].

In addition, traditional trade theory maintains that scale economy is the basis for trade benefits and that differentiated demand is the driving force of trade [53]. Consumers tend to diversify their demand when facing similar financial constraints. Under a specific demand, manufacturers will reduce the production of the same product if they increase product variety, thereby diminishing the scale economy. The expansion of the international market can improve the demand level, which creates a scale economy. The application of digital technology has significantly altered traditional trade in two ways. First, digital technology effectively matches a scale economy with distinct needs [24], enabling large-scale differentiated production. The scale economy has been transformed from standardization to differentiation. The PEVC is encouraged to take the CEVC as the core and implement the strategy of geographic proximity to the consumer. Demand differentiation has become an essential factor in attracting the PEVC. Second, manufacturers expand product categories under specific demand conditions to increase the scope economy. In the region, a scope economy can be realized through digital technology. Additionally, production in the region is advantageous due to its proximity to market demand and the low production risk. In contrast, transregional production faces trust, language, and payment system barriers, reducing the producers’ ability to develop transregional markets, which causes the PEVC to focus on the CEVC in the region.

The transformation of the PEVC is ultimately influenced by digital technology. First, network externality highlights the first-mover advantage. It is easier to obtain market resources and establish a scale economy if enterprises occupy the market in advance, thus creating a higher market threshold for latecomers [54]. To obtain a time-competitive advantage, businesses must quickly succeed in data collection, product development, and customer service, which compels them to be geographically close to the consumer market, establish a consumer value feedback mechanism, and quickly occupy the market to achieve a scale economy. Second, digital technology enables the time factor to be recorded. As it is technically difficult to measure, the time factor has been neglected in traditional trade. As the GVC division deepens, coordination and rapid feedback become important, and the ability to respond rapidly becomes crucial in the GVC division system’s layout. Digital technology has made it possible to measure response times. The Internet of Things (IoT) has enabled the real-time tracking of each link of the division and coordinates the flow of various production factors [3]. Close geographical proximity to the CEVC enables effective real-time control and enhances the capacity for timely responses, making the CEVC a critical factor in attracting the PEVC.

From the CEVC, the data factor establishes the consumer’s market dominance. In conventional economic theory, because resource scarcity restricts production supply, producers dominate the market, and consumers are primarily passive recipients of commodities [12]. With the proliferation of digital technology, information regarding supply and demand is more balanced. The extensive involvement of consumers in the production process has profoundly altered the supply side, fostering high cohesion and precise matching between supply and demand, accelerating the rate of new product R&D, reducing the uncertainty risk of new product launches, and enhancing enterprises’ innovation efficiency and anti-risk ability. In comparing supply and demand forces, the characteristics of market demand and the logic of value creation have shifted, indicating a trend of consumption-end power consolidation [55]. Consumer demand has become an increasingly dominant value force, giving rise to the logic of demander predominance [56]. To increase market competitiveness and enhance consumer value, businesses implement a series of strategic measures, such as product R&D and technological upgrading, resulting in the formation of a new production ecosystem composed of actively pursuing consumers. In addition, the data factor’s low marginal cost rapidly increases the market size, lowers the market entry barriers, and intensifies the market competition. As a result, the enterprise’s marketing strategy shifts from low price to differentiation, highlighting the significance of acquiring timely information regarding shifting consumer demand. Close geographic proximity to consumers permits more precise demand data thorough product R&D, rapid market penetration, and first-mover advantages. Therefore, the CEVC becomes a powerful force that attracts the PEVC.

Accordingly, we propose the following hypothesis:

Under the data factor flow, at the PEVC, the law of increasing returns to scale reconciles the scale economy and production differentiation. At the CEVC, the data factor establishes the consumer’s dominant position and encourages production differentiation through intensified competition. As a result, the CEVC becomes a powerful driving force to attract the PEVC, promoting production agglomeration in the region on a large market scale and forming the SE to realize greater value.

Changes in the PEVC and CEVC indicate that the faster the data factor flows, the greater the attraction of the CEVC to the PEVC, which makes it easier for the profit-driven PEVC to modify the VC layout.

In the RCEP region, the increasing division of the VC has increased the degree of production linkage between member nations and has ensured the optimization and stability of the VC, laying the groundwork for the widespread adoption of digital technology. Furthermore, the extensive data factor flow in the region reinforces the data factor’s scale and scope economies. Theoretically, RCEP member countries show differences in their economies, systems, cultures, geographies, etc., making it easier to conduct division and cooperation, form a scope scale in the region, and drive the PEVC to the CEVC to enable efficient production. With the launch of the RCEP and the future improvement in the integration level, the cross-border data factor flow will strengthen, thus further releasing the SE. In addition, due to the unique characteristics of each country and industry, the SE will be heterogeneous.

3. Materials and Methods

3.1. Variables and Data

The explained variable in this paper is the PEVC. On the basis of the regional agglomeration trend in the VC, in order to describe the SE of the CEVC attracting the PEVC, the quantitative indicator of the PEVC should exhibit production agglomeration characteristics. Therefore, following Yang [57] and Shi and Wang [58], in this paper, we measured the agglomeration degree using the location quotient index method. The adoptive data are the vertically specialized indicator (Vs1) in the VC index system. Vs1 refers to the total export that is used by importing countries for export production again. It is a comprehensive statistical indicator used to measure the VC production division, and it reveals the supply of export production. Consequently, this indicator is used to measure the distribution of the PEVC. PEVC (Vcaggvs1) is subsequently constructed as follows:

Vcaggvs1_ijt = (Vs1_ijt/Vs1_it)/(Vs1_wjt/Vs1_wt)

(1)

where i, j, t, and w represent country, industry, year, and world, respectively.

The core explanatory variable is the CEVC. As mentioned in Section 3.2., the pure domestic VC length introversion (Invc_l_d) is used as a proxy variable for the CEVC.

