Deconstructing the Digital Economy: A New Measurement Framework for Sustainability Research

Abstract

Empirical research on the impact of the digital economy on sustainable development is hampered by severe methodological challenges. Discrepancies in the theoretical foundations and construction logic of measurement frameworks have led to diverse and often conflicting conclusions, hindering the systematic accumulation of knowledge. This study aims to address this critical gap by proposing a new, logically consistent measurement framework. To overcome the existing limitations, we construct a functional deconstruction framework grounded in General-Purpose Technology (GPT) theory and a “stock–flow” perspective. This framework deconstructs the digital economy into a neutral “digital infrastructure” (stock platform) and two forces reflecting its inherent duality: a “consumption force” (digital industrialization) and an “empowerment force” (industrial digitalization). Based on this, we develop a measurement system adhering to the principle of “logical purity” and apply a “two-step entropy weighting method with annual standardization” to assess 30 provinces in China from 2012 to 2023. Our analysis reveals a multi-scalar evolution. At the micro level, we identified four distinct provincial development models and three evolutionary paths. At the macro level, we found that the overall inter-provincial disparity followed an inverted U-shaped trajectory, with the core contradiction shifting from an “access gap” to a more profound “application gap.” Furthermore, the primary driver of this disparity has transitioned from being “empowerment-led” to a new phase of a “dual-force rebalancing.” The main contribution of this study is the provision of a new analytical tool that enables a paradigm shift from “aggregate assessment” to “structural diagnosis.” By deconstructing the digital economy, our framework allows for the identification of internal structural imbalances and provides a more robust and nuanced foundation for future causal inference studies and evidence-based policymaking in the field of digital sustainability

1. Introduction

Globally, a profound transformation driven by digital technology is reshaping the economic and social landscape. As the world’s second-largest economy undergoing a rapid and large-scale digital transformation, China presents a unique and critical “natural laboratory” for examining these dynamics. The profound regional imbalances within China further offer an ideal setting to study how digitalization interacts with pre-existing inequalities, a question of global relevance. As the core vehicle of this transformation, the digital economy has rapidly emerged as a critical frontier research topic in economics, management, and related social sciences. The breadth and depth of its impact are often compared to the historical revolutions of the steam engine and electricity. Particularly under the global consensus on achieving sustainable development, as encapsulated in the Sustainable Development Goals (SDGs), exploring the complex relationship between the digital economy and green transition has become one of the most active and controversial research focuses in the field. Understanding the dualistic nature of this relationship is critical for harnessing the potential of digitalization to contribute to goals like SDG 9 (Industry, Innovation, and Infrastructure) and SDG 12 (Responsible Consumption and Production), while mitigating its negative externalities.

However, behind this burgeoning area of research, a fundamental methodological obstacle is becoming increasingly prominent, severely eroding the very foundation of knowledge accumulation in this field. This obstacle is centrally manifested in the lack of a recognized and logically consistent measurement framework for the core explanatory variable—the “digital economy” itself. It is precisely this lack of a unified “ruler” that has directly led to diverse and even systematically contradictory research conclusions. On the issue of sustainable development, the literature presents co-existing but seemingly conflicting findings, including linear promotion [1,2,3], linear inhibition [4,5,6,7], non-linear relationships [8,9], and even conditional effects [10,11]. This “jungle of conclusions” makes it difficult to ascertain whether these discrepancies reflect the objective complexity of economic reality or are merely artifacts of differing measurement methods.

A deeper analysis reveals that the current methodological predicament stems from a systemic chain of problems that cascades from the conceptual to the operational level. The root cause lies in conceptual ambiguity: a global consensus on the definition of the “digital economy” has yet to be formed, with its boundaries exhibiting significant elasticity between the “core ICT industry” and the “entire digitally touched national economy.” This foundational uncertainty inevitably propagates to the meso level, manifesting as a lack of unified theoretical guidance in the top-level design of measurement frameworks, where dimensional divisions are often based on empirical induction rather than rigorous theoretical deduction. Ultimately, this problem chain is reflected at the micro-operational level in the paradoxical selection of indicators, such as conflating infrastructure as an input with economic activities as an output, or internalizing the “development environment,” which should be treated as an external factor.

This interlocking chain of problems makes clarifying the measurement issue an urgent priority. Therefore, a deeper and more pressing question to be answered is the following: how can we transcend the current unsystematic state of measurement to construct an analytical framework that can clearly reveal the internal structure of the digital economy and its impact on sustainable development? Clarifying this core methodological question is a key step in advancing research in this field from “conclusion divergence” to “knowledge accumulation.”

The core objective of this study is to directly address the root of this predicament by committing to the construction of a new digital economy measurement framework with a solid theoretical foundation and clear internal logic. To achieve this, we follow a deductive path originating from first principles. We first define the essence of the digital economy as a profound technological revolution and identify General-Purpose Technology (GPT) theory [12,13] as the most appropriate theoretical lens for explaining its regularities. The core mechanisms of GPT theory—“innovation complementarity” and “adjustment costs”—profoundly reveal that any technological revolution inevitably entails a duality of efficiency enhancement and cost consumption. It is precisely this theoretical insight that provides us with a theoretical starting point to move beyond aggregate assessment and toward structural diagnosis.

Building on this theoretical cornerstone and integrating the “stock–flow” analytical logic, we propose the core contribution of this paper—a “functional deconstruction” framework for sustainability-oriented research. This framework operationalizes the deconstruction of the digital economy into three functional dimensions: “digital infrastructure” (platform) as a neutral capital stock, and two dynamic flows that embody its inherent duality—the “consumption force” (digital industrialization), representing GPT adjustment costs and environmental burdens, and the “empowerment force” (industrial digitalization), reflecting GPT’s enabling effects and green dividends. By providing a “calibrated ruler” for the field, this framework aims to offer a reliable methodological foundation for clarifying existing controversies. In doing so, this study seeks to contribute not only to the empirical application of GPT theory but also to the broader academic discourse on regional inequality and sustainability transitions in the digital age.

The remainder of this paper is organized to systematically unfold this research agenda, as illustrated in the flowchart in Figure 1. Section 2 provides a literature review on the current state of measurement. Section 3 presents our theoretical framework and the design of the indicator system. Section 4 discusses the methods and data used. Section 5 presents the empirical results of our multi-scalar analysis. Section 6 discusses the policy implications, China’s case in a global context, and limitations of the study. Finally, Section 7 concludes the paper.

2. Literature Review

Conducting scientific and consistent measurements of the digital economy is a core methodological challenge currently facing relevant research fields. This challenge is prevalent worldwide, rooted in the lack of consensus on the definition of core concepts (Oloyede et al., 2023) [14], which directly leads to the absence of a unified paradigm for constructing subsequent measurement frameworks. This inconsistency in measurement methods is considered one of the key reasons for the highly divergent conclusions of subsequent empirical studies. This is particularly true when exploring complex relationships such as those between the digital economy and sustainable development, where different measurement “yardsticks” are likely to point to entirely different results. A large body of empirical literature on China provides rich examples of this phenomenon. The proliferation of these studies has been accompanied by a diversification and lack of standardization in measurement methods, making it a typical field where current methodological shortcomings are exposed. Therefore, this section will systematically review existing measurement practices from three dimensions—conceptual definition, theoretical basis, and indicator construction—to provide a critical commentary, with the aim of revealing the deep methodological roots of the current research dilemma.

2.1. Conceptual Controversies and Ambiguous Boundaries

The root of all measurement difficulties lies in the definition of the core concept of the “digital economy,” which remains highly contested globally. This foundational uncertainty is the starting point for all subsequent measurement problems. Surveying its evolution, as illustrated in Figure 2, the definition of the digital economy has roughly followed a clear evolutionary path of expansion from the core ICT sector towards the entire national economy.

The starting point of this evolutionary path is the “narrow-scope” definition, centered on the core ICT sector. Early definitions were closely tied to the Information and Communication Technology (ICT) industry, emphasizing its role as a vehicle for the retail of tangible and intangible goods [15].

With the permeation of digital technologies, the definition evolved toward a “medium-scope” conceptualization, framed around the broad digital sector. A representative example is the more structured, three-tiered framework proposed by the U.S. Bureau of Economic Analysis (BEA), which encompassed digital-enabling infrastructure, e-commerce, and digital media [16], providing a crucial foundation for subsequent accounting.

