Empirical Analysis of Digital New-Quality Productive Forces Driving Sustainable Industrial Structural Upgrading in China

Abstract

In response to global sustainability challenges, this study investigates how Digital New-Quality Productive Forces (DNQPF), which integrate digitalization with green innovation, contribute to Sustainable Industrial Structural Upgrading (SISU) in China. Using panel data from 30 provinces spanning 2011–2023, a multidimensional DNQPF index was constructed, and a comprehensive econometric framework was applied, including two-way fixed effects, mediation and moderation analyses, Hansen threshold models, and Spatial Durbin models. The results indicate that DNQPF significantly enhances SISU (β = 0.291, p < 0.01), with household consumption upgrading serving as the key mediating channel. Regional heterogeneity is evident: Eastern provinces show strong effects (β = 0.295, p < 0.01) and central provinces exhibit catch-up potential (β = 0.467, p < 0.10), while the Western and Northeastern regions display insignificant effects due to digital infrastructure disparities. The threshold effects reveal diminishing returns beyond a DNQPF level of 0.239 (coefficient decline from 0.518 to 0.323, p < 0.01), a marketization level of 6.181, and an innovation level of 9.520. Spatial analysis further confirms positive spillovers (direct effects = 0.282–0.320; indirect effects = 0.260–1.317; p < 0.05). These findings enrich endogenous growth theory by integrating digital and green development into emerging economies and underscore DNQPF’s role in advancing SDG 9 (Industry, Innovation, and Infrastructure) and SDG 12 (Responsible Consumption and Production). Coordinated digital–green strategies, institutional reforms, and inclusive infrastructure are therefore critical for achieving sustainable industrial transformation in China and beyond.

1. Introduction

The rapid advancement of digital technologies, including artificial intelligence (AI), big data, and 5G, is reshaping global economic systems. These technologies offer new opportunities for sustainable development by improving resource efficiency, fostering green innovation, and promoting inclusive growth [1,2,3]. Under the dual imperatives of carbon neutrality and the United Nations Sustainable Development Goals (SDGs), digitally driven productivity has become a critical engine for low-carbon and equitable industrial upgrading, especially in emerging economies such as China. In particular, this study aligns with SDG 9 (Industry, Innovation, and Infrastructure) and SDG 12 (Responsible Consumption and Production), as it investigates how digital innovation and responsible consumption jointly promote sustainable transformation. This study directly contributes to the United Nations 2030 Agenda by engaging with Sustainable Development Goal 9 (Industry, Innovation, and Infrastructure) and Goal 12 (Responsible Consumption and Production). Specifically, SDG 9 emphasizes the need to develop resilient infrastructure and promote inclusive and sustainable industrialization through innovation. By constructing and empirically testing a multidimensional index of Digital New-Quality Productive Forces (DNQPF), this study captures the pathways through which digital infrastructure and data-driven tools facilitate industrial transformation in China. SDG 12 encourages the transition toward responsible production and consumption patterns. The mediating role of household consumption upgrading in this study shows how digital technologies reshape demand structures, thereby promoting sustainable, environmentally conscious consumption behaviors. Through this dual perspective, this study bridges digital development with sustainability goals, providing evidence-based insights into policy pathways for green transformation.

Within this context, the concept of New-Quality Productive Forces (NQPF) has emerged as a paradigm emphasizing technological innovation, institutional efficiency, and human capital as key drivers of structural transformation [4,5,6]. This study focuses on the digital dimension of Digital New-Quality Productive Forces (DNQPF), which integrates digital labor, tools, and infrastructure to enhance efficiency, scalability, and inclusiveness [7,8,9]. Given the availability of provincial data from 2011 to 2023, DNQPF is measured using representative indicators such as Internet penetration and digital finance, which collectively capture the impacts of AI, Internet of Things (IoT), and big data [3]. This construct is grounded in endogenous growth theory and the national innovation system, linking digitalization to industrial upgrading, green productivity, and trade expansion [10,11,12,13].

Despite the growing interest in digital transformation, several knowledge gaps remain regarding how DNQPF contributes to Sustainable Industrial Structural Upgrading (SISU). Existing studies on the digital economy and Industry 4.0 often rely on aggregate indicators or automation-focused perspectives, neglecting provincial heterogeneity, dynamics of factor reallocation, environmental externalities, and social equity dimensions within China [14,15,16]. Moreover, few studies systematically examine the internal transmission mechanisms—such as household consumption upgrading or institutional frictions—and their nonlinear threshold and spatial spillover effects within a unified framework [17,18,19,20,21,22]. In addition, a comprehensive DNQPF index that integrates the technological, institutional, and human capital dimensions remains underdeveloped.

To address these gaps, this study developed a multidimensional DNQPF index and applied an integrated econometric framework. The framework combines two-way fixed effects, causal mediation analysis (via household consumption upgrading), moderation analysis (institutional and human capital factors), Hansen’s threshold regression, and a Spatial Durbin Model (SDM). This approach enables us to capture the direct, mediated, conditional, nonlinear, and spatial effects of DNQPF on SISU, thereby deepening our understanding of digital–real economy integration and sustainable industrial transformation.