The moderating variables include data researcher flow (Dpe) and data capital flow (Dca). In this paper, we did not emphasize the comprehensive influence of the data factor flow on the SE, but we focused on whether the different data factor flows had the same effect on the SE, and whether this effect was stable. Therefore, in this paper, we divided the data factor into the data researcher and data capital for analysis. Additionally, the refinement of the data factor can help overcome the bias of empirical aggregation to draw more comprehensive and reliable conclusions. Referencing Bai et al. [23] and Shi and Wang [58], the output-constrained double logarithmic gravity model of the attraction variable was adopted to measure the data factor flow between countries. According to the rational man hypothesis in economics, a data researcher will relocate for superior living and working conditions. Therefore, the difference between the number of patent applications (Pate) and the difference between the per capita GDPs of employed populations (Epgdp) were selected as attractiveness variables. The data researcher flow from country i to country k was as follows (data factor will flow only if other countries are attractive to the data factor of the home country; therefore, the difference in the formula only takes the values of countries that are greater than that of the home country, that is, the difference should be greater than 0):

Dpe_ikt = Rd_it (log) ·(Pate_kt − Pate_it)(log) ·(Epgdp_kt − Epgdp_it)(log) ·D_ik⁻¹

(2)

where Rd represents the number of R&D personnel, while D is the distance between two capitals.

According to the profit-seeking nature of capital, market size and production correlation will impact data capital. As attraction variables, we used the difference between the per capita GDP (Pgdp) and the difference between the correlation degree of the VC (Vcli) (Vcli= Vcli_b + Vcli_f, Vcli_b = FGY_GVC/FGY, Vcli_f = VA_GVC/VA; Vcli_b and Vcli_f denote upstream and downstream correlation degrees of VC, respectively; FGY_GVC represents the added value from intermediate imports in the production of final products; VA represents the final products and services; VA_GVC denotes the added value used in the production of intermediate exports; VA denotes the gross added value). Hence, the data capital flow from industry j in country i to industry j in country k is

Dca_ijkt = Ca_it (log) ·(Pgdp_kt − Pgdp_it)(log) ·(Vcli_kjt − Vcli_ijt)(log) ·D_ik⁻¹

(3)

where Ca represents data capital; considering that the internet will affect countries’ revenue from cross-border data flow, secure internet servers (per million people) are used to represent data capital.

The control variables included infrastructure, openness, and market size. The variables are described in

Table 1. Variable declaration.

Variable	Symbol	Measurement Index
Production end of value chain	Vcaggvs1	vertical specialization agglomeration degree
	Vcaggvax	value-added export agglomeration degree
Data researcher flow	Dpe	data researcher flow (calculated by R&D researchers per million people)
Data capital flow	Dca	data capital flow (calculated by secure internet servers per million people)
Infrastructure	Inf	frequency of receiving goods within a predetermined time (1–5 from lowest to highest)
	Inf2	infrastructure quality (1–5 from lowest to highest)
Openness	Open	goods trade as a share of GDP
	Open2 (log)	net inflow of foreign direct investment (current USD)
Market size	Gdp (log)	GDP (current USD)
	Pgdp (log)	per capita GDP (current USD)
	Inva (log)	industrial value-added (current USD)
Consumption end of value chain	Invc	pure domestic value chain length introversion

Notes: (log) means taking the natural logarithm of a variable.

. In this paper, the VC indicators (Vs1, Vax, VA_GVC, VA, FGY_GVC, FGY, Plv_d, Plv, Ply_d, and Ply) were obtained from the UIBE_GVC database, and the distance between two capitals (D) was derived from the Cepii database and www.indo.com, accessed on 3 October 2022, while the other indicators were obtained from the World Bank database. The data covered 35 industries in 10 countries from 2010 to 2016. However, due to the absence of data for specific countries and years, there were 46 national and 1610 industrial samples. The 10 countries were China, Japan, South Korea, Australia, and the ASEAN 6 (Malaysia, Singapore, Indonesia, Thailand, the Philippines, and Vietnam). The 35 industries were c1—agriculture, hunting, forestry, and fishing; c2—mining and quarrying; c3—food, beverages, and tobacco; c4—textiles and textile products; c5—leather, leather goods, and footwear; c6—wood and products of wood and cork; c7—pulp, paper, printing, and publishing; c8—coke, refined petroleum, and nuclear fuel; c9—chemicals and chemical products; c10—rubber and plastics; c11—other non-metallic mineral; c12—basic metals and fabricated metal; c13—machinery, nec; c14—electrical and optical equipment; c15—transport equipment; c16—other manufacturing; c17—electricity, gas, and water supply; c18—construction; c19—sale, maintenance, and repair of motor vehicles and motorcycles, and retail sale of fuel; c20—wholesale trade and commission trade, except of motor vehicles and motorcycles; c21—retail trade, except of motor vehicles and motorcycles, and repair of household goods; c22—hotels and restaurants; c23—inland transport; c24—water transport; c25—air transport; c26—other supporting and auxiliary transport activities; activities of travel agencies; c27—post and telecommunications; c28—financial intermediation; c29—real estate activities; c30—renting of m&eq and other business activities; c31—public administration and defense, compulsory social security; c32—education; c33—health and social work; c34—other community, social, and personal services; c35—private households with employed persons.

3.2. Research Methods

Using the method of Ma et al. [28] to measure the regionalization of the VC and the method of Liu and Zhao [29] to measure the dependence of the VC, in this paper, we constructed the CEVC indicator according to the VC index’s meaning. VC production length refers to the number of production stages in the VC, which reflects the complexity of the production process or the closeness of the relationships between industries. Forward (downstream) production length and backward (upstream) production length can be distinguished. The former represents the total output induced by unit added value, while the latter denotes the total output drawn by unit final goods [59]. In this paper, consumption refers to domestic consumption, as all production must eventually be consumed. As the pure domestic VC length can be understood as circuitous production in the country to satisfy final domestic demand, it can be considered as the consumption portion of the country. It reflects the close level of linkage between the VC production and its own consumption. The pure domestic VC length ratio represents the proportion of the closeness of the linkage between the VC production and its own consumption. This proportion reflects the consumption level in the country. Consequently, it can be considered as the CEVC (according to the upstream and downstream of VC, the consumption end index of upstream and downstream VC is constructed; these values refer to the consumption at the upstream and downstream VC links, respectively). Accordingly, the pure domestic VC length introversion (Invc_l_d) is constructed as a proxy variable for the CEVC, which is defined as follows:

Invc_l_d = Invc_l_f_d + Invc_l_b_d

(4)

The downstream pure domestic VC length introversion (Invc_l_f_d) is defined as

Invc_l_f_d = Plv_d/Plv

(5)

Meanwhile, the upstream pure domestic VC length introversion (Invc_l_b_d) is defined as

Invc_l_b_d = Ply_d/Ply

(6)

where Plv_d and Ply_d represent the downstream and upstream pure domestic production lengths, respectively, while Plv and Ply are the total downstream and upstream production lengths, respectively.