However, the recent trend has been to dramatically expand the boundaries of the digital economy, ultimately leading to the current “broad-scope” definition, which substantively views the digital economy as a digitally integrated economy. This perspective emphasizes its pervasiveness and integration as a General-Purpose Technology (GPT). For instance, some scholars view it as the integration of multiple GPTs with all socio-economic activities [17], or emphasize its “hyperconnective” nature [18]. The latest OECD definition takes this to an extreme, considering the digital economy to encompass all economic activities that are significantly enhanced by digital inputs (including technology, infrastructure, services, and data) [19]. This broad, integration-focused perspective is also more explicitly and systematically reflected in China’s mainstream definitions. For example, the National Bureau of Statistics (NBS) of China defines it as “a series of economic activities with data resources as the key element, information networks as the carrier, and ICT technology as the driving force” [20]. The China Academy of Information and Communications Technology further emphasizes that the core of the digital economy lies in the “deep integration of digital technology with the real economy,” with the goal of “accelerating the restructuring of economic development and governance models” [21]. This emphasis on “integration” has also been echoed in academic research, such as Wang Jun et al. [22], who advocate for taking infrastructure as the “ballast stone” and focusing on the deep integration of industrial digitalization.

Thus, it is evident that in both international trends and Chinese practice, the definition of the digital economy possesses significant elasticity, stretching from the core ICT industry (narrow scope), through the broad digital sector (medium scope), to nearly the entire digitally touched national economy (broad scope). This conceptual uncertainty blurs the line between the traditional and digital economies, creating a natural “fuzzy boundary.” It is this fuzzy boundary that sets the stage for the inconsistencies in theoretical foundations, dimensional divisions, and indicator selection in subsequent measurement frameworks.

2.2. Lack of Theory-Driven Dimension Design

The ambiguity of concepts at the source inevitably leads to a lack of theoretical foundation for the measurement framework at the top-level design stage. A logically rigorous measurement framework should divide its first-level dimensions based on a clear economic theory to clarify the internal relationships between the dimensions. However, a review of the existing literature reveals that the design of the dimensions in most measurement frameworks is not derived from theoretical reasoning, but rather from empirical induction or a simple combination of specific objectives.

International organizations have made some pioneering explorations in this area. For example, the U.S. Bureau of Economic Analysis (BEA) has developed a three-dimensional accounting framework within its “supply-use” framework, which includes “digital enabling infrastructure,” “e-commerce,” and “digital media content” [16]. While this framework is highly operational at the official statistical level, its theoretical foundation is relatively simplified and fails to deeply reveal the dynamic relationships between different functional components of the digital economy. As a result, it has not been widely adopted or applied in academic research.

More international frameworks, such as the G20’s Digital Economy Assessment Toolkit, the EU’s Digital Economy and Society Index (DESI), and the Network Readiness Index (NRI), adopt a “multi-objective checklist” design approach [23,24,25]. They typically list multiple objectives such as “infrastructure,” “human capital,” “social empowerment,” “technology adoption,” and “governance” as first-level dimensions based on policy concerns. The advantage of this design lies in its comprehensive coverage, but its fundamental flaw is that these dimensions are treated as parallel relationships rather than causal ones. Why these dimensions and not others? What is the underlying theoretical logic behind their division? These frameworks fail to provide clear answers to these fundamental questions, rendering them more akin to checklists for policy evaluation than theoretically coherent frameworks with internal logical consistency.

This “theoretical vacuum” is particularly pronounced in academic research with China as the application context, giving rise to diverse dimension design schemes. Due to the lack of a unified theoretical foundation, scholars often propose different measurement frameworks based on their own understanding of the digital economy and the availability of data, drawing on existing research. These frameworks vary in the number of dimensions they include, ranging from one to four, and their internal structures exhibit significant heterogeneity. Within the four-dimensional framework, the combination of dimensions is highly variable. For example, some studies combine “infrastructure, talent investment, industrial development, and inclusive finance” [26], while others replace “talent investment” with “innovation capacity” [27], or adopt entirely different dimensions, such as “infrastructure, digital application, digital innovation, and digital benefits” [3], or the more classic combination of “infrastructure, digital industrialization, industrial digitization, and development environment” [22]. Meanwhile, there are also more streamlined three-dimensional frameworks that typically center around the consensus dimension of “infrastructure,” paired with “industry scale” and “user scale” [28] or “R&D investment” [29]. Additionally, some studies adopt a two-dimensional framework, focusing exclusively on “digital industrialization” and “industrial digitization,” which are considered core economic activities [21,30,31,32]. Currently, some scholars are still using single indicators to measure the level of digital economic development, such as digital infrastructure [33] or the digital finance index [34].

This high degree of diversity in dimensional design clearly exposes two serious consequences: First, the arbitrariness of dimension establishment. Researchers can arbitrarily add, remove, or reorganize dimensions based on their preferences, causing the framework to lose stability and replicability. Second, the conclusions drawn from research become incomparable. Different dimension designs essentially represent different “slices” and “projections” of the complex phenomenon of the “digital economy.” When different studies even differ in the angles of their “projections,” the comprehensive indices they ultimately derive naturally become “same in name but different in substance,” rendering any cross-study comparisons based on these indices meaningless.

2.3. Heterogeneity and Logical Paradoxes in Indicator Selection

2.3.1. Significant Heterogeneity in the Selection of Secondary Indicators

Due to the lack of a unified theoretical framework to guide them, researchers often use different standards when selecting proxy variables for the same conceptual dimension, relying primarily on data availability rather than theoretical rigor. Take the “digital infrastructure” dimension, which is widely recognized in academia, as an example. Different studies show systematic differences in their measurement focus and conceptual boundaries. Upon closer examination, existing measurement practices generally reflect an evolutionary path from simple to complex and from single-dimensional to multi-dimensional, while also exposing serious logical confusion.

Specifically, first, there are significant differences in the breadth and depth of the dimensions used to measure digital infrastructure. The measurement system has gradually expanded from the most basic user-end access indicators to network virtual resources and core physical hardware. For example, the system proposed by Huang, Jie et al. is the most concise, containing only three indicators: internet broadband access port density, internet penetration rate, and mobile phone penetration rate [35]. Pan et al. built upon this framework by adding “CN domain name count” and “IPv4 count,” extending the measurement scope to the network virtual resource layer [36]. In contrast, Wang et al. and Huang and Lin developed more comprehensive frameworks that not only cover access and virtual resources (such as domain names and web pages) but also include key physical hardware indicators such as “length of long-haul optical fiber core lines” and “mobile communication switch capacity,” thereby providing a more comprehensive characterization of infrastructure carrying capacity [37,38]. Second, there are intrinsic structural differences in digital infrastructure indicator systems. Faced with increasingly complex indicator sets, some studies have attempted to enhance their structural integrity by constructing internal dimensions. For example, Wei et al. categorized indicators into two major categories: “internet communication” and “internet infrastructure” [39]; Wang et al. divided them into “internet development level” and “growth-oriented infrastructure” [37]; while Huang and Lin adopted a classification based on three technological modalities: “internet,” “mobile phones,” and “television broadcasting” [38]. While these attempts reflect scholars’ pursuit of structure, the classification criteria themselves remain inconsistent, further highlighting the absence of theoretical guidance. Third, a more serious issue lies in the blurred or even breached conceptual boundaries. The most typical example is the indicator system proposed by Hao et al., which includes both hardware indicators such as fiber-optic length and indicators like “total telecommunications services volume” and “value added of the tertiary industry” [40]. This confuses infrastructure as an input factor with the economic output it generates, resulting in “causal inversion” and conceptual contamination. This practice of blurring the distinction between infrastructure and the economic activities it supports fundamentally undermines the validity of measurement indicators.

This methodological divergence and confusion is by no means a manifestation of healthy academic diversity, but rather a symptom of a deeper conceptual crisis. It renders cross-research synthesis and meta-analysis nearly impossible, creating an illusion of “knowledge accumulation”—the volume of literature is growing, but consensus on basic facts remains elusive. More critically, when the validity and reliability of core explanatory variables cannot be guaranteed from the outset, any quantitative analysis based on them loses its scientific foundation. Causal inferences and policy recommendations derived from such ambiguous measurements are akin to a “house of cards built on sand,” not only severely misleading academic understanding but also potentially leading to erroneous policy practices. It is important to emphasize that this measurement dilemma is not unique to the “digital infrastructure” dimension. Similar issues of indicator misuse and conceptual ambiguity are equally prevalent in the measurement of other core dimensions such as “digital industrialization” and “industrial digitalization.” This makes the construction of a logically consistent and structurally clear new measurement framework all the more urgent.

2.3.2. Logical Paradox of Indicators

Beyond the heterogeneity in indicator choice, current measurement practices also face severe logical challenges, with many indicator systems exhibiting inherent structural flaws. As summarized and illustrated in

Table 1. Comparison of measurement practices regarding logical paradoxes.