While both environmental and economic mechanisms are theoretically relevant, empirical analysis is constrained by the availability of provincial data. Consistent environmental indicators, such as carbon emissions, energy intensity, and green total factor productivity (GTFP), were incomplete for 2011–2023. Consequently, this study focuses on the demand-side mediation mechanism through household consumption upgrading, a well-measured channel linking digitalization to sustainable transformation. Environmental mediation has been conceptually discussed and identified as a priority for future empirical testing.

Accordingly, this study had five objectives. (1) To estimate the direct impact of DNQPF on SISU; (2) to assess household consumption upgrading as a demand-side mediating mechanism; (3) to evaluate how capital misallocation, innovation capacity, entrepreneurial vitality, and income inequality moderate this relationship; (4) to identify nonlinear threshold effects and regional heterogeneity; and (5) to propose policy implications for sustainable digital transformation across Chinese provinces. This study makes three major contributions to the literature. First, it constructs a novel DNQPF index that integrates technological, institutional, and human capital dimensions, extending beyond the traditional digital economy or Industry 4.0. Second, it proposes a unified empirical framework that captures the direct, mediated, moderated, threshold, and spatial mechanisms. Third, the 2011–2023 panel of 30 Chinese provinces reveals significant regional heterogeneity, offering actionable insights for policymakers in China and other emerging economies pursuing SDG-aligned digital and green development.

The remainder of this paper is organized as follows. Section 2 presents a literature review, theoretical framework, and research hypotheses. Section 3 introduces the data and econometric models. Section 4 reports the empirical results, including the baseline, the mechanism, the threshold, and the spatial analysis. Section 5 concludes with policy implications and directions for future research aligned with SDGs.

2. Theoretical Framework, Literature Review, and Research Hypotheses

This study investigates how Digital New-Quality Productive Forces (DNQPF), which integrate data-driven technologies such as artificial intelligence (AI), big data, the Internet of Things (IoT), and the industrial internet, drive Sustainable Industrial Structural Upgrading (SISU) in China. To ensure theoretical grounding and problem orientation, this section first reviews the relevant literature, develops an integrative theoretical framework, and formulates research hypotheses that guide the empirical analysis.

2.1. Literature Review

The rapid advancement of digital technologies has reshaped global production systems and created new opportunities for sustainable development through improved resource efficiency, green innovation, and inclusive growth. This review synthesizes the existing research on DNQPF and SISU, focusing on direct effects, mediating and moderating mechanisms, nonlinear dynamics, and spatial spillovers, and identifies the knowledge gaps this study seeks to address.

Empirical evidence has consistently shown that digitalization substantially promotes industrial upgrading. Toader et al. (2018) demonstrated that information and communication technology (ICT) infrastructure enhances productivity and economic growth by improving efficiency [1]. Similarly, Liu et al. (2022) find that digital economy development raises green total factor productivity across Chinese cities, while Cai and Cao (2023) highlight digital transformation as a catalyst for manufacturing upgrading via technological innovation [3,7]. However, these studies often rely on single-dimensional indicators (e.g., Internet penetration) and fail to capture the multidimensional construct of the DNQPF, which integrates digital labor, tools, and infrastructure to support green and inclusive development, particularly within China’s provincial context.

The mechanisms linking digitalization and industrial upgrading involve both demand-side and institutional factors. Solow’s (1956) neoclassical growth theory suggests that consumption upgrading enabled by digital platforms such as e-commerce and live-streaming incentivizes firms to shift toward higher value-added, sustainable production [23]. Zhu (2022) further found that technological innovation mediates the relationship between green finance and industrial upgrading [17]. Institutional factors moderate these relationships: capital misallocation undermines efficiency gains (Hsieh & Klenow, 2009) [24], innovation ecosystems strengthen absorptive capacity [25], and income inequality constrains inclusive digital diffusion [18]. However, prior studies have rarely analyzed these mechanisms jointly within a unified empirical structure, limiting the understanding of the complex dynamics of sustainable digital transformation.

Digital transformations may exhibit nonlinear behavior. Hansen’s (1999) panel threshold model identifies regime-switching patterns in economic processes [26]. Zhu (2022) shows that environmental regulation affects industrial upgrading nonlinearly because of institutional constraints [17], and Guan et al. (2022) observe diminishing returns to digitalization when infrastructure or markets reach saturation [19]. Nevertheless, most studies assess single thresholds in isolation, neglecting integrated examinations of digitalization, marketization, and innovation thresholds within China’s green development framework.

Spatial interactions are critical for understanding digitally driven transformations. Krugman’s (1991) new economic geography theory highlights factor mobility and knowledge diffusion as key drivers of regional interdependence [27]. Empirical studies demonstrate that digital infrastructure enhances the efficiency of green innovation and inter-provincial technology transfer across Chinese regions [8,11]. However, few studies have examined whether DNQPF, as a systemic construct, produces spatial spillovers that promote sustainable and regionally coordinated industrial upgrades.

Despite these insights, significant gaps remain: (1) lack of a multidimensional DNQPF index capturing digital labor, tools, and infrastructure for green development; (2) limited integration of mediation, moderation, threshold, and spatial effects in a unified framework; and (3) insufficient focus on China’s provincial heterogeneity in sustainable industrial transformation. This study addressed these gaps by constructing a comprehensive DNQPF index and employing an integrated econometric approach.