In this paper, we aimed to investigate whether the VC production system in the RCEP region had been reshaped under the conditions of digital technology, giving rise to the SE characteristic of the CEVC attracting the PEVC. Accordingly, during the empirical analysis, we intended to examine the role of data factor flow in the impact of the CEVC on the PEVC, that is, whether the data factor flow promotes the SE of the CEVC attracting the PEVC. Therefore, in this paper, we adopted the following interaction term regression model for analysis:

Vcaggvs1_ijt = α₀ + α₁Dpe_it + α₂Invc_ijt + α₃ (Dpe_it × Invc_ijt) + α₄Cont + v_i + v_j + v_t + ε_ijt

(7)

Vcaggvs1_ijt = β₀ + β₁Dca_it + β₂Invc_ijt + β₃ (Dca_it × Invc_ijt) + β₄Cont + η_i + η_j + η_t + μ_ijt

(8)

where i, j, and t represent country, industry, and year, respectively; Vcaggvs1 is the PEVC; Dpe and Dca denote data researcher flow and data capital flow, respectively; Invc is the CEVC; Cont expresses the control variable; v_i and η_i, v_j and η_j, and v_t and η_t are country, industry, and year fixed effects, respectively; and ε_ijt and μ_ijt are the random error terms. The interaction terms of Invc interacting with Dpe and Dca are introduced into the model individually.

4. Results and Discussion

4.1. Regressing the Siphon Effect

4.1.1. Regression Results of the Siphon Effect and Discussion on Endogeneity

In

Table 2. The siphon effect (SE) regression of the data factor flow.

	7-1	7-2	7-3	7-4	8-1	8-2	8-3	8-4
Dpe	−1.3100 ***	−1.3122 ***	−1.2999 ***	−1.3026 ***
	(0.2804)	(0.2807)	(0.2854)	(0.2861)
Dca					−6.2799 ***	−6.2819 ***	−6.2859 ***	−6.2967 ***
					(1.0768)	(1.0772)	(1.0776)	(1.0793)
Invc_l_d	0.1449 **	0.1444 **	0.1445 **	0.1448 **	0.1144 *	0.1140 *	0.1138 *	0.1138 *
	(0.0614)	(0.0614)	(0.0615)	(0.0615)	(0.0625)	(0.0626)	(0.0626)	(0.0626)
Dpe × Invc_l_d	0.8529 ***	0.8541 ***	0.8535 ***	0.8530 ***
	(0.1658)	(0.1660)	(0.1661)	(0.1662)
Dca × Invc_l_d					3.2203 ***	3.2209 ***	3.2213 ***	3.2310 ***
					(0.6767)	(0.6770)	(0.6772)	(0.6792)
Control variables	No	Yes	Yes	Yes	No	Yes	Yes	Yes
Country-fixed, year-fixed, and industry-fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	1610	1610	1610	1610	1610	1610	1610	1610
R2	0.1750	0.1750	0.1750	0.1750	0.1899	0.1899	0.1900	0.1900

Notes: ***, **, and * represent significance at the 1, 5, and 10 percent levels, respectively. The numbers in parentheses are standard errors. Due to space limitation, constant terms and control variables are not reported. The following tables are the same.

, 7-1 to 7-4 and 8-1 to 8-4 present the regression results of models (7) and (8) obtained by successively adding the control variables Open, Inf, and Gdp (log). The interaction term coefficients were all significantly positive, indicating that the data factor flow promoted the effect of the CEVC on the PEVC, thereby forming the SE. In terms of the PEVC, the data factor’s law of increasing returns to scale made the scale economy an essential source of production benefits. It motivates producers to seek out regions with large market sizes for production, resulting in a production agglomeration in the regions with large market sizes. In terms of the CEVC, the data factor established the consumer’s dominant market position and substantially impacted R&D, production, marketing, and other enterprise links. Moreover, the low cost of the data factor not only enlarged the market but also intensified competition and encouraged competition from price to product differentiation, making it necessary for businesses to promptly control information at the CEVC. Thus, the proximity of a region to customers with a vast market potential becomes a crucial factor in the formation of the SE. The hypothesis is thus confirmed.

In order to ensure the reliability of the above-mentioned conclusion, a robustness test was performed on the alternative indicators.

Table 3. Alternative indicators: the SE of the data researcher flow.

	7-1	7-2	7-3	7-4	7-5	7-6	7-7
Dpe	−1.5366 ***	−1.3056 ***	−1.3103 ***	−1.3046 ***	−1.0682 ***	−0.5521 **	−0.0104
	(0.3117)	(0.2860)	(0.2816)	(0.2864)	(0.2962)	(0.2528)	(0.1369)
Invc_l_d	0.1527 **	0.1455 **	0.1446 **	0.1449 **
	(0.0670)	(0.0615)	(0.0615)	(0.0615)
Invc_l_b_d					1.6407 ***
					(0.2363)
Invc_l_f_d						0.0519
						(0.0684)
Dinvc_l_d							0.0936
							(0.0566)
Dpe × Invc_l_d	0.9851 ***	0.8513 ***	0.8532 ***	0.8528 ***
	(0.1811)	(0.1661)	(0.1661)	(0.1661)
Dpe × Invc_l_b_d					1.4367 ***
					(0.3528)
Dpe × Invc_l_f_d						0.7617 ***
						(0.2818)
Dpe × Dinvc_l_d							0.8637 ***
							(0.1633)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Country-fixed, year-fixed, and industry-fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	1610	1610	1610	1610	1610	1610	1604
R2	0.1278	0.1750	0.1750	0.1751	0.2159	0.1574	0.1974

Notes: *** and ** represent significance at the 1 and 5 percent levels, respectively.

and

Table 4. Alternative indicators: the SE of the data capital flow.