Paradox Type	Flawed Practice (With Cited Examples)	Principled Approach (This Study)
Conflating Input, Process, and Output Indicators (An index simultaneously includes all three)	“fixed investment” (Input) and “information service revenue” (Output ) [22].	Strictly separates dimensions based on a causal chain: “Digital Infrastructure” (Capital Stock/Input), “Digital Industrialization,” and “Industrial Digitalization” (Economic Activities/Flows). Adheres to a results-oriented principle.
	“internet penetration rate” (Process) and “operating revenue” (Output) and “number of employees” (Input) [38].
	“broadband penetration” (Process) and “employment ratio” (Input) and “telecom services per capita” (Output) [41,42].
	“R&D personnel” (Input) and “value added of the tertiary industry” (Output) [36].
Internalizing External Factors (An index incorporates them as core components)	“digital economic development environment” [22].	Treats “environment,” “regulation,” and “outcomes” as external explanatory, control, or dependent variables, ensuring the logical purity of the digital economy index as an independent variable for causal inference.
	“social impact” and “economic growth” [37].
	“government regulation” [38].
	“digital governance” [43].

, these paradoxes are primarily manifested in two aspects.

The first is the conflation of input, process, and output indicators. Scientific measurement should clearly reflect a causal chain and distinguish between different stages of development. However, a common tendency in existing research is to synthesize, without distinction, input-based indicators that characterize the developmental foundation, process-based indicators that reflect behavioral aspects, and output-based indicators that measure developmental outcomes. As shown in Table 1, this practice blurs the line between “what has been done to develop the digital economy” (inputs and processes) and “what the digital economy has achieved” (outputs), resulting in a composite index that is logically inconsistent and ambiguous in its economic meaning.

The second, a more subtle yet equally critical issue, is the internalization of external factors. Some studies erroneously include indicators that should be treated as external “explanatory variables” (to analyze the impact of the digital economy) or “control variables” (to exclude confounding factors) in the core digital economy index itself (the independent variable). As the examples in Table 1 demonstrate, the essence of this practice is to confuse the “environment,” “outcomes,” or “governance measures” of the digital economy with its “core” activities. Methodologically, embedding these external or consequential dimensions within the index constitutes a circular argument of “defining the cause by its effect.” Such a design fundamentally weakens the logical foundation of any subsequent empirical analysis and severely limits the possibility and credibility of drawing causal inferences based on such indices.

In summary, current empirical research on the digital economy, especially when exploring its environmental effects, faces a systemic methodological bottleneck. This bottleneck stems from vague definitions of top-level concepts, is exacerbated by the lack of a middle-level theoretical framework, and ultimately manifests itself in the logical confusion of the selection of bottom-level indicators. This problem persists largely because of path dependence in academic research: many subsequent studies tend to incrementally modify existing indicator systems rather than carefully reconstruct them from first principles. This academic inertia has led to the inherent conceptual ambiguities and logical flaws in early studies being perpetuated and even amplified, ultimately forming a paradigm lock that is difficult to break in the current field and fundamentally eroding the scientific basis for the effective accumulation of knowledge. Therefore, to break out of the current research dilemma, it is necessary to achieve a paradigm shift from “aggregative” measurement to “deconstructive” measurement. This requires returning to the starting point of measurement, abandoning the path of patchwork, and instead constructing a new analytical framework that is more theoretically sound and logically pure. The rest of this paper is a direct response to this fundamental challenge.

The core academic contribution of this paper lies in its systematic reconstruction of the research paradigm on the environmental effects of the digital economy, achieving a fundamental leap from “aggregate assessment” to “structural diagnosis.” Its innovations are seamlessly integrated across three dimensions: theory, methodology, and empirical analysis. Theoretically, this study abandons the traditional perspective of treating the digital economy as a homogeneous whole. Based on the theory of General-Purpose Technology (GPT) and the “stock–flow” logic, it proposes for the first time a functional deconstruction framework that innovatively interprets the digital economy as neutral “digital infrastructure” (platforms), as well as the “enabling force” (industrial digitization) representing green dividends and the “consuming force” (digitalization of traditional industries) that triggers environmental burdens (digital industrialization), thereby revealing its inherent dualistic nature. To ensure the practical applicability of this theoretical framework, this study has developed a new measurement system based on the principle of “logical purity.” By strictly adhering to the separation principle of “input–process–output,” existing measurement practices have been systematically corrected. Ultimately, supported by this new theoretical and methodological framework, this study empirically reveals a series of structural facts obscured by traditional aggregate analysis: through a “bottom-up” diagnostic paradigm, this study not only identified four typical patterns and three evolutionary paths of provincial-level development in China for the first time but also confirmed that the “application gap” has replaced the “access gap” as the dominant contradiction driving regional imbalances, providing new, in-depth evidence for understanding the complex dynamics of China’s digital economy.

3. A New Framework for Functional Decomposition: Theoretical Basis and Indicator System Design

In response to the methodological challenges outlined above, this section aims to construct a more theoretically robust and logically pure analytical framework. Figure 3 provides a conceptual map of this framework. Its construction follows a rigorous deductive approach from macro theory to micro indicators: first, an operational definition for sustainability research is proposed; second, through functional deduction based on the theory of General-Purpose Technology (GPT), we derive three core measurement dimensions—“platform,” “consumption,” and “empowerment”—that are logically interconnected. Finally, under the overarching framework of these dimensions and adhering to principles such as logical purity, we systematically identify and structurally design specific indicators for digital infrastructure, digital industrialization, and industrial digitalization, ultimately establishing a comprehensive measurement system.

3.1. Operational Definition and Analytical Perspectives of the Digital Economy

The scientific measurement of the “digital economy” begins with a clear and operational definition of the concept. Existing definitions oscillate between a “narrow scope” (ICT industry) and a “broad scope” (the entire national economy), leading to the problem of “blurred boundaries” in measurement. To overcome this shortcoming and establish a logical starting point for subsequent functional decomposition, this study defines the digital economy as a complex ecosystem supported by a capital stock platform of “digital infrastructure” and composed of two core economic activities: “digital industrialization” and “industrial digitalization.”

This definition has three core characteristics: First, it adopts a “stock–flow” perspective. It clearly separates infrastructure as capital stock from economic activities as flow, addressing the logical flaw in existing research that confuses input indicators (such as infrastructure investment) with output indicators (such as business revenue), and laying the foundation for constructing a measurement system with clear causal chains. Second, it incorporates distinctions based on heterogeneity. It clearly distinguishes between two economically distinct activities—“digital industrialization” and “industrial digitalization”—thereby creating the possibility of penetrating the “black box” and exploring its internal structural tensions. Third, it follows a functional orientation. The classification criteria are not based on traditional industrial classifications but on the unique roles of each component within the economic system. This enables the definition to directly serve subsequent analyses of the structural impacts of the digital economy, particularly its complex role in sustainability issues.

3.2. Functional Decomposition and New Measurement Framework Based on GPT Theory

3.2.1. Functional Deconstruction Based on GPT Theory

The theoretical foundation of this study is the General-Purpose Technology (GPT) theory [12,44]. This theory transcends traditional growth models that treat technology as an exogenous shock, instead viewing it as an endogenous force capable of triggering sustained, widespread downstream innovation and reshaping economic structures. As the most profound GPT since the steam engine and electricity, digital technology’s impact is a complex structural transformation rather than a single growth driver. This provides the most appropriate theoretical lens for understanding the multifaceted, even contradictory, impacts of the digital economy on sustainable development issues.

To translate the general principles of GPT into an operational measurement framework, we have conducted a functional deduction focused on sustainability issues. Its core logic lies in the fact that any GPT revolution will inevitably give rise to two core dynamic forces of opposite natures. This stems from the functional translation of the two core mechanisms in GPT theory: “innovation complementarity” and “adjustment costs.”

The first is the “empowerment force,” which originates from the “innovation complementarity” of GPT. As an enabling technology, digital technology penetrates all sectors of the national economy, stimulating complementary innovations in downstream industries (such as smart manufacturing and precision agriculture) and systematically improving resource allocation efficiency and total factor productivity [45]. In a sustainability context, this force exerts a positive structural impact on the green transition of the economic system through mechanisms such as production optimization, reduction of mismatches, and promotion of clean technology diffusion. This theoretical mechanism is empirically validated by a growing body of literature finding that digitalization can enhance energy efficiency, reduce pollution, and promote green growth [1,3]. The core vehicle for this force is “industrial digitalization.”

The second is the “consumption force,” derived from GPT’s “adjustment costs” and the emergence of new sectors. The widespread adoption of GPT requires the construction of a vast new infrastructure network (data centers, 5G base stations) and the emergence of resource-intensive new core industries (ICT manufacturing). These activities are significant sources of energy and resource consumption and generate new environmental pressures (such as carbon emissions and electronic waste), a fact empirically supported by numerous sustainability studies that link ICT development to increased electricity consumption and environmental burdens [4,6]. Within a sustainability analysis framework, their functional role must be identified as the “consumption force” with negative externalities, with “digital industrialization” as its core carrier.