2.2. Theoretical Framework

Building on the literature gaps identified in Section 2.1, we develop an integrated framework rooted in endogenous growth theory, new institutional economics (Lundvall, 1992), and spatial economics [13,27,28,29]. Digital new quality productive forces (DNQPF), defined as data-driven labor, tools, and content enabled by AI, big data, IoT, and the industrial Internet, reshape value chains by enhancing resource efficiency, fostering green innovation, and promoting inclusive growth. These forces drive sustainable industrial structural upgrading (SISU) through (1) technological penetration, enabling precision manufacturing and green agglomeration, (2) factor reallocation, shifting from high-to efficiency-oriented growth, and (3) institutional enhancements via digital governance. The framework examines whether DNQPF promotes SISU (H1), how (via consumption upgrading, H2a), under what conditions (via institutional moderators and thresholds, H2b–d, H3a–c), and within what spatial context (via regional spillovers, H4). Causal mediation theories support this framework to ensure robust analysis of transmission mechanisms [29].

The variable model design was grounded in theory. The DNQPF index, which integrates digital labor, objects, and instruments, aligns with endogenous growth theory by treating data as a production factor that drives innovation [28]. The mediating variable, household consumption upgrading (Consu), reflects demand-side dynamics in neoclassical growth theory [23]. Moderating variables—capital misallocation, innovation vitality, and income inequality–are drawn from new institutional economics, capturing allocation inefficiencies and absorptive capacity [24,25]. The threshold and spatial variables extend the spatial innovation diffusion theory, accounting for nonlinear dynamics and regional spillovers [23,27]. This study contributes to these theories by integrating digital-green development in the Chinese provincial context, thereby filling critical gaps in digital–real economy integration and advancing the understanding of new-quality productive forces in sustainable transformation.

Moreover, the theoretical framework developed in this study aligns closely with the ambitions of SDG 9 and SDG 12. By emphasizing digital infrastructure, innovation capacity, and institutional coordination, the DNQPF concept operationalizes SDG 9. It advances the understanding of how intelligent technologies and digital capital can serve as enablers of industrial efficiency and regional coordination. Additionally, the mediation hypothesis (H2a) foregrounds the role of household consumption upgrading as a channel for transforming demand patterns, which resonates with SDG 12’s target of promoting sustainable consumption behaviors. Thus, this study not only tests the economic mechanisms of digital transformation but also advances the normative agenda of sustainability.

2.3. Research Hypotheses

Based on the literature review and theoretical framework, we formulate problem-oriented hypotheses to clarify how the DNQPF drives sustainable industrial transformation in China.

H1 (direct effect).

To address the problem of digital-green integration, DNQPF significantly enhances sustainable industrial structural upgrading by improving resource efficiency and fostering green innovation, as supported by Liu et al. (2022) [3] and Cai and Cao (2023) [7].

H2a (mediation).

To explore transmission mechanisms, household consumption upgrading mediates the DNQPF–SISU relationship, fostering sustainable consumption patterns through digital platforms, consistent with Solow (1956) [23] and Zhu (2022) [17].

H2b (Moderation—capital misallocation).

To examine contextual conditions, capital misallocation weakens the DNQPF–SISU effect by diverting resources to less-productive sectors, hindering resource efficiency, as per Hsieh and Klenow (2009) [24].

H2c (Moderation—innovation & entrepreneurship).

Regional innovation and entrepreneurial vitality strengthen the DNQPF–SISU effect by enhancing technology adoption and green innovation, supported by Cohen and Levinthal (1990) [25].

H2d (Moderation—income inequality).

Urban–rural income disparities weaken the DNQPF–SISU effect by suppressing rural demand and inclusive digital diffusion, as noted by Zhu et al. (2025) [18].

H3a (Threshold—DNQPF level).

To address heterogeneous impacts, the marginal effect of DNQPF on SISU diminishes beyond a critical digitalization level owing to infrastructure saturation, as per Hansen (1999) [26] and Guan et al. (2022) [19].

H3b (Threshold—factor marketization).

The DNQPF–SISU effect varies across factor marketization regimes, reflecting institutional constraints, consistent with Zhu (2022) [17].

H3c (Threshold—technological innovation).

The DNQPF–SISU effect varies across technological innovation capacity levels, indicating nonlinear dynamics, as supported by Hansen (1999) [26].

H4 (Spatial effects).

To examine regional coordination, DNQPF exhibits positive spatial dependence, generating spillovers that enhance SISU across provinces through technology transfer and market linkages as per Krugman (1991) [27] and Li et al. (2023) [8].

Together, these hypotheses form a multidimensional analytical framework that captures the direct, indirect, conditional, nonlinear, and spatial pathways through which the DNQPF fosters sustainable, inclusive, and regionally coordinated industrial transformation in China.

3. Methodology

This study develops a multi-tier econometric framework to evaluate the impact of Digital New-Quality Productive Forces (DNQPF) on sustainable industrial structural upgrading (SISU) in China, aligning with the hypotheses (H1, H2a–d, H3a–c, H4) in Section 2. This framework integrates direct, mediating, moderating, threshold, and spatial analytical perspectives to identify causal mechanisms and contextual heterogeneity.