	8-1	8-2	8-3	8-4	8-5	8-6	8-7
Dca	−6.5961 ***	−6.2996 ***	−6.2908 ***	−6.3020 ***	−3.5242 ***	−4.2872 ***	−1.5440 ***
	(1.1784)	(1.0793)	(1.0792)	(1.0793)	(1.2387)	(0.9324)	(0.2876)
Invc_l_d	0.1300 *	0.1142 *	0.1139 *	0.1138 *
	(0.0684)	(0.0626)	(0.0626)	(0.0626)
Invc_l_b_d					1.7295 ***
					(0.2593)
Invc_l_f_d						0.0297
						(0.0680)
Dinvc_l_d							0.0368
							(0.0572)
Dca × Invc_l_d	3.3607 ***	3.2336 ***	3.2291 ***	3.2356 ***
	(0.7416)	(0.6791)	(0.6792)	(0.6791)
Dca × Invc_l_b_d					3.0706 *
					(1.5685)
Dca × Invc_l_f_d						3.8189 ***
						(1.1579)
Dca × Dinvc_l_d							4.6379 ***
							(0.7407)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Country-fixed, year-fixed, and industry-fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	1610	1610	1610	1610	1610	1610	1604
R2	0.1404	0.1900	0.1899	0.1900	0.2231	0.1772	0.2203

Notes: *** and * represent significance at the 1 and 10 percent levels, respectively.

itemize the results of model (7) and model (8), respectively. In model 7-1 and model 8-1, in reference to Shi and Wang [58], Vcaggvax (Vcaggvax_ijt = (Vax_ijt/ Vax_it)/(Vax_wjt/Vax_wt), where Vax denotes value-added export) replaced Vcaggvs1. The net inflow of foreign direct investment (Open2(log)) replaced Open in models 7-2 and 8-2. Infrastructure quality (Inf2) substituted Inf in models 7-3 and 8-3. Per capita GDP (Pgdp (log)) replaced Gdp (log) in models 7-4 and 8-4. In models 7-5 and 8-5, upstream pure domestic VC length introversion (Invc_l_b_d) replaced Invc_l_d. In models 7-6 and 8-6, the downstream pure domestic VC length introversion (Invc_l_f_d) substituted Invc_l_d. In models 7-7 and 8-7, Invc_l_d and Vcaggvs1 were replaced by their respective first-order differences to test the data factor flow’s contribution to the effect of a change in Invc_l_d (Dinvc_l_d) on the change in Vcaggvs1 (Dvcaggvs1). The outcomes showed that the significance and direction of the interaction term coefficients remain unchanged, manifesting the robustness of the preceding conclusion.

In terms of the regression results, the issue of endogeneity should also be focused on. Generally, the model may be affected by some unobserved factors, thus leading to the endogeneity issue. In this subsection, similar to the studies by Shi [60] and Liu and Gu [54], historical data are utilized as the instrumental variable (IV), and the two-stage least squares (2SLS) method was adopted to overcome the endogeneity issue. Columns 1–3 of

Table 5. Test of endogeneity: the SE of the data factor flow.

Model (7)	7-1	7-2	7-3	7-4
Dpe	−1.2827 ***	−1.1809 ***	−1.2693 ***	−1.3024 ***
	(0.2868)	(0.2890)	(0.2873)	(0.2861)
Invc_l_d	0.1396 **	0.1448 **	0.1413 **	0.1447 **
	(0.0617)	(0.0616)	(0.0616)	(0.0615)
Dpe × Invc_l _d	0.8643 ***	0.8494 ***	0.8596 ***	0.8530 ***
	(0.1666)	(0.1665)	(0.1664)	(0.1662)
R2	0.1728	0.1723	0.1734	0.1750
Model (8)	8-1	8-2	8-3	8-4
Dca	−6.2803 ***	−6.2621 ***	−6.1969 ***	−6.3046 ***
	(1.0811)	(1.0811)	(1.0826)	(1.0794)
Invc_l_d	0.1106 *	0.1150 *	0.1139 *	0.1138 *
	(0.0628)	(0.0627)	(0.0627)	(0.0626)
Dca × Invc_l_d	3.2101 ***	3.2205 ***	3.1417 ***	3.2381 ***
	(0.6804)	(0.6803)	(0.6825)	(0.6793)
R2	0.1878	0.1879	0.1884	0.1900
Control variables	Yes	Yes	Yes	Yes
Country-fixed, year-fixed, and industry-fixed effects	Yes	Yes	Yes	Yes
Observations	1609	1609	1609	1610

Notes: Phase I results were not reported due to space limitations and are available in the annex. Notes: ***, **, and * represent significance at the 1, 5, and 10 percent levels, respectively.

used Open, Inf, and Gdp (log) lag periods, respectively, as their own IV, while column 4 considered industrial value-added (Inva (log)) as the IV of Gdp (log). These IVs were strongly correlated with their own endogenous explanatory variables, but they did not directly affect the PEVC, satisfying the correlation and exogenous requirements of IV [61,62]. The results show that by controlling for the endogeneity issue, the significance and direction of the interaction term coefficients did not change, thus further supporting the robustness and validity of the estimation mentioned above.

4.1.2. Regression Results and Discussion of the Siphon Effect Upstream and Downstream

Different positions in the upstream and the downstream of the GVC had different knowledge densities, organizational structures, and resource openness; undertake different production links; and participate in different divisions. R&D, design, and other links are mainly conducted upstream of the GVC, with high knowledge density and limited resource openness. Processing, assembly, and other links are mainly conducted downstream of the GVC, with low knowledge density and high resource openness. These differences will affect the data factor flow, and the data factor flow brings about a different effect of the CEVC on PEVC, which affects the SE. Therefore, in this paper, we further explored whether the structural differences between the upstream and downstream of the VC had differential effects on the SE.

First, Invc_l_f_d was used as the proxy variable for the downstream CEVC. In

Table 6. The SE regression of the data factor flow: Invc_l_f_d.

	7-1	7-2	7-3	7-4	8-1	8-2	8-3	8-4
Dpe	−0.5627 **	−0.5635 **	−0.5488 **	−0.5521 **
	(0.2470)	(0.2472)	(0.2517)	(0.2528)
Dca					−4.2845 ***	−4.2874 ***	−4.2873 ***	−4.2872 ***
					(0.9305)	(0.9308)	(0.9311)	(0.9324)
Invc_l_f_d	0.0521	0.0517	0.0518	0.0519	0.0306	0.0299	0.0297	0.0297
	(0.0683)	(0.0683)	(0.0684)	(0.0684)	(0.0679)	(0.0679)	(0.0680)	(0.0680)
Dpe × Invc_l_f_d	0.7612 ***	0.7619 ***	0.7621 ***	0.7617 ***
	(0.2815)	(0.2816)	(0.2817)	(0.2818)
Dca × Invc_l_f_d					3.8214 ***	3.8231 ***	3.8190 ***	3.8189 ***
					(1.1538)	(1.1541)	(1.1546)	(1.1579)
Control variables	No	Yes	Yes	Yes	No	Yes	Yes	Yes
Country-fixed, year-fixed, and industry-fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	1610	1610	1610	1610	1610	1610	1610	1610
R2	0.1574	0.1574	0.1574	0.1574	0.1771	0.1772	0.1772	0.1772

Notes: *** and ** represent significance at the 1 and 5 percent levels, respectively.