These two dynamic “flow” forces rely on a “stock” platform—“digital infrastructure.” In economic theory, infrastructure is conceptualized as a fundamental capital stock that serves as a neutral enabling platform for a wide range of economic activities, rather than being an activity itself. Seminal works have long established infrastructure’s role as a critical input that enhances the productivity of other factors with modern research extending this logic to telecommunications infrastructure [46,47]. Functionally, it is the common “field” or “multiplier” upon which both “empowerment” and “consumption” occur, with its quality and coverage regulating the conversion efficiency of the two forces.

Through this functional translation, we reconceptualize the impact of the digital economy on sustainable development from a vague linear relationship into the net outcome of a dynamic interplay between its internal “empowerment force” and “consuming force” on a specific “foundational platform,” thereby revealing its inherent structural tension.

3.2.2. Proposal of a New Measurement Framework

Based on the above theoretical reasoning, this study formally proposes a three-dimensional measurement framework of “platform–consumption–empowerment.” Each dimension of this framework has a clear functional definition and follows a unified “results-oriented” principle, systematically addressing the aforementioned methodological issues.

First, digital infrastructure (platforms) is strictly defined as neutral capital stock. Measurement indicators (such as the number of ports and base stations) are intended to measure actual service capacity at a specific point in time, rather than inputs during the construction process. This positioning focuses on “result-based stock” and avoids confusion between “input” and “output” indicators at different stages, providing a cleaner basis for subsequent analysis.

Second, digital industrialization (consumption power) is functionally defined as economic activity flows that cause environmental burdens. This aims to challenge the widespread tendency to view all parts of the digital economy as contributing to homogeneous growth. We believe that the scale expansion of the digital industry itself brings growth, but is also accompanied by resource consumption and environmental pressures. Clarifying its “consumption” role is key to capturing the inherent contradictions of the digital economy and comprehensively assessing its net sustainability effects.

Third, industrial digitization (enabling power) is defined as economic activity flows that enhance green efficiency. It aims to measure the structural outcomes of the deep integration of digital technology and the real economy (such as intelligent production and online transactions). This dimension strictly focuses on the core processes and outcomes of digital technology’s impact on the real economy, treating factors such as “development environment” and “government support” as exogenous variables. This “internal–external separation” design enhances the logical purity of the index, thereby better supporting subsequent causal inference.

3.3. Construction of a Three-Level Indicator System for the Digital Economy

Based on the “platform–consumption–empowerment” theoretical framework, and adhering to the core principles of logical purity, result orientation, and representativeness, we have constructed an indicator system for the digital economy. This system is designed to precisely serve the sustainability analysis objectives of this study. The final indicator system is presented in

Table 2. Digital economy indicator system.

Primary Indicator	Secondary Indicator	Calculation Method	Unit
Digital infrastructure	Fixed network access density	Number of internet broadband access ports/year-end resident population	ten thousand/ten thousand people
	Mobile network coverage strength	Number of mobile phone base stations/permanent resident population at year-end	ten thousand/ten thousand people
	Mobile network processing capacity intensity	Mobile phone exchange capacity/year-end resident population	ten thousand households/ten thousand people
Digital industrialization	Scale of the electronic information manufacturing industry	Revenue from the electronic information manufacturing industry/Revenue from large-scale industrial enterprises	billion CNY
	Communication network services intensity	Total telecommunications services/permanent resident population at year-end	billion CNY per 10,000 people
	Software and Information Services Industry Intensity	Software business revenue/year-end resident population	billion CNY per 10,000 people
Industrial digitalization	Digitalization of production methods	Industrial robot installation density, calculated using reference methods	units per 10,000 people
	Digitalization of commerce and trade	E-commerce transaction volume/year-end resident population	billion CNY per 10,000 people
	Digitalization of financial services	Peking University Digital Financial Inclusion Index
	Digitalization of physical circulation	Per capita express delivery volume	piece/person

, and the specific design rationale for each dimension is explained below.

3.3.1. Digital Infrastructure (The Platform)

Our construction of the Digital Infrastructure indicators is guided by a comprehensive assessment of its overall service capacity. We posit that contemporary digital access is characterized by the complementary nature of fixed and mobile networks. This dual focus is consistent with the foundational components identified in mainstream measurement literature [37,38]. Our primary innovation, however, lies in the principled screening process that follows. In the screening process, we strictly adhered to the principle of “logical purity.” A critical decision was to exclude demand-side adoption metrics like “internet penetration rate,” which we classify as a socio-economic outcome of digitalization, to avoid causal confusion. Following this logic, we selected three results-oriented indicators. For fixed networks, Fixed network access density (per capita internet broadband access ports) is used to measure the “depth” of access. For mobile networks, two complementary indicators were constructed: Mobile network coverage strength (per capita mobile phone base stations) measures the “breadth” of coverage, while Mobile network processing capacity intensity (per capita mobile phone exchange capacity) characterizes the “intensity” of the core network’s processing power. It is worth emphasizing that the uniform use of “per capita” for standardizing all scale-based indicators is a critical methodological decision. From a statistical standpoint, this approach is essential for controlling for the significant confounding effects of population size. Without such normalization, provinces with larger populations would mechanically exhibit higher absolute values, introducing a severe scale bias that would distort the true intensity and availability of digital resources. Conceptually, this choice embodies a “people-centered” measurement philosophy. It shifts the analytical focus from the absolute scale of digital assets to the per capita dividend and social pervasiveness of digital development. This makes the measurement results a more accurate reflection of how equitably digital benefits are distributed across society, which is of paramount importance for sustainability-oriented research.

3.3.2. Digital Industrialization (The “Consumption Force”)

The measurement of digital industrialization faces the challenge of unavailable provincial-level value-added data for specific digital industries. To overcome this, we proposed a supply-chain deconstruction framework comprising three pillars: hardware manufacturing (the physical foundation), communication services (the network), and software and information services (the “brain”). For Scale of the electronic information manufacturing industry, we use the ratio of its operating revenue to that of all large-scale industrial enterprises, accurately reflecting its structural position. For Communication network services intensity, we use per capita total telecommunications services volume, a robust physical quantity measure that avoids the data continuity issues of revenue data. Finally, for Software and information services industry intensity, we use per capita software business revenue, which effectively gauges the absolute economic strength of a region’s software industry cluster.

3.3.3. Industrial Digitalization (The “Empowerment Force”)

Rejecting the traditional industry-based paradigm due to data scarcity and its inability to capture cross-sector integration, we adopt a function-centric value chain strategy. This approach is not only a pragmatic response to data constraints but is also well-grounded in sustainability research. Analyzing economic activities along a value chain—from production and transaction to logistics—is a standard framework in environmental economics and sustainable supply chain management for identifying environmental impacts and opportunities for efficiency gains [48,49]. Our strategy thus deconstructs the empowerment process into an ecosystem comprising four functional pillars: production (value creation), transaction (value realization), finance (value circulation), and logistics (value delivery). For Digitalization of production methods, we use industrial robot installation density as the best feasible proxy for intelligent transformation under current data constraints, calculated following Acemoglu et al. [50]. For Digitalization of commerce and trade, we use per capita e-commerce transaction volume, pragmatically adopting official data based on principles of transparency and risk diversification, despite potential component-mixing issues. For Digitalization of financial services, we employ the widely recognized Peking University Digital Financial Inclusion Index for its authoritativeness. Lastly, for Digitalization of physical circulation, we use per capita express delivery volume, a metric that excludes price disturbances to focus on the real scale of digitally driven economic activities.

4. Methods and Data

4.1. Index Synthesis Method

This study employs the entropy weight method to assign weights and synthesize the indicator system, ensuring the objectivity of the measurement process. For panel data, the application path of the entropy weight method endows the final index with distinct economic implications.

A common approach is “panel-wide standardization,” which treats all i provinces and t years as a single entity for processing.The advantage of this method is its ability to assess and compare the absolute levels of all observed objects annually. However, this study’s core objective is to deconstruct the dynamic evolution and structural transformation of the digital economy, which requires a method more sensitive to inter-annual shifts in the national landscape. Therefore, we selected the “two-step entropy weighting method with annual standardization,” which is more aligned with our research objectives. This method conducts standardization and weighting independently within each year t. The normalization formula is:

$$x_{it}^{\prime }={\displaystyle \frac{x_{it}-\min (x_{i\_t})}{\max (x_{i\_t})-\min (x_{i\_t})}}$$

(1)

where $\max (x_{i\_t})$ and $\min (x_{i\_t})$ are the maximum and minimum values of indicator x across all provinces within a specific year t. The unique advantages of this method are manifested on two levels:

First, at the methodological level, it achieves a unity of dynamism and objectivity in weights. By standardizing and weighting independently within each year, this method allows the weights of various indicators to evolve dynamically over time. This not only avoids the arbitrariness of subjective weighting but, more importantly, objectively reveals the shifting relative importance of various driving forces (such as infrastructure, core industries, and integrated applications) at different stages of digital economic development.