3.1. Model Design and Analytical Strategy

This study integrated six econometric techniques into a coherent framework to address the research questions (RQ1–RQ4) (Figure 1).

RQ1 (Does DNQPF enhance SISU?) was tested via two-way fixed effects (Equation (1), H1), controlling for unobserved heterogeneity. RQ2 (Does household consumption mediate?) employs causal mediation analysis (Equations (2) and (3), H2a) to ensure robust transmission mechanisms. RQ3 (What conditions moderate or threshold the effect?) uses moderation (Equation (4), H2b–d) and Hansen threshold models (Equation (5), H3a–c) to capture heterogeneity. RQ4 (Are there spatial spillovers?) applies the Spatial Durbin Model (Equation (6), H4) to quantify regional dynamics. Grounded in the endogenous growth theory, this framework comprehensively examines DNQPF’s impact on SISU. The empirical strategy comprises five components to test the hypotheses proposed in Section 2.

• Baseline Estimation (H1): A TWFE model (Equation (1)) estimates the direct effect of DNQPF on SISU, accounting for unobserved regional and time-specific factors.
• Mediation Analysis (H2a): A sequential regression framework, supplemented by causal mediation analysis [30], tests the mediating role of household consumption upgrading, addressing confounding bias.
• Moderation Analysis (H2b–d): Interaction terms (Equation (4)) Examine how capital misallocation, regional innovation, and income inequality affect the impact of DNQPF.
• Threshold Regression (H3a–c): Hansen’s (1999) [26] panel threshold model (Equation (5)) Detect nonlinear regime shifts across DNQPF levels, marketization, and technological innovation capacity.
• Spatial Spillover Analysis (H4): SDM (Equation (6)) quantifies spatial dependence and interregional spillovers. Owing to provincial data limitations (2011–2023, e.g., incomplete 2023 environmental data), additional analyses (e.g., PCA, IV) are not included, but existing models ensure robustness.

These methods address the following research questions (RQs).

• RQ1: Does the DNQPF significantly enhance regional sustainable industrial upgrading? (H1)
• RQ2: Does household consumption upgrading mediate the DNQPF–SISU relationship (H2a) and do institutional factors (capital misallocation, innovation vitality, income inequality) moderate this effect? (H2b–d)
• RQ3: Are there nonlinear threshold effects in the DNQPF–SISU relationship across digitalization, marketization, and technological innovation levels? (H3a–c)
• RQ4: Does the DNQPF generate spatial spillovers and enhance SISU across regions? (H4)

3.2. Theoretical Justification

The empirical strategy of this study is grounded in endogenous growth theory, new structural economics, and spatial economics. These frameworks jointly emphasize that productivity improvements from digital technological change can occur through direct efficiency gains, mediated demand-side effects, institutionally conditioned outcomes, nonlinear dynamics, and spatial spillovers. Accordingly, we employ a multidimensional econometric framework corresponding to Hypotheses H1–H4.

3.2.1. Fixed Effects Model—Direct Effects (H1)

To test the direct impact of Digital New-Quality Productive Forces (DNQPF) on sustainable industrial structural upgrading (SISU) (H1), we estimate a two-way fixed-effects (TWFE) model controlling for province-specific (${\mu }_i$) and time-specific (${\lambda }_t$) unobserved heterogeneity:

$${\mathrm{SISU}}_{it}={\beta }_0+{\beta }_1{\mathrm{DNQPF}}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(1)

where ${\mathrm{SISU}}_{it}$ denotes the sustainable industrial structure index, ${\mathrm{DNQPF}}_{it}$ measures digital new-quality productive forces, and ${\epsilon }_{it}$ is the error term. This specification exploits within-province variation to mitigate omitted variable bias and provides a robust baseline test for H1. Control variables, such as government intervention and urbanization, were introduced in later models (for example, Equation (3)) to avoid post-treatment bias in the mediation analysis.

3.2.2. Mediation Analysis—Transmission Mechanisms (H2a)

To examine whether DNQPF promotes SISU through household consumption upgrading (H2a), we adopt a sequential regression framework to mitigate the potential endogeneity bias. This framework follows the classical mediation logic of Baron and Kenny (1986) and its panel-data extension by Imai et al. (2011), which decomposes the total effect of the DNQPF into direct and indirect components [30,31].

First, we estimated the total effect of the DNQPF on SISU without controls (Equation (1)). Second, we estimated the mediator equation, assessing how DNQPF affects household consumption upgrading, while excluding controls to avoid post-treatment bias:

$${\mathrm{Consu}}_{it}={\gamma }_0+{\gamma }_1{\mathrm{DNQPF}}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(2)

Third, we included mediator and control variables in the outcome model.

$${\mathrm{SISU}}_{it}={\delta }_0+{\delta }_1{\mathrm{DNQPF}}_{it}+{\delta }_2{\mathrm{Consu}}_{it}+{\delta }_3{X}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(3)

where $X_{it}$ represents the control variables (government intervention, urbanization, trade openness, infrastructure, and financial development), a significant ${\delta }_2$ indicates that household consumption upgrading mediates the DNQPF–SISU relationship. To ensure robustness, we implemented the causal mediation approach of Imai et al. (2011) to estimate the Average Causal Mediation Effect (ACME) and the Average Direct Effect (ADE) [30]. Sensitivity tests using the correlation parameter (ρ) assess robustness to unmeasured confounding factors. This design accurately identifies the consumption upgrading channel while addressing the biases associated with earlier specifications.