, columns 7-1 through 7-4 and 8-1 through 8-4 display the regression results of models (7) and (8), respectively, which were obtained by successively adding the control variables Open, Inf, and Gdp (log). The interaction term coefficients were significantly positive, indicating that the data factor flow enhanced the effect of the downstream CEVC on the PEVC, thereby producing the SE.

Digital technology is embedded in production manufacturing and drives the digital participation of conventional manufacturing. (i) Digital technology automation and intelligentization increases production efficiency, boosts returns on the scale, and entices more production factors to assemble at the CEVC, thereby forming the SE. (ii) Digitalization reinforces consumers’ sovereignty and encourages production from standardization to dynamic customization, resulting in a competitive advantage derived from price to product differentiation. Product differentiation has emerged as a significant driving force in the SE. The ability to meet customized needs dominates the competitive orientation, which drives the PEVC towards the CEVC and forms the SE. The hypothesis is thus confirmed.

Second, Invc_l_b_d was used as the proxy variable for the upstream CEVC. In

Table 7. The SE regression of the data factor flow: Invc_l_b_d.

	7-1	7-2	7-3	7-4	8-1	8-2	8-3	8-4
Dpe	−1.0659 ***	−1.0709 ***	−1.0637 ***	−1.0682 ***
	(0.2906)	(0.2909)	(0.2960)	(0.2962)
Dca					−3.5129 ***	−3.4980 ***	−3.5087 ***	−3.5242 ***
					(1.2363)	(1.2370)	(1.2376)	(1.2387)
Invc_l_b_d	1.6333 ***	1.6338 ***	1.6342 ***	1.6407 ***	1.7247 ***	1.7290 ***	1.7286 ***	1.7295 ***
	(0.2356)	(0.2357)	(0.2358)	(0.2363)	(0.2589)	(0.2592)	(0.2592)	(0.2593)
Dpe × Invc_l_b_d	1.4403 ***	1.4456 ***	1.4443 ***	1.4367 ***
	(0.3518)	(0.3520)	(0.3523)	(0.3528)
Dca × Invc_l_b_d					3.0541 *	3.0320 *	3.0401 *	3.0706 *
					(1.5642)	(1.5654)	(1.5659)	(1.5685)
Control variables	No	Yes	Yes	Yes	No	Yes	Yes	Yes
Country-fixed, year-fixed, and industry-fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	1610	1610	1610	1610	1610	1610	1610	1610
R2	0.2157	0.2158	0.2158	0.2159	0.2228	0.2229	0.2230	0.2231

Notes: *** and * represent significance at the 1 and 10 percent levels, respectively.

, columns 7-1 to 7-4 and 8-1 to 8-4 display the regression results of models (7) and (8), respectively, which were obtained by successively adding the control variables Open, Inf, and Gdp (log). Significantly positive interaction term coefficients indicated that the data factor flow facilitated the effect of the upstream CEVC on the PEVC, resulting in the SE. The upstream VC length expanded the demand of the production end through the total output induced by the unit final product; via the feedback and conduction of the VC, the demand for intermediate products grew, giving rise to a bullwhip effect (an economic term for amplifying demand within the supply chain—when demand shifts in international trade, the transmission of the value chain causes a remarkable shift in intermediate product demand) in trade. The increasing returns to scale created by demand multiplication promoted a concentration from the PEVC to the CEVC, thereby forming the SE. When digital technology is integrated into the R&D link of the VC, it is primarily utilized for model design or proprietary equipment development. This will accelerate the R&D application process and decrease the time required to deliver products to consumers. In a large market, close proximity to the CEVC will augment this advantage. The traditional VC, dominated by commodity quality and quantity, has shifted in terms of time and location. Value increases in design, brand, consumer interaction, and other areas. This bolsters the requirement for consumer information management, thereby forming an SE based on geographical proximity and time preference. Hence, the validity of the hypothesis was verified.

The conclusions described above show that, under the data factor flow, the CEVC did not change the SE on the PEVC due to the different structures of the upstream and the downstream, and the bidirectional drive of the upstream and downstream CEVC formed a strong SE.

4.2. Grouping Regression of the Siphon Effect

In terms of the data factor flow, the scale economy was prominent, and it was related to the economy size. The economy sizes of RCEP member countries vary widely. Additionally, the digital economy centered on the data factor is essentially a knowledge economy, and industries with distinct factor intensities have different knowledge densities. Thereby, the characteristics of different economic sizes and industrial factor densities will affect the data factor flow; furthermore, the data factor flow determines the effect of the CEVC on the PEVC and affects the formation of the SE. Therefore, in this paper, we conducted group regression according to a country’s economic size and its industries’ factor intensity in order to discover the heterogeneity characteristics that affect the SE.

At the national level, in terms of economic size, China, Japan, and South Korea are the main countries with large economies in the RCEP. At the industrial level, according to the Organization for Economic Co-operation and Development’s (OECD’s) classification of industries by factor intensity, in this paper, we classified industries as technology-, capital-, or labor-intensive.

4.2.1. Grouping Regression Results and Discussion of the Siphon Effect

First, the SE grouping regression of the data researcher flow was investigated. Columns (1) and (2) of the upper portion of

Table 8. The grouping regression of the SE.

The Grouping Test of Dpe	Country-Economy Size		Industry-Factor Intensity
	Large	Small	Technology	Capital	Labor
	(1)	(2)	(3)	(4)	(5)
Dpe	−0.6504	−1.3890 ***	3.9002 ***	−3.3469 **	0.2187
	(2.2818)	(0.4378)	(0.7785)	(1.6852)	(2.2809)
Invc_l_d	0.9001 ***	0.0424	−0.0527	−1.5680 **	−0.0167
	(0.0923)	(0.0825)	(0.0360)	(0.7872)	(0.4058)
Dpe × Invc_l_d	0.2261	0.9965 ***	−2.5977 ***	2.0964 **	−0.0943
	(0.1976)	(0.2294)	(0.5212)	(1.0518)	(1.3970)
Control variables	Yes	Yes	Yes	Yes	Yes
Country-fixed, year-fixed, andindustry-fixed effects	Yes	Yes	Yes	Yes	Yes
Observations	735	875	184	276	184
R2	0.5269	0.1999	0.6376	0.1463	0.0576
The grouping test of Dca	Country-Economy Size		Industry-Factor intensity
	Large	Small	Technology	Capital	Labor
	(1)	(2)	(3)	(4)	(5)
Dca	−3.7270 ***	−1.4005 ***	0.1556 ***	−0.0961	−0.0137
	(0.7483)	(0.2951)	(0.0531)	(0.0595)	(0.0918)
Invc_l_d	0.7687 ***	0.1073	−0.0499	−1.0249	−0.0260
	(0.1065)	(0.0805)	(0.0374)	(0.6674)	(0.3926)
Dca×Invc_l_d	0.0406 **	0.0283	−1.8339 ***	1.1696 **	0.4236
	(0.0193)	(0.0220)	(0.5425)	(0.5009)	(0.5509)
Control variables	Yes	Yes	Yes	Yes	Yes
Country-fixed, year-fixed, and industry-fixed effects	Yes	Yes	Yes	Yes	Yes
Observations	735	875	184	276	184
R2	0.5428	0.2041	0.6095	0.1516	0.0609