Second, at the level of explanatory power, it shifts the measurement focus from “absolute level” to “relative position.” The index obtained through this method reflects a province’s relative standing within the national landscape in a given year. While this limits the direct temporal comparison of the index values, it endows the research with deeper insight: by analyzing annual rankings, internal structures, and changes in dynamic weights, we can precisely characterize dynamic issues such as regional convergence, competitive catch-up, and the evolution of the “access gap” and “application gap.” This makes it the most appropriate measurement tool for revealing the structural facts and evolutionary patterns of China’s digital economic development.

4.2. Data Sources and Processing

To clarify our units of analysis, it is important to explain the administrative divisions used in this study. Our sample consists of 30 “provinces,” which is a collective term for China’s provincial-level administrative regions. This includes not only provinces in the traditional sense but also four municipalities directly under the central government (Beijing, Shanghai, Tianjin, and Chongqing) and five autonomous regions. In China’s statistical and administrative system, these different types of units are treated as peers. The “regions” or “blocs” (e.g., eastern, central) discussed later in the paper refer to official, geographically based groupings of these provincial-level units used for macro-level analysis.

The sample for this study covers 30 provincial-level administrative regions in China (excluding Tibet, Hong Kong, Macao, and Taiwan due to data availability), spanning from 2012 to 2023. This period was deliberately chosen as it aligns with a critical phase of China’s economic transformation into a “New Normal,” where the digital economy emerged as a key driver of sustainable development. While our analytical framework is designed for general applicability, this 12-year timeframe provides a sufficiently long and complete dataset to capture the dynamic evolution of China’s digital economy. All original data were sourced from official or authoritative channels, including the China Statistical Yearbook, the China Industrial Statistics Yearbook, the China Electronics and Information Industry Statistics Yearbook, provincial and municipal statistical yearbooks, the Peking University Digital Financial Inclusion Index Report, the CEIC database, and the International Federation of Robotics.

To ensure the integrity and comparability of the data, we conducted systematic processing of the raw data. First, we addressed data applicability and missing values. The data for the vast majority of our indicators were available for the entire sample period. For the few sporadic missing data points, we employed linear interpolation, a reasonable and standard solution for ensuring the continuity of panel data with clear trends. Second, we clarified the units of measurement. The specific original units for each indicator (e.g., “ten thousand ports,” “billion CNY”) are detailed in Table 2. As a primary standardization step, all absolute scale indicators were converted into per capita terms to control for population-based scale bias. Third, we conducted strict price deflation for all value-type variables, with the base period uniformly set as 2012. The selection of deflators adhered to the principle of high alignment with the economic content of each indicator: Although “telecommunications service volume” is a physical quantity index, to ensure a unified base period, it was standardized using each province’s “communications CPI”; “Software business revenue,” as an important intermediate input, was deflated using each province’s “Industrial Producer Price Index (PPI).” For “E-commerce transaction volume,” which includes both consumer-end (B2C) and production-end (B2B) transactions, the most macro and comprehensive “Regional Gross Domestic Product (GDP) deflator index” was selected for deflation to maximize the accuracy of its actual transaction scale. This series of cautious data processing steps aimed to provide a robust and reliable data foundation for subsequent index calculations and empirical analysis.

5. The Evolution of China’s Digital Economy: From Micro-Level Differentiation to Macro-Level Convergence

This section applies the new measurement framework to conduct a multi-scalar systematic diagnosis of the development of China’s digital economy. To ensure clarity, we define our analytical scales as follows: the “micro level” refers to the analysis of individual provincial-level units, focusing on their internal structural models and evolutionary paths. The “macro level” pertains to the analysis of aggregate national trends and disparities across all 30 provinces. Bridging these is the “meso-level” analysis of regional blocs. Following the “bottom-up” paradigm established in our introduction, our analysis will proceed from micro-level differentiation to macro-level convergence.

This section aims to apply the new measurement framework constructed in the preceding section to conduct a multi-scale systematic diagnosis of the development of China’s digital economy. In terms of analytical paradigm, it abandons the traditional deductive approach from national macro to local micro levels, and instead adopts a “bottom-up” analytical strategy that starts from the provincial micro level and ultimately summarizes the national macro pattern. This shift in strategy is rooted in a return to the “first principles” of analysis: the essence of any macro trend is the statistical aggregation and structural emergence of heterogeneous developments at the provincial “micro level.” This has enabled this study to achieve a paradigm shift from “total assessment” to “structural diagnosis”—by identifying, at the micro level, the heterogeneous provincial development models and evolutionary paths (such as “empowerment-driven” or “consumption-driven”) that have been obscured by traditional total analysis, we are able to make a more causally explanatory systemic diagnosis of the evolution trajectory, source structure, and core driving forces of national imbalances based on solid micro-level evidence. Ultimately, macro phenomena such as the “inverted U-shaped” evolution at the national level will no longer be isolated statistical findings, but rather traceable and logically consistent inevitable outcomes.

5.1. Micro-Differentiation at the Provincial Level

5.1.1. Overall Evolution of Spatial Patterns

Classification and Changes in Membership

To illustrate the spatial evolution of China’s digital economy development, this study categorizes provinces into four development tiers based on their quartile rankings for the total index scores in 2012 and 2023: the “Leading Zone” (top 25%), the “Advanced Zone” (25–50%), the “Catching-Up Zone” (50–75%), and the “Lagging Zone” (bottom 25%) (as shown in

Table 3. Divergence of the total digital economy index, 2012 and 2023.

Rank	2012 Year	2023 Year
Leading Zone	Beijing, Shanghai, Guangdong, Zhejiang, Jiangsu, Fujian, Tianjin, Shandong	Beijing, Shanghai, Guangdong, Zhejiang, Jiangsu, Chongqing, Tianjin, Fujian
Advanced Zone	Liaoning, Chongqing, Sichuan, Hubei, Hainan, Jilin, Hebei, Hunan	Shandong, Hubei, Sichuan, Henan, Anhui, Shaanxi, Hainan, Liaoning
Catching-up zone	Anhui, Henan, Shaanxi, Ningxia, Inner Mongolia, Xinjiang, Shanxi, Jiangxi	Jiangxi, Ningxia, Hunan, Inner Mongolia, Hebei, Guangxi, Xinjiang, Qinghai
Late-developing area	Guangxi, Heilongjiang, Yunnan, Guizhou, Qinghai, Gansu	Yunnan, Guizhou, Jilin, Shanxi, Gansu, Heilongjiang

). It is important to note that since the index is constructed using a “year-on-year standardized” method, its core essence lies in the relative national rankings of provinces in a specific year.

In 2012, the “leading zone” consisted of eight eastern provinces: Beijing, Shanghai, Guangdong, Zhejiang, Jiangsu, Fujian, Tianjin, and Shandong. The “advanced zone” included eight provinces and municipalities: Liaoning, Chongqing, Sichuan, Hubei, and others. The “catching-up zone” comprised eight provinces: Anhui, Henan, Shaanxi, and others. The “lagging zone” consisted of six provinces: Guangxi, Heilongjiang, Yunnan, and others.

By 2023, there had been significant changes in the rankings of members in each tier. Chongqing successfully rose to replace Shandong Province in the “leading zone.” The “leading zone” absorbed Shandong Province, as well as Henan Province and Anhui Province, which rose strongly from the “catching up zone.” Jilin Province, which fell sharply from the “leading zone,” was added to the “lagging zone.” The reshuffling of these members heralds a profound restructuring of the national competitive landscape.

Main Characteristics of the Evolution of the Pattern

Based on the above classification and changes in membership, three core characteristics of China’s digital economy landscape can be identified for the past 12 years:

First, the eastern coastal regions have consistently led the “leading zone,” but internal divisions are becoming more pronounced. Eastern provinces have consistently occupied the majority of seats in the leading zone in both 2012 and 2023, maintaining their dominant position. However, Shandong Province’s decline from the leading zone, coupled with the drop in rankings of some provinces (such as Liaoning) within the leading zone, reveals intensifying internal competition and structural adjustments within the eastern bloc.

Second, the non-coastal regions have undergone dramatic restructuring, with the central region rising strongly and the western region breaking out through differentiation. The most notable change is that Chongqing, as the only western province, has risen to the “leading zone.” At the same time, the rise of the central provinces is particularly noteworthy, with Henan and Anhui provinces moving from the “catching-up zone” to the “leading zone” and joining Hubei province, which has maintained its position in this camp, to form the rising force of the “central bloc.”

Third, the catch-up group faces both the risk of consolidation and the risk of deceleration. The list of “lagging regions” exhibits strong signs of consolidation, with some provinces remaining at the bottom of the catch-up pack for an extended period. However, more concerning is the “stagnation” of certain provinces, such as Jilin Province, which has slipped from the “leading region” to the “lagging region.” This highlights that in the competition for digital economic development, some regions not only face the challenge of weak catch-up efforts but also the risk of further marginalization.