3.2.3. Moderation Analysis—Conditional Effects (H2b–d)

To evaluate whether institutional and structural factors condition the effectiveness of DNQPF, we introduce interaction terms with moderators $Z_{it}$ such as capital misallocation, innovation and entrepreneurial vitality, and urban–rural income inequality:

$${\mathrm{SISU}}_{it}={\beta }_0+{\beta }_1{\mathrm{DNQPF}}_{it}+{\beta }_2\left(Z_{it}\times {\mathrm{DNQPF}}_{it}\right)+{\beta }_3{X}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(4)

The interaction coefficient ${\beta }_2$ captures how the marginal effect of DNQPF varies with $Z_{it}$, consistent with the absorptive-capacity perspective and new institutional economics, which emphasize that digital productivity gains depend on local institutional quality and resource-allocation efficiency. This specification jointly tests H2b, H2c, and H2d.

3.2.4. Threshold Regression—Nonlinear Dynamics (H3a–c)

To capture potential nonlinear effects, we employ Hansen’s (1999) panel threshold regression framework [26], which identifies regime-switching behavior governed by the threshold variables ($q_{it}$):

$${\mathrm{SISU}}_{it}={\beta }_0+{\beta }_1{\mathrm{DNQPF}}_{it}\times I\left(q_{it}\le \gamma \right)+{\beta }_2{\mathrm{DNQPF}}_{it}\times I(q_{it}>\gamma )+{\beta }_3{X}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(5)

where $q_{it}$ is the threshold variable (DNQPF level for H3a, marketization for H3b, and technological innovation for H3c), $\gamma$ is the estimated threshold, and I($·$) is an indicator function. This model identifies regime-switching dynamics consistent with staged technological diffusion and institutional thresholds, thereby testing H3a–c.

3.2.5. Spatial Durbin Model—Spillover Effects (H4)

Finally, to explore whether DNQPF generates spatial spillovers across provinces, we apply the Spatial Durbin Model (SDM) [32]:

$${\mathrm{SISU}}_{it}={\beta }_0+\rho W{(\mathrm{SISU}}_{it})+{\beta }_1{\mathrm{DNQPF}}_{it}+{\beta }_{2}W{(\mathrm{DNQPF}}_{it})+{\beta }_3{X}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(6)

where $W$ is the row-standardized spatial weight matrix, $\rho$ captures endogenous spatial dependence, and ${\beta }_2$ measures exogenous spillovers from neighboring provinces. This specification aligns with the spatial growth theory, which posits that digital technologies diffuse through inter-regional factor mobility, knowledge spillovers, and network linkages, thereby testing H4.

By combining fixed-effects estimation, mediation, moderation, threshold regression, and spatial econometrics, this integrated framework captures the direct, indirect, conditional, nonlinear, and spatially contingent effects of the DNQPF on SISU. This multidimensional design ensures theoretical coherence and empirical rigor in evaluating Hypotheses H1–H4.

3.3. Variable Selection

To assess the relationship between DNQPF and SISU rigorously, this study constructs a comprehensive set of variables grounded in economic and management theories, including endogenous growth theory, new institutional economics, neoclassical growth theory, and spatial economics [23,25,27,28,33]. The system includes the dependent variable, core explanatory variable, mediating, moderating, threshold, and control variables designed to test hypotheses H1, H2a–d, H3a–c, and H4.

3.3.1. Dependent Variable: Industrial Structural Upgrading

The dependent variable measures the degree of structural upgrading of industries across the Chinese provinces by capturing the hierarchical advancement of industrial composition. Following the established practice in structural economics, the SISU index is constructed as

$${\mathrm{SISU}}_{it}={{\displaystyle \sum }}_{k=1}^{3}k\frac{GD{P}_{kit}}{GD{P}_{it}}$$

(7)

where $GD{P}_{kit}/GD{P}_{it}$ epresents the GDP share of sector (k) (primary, secondary, tertiary) in province (i) and year (t). A higher ${\mathrm{SISU}}_{it}$ indicates greater sustainable industrial upgrading, consistent with structural economics [3].