Notes: Countries with large economies include China, Japan, and South Korea. Countries with small economies include Australia and the ASEAN 6 (Malaysia, Singapore, Indonesia, Thailand, the Philippines, and Vietnam). According to the industry classification description of ADB-MRIO2018 in UIBE_GVC, technology-intensive industries in this paper include c9—chemical and chemical products; c13—machinery nec; c14—electrical and optical equipment; and c15—transport equipment. Capital-intensive industries include c3—food, beverages, and tobacco; c7—pulp, paper, printing, and publishing; c8—coke, refined petroleum, and nuclear fuel; c10—rubber and plastics; c11—other non-metallic minerals; and c12—basic metals and fabricated metal. Labor intensive industries include c4—textiles and textile products; c5—leather, leather goods, and footwear; c6—wood and products of wood and cork; and c16—other manufacturing. Notes: *** and ** represent significance at the 1 and 5 percent levels, respectively.

are grouped by country. The results show that the interaction term coefficient of countries with large economies was not statistically significant, while that of countries with small economies was significantly positive. This implies that the data researcher flow promotes the effect of the CEVC on the PEVC, forming the SE for countries with small economies. A comparison of small economies’ economic development levels and cultural institutions revealed substantial demand disparities. Differentiated demands encouraged diversified new product R&D; enhanced the data researcher flow; precisely matched supply and demand; more effectively realized differentiated demand and scope economy unification; and fostered the expansion of production enterprises, resulting in the SE. Columns (3)–(5) are grouped by industry. The results indicate that the interaction term coefficient of technology-intensive industries was significantly negative, that of capital-intensive industries was significantly positive, and that of labor-intensive industries was insignificant. Data researchers in technology-intensive industries are the owners of core technologies, and the flow of these technologies is strictly regulated, which hinders the effect of knowledge diffusion and impedes SE. In capital- and labor-intensive industries, the data researcher flow provides technical assistance for capital to replace labor. An increase in consumption will strengthen the growth of capital-intensive industries while gradually weakening the latter. Therefore, the data researcher flow promotes the SE in capital-intensive industries but fails to form the SE in labor-intensive industries.

Second, the SE grouping regression of data capital flow was investigated. Columns (1) and (2) in the lower part of Table 8 are grouped by country. The results show that the interaction term coefficient of countries with large economies was significantly positive, whereas that of countries with small economies was insignificant. The apparent market demand of large-economy nations was consistent with the scale economy of data capital flow, promoting the SE. Columns (3)–(5) were grouped by industry. The results indicated that the interaction term coefficient of technology-intensive industries was significantly negative, that of capital-intensive industries was significantly positive, and that of labor-intensive industries was insignificant. As technology-intensive industries involve national core technologies, subjects with core technologies implement stringent technology protection measures to prevent the leakage of core technologies, thereby restricting the SE. Compared with labor-intensive industries, the digitization of capital-intensive industries causes the data capital factor to replace the labor factor. Automation and artificial intelligence (AI) are replacing manual labor, increasing the precision of manufacturing processes and facilitating the SE.

4.2.2. Grouping Regression Results and Discussion of the Siphon Effect Upstream and Downstream

In the downstream CEVC (Invc_l_f_d), the SE grouping regression of the data researcher flow was initially implemented. Columns (1) and (2) of the upper part of

Table 9. The grouping regression of the SE: Invc_l_f_d.

The Grouping Test of Dpe	Country-Economy Size		Industry-Factor Intensity
	Large	Small	Technology	Capital	Labor
	(1)	(2)	(3)	(4)	(5)
Dpe	−0.5673	−0.6239	2.3989 ***	0.0029	−1.3812
	(0.3404)	(0.4013)	(0.5693)	(1.2930)	(1.2985)
Invc_l_f_d	1.6135 *	0.0052	−0.0527	−2.7264 **	−0.4997
	(0.5482)	(0.0862)	(0.0366)	(1.1926)	(0.5844)
Dpe × Invc_l_f_d	0.1209	0.9707 **	−3.0228 ***	0.0247	1.7284
	(0.2601)	(0.3868)	(0.7185)	(1.6040)	(1.4856)
Control variables	Yes	Yes	Yes	Yes	Yes
Country-fixed, year-fixed, and industry-fixed effects	Yes	Yes	Yes	Yes	Yes
Observations	735	875	184	276	184
R2	0.5156	0.1849	0.6223	0.1593	0.0666
The grouping test of Dca	Country-Economy Size		Industry-Factor Intensity
	Large	Small	Technology	Capital	Labor
	(1)	(2)	(3)	(4)	(5)
Dca	−3.8438 ***	−3.0055 **	0.4371 ***	−0.0612	−0.0090
	(0.7637)	(1.3729)	(0.0894)	(0.0594)	(0.0918)
Invc_l_f_d	1.2794 ***	0.0176	−0.0531	−2.9644 ***	−0.2699
	(0.2060)	(0.0850)	(0.0359)	(0.9159)	(0.5466)
Dca×Invc_l_f_d	0.0892 **	2.0563	−0.4865 ***	1.4814	0.8020
	(0.0382)	(1.7148)	(0.0972)	(0.9799)	(1.0578)
Control variables	Yes	Yes	Yes	Yes	Yes
Country-fixed, year-fixed, and industry-fixed effects	Yes	Yes	Yes	Yes	Yes
Observations	735	875	184	276	184
R2	0.5328	0.2016	0.6378	0.1671	0.0620