The above observations based on the composite index paint a comprehensive picture of the evolution. However, what are the underlying structural drivers behind these macro-level changes? What are the differences in internal development models that have led to Chongqing’s “leap forward,” the “rise” of central China, and the “slowdown” of some provinces? To answer these questions, we must delve into the structural level of the digital economy and conduct a typological diagnosis.

5.1.2. Four Types of Development Models and Three Pathways of Dynamic Evolution

To explore the structural drivers behind the macro-level evolution, this section delves into the internal structure of the digital economy to conduct a typological diagnosis of provincial development models. As depicted in the scatter plots of Figure 3 and Figure 4, a clear empirical pattern emerges: a dense cluster of provinces is concentrated near the origin in both years, while a few leading provinces diverge distinctly towards the upper-right, upper-left, and lower-right regions. This visual evidence suggests the existence of heterogeneous development models.

Classification into Four Types and Changes in Members

Based on this method, using the national average for each year as a reference, the 30 provinces are divided into four ideal models (as shown in Figure 4 and Figure 5), defined as follows: First quadrant (upper-right): “Balance-driven type,” where both “consumption power” and “enabling power” are higher than the national average, representing the highest stage of development. Second quadrant (upper-left): “Empowerment-Driven Type,” characterized by “empowerment power” exceeding “consumption power,” reflecting a growth path centered on endogenous application demand. Third quadrant (lower-left): “Structural Lag Type,” where both core indices are below average, facing dual development challenges. Fourth Quadrant (bottom-right): “Consumption-Dominated Type,” where the “consumption force” is stronger than the “empowerment force,” revealing a development model with risks of “structural decoupling.”

The specific distribution of types across provinces in 2012 and 2023 is as follows: In 2012, the “balance-driven type” included seven provinces—Beijing, Tianjin, Shanghai, Jiangsu, Zhejiang, Fujian, and Guangdong—which almost encompassed all the economically strong provinces in eastern China. The “empowerment-driven type” included only Shandong Province. The “consumption-driven type” included three provinces: Liaoning, Chongqing, and Sichuan. The remaining 19 provinces, including Hebei, Shanxi, Inner Mongolia, Jilin, Heilongjiang, Anhui, Jiangxi, Henan, Hubei, Hunan, Guangxi, Hainan, Guizhou, Yunnan, Shaanxi, Gansu, Qinghai, Ningxia, and Xinjiang, were all classified as “structurally lagging.” By 2023, there had been a significant restructuring of the types of provinces. In the “balance-driven” category, Fujian Province withdrew, reducing the number of members to six (Beijing, Tianjin, Shanghai, Jiangsu, Zhejiang, and Guangdong). The “empowerment-driven” category welcomed Fujian Province, expanding the number of members to two (Fujian and Shandong). The “consumption-driven” camp sharply contracted, with Liaoning and Sichuan provinces withdrawing, leaving only Chongqing municipality. Correspondingly, the “structurally lagging” camp absorbed Liaoning and Sichuan provinces, expanding its total membership to 21.

Three Types of Dynamic Evolution Paths

A dynamic comparison of membership changes between 2012 and 2023 reveals that China’s provincial digital economies have not converged over the past 12 years, but have undergone profound structural differentiation along several distinct paths. The evolutionary trajectories of each province exhibit strong path-dependency characteristics, ultimately converging into three logically distinct “evolutionary clubs.”

The first path is one of advantage consolidation and functional specialization, primarily led by provinces that have long been in a “balance-driven” state. Its core characteristic is the continued consolidation of “first-mover advantages,” evolving from universal leadership in 2012 to a structurally heterogeneous, functionally complementary system in 2023. This path is characterized by profound functional differentiation. Beijing’s evolutionary trajectory involves the continuous consolidation of its “all-round advantage,” with its “consumption power” (derived from high-end software and information services) and “enabling power” consistently maintaining top-tier national levels, supported by the largest infrastructure platform, forming an extreme synergy model that signifies its strengthened status as a point of technology origin and the top of the value chain. In contrast, Shanghai, Guangdong Province, and Zhejiang Province are increasingly specializing as “enabling centers,” with their development models exhibiting a high degree of dependence on robust infrastructure and efficient conversion capabilities. Guangdong Province, in particular, has leveraged its above-average infrastructure to drive top-tier digital applications nationwide, revealing an extremely high “platform conversion efficiency,” which may be closely related to its highly market-oriented economic environment and massive demand for manufacturing digitization. Tianjin and Jiangsu exhibit a “threshold-type” leading model, with both indices slightly above the national average. However, Tianjin shows signs of lag in converting platform advantages into economic momentum, while Jiangsu adopts a more balanced “follow-type” leading model. These trajectories indicate that the “path dependence” of leading provinces is not a simple linear growth process but a dynamic structural lock-in process characterized by increasingly specialized functions. This “lock-in” is rooted in the classic theories of path dependence, where historical advantages in technology, capital, and networks create powerful self-reinforcing mechanisms that constrain provinces to their existing, high-level but specialized development trajectories [51,52].

The second type of path is an enabler-driven structural leap, representing the most successful structural upgrading model among catching-up provinces. Its core driving force lies in the significant enhancement of “enabling power.” Fujian Province is a typical example of this path. By actively adjusting its development strategy, it has maintained a high level of “empowerment power” while relatively reducing “consumption power,” ultimately transitioning successfully from a “balance-driven type” to an “empowerment-driven type.” The focus of its development has shifted from hardware manufacturing to more sustainable software services and integrated applications, achieving a successful structural leap. Corresponding to Fujian Province’s “quadrant leap,” some central provinces have experienced “position shifts” within the “structural lagging type.” Take Anhui Province, Henan Province, and Hubei Province as examples. They have moved significantly upward and to the right on the two-dimensional spectrum, particularly in the “empowerment” dimension, making them the most promising candidates for achieving a typological breakthrough within this group. This also perfectly explains the “central rise” phenomenon observed in the previous section. Shandong Province, leveraging its robust industrial foundation, has continued to deepen industrial digital transformation, solidifying its position within the “enabler-driven” category. This collectively validates that “enabler-driven” is an effective path for lagging regions to achieve structural optimization.

The third path is structural lock-in and relative decline, revealing the predicament faced by some provinces that have failed to adapt to the new competitive paradigm. The most cautionary examples of this path are the “structural decline” experienced by Liaoning Province and Sichuan Province. Both provinces have completely regressed from a “consumption-led” model to a “structurally lagging” model. This confirms the inherent fragility of the “resource-driven” model: for regions lacking core technologies and strong domestic markets, this path is highly unstable. Once external conditions change or traditional advantages weaken, their fragile industrial foundations easily shrink, leading to stagnation or even regression in development. Another form of “structural lock-in” is evident in Chongqing Municipality and most late-developing provinces. Chongqing has long been entrenched in a “consumption-driven” model, failing to effectively convert its industrial scale advantage into economic efficiency advantages, while the vast majority of provinces that have remained stuck in the “structural lag” category reflect the dilemma of being unable to break free from low-level equilibrium, highlighting the arduous nature of digital transformation for lagging regions.

It is crucial to emphasize that our framework diagnoses the structural characteristics of a province’s digital economy, rather than claiming digitalization as the sole driver of its development trajectory. The observed changes in provincial rankings are generated by a combination of factors. The “empowerment-driven structural upgrading” of central provinces, for example, is likely the result of their digital strategies effectively integrating with other key conditions, such as a solid industrial foundation, favorable industrial transfer policies, and a proactive innovation environment. Our analysis serves to illuminate the digital dimension of this complex interplay.

5.2. Macro Convergence at the National Level

5.2.1. The “Inverted U-Shaped” Trajectory of Overall Gaps

General Characteristics and Structural Facts

To grasp the overall characteristics of China’s digital economy development from a macro perspective, we first conducted descriptive statistical analysis of provincial panel data from 2012 to 2023 (see

Table 4. Descriptive statistics of digital economy composite index and sub-indexes (2012–2023).

Variable	Obs	Mean	Std.Dev.	Min	Max
Infra_Index	360	0.0490	0.0292	0.0045	0.1526
Ind_Index	360	0.0583	0.0680	0.0003	0.3724
Digi_Index	360	0.0798	0.0941	0.0000	0.4396
Total_Index	360	0.0667	0.0708	0.0095	0.3438

). The results of this statistical analysis reveal three preliminary but crucial diagnostic conclusions.

First, China’s digital economy development exhibits significant and widespread regional imbalances. As shown in Table 4, both the overall index and the three sub-indices have large standard deviations, with the coefficient of variation for the overall index reaching 1.06, far exceeding the generally accepted threshold for high variability. The enormous gap of nearly 36 times between the maximum and minimum values further confirms the high heterogeneity of inter-provincial development levels, providing a solid foundation for subsequent exploration of the dynamic evolution of regional differences.