3.3.2. Core Explanatory Variable: Digital New-Quality Productive Forces (DNQPF)

This study refines the Digital Emerging Productive Forces framework into Digital New-Quality Productive Forces (DNQPF), emphasizing the transformation of labor, instruments, and objects through digital technologies to promote innovation-driven, sustainable, and low-carbon industrial development. The DNQPF index, grounded in endogenous growth theory as data-driven production factors [28], is constructed based on three criteria: (1) theoretical relevance to green industrial upgrading and resource reallocation, (2) consistent data coverage across 30 Chinese provinces (2011–2023), and (3) low pairwise correlations (<0.80) to minimize multicollinearity. The index comprises three dimensions: Digital Laborers, capturing workforce quality and green skills; Digital Labor Objects, measuring economic and environmental contributions of digital content; and Digital Labor Instruments, reflecting investments in sustainable digital infrastructure. Indicator selection (e.g., Internet penetration, digital finance, green skills) demonstrates the role of data in enhancing innovation and resource efficiency, as digital inputs drive productivity and green transformation [28]. The entropy weighting method (Equations (8)–(11)) ensures objectivity by minimizing subjective biases, aligning with data-driven methodologies [34]. Owing to provincial data limitations (2011–2023, e.g., incomplete AI, IoT, or Big Data metrics), additional validation analyses (e.g., PCA) were not conducted, but the robustness of the entropy method aligns with provincial studies [3]. Owing to provincial data limitations, the index uses broad digitalization proxies (e.g., Internet penetration, digital finance) rather than direct measures of AI, IoT, or Big data, aligning with empirical literature [3,7]. This approach captures the aggregated impact of digital technologies on green transformation, with future micro-level data suggested for refinement.

All tertiary indicators (

Table 1. Indicator system for the Digital New-Quality Productive Forces (DNQPF) index.

Dimension	Level-2 Indicator	Level-3 Indicator	Definition/Computation	Unit	Attribute
Digital Workers	Quality of Digital Workers	Educational Attainment	Average years of schooling per capita	years	+
		Share of Digital Service Employment	Employment in information transmission, software and IT services as a share of total employment	%	+
		R&D Workforce Intensity	Employment in scientific research and technical services as a share of total urban employment	%	+
		Higher Education Enrollment	Number of students in higher education per 100,000 population	persons/100,000	+
	Innovation Investment in Labor	Investment in Innovative Human Capital	Full-time equivalent (FTE) of R&D personnel in industrial enterprises above designated size (per 10,000 industrial employees)	FTE/10,000 employees	+
Digital Labor Objects	Intelligence	Robot Density	Industrial robots in operation per 10,000 manufacturing employees $R{D}_{it}=10,000\times {\mathrm{RobotStock}}_{it}/{\mathrm{Manufacturing}\mathrm{Employment}}_{it}$	units/10,000 workers	+
	Informatization	Enterprise Informatization	Firms with e-commerce transactions as a share of total enterprises	%	+
	Digital Business	Postal Service Intensity	Postal service output as a share of GDP	% of GDP	+
		E-commerce Transaction Level	Total e-commerce sales as a share of GDP	% of GDP	+
		Telecommunications Output	Total telecom business as a share of GDP	% of GDP	+
		Software Business Output	Software business revenue as a share of GDP	% of GDP	+
Digital Labor Instruments	Tangible Digital Capital	Mobile Penetration Rate	Mobile phone subscriptions per 100 persons	subs/100 persons	+
		Broadband Access Port Density	Number of fixed-broadband access ports per 100 persons	ports/100 persons	+
		Fiber Optic Infrastructure Density (FOID)	Route length of optical fiber cable per 100 km2 of land area $FOI{D}_{it}=100\times {\displaystyle \frac{{\mathrm{FiberLength}}_{it}\left(\mathrm{km}\right)}{{\mathrm{LandArea}}_i\left({\mathrm{km}}^2\right)}}$	km/100 km2	+
	Intangible Digital Capital	Inclusive Digital Finance—Breadth	Coverage breadth index	index (0–100)	+
		Inclusive Digital Finance—Depth	Depth of use index	index (0–100)	+
		Inclusive Digital Finance—Digitization	Overall digitization index	index (0–100)	+

) showed positive correlations. The cost-type variables, if any, are transformed as follows:

$$X_{jit}^{\prime }={\displaystyle \frac{1}{X_{jit}}}$$

(8)

before normalization.

Standardization step: Indicators were min-max scaled.

$$X_{jit}^{″}={\displaystyle \frac{X_{jit}^{\prime }-\min \left(X_{jit}^{\prime }\right)}{\max \left(X_{jit}^{\prime }\right)-\min \left(X_{jit}^{\prime }\right)}}$$

(9)

where $X_{jit}^{″}\in \left[0,1\right]$ for all variables.

Weighting step: Weights were computed using the entropy method.

$$w_j={\displaystyle \frac{(1-e_j)}{{\displaystyle {\sum }_{(j=1)}^m(1-e_j)}}},e_j=-{\displaystyle \frac{1}{ln(n)}}{\displaystyle {\sum }_{(i=1)}^n{p}_{ijt}}\ln (p_{ijt}),p_{ijt}={\displaystyle \frac{X_{jit}^{\prime }}{{\displaystyle {\sum }_{(i=1)}^n{X}_{jit}^{\prime }}}}.$$

(10)

where $e_j$ is the entropy of indicator $j_{,} p_{ijt}$ is its proportion for region $i$, and $w_j$ is the normalized weight.

Aggregation step: The DNQPF index is computed as follows:

$$\mathrm{D}\mathrm{N}\mathrm{Q}\mathrm{P}{F}_{it}={\displaystyle {\sum }_{j=1}^m{w}_j}\times X_{jit}^{\prime }$$

(11)

The index was rescaled to [0, 1] for interpretability of the regression model. This variable design contributes to digital-real economy integration by capturing China’s provincial dynamics of green transformation.