Notes: ***, **, and * represent significance at the 1, 5, and 10 percent levels, respectively.

are grouped by country. The results show that the interaction term coefficient of countries with large economies was insignificant, while that of countries with small economies was significantly positive. Downstream, countries with small economies had diverse intermediate goods production technology levels, resulting in various production demands. The data researcher flow will promote the unification of customized production and differentiated demand through technical support, as well as improve the scope economy of production, thus promoting the SE. Columns (3)–(5) are grouped by industry. The results show that the interaction term coefficient of technology-intensive industries was significantly negative, while that of capital- and labor-intensive industries was insignificant. Data researchers in technology-intensive industries possess core technologies and are subject to stringent government regulations. The greater the demand for downstream consumption ends, the more constrained the flow of data researchers in these industries, thereby limiting the SE produced by the knowledge spillover of the data researcher flow. In capital-intensive industries, generally, the research environment downstream of the VC cannot effectively stimulate a data researcher’s innovation potential and hinders improvements in the production efficiency. Consequently, the SE is not obvious. Due to the substitution of capital for labor, labor-intensive industries decline gradually, and the data researcher flow fails to effectively promote the SE.

Second, the SE grouping regression of data capital flow was conducted. Columns (1) and (2) in the lower part of Table 9 are grouped by country. The results indicate that the interaction term coefficient of countries with large economies was significantly positive, whereas that of countries with small economies was insignificant. The expansion of the downstream data capital flow will increase the total output induced by unit added value in countries with larger economies, thereby extending the production chain, promoting the effect of the CEVC on the PEVC, and forming the SE. Columns (3)–(5) are grouped by industry. The results indicated that the interaction term coefficient of technology-intensive industries was significantly negative, whereas it was not significant for capital- and labor-intensive industries. The data capital of technology-intensive industries was composed of numerous core technologies and was subject to extensive state intervention. The expansion of the downstream data capital flow will increase the government’s sway over the sector, thereby hindering the SE. In capital-intensive industries, the expansion of downstream data capital flows can drive production growth. However, because the industry is in the early stages of digitalization, the promotion effect of the CEVC on the PEVC has not played a significant role, and the SE is insignificant. Due to the massive substitution of capital for labor factors, labor-intensive industries gradually lose their competitive edge. Consequently, the SE is inconsequential.

Within the upstream CEVC (Invc_l_b_d), the SE grouping regression of the data researcher flow was initially conducted. Columns (1) and (2) of the upper part of

Table 10. The grouping regression of the SE: Invc_l_b_d.

The Grouping Test of Dpe	Country-Economy Size		Industry-Factor Intensity
	Large	Small	Technology	Capital	Labor
	(1)	(2)	(3)	(4)	(5)
Dpe	−0.4176	−0.9694 **	2.5610 **	−3.5166 ***	7.8108 ***
	(0.5126)	(0.4864)	(1.2744)	(1.1281)	(2.7805)
Invc_l_b_d	1.8012 *	1.6026 ***	−3.6839 *	0.1420	1.3029
	(0.4890)	(0.4786)	(1.9387)	(1.1525)	(1.1090)
Dpe × Invc_l_b_d	0.4861	1.5581 ***	−3.5216 *	4.3925 ***	−10.0470 ***
	(0.9395)	(0.5432)	(1.7999)	(1.3990)	(3.5779)
Control variables	Yes	Yes	Yes	Yes	Yes
Country-fixed, year-fixed, and industry-fixed effects	Yes	Yes	Yes	Yes	Yes
Observations	735	875	184	276	184
R2	0.5286	0.2266	0.6313	0.1718	0.1051
The grouping test of Dca	Country-Economy Size		Industry-Factor Intensity
	Large	Small	Technology	Capital	Labor
	(1)	(2)	(3)	(4)	(5)
Dca	−3.8105 ***	−2.0807	−0.7480	−0.1105 *	−0.0249
	(0.7348)	(1.8087)	(0.8227)	(0.0583)	(0.0914)
Invc_l_b_d	1.5564 ***	2.1647 ***	−6.4387 ***	0.9899	0.9242
	(0.2081)	(0.4555)	(1.4554)	(1.0706)	(1.1322)
Dca×Invc_l_b_d	0.0854 **	1.0946	0.0745	2.6912 ***	1.0146
	(0.0386)	(2.2994)	(0.0631)	(0.9774)	(1.1484)
Control variables	Yes	Yes	Yes	Yes	Yes
Country-fixed, year-fixed, and industry-fixed effects	Yes	Yes	Yes	Yes	Yes
Observations	735	875	184	276	184
R2	0.5455	0.2350	0.6251	0.1650	0.0651

Notes: ***, **, and * represent significance at the 1, 5, and 10 percent levels, respectively.

are grouped by country. The results show that the interaction term coefficient of countries with large economies was insignificant, while that of countries with small economies was significantly positive. Compared to large economies, the increased flow of data researchers provided robust technical support for production and satisfied diverse demands more effectively. Therefore, the flow of data researchers facilitated the SE. Columns (3)–(5) are grouped by industry. According to the results, the interaction term coefficient of technology-intensive industries was significantly negative. At the same time, it was significantly positive for capital-intensive industries and significantly negative for labor-intensive industries. As data researchers in technology-intensive industries control core technologies, the flow of data researchers is strictly regulated, which impedes the SE generated by knowledge spillover. In addition, the labor factor has been largely replaced by data capital. Therefore, the expansion of the data researcher flow encourages the SE in capital-intensive industries, but this expansion discourages it in labor-intensive industries.

Second, SE grouping regression of the data capital flow was investigated. Columns (1) and (2) of the lower part of Table 10 are grouped by country. The results indicate that the interaction term coefficient of countries with large economies was significantly positive, whereas that of countries with small economies was insignificant. Compared to countries with small economies, upstream consumption end growth induced unit final goods in order to draw more total output from countries with large economies. Data capital flow advanced the SE. Columns (3)–(5) are grouped by industry. The results show that the interaction term coefficient of technology-intensive industries was insignificant, that of capital-intensive industries was significantly positive, and that of labor-intensive industries was insignificant. For technology-intensive industries, technology diffusion was restricted to a specific range that retained key technologies and ensured its own production needs, thereby forming technology diffusion with locked-down core technologies. Consequently, the SE was not obvious. Due to the digital input of capital-intensive industries, productivity has been significantly improved; thus, the SE is significant. Labor-intensive industries gradually shrink due to the substitution of capital for the labor factor, and thus, the SE is not obvious.