Second, the “application gap” is the primary driver of current regional disparities. This is the most core structural fact revealed by this descriptive statistical analysis. By comparing the mean values of the three sub-indices, a clear gradient difference can be observed: the mean value of the Industrial Digitization Index (Digi_Index) (0.0798) is significantly higher than the mean value of the Digital Industrialization Index (Ind_Index) (0.0583) and the mean value of the Digital Infrastructure Index (Infra_Index) (0.0490). Since the indices in this study measure relative differences between provinces, this result profoundly reveals that, on a national scale, the disparity in provinces’ capabilities to effectively utilize digital technology to create value (i.e., industrial digitization) is the primary contradiction driving the varying levels of digital economic development.

Third, there are fundamental differences in development logic and growth potential across different dimensions. Observing the maximum values, it can be seen that the peak levels of industrial digitization (0.4396) and digital industrialization (0.3724) are significantly higher than those of digital infrastructure (0.1526). This suggests, to some extent, the differing economic attributes of the three: digital infrastructure is closer to a public good, with its development facing certain “ceiling” constraints, and inter-provincial disparities tend to converge after high-level universal coverage has been achieved; in contrast, digital industrialization and industrial digitization are more akin to market-driven competitive sectors, where leaders can leverage positive feedback loops to continuously widen the gap, thereby establishing extremely high development ceilings and more enduring heterogeneity.

The Formation of an Inverted U-Shaped Trajectory and Structural Imbalances: An Analysis Based on Time-Series Averages

Figure 6 shows the annual average changes in the national composite index and the three sub-indices. This figure not only reveals the overall trend, but, more importantly, provides key evidence for diagnosing regional competitive patterns, changes in development momentum, and potential structural imbalances.

First, the “inverted U-shaped” evolution trajectory of the composite index accurately captures the grand process of regional disparities shifting from “divergence” to “convergence,” and reveals the structural significance of the period around 2017 as a critical turning point. From 2012 to 2016, the steady upward trend in the average value of the composite index reflected the intensification of regional disparities. Since 2017, the average value of the index has begun to fluctuate downward, which does not indicate a stagnation in development but rather a clear signal of regional convergence: the accelerated catch-up of lagging regions has led to a narrowing of relative gaps across the country.

Second, by analyzing the “shape” and “rank” of the three sub-index curves, we can identify the core contradictions and internal tensions driving development. First, the solidification and dominance of the “application gap.” The industrial digitization index not only consistently ranks highest in numerical terms but also exhibits the steepest decline, indicating that disparities in application and enabling capabilities constitute the deepest and most stable structural divide among regional differences. Second, the cyclical evolution of the role of digital infrastructure. This index exhibits the steepest trajectory, having undergone a complete evolutionary cycle from a “development lag” to a “competitive peak” and then to a “universal platform.” Third, the relative convergence of digital core industries. The long-term decline of the digital industrialization index suggests an industrial landscape characterized by “head solidification and middle convergence.”

Finally, the “speed differential” in the evolution of the three major indices reveals potential risks of “structural imbalances.” Observing the period after 2017, a concerning trend has emerged: the average values of the infrastructure and industrial digitization indices have both shown a noticeable downward trend, while the digital industrialization index, after an initial decline, has stabilized in recent years and even shown slight rebounds. This may indicate that the relative gaps between regions in “platform construction” (access gap) and “integration and application” (application gap) are narrowing, while competition surrounding “consumption-oriented” digital core industries may remain intense. The trend of “enabling-type differences” converging faster than “consumption-oriented differences” is a structural risk signal that requires urgent attention as the digital economy transitions toward high-quality, sustainable development.

5.2.2. Quantitative Measurement of Imbalance and Source Decomposition

To conduct a more rigorous quantitative test of the “inverted U-shaped” evolutionary trend observed in the previous section and to structurally deconstruct its origins, this study further employs the Gini coefficient, a classic indicator, to measure the inter-provincial differences in the total index and its sub-indices from 2012 to 2023 (see Figure 7). As a standardized tool for measuring inequality, the Gini coefficient’s dynamic evolution provides more robust statistical evidence to support the aforementioned conclusions.

First, the evolution of the Gini coefficient of the composite index provides decisive quantitative support for the conclusion that inter-provincial disparities followed a pattern of “initial divergence followed by convergence.” As shown in Figure 6, the coefficient rose from 0.501 in 2012 to a peak of 0.527 in 2014, before entering a long-term, fluctuating decline, falling back to 0.465 by 2023. This clear “inverted U-shaped” trajectory statistically confirms that inter-provincial imbalances in China’s digital economy reached their peak around 2014 and subsequently entered a significant convergence phase, thereby providing a more robust methodological validation of the preliminary conclusions drawn in Section 5.2.1 based on mean value analysis.

Second, by breaking down the Gini coefficients of the three major sub-indices, we can accurately identify the structural sources of inter-provincial imbalances and their dynamic evolution, and clearly validate the dominant role of the “application gap.” The Gini coefficient of the “industrial digitization” index has remained at the highest level in most years, consistently above 0.5, revealing that the “application gap” is the primary source of inter-provincial imbalances. This profoundly indicates that the unequal capabilities among provinces in “how to transform digital technology into productivity” at the application level is the most significant and persistent contradiction currently shaping the overall gap in China’s digital economy. In contrast, the evolution logic of differences in the other two dimensions is entirely different. The Gini coefficient of the “digital industrialization” index exhibits a complex trend of “initial convergence followed by stabilization.” After a significant decline in the early period (2012–2018), it stabilized and began to rise again starting in 2018, potentially indicating that digital core industries are re-concentrating in a few advantageous regions, posing a risk of widening gaps once again. In contrast, the Gini coefficient of the “digital infrastructure” index is the only indicator showing a long-term, one-sided decline trend, and it has consistently remained at the lowest level. This provides irrefutable evidence of China’s significant achievements in bridging the digital “access gap.”

5.2.3. Phase Transition of Driving Forces

The “two-step entropy weighting method with annual standardization” adopted by this study has the unique advantage of revealing how the relative importance (i.e., weights) of the three sub-indices that constitute the composite index changes over time. Figure 8 illustrates these three dynamic weights to gain insights into the structural changes in the core drivers of inter-provincial differences in China’s digital economy. As shown in the figure, this driver structure is not static but has undergone a profound transformation that can be clearly divided into three distinct phases.

Phase One (2012–2016): A dual-driven phase dominated by “empowerment” and “consumption.” In the early stages of digital economic development, the weight of industrial digitization (“empowerment”) and digital industrialization (“consumption”) was significantly higher than that of digital infrastructure. The two alternated in leading the way, jointly constituting the dominant force driving inter-provincial differences. Among them, the weight of industrial digitization reached a stage peak in 2014, reflecting that in the early stages of the rise of mobile internet, the digitization process represented by consumer-end applications first became the key factor in widening inter-provincial gaps. Meanwhile, the weight of digital industrialization remained consistently high. During this phase, inter-provincial differences were primarily defined by the “dual advantages” of a few leading provinces in core industries and integrated applications, while the relative weight of infrastructure was relatively low, indicating that it had not yet become the primary focal point of regional competition at that time.

Phase II (2017–2020): The period of establishing the dominant role of “empowerment.” The year 2017 marked a critical turning point in the evolution of the development momentum structure. As shown in Figure 7, in this year, the weight of industrial digitization reached its highest point in the entire sample period, while the weight of digital industrialization reached its lowest point, with the relative importance gap between the two reaching its maximum. This significant weight “scissors gap” marked a shift in the core driving force behind inter-provincial differences, from the “dual-wheel drive” of the previous stage to a single core drive dominated by “empowerment” in the new stage. This shift aligns closely with the concurrent macro-level technological and market environment, particularly the widespread adoption of 4G networks, the deep penetration of mobile payments, and the nationwide expansion of various new economic models. Against this backdrop, a region’s ability to effectively apply digital technologies to transform traditional industries and services—i.e., the strength of its “empowering force”—became the most critical and distinctive capability determining its relative position within the national landscape.