The construction of the DNQPF index also aligns closely with key dimensions of the United Nations Sustainable Development Goals (SDGs). In particular, the selected indicators reflect measurable pathways toward achieving SDG 9 (Industry, Innovation, and Infrastructure) and SDG 12 (Responsible Consumption and Production). For instance, Internet penetration, e-commerce transaction levels, and broadband access density directly support SDG targets 9.1 and 9.c by enhancing access to quality infrastructure and information and communication technology. Similarly, the inclusion of green-skilled labor, innovation investment in human capital, and inclusive digital finance reflects progress toward SDG 9.5, which emphasizes research and innovation. On the demand side, indicators such as the share of digital services and the transformation of retail consumption align with SDG 12.2 and 12.a by promoting responsible consumption behaviors and enabling sustainable production models. By embedding these dimensions within the DNQPF framework, this study operationalizes the sustainable transformation agenda empirically and contributes to the measurement and advancement of global sustainability commitments.

3.3.3. Mediating Variable: Social Consumption Level

To explore whether domestic demand expansion mediates the DNQPF–SISU relationship (H2a), we introduce the social consumption level (Consu), measured as the ratio of total retail sales of consumer goods to regional GDP. This reflects purchasing power and market potential, incentivizing firms to shift toward high-value, sustainable production, grounded in neoclassical growth theory.

3.3.4. Moderating Variables

To capture contextual heterogeneity, three moderating variables are introduced, aligned with new institutional economics and absorptive capacity theory:

• Capital Misallocation Index (MISK): Adapted from Hsieh and Klenow (2009), this index measures deviations between actual and optimal capital allocations, reflecting institutional inefficiencies that hinder resource reallocation (H2b) [24].

$$\begin{array}{c}MIS{K}_{it}={\displaystyle \sum }_s\left|{\displaystyle \frac{K_{sit}}{Y_{sit}}}-{\displaystyle \frac{{\alpha }_s}{{\beta }_s}}\right|,\end{array}$$

(12)

where $K_i$, $s_i$ and ${\beta }_{Ki}$ denote capital share, output share, and output elasticity of province i, respectively. where $K_{sit}$, $Y_{sit}$, and ${\alpha }_s/{\beta }_s$ denote capital share, output share, and output elasticity for province (i), respectively.

• Urban-Rural Income Inequality (Inequality): Measured by the Theil index, capturing disparities that suppress inclusive digital diffusion, consistent with institutional constraints (H2d) [18].

  $${\mathrm{Inequality}}_{it}={{\displaystyle \sum }}_j\left({\displaystyle \frac{Y_{jit}}{Y_{it}}}\right)\ln \left({\displaystyle \frac{Y_{jit}/P_{jit}}{Y_{it}/P_{it}}}\right)$$

  (13)

  where j represents urban and rural sectors, $Y_{jit}$ denotes income, and $P_{jit}$ denotes population.
• Regional Innovation and Entrepreneurship Capacity (IRIEC): Drawing from the Peking University Innovation Index, this measures technological and entrepreneurial vitality, supporting absorptive capacity theory (H2c) [25].

3.3.5. Threshold Variables

These variables enable the examination of nonlinear effects and assess how regional readiness shapes SISU. Three threshold variables capture the nonlinear dynamics in the DNQPF–SISU relationship (H3a–H3c) and reflect contingencies for sustainable transformation. The marketization level (mark) in Fan Gang’s Marketization Index measures institutional development and resource allocation efficiency, both of which are critical for green transition (H3b). Technological innovation (tec), the natural logarithm of granted invention patents, assesses green innovation capacity aligned with endogenous growth theory (H3c). The DNQPF level tests variations across digital development stages, revealing the thresholds for sustainable productivity (H3a).

3.3.6. Control Variables

Five control variables minimize the omitted-variable bias, reflecting the determinants of sustainable regional economic performance based on regional economic theory [28]. Government intervention (gov), measured by the fiscal expenditure-to-GDP ratio, captures the influence of the public sector on green policy. Urbanization (Urb), the proportion of the urban population, reflects demographic shifts and sustainable agglomeration. Trade openness (Tra), or the trade-to-GDP ratio, indicates the degree of global integration. Infrastructure development (Infra), proxied by road length per unit area and including digital infrastructure (e.g., broadband), supports connectivity for sustainable growth, consistent with digitalization proxies due to data limitations. Financial development (Fin), measured by deposits and loans as a percentage of GDP, including digital finance, reflects low-carbon investment capacity. These variables contribute to digital-real economy integration, addressing China’s provincial heterogeneity.

3.4. Data Source and Preprocessing

This study constructs a balanced panel dataset covering 30 Chinese provinces (excluding Hong Kong, Macao, Taiwan, and Tibet) from 2011 to 2023, enabling a robust analysis of Digital New-Quality Productive Forces (DNQPF) and sustainable industrial structural upgrading (SISU). Data are sourced from the China Statistical Yearbook, China Economic Census Yearbook, China Industrial Statistical Yearbook, Wind Database, and National Bureau of Statistics (NBS) Bulletin. Innovation and marketization indicators were derived from the Peking University China Regional Innovation and Entrepreneurship Index and Fan Gang’s Marketization Index, supporting the analysis of green and inclusive development.