4.3. Discussion of the Limitations

In this study, we empirically examined, in terms of the “materialization” of the data factor flow carrier, the effect of the CEVC on the PEVC, that is, whether the data factor flow promotes the SE of the CEVC, thus attracting the PEVC. The results of this study assist us in gaining a comprehensive understanding of the impact of the data factor flow on reshaping the characteristics of the RCEP region’s VC, elucidating the driving force of the RCEP region’s VC, resolving the issues related to the traditional development mode, and promoting the sustainable development of the RCEP region’s economy. However, this study still has the following limitations:

(i)

Regarding data selection, the VC data in this paper were limited to the industrial level due to data availability. In the future, with improvements in data availability, data can be detailed at the enterprise level to describe the facts and characteristics ignored by the PEVC from a more nuanced standpoint.

(ii)

Heterogeneity analysis should not be limited to the characteristics of countries and industries, but it should be expanded to include additional dimensions. For instance, differences in countries’ openness and industrial structures will influence the relationship between data factor flow and SE in the RCEP region, thereby enhancing and expanding the study in this regard.

(iii)

Regarding the research method, as the number of measurement methods increases, more measurement methods that reflect SE can be selected to further analyze the correlation between the data factor flow and SE in the RCEP region.

5. Conclusions

Previous research has highlighted the impact of conventional production factors on the VC. In this paper, we further deconstructed the data factor into data capital and data researchers by sorting out the “materialization” evolution trend of data factor extension. Furthermore, we interpreted the SE of the CEVC attracting the PEVC by investigating its liquidity, which will aid in our understanding of the cause and characteristics of the current reshaping of the VC.

In terms of quantitative indicators, in existing studies, the analysis of the VC’s characteristics is frequently based on a unilateral analysis of VC data, ignoring the description of its characteristics. In this paper, we comprehensively applied VC data and the agglomeration method to create a PEVC measurement indicator. This indicator incorporates the VC and its characteristics. Additionally, a measurement indicator for the CEVC was developed. It consists of the pure domestic VC length introversion and upstream and downstream pure domestic VC length introversion. Thus, it aided in a thorough investigation into the structural difference of the CEVC, yielding reasonable and detailed research conclusions for the empirical analysis of the SE.

In the mechanism analysis, the following hypothesis was proposed:

Under the data factor flow, at the PEVC, the law of increasing returns to scale reconciles a scale economy and production differentiation. At the CEVC, the data factor establishes the consumer’s dominant position and encourages production differentiation through intensified competition. As a result, the CEVC becomes a powerful driving force to attract the PEVC, promoting production agglomeration in the region on a large market scale and forming the SE to create more value.

In this study, we conducted empirical research into the role of the data factor flow in the impact of the CEVC on the PEVC in the functional integration stage of the RCEP region’s VC, and we tested our hypothesis.

In general, via the “materialization” carriers (data researcher flow and data capital flow), the data factor flow promoted the effect of the CEVC on the PEVC, thereby forming the SE, and thus the hypothesis was confirmed. According to the robustness test and endogeneity discussion, we determined the reliability of the conclusion. Furthermore, the data factor flow drove the effect of the downstream and upstream CEVCs on the PEVC, which indicated that the SE was unaffected by the upstream and downstream structures of VC, driving the SE of the VC in the RCEP region at two levels. The validity of the hypothesis was verified.

There are obvious differences at the national and industrial levels. At the national level, countries with large economies exhibited SEs in data capital flows based on scale economies. In contrast, countries with small economies exhibited SEs in data researcher flows based on scope economies. Countries with large economies have larger market sizes and scale economies that are closely related to data capital. Countries with small economies are quite different in various aspects, and it is easy to generate diversified needs, increase the new product R&D, and increase the data researcher flow to achieve the unity of differentiated needs and the scope economy more effectively.

At the industrial level, technology-intensive industries mainly restrain the SE as core technologies are controlled. However, in the upstream CEVC, technology diffusion is restricted to a specific range that retains key technologies and ensures production needs, forming technology diffusion with locked-down core technologies. Therefore, the SE in the data capital flow is not obvious. In capital-intensive industries, due to the fusion of digital technology and traditional industry fostered by the data factor flow, the SE is significant in most cases. Only under the downstream CEVC, because of the general lack of a good research environment and the immaturity of digitalization, is the SE insignificant. In labor-intensive industries, the substitution of capital for labor caused by digital input weakens or even inhibits the SE. This inhibitory effect is more pronounced in the data researcher flow under the upstream CEVC.

The findings mentioned above have the following policy implications: First, the current RCEP agreement, which includes tariff concessions, only represents the initial phase of regional integration. Future RCEP members should follow the trend of digital trade and actively participate in cooperative agreements for the data factor flow. Specialized agreements such as the Digital Economy Partnership Agreement (DEPA) will form a new trend in the development of digital trade rules [63]. Singapore has signed the most digital trade rules in Asia, and China also formally applied to join the DEPA to reduce digital trade barriers in November 2021. In the future, China should actively connect with regional and bilateral high-level digital trade rules, seizing the opportunity to apply for DEPA membership [64]. This will undoubtedly contribute to the consolidation and cooperation of the VC within the region, the reduction in excessive dependence on external demand, and the promotion of regional sustainable development.

Second, as countries with large economies benefit greatly from the data capital flow, restricting the cross-border data factor flow will undermine business competitiveness and economic efficiency [65], and data localization measures will increase the cost of data usage and diminish the benefits of digital trade [66]. Therefore, China, Japan, and South Korea should promote the construction of the China–Japan–South Korea Free Trade Agreement (FTA) under the RCEP framework; reconstruct a more resilient and dynamic VC system between the three countries and throughout Asia; facilitate the steady development of the RCEP region; and propel the better alignment of China’s “One Belt, One Road” initiative, Japan’s “High Quality Infrastructure Partnership Program”, and South Korea’s “New Southern Policy” within the RCEP framework [67]. Thus, the scale economy’s advantage in terms of the data factor flow can be exerted to strengthen the SE. For countries with small economies, R&D input must be further strengthened to enable data researcher aggregation, develop differentiated products, and strengthen the scope economy, thereby forming an SE. Moreover, it is necessary to encourage cooperation in the field of high-tech industries, strive to reach a consensus regarding the principle of fairness in the technology market within the RCEP framework, strengthen the digital input of capital-intensive industries, increase the education and training of information personnel, propel the digital participation of labor-intensive industries, and improve the efficiency of resource allocation.