Phase Three (2021 to present): The Re-emergence of the “Consumption Force” and a New “Dual-Force Rebalancing”. Beginning in 2021, an extremely important new trend emerged, marking another fundamental transformation in the structure of driving forces. We define this as a “dual-force rebalancing”: a phase where the relative importance (weight) of the “consumption force,” after a long period of decline, rebounds to match and eventually surpass that of the “empowerment force”—a first in twelve years as observed in 2022. This handover of weights signifies that the primary driver of inter-provincial disparity has shifted from being singularly dominated by “empowerment” back to a new dynamic equilibrium where both forces are critically important. This shift may reflect complex economic logic: on the one hand, as the benefits of the consumer internet era gradually peak, the marginal differences that can be achieved through pure application model innovation are narrowing; on the other hand, “hardcore” technological competition and new infrastructure development, represented by initiatives such as “East Data, West Computing,” artificial intelligence, advanced semiconductors, and industrial internet, are becoming the focal points of the next round of regional competition. These areas inherently fall more within the realm of “consumption-driven forces.” This “rebalancing” of priorities signals that China’s digital economy is transitioning from an era of “model innovation” to a new phase characterized by a greater emphasis on underlying technology and industrial system construction, which holds greater strategic significance. This transformation also presents new and more challenging questions for this study as it seeks to explore the implications for sustainable development.

6. Discussion

This study set out to address the systemic methodological challenges in the measurement of the digital economy. By developing and applying a “platform–consumption–empowerment” framework, our empirical findings offer several significant implications for theory and practice. The deconstruction of the digital economy reveals that its impact on sustainability is not monolithic but is the net outcome of a dynamic interplay between a resource-intensive “consumption force” and an efficiency-enhancing “empowerment force.” Our multi-scalar analysis further demonstrates that the evolution of this interplay is highly path-dependent and structurally heterogeneous across regions. These insights challenge the conventional “black box” approach and underscore the necessity of a structural diagnostic perspective for both academic research and policymaking.

6.1. Policy Implications

The findings of this study yield important policy implications for promoting high-quality, sustainable, and coordinated development of China’s digital economy.

First, the policy focus should shift from “universal access” to “precision empowerment” to bridge the “application gap.” Given that the “application gap” has become the dominant contradiction in inter-provincial imbalances, future policies should move beyond mere infrastructure construction. The emphasis must be on enhancing the capacity of lagging regions to transform digital technologies into tangible productivity gains. This requires the implementation of more targeted industrial digital transformation support plans, the promotion of mature application scenarios, and the strengthening of digital capacity training for local governments and enterprises.

Second, regional development strategies should pivot from “total volume catch-up” to “structural optimization” with differentiated guidance. Policy formulation must fully recognize the heterogeneity of provincial development models and avoid a “one-size-fits-all” approach. For “balance-driven” provinces, they should be encouraged to advance to higher tiers of the global value chain. For “empowerment-driven” provinces, support should be provided to deepen their application and establish replicable industry benchmarks. For “consumption-driven” provinces, efforts should focus on strengthening inter-industrial linkages and green regulation to mitigate the risks of a “structural digital divide.” For the vast majority of “structurally lagging” provinces, while addressing industrial shortcomings, greater emphasis must be placed on cultivating their internal digital application markets and capabilities.

Third, macroeconomic regulation should address the new risks arising from the “rebalancing of the two forces” to guide sustainable development. In the face of the emerging trend of the “consumption force” regaining its importance, macroeconomic policies must remain vigilant to prevent a new round of “hardcore” technological competition from devolving into high-energy-consuming, extensive growth. Market-based measures such as green finance, carbon pricing, and environmental regulations should be employed to guide the green and intensive development of industries like data centers and artificial intelligence, ensuring that the growth of the “consumption force” aligns with the enhancement of the “empowerment force” to achieve a higher-level dynamic balance.

6.2. China’s Case in a Global Context

A key question is the utility of this study’s findings for a global audience. While empirically tested on Chinese data, our framework and the structural phenomena it reveals have broader international relevance.

The emergence of the “application gap” as the primary driver of inequality, for instance, is likely a universal challenge. In developed economies like the United States and the European Union, while basic digital access is widespread, significant disparities persist in how effectively different regions and industries leverage digital technologies for productivity gains. Our framework provides a tool for diagnosing such structural imbalances, which are often masked by aggregate national statistics.

Furthermore, the “dual-force rebalancing” observed in China since 2021—a renewed emphasis on the “consumption force” (core technology and industry)—may offer a preview of challenges other nations will face. As consumer-facing digitalization matures, strategic competition increasingly shifts to “hardcore” technologies like AI and semiconductors, a trend visible in policy initiatives of both the US (e.g., the CHIPS Act) and the EU. Our framework highlights the critical sustainability challenge this poses: ensuring that this new wave of resource-intensive digital industrialization is met with a commensurate increase in economy-wide empowerment.

For large developing economies like India, our findings are also highly pertinent. Given its immense regional diversity, a successful digital strategy would require moving beyond infrastructure rollout to actively fostering the “empowerment force” within its vast traditional economy. This aligns with the principle that effective corporate involvement is key to achieving the SDGs, as businesses are the primary actors in leveraging digital tools for more sustainable practices [53]. By viewing the Chinese case through our deconstruction framework, it serves not as an isolated example, but as a powerful illustration of the structural tensions inherent in digital transformation globally.

6.3. Research Limitations and Outlook

While this study makes several contributions, we acknowledge certain limitations that open avenues for future research. At the data level, the primary limitations stem from data sparsity and harmonization challenges at the provincial level. For instance, in measuring the “Digitalization of production methods,” the lack of granular and consistent data for the agricultural and service sectors necessitated our use of “industrial robot installation density” as a proxy. While this is the most feasible measure currently, it provides an incomplete picture of economy-wide production transformation. Similarly, the “e-commerce transaction volume” indicator, sourced from official statistics, likely suffers from a component-mixing issue, blending transactions from core digital industries with those from newly digitized traditional firms. Disentangling these would require more fine-grained, firm-level data, which presents a direction for future data collection efforts. At the methodological level, the core task of this study was measurement and fact-finding, primarily employing descriptive and correlational analysis. We have focused on “what” the structural patterns are, rather than conducting rigorous causal inference on “why” they exist or “what” their net effects are.

Looking ahead, this study and its framework open up broad prospects for future exploration, many of which align with the reviewer’s insightful suggestions. We propose the following key directions:

First, deepening the empirical validation. The next logical step is to apply this framework to different contexts to test its robustness and practical utility. This could involve multi-level case studies of specific cities or industrial parks, sectoral analyses focusing on high-impact industries like manufacturing or agriculture, or ambitious cross-country comparative studies to explore how institutional differences shape the balance of “consumption” and “empowerment” forces.

Second, incorporating dynamic perspectives. The digital economy is in constant flux. Future research should extend the framework to capture emerging trends and their sustainability implications. For instance, how does the widespread adoption of Artificial Intelligence (AI) alter the structure of the “empowerment force”? How do platform economies reshape the relationship between digital infrastructure and value creation? And how can circular digital practices (e.g., hardware refurbishment) mitigate the “consumption force”?

Third, expanding the application to policy evaluation. Beyond diagnosing patterns, the framework can be developed into a powerful tool for policy assessment. Future studies could use the “consumption” and “empowerment” indices as outcome variables to evaluate the effectiveness of specific digital economy policies (e.g., the “East Data, West Computing” initiative) in advancing the Sustainable Development Goals (SDGs).

Finally, the most critical next step remains conducting rigorous causal inference. The structured indices developed herein can serve as core explanatory or treatment variables to test their net causal effects on a range of sustainability outcomes, including economic growth, energy efficiency, and social welfare, which is the ultimate goal of our research endeavor.

7. Conclusions

The empirical analysis in this study reveals the complex dynamics of China’s digital economy development. Its core conclusions, mutually corroborated at the micro, meso, and macro levels, can be summarized as follows:

First, at the micro level, there are profound structural diversifications and path dependencies. The development of the digital economy in China’s provinces is not a single model but can be clearly diagnosed into four typical patterns: “balance-driven,” “empowerment-driven,” “consumption-driven,” and “structural lag.” Dynamically, provinces have embarked on three distinct evolutionary paths based on their initial endowments and strategic choices: “consolidation of advantages,” “empowerment-driven structural upgrading,” and “structural lock-in and relative decline.” This micro-level heterogeneity serves as the fundamental starting point for understanding all macro-level trends.

Second, at the macro and meso levels, there is an overall trend toward convergence in the shape of an inverted U, but the underlying contradictions have undergone a profound shift. Overall inter-provincial disparity expanded briefly around 2014 before entering a long-term convergence trend, driven primarily by the narrowing gap between the eastern and central/western regions. However, this convergence masks a shift in contradictions: the “access gap” caused by uneven digital infrastructure has been significantly narrowed, while the “application gap” stemming from disparities in industrial digitalization capabilities has become the primary and most persistent source of imbalance.

Third, the core driving forces of national disparity have undergone a structural transformation. Through an analysis of dynamic weights, we found that the core driver behind inter-provincial differences has shifted from a phase dominated by the “empowerment force” (2017–2020) to a new phase of “rebalancing” between the “consumption force” and the “empowerment force” since 2021. This signals the onset of a new round of regional competition characterized by the development of core technologies and industrial systems, presenting new challenges for sustainable development.