Preprocessing enhances analytical rigor: sporadic missing values are interpolated, whereas structural gaps are excluded to maintain robustness. Non-ratio indicators were standardized to z-scores to ensure comparability. The tertiary indicators of the DNQPF index (digital laborers, objects, and instruments) were normalized and aggregated using entropy weighting, thereby minimizing the subjective bias. A row-standardized inverse-distance matrix captures spatial dependence in the Spatial Durbin Model. Monetary variables (e.g., GDP and financial indicators) are deflated (base year: 2011) and skewed variables (e.g., patents and retail sales) are log-transformed. Despite comprehensive coverage, environmental data (e.g., carbon emissions) are absent, suggesting their inclusion in future analyses to strengthen sustainability analyses.

4. Empirical Analysis and Discussion

To investigate how Digital New-Quality Productive Forces (DNQPF) drive sustainable industrial structural upgrading (SISU) across Chinese provinces, this study employs a multilayered econometric framework that aligns with the methodology presented in Section 3 and supports green and inclusive development. The analysis tests four hypotheses (H1–H4) through six components: (1) baseline estimation, using two-way fixed-effects models to assess DNQPF’s direct impact on SISU (H1); (2) endogeneity correction, applying lagged variable models and instrumental variable (IV) estimations for robustness; (3) mediation analysis, testing household consumption as a channel for SISU (H2a); (4) moderation analysis, evaluating how capital misallocation (H2b), income inequality (H2c), and innovation capacity (H2d) shape the DNQPF–SISU relationship; (5) threshold regression, using Hansen’s model [26] to capture nonlinear dynamics across DNQPF intensity (H3a), marketization (H3b), and technological innovation (H3c); and (6) spatial spillover analysis, employing the Spatial Durbin Model with Moran’s tests to assess regional diffusion (H4). These strategies provide rigorous evidence for inclusive, green and regionally coordinated industrial transformation policies.

4.1. Data Sources and Descriptive Statistics

This study utilizes a balanced panel dataset covering 30 Chinese provinces (excluding Tibet, Hong Kong, Macao, and Taiwan) from 2011 to 2023, enabling a robust analysis of Digital New-Quality Productive Forces (DNQPF) and sustainable industrial structural upgrading (SISU). The data sources and preprocessing are detailed in Section 3.3.

Table 2. Descriptive statistics of key variables.

Variable Name	Symbol	Obs.	Mean	Std. Dev.	Min	Max
Digital New-Quality Productive Forces	DNQPF	390	0.227	0.129	0.089	0.669
Industrial Structural Upgrading	SISU	390	2.403	0.123	2.132	2.846
Government Intervention Level	Gov	390	0.257	0.111	0.105	0.758
Marketization Level	mark	390	8.271	1.989	3.359	13.360
Urbanization Level	Urb	390	0.606	0.120	0.350	0.896
Technological Progress	tec	390	10.350	1.451	6.219	13.680
Degree of Openness	Tra	390	0.108	0.155	0.000	0.944
Infrastructure Level	Infra	390	2.423	1.045	0.494	5.607
Financial Development	Fin	390	3.484	1.101	1.688	8.164
Social Consumption Level	Consu	390	0.389	0.067	0.180	0.610
Capital Misallocation Index	MISK	390	0.226	0.163	0.001	1.047
Urban–Rural Income Inequality	Inequality	390	0.086	0.042	0.016	0.227
Innovation–Entrepreneurship Index	IRIEC	390	82.800	14.570	23.000	117.600

presents descriptive statistics that reveal the distributions and variability of the key variables related to sustainable transformation.

The mean DNQPF index was 0.227 (SD = 0.129, range: 0.089–0.669), reflecting significant regional disparities in green digitalization potential. The SISU index, with a mean of 2.403 (SD = 0.123, range: 2.132–2.846), indicates a sustainable transition toward high-value, low-carbon sectors. Control variables, including marketization (range: 3.359–13.360) and urbanization (35–90%), exhibit variability, capturing institutional and structural diversity critical for green and inclusive growth. Indicators such as capital misallocation and innovation capacity further support the sustainability dynamics analysis. These statistics, which are free of ceiling or floor effects, validate the use of fixed effects, interaction, threshold, and spatial models. The future inclusion of environmental indicators (e.g., carbon emissions) could enhance sustainability insights.

4.2. Multicollinearity Diagnostics

To ensure a robust estimation of the impacts of Digital New-Quality Productive Forces (DNQPF) on sustainable industrial structural upgrading (SISU), a Variance Inflation Factor (VIF) test was used to assess multicollinearity among explanatory variables, thereby supporting reliable green and inclusive policy inferences. High multicollinearity can inflate standard errors, compromising statistical reliability.

Table 3. Variance Inflation Factor (VIF) results.

Variable	VIF	1/VIF
Digital New-Quality Productive Forces (DNQPF)	4.070	0.246
Social Consumption Level (Consu)	1.200	0.833
Government Intervention (Gov)	2.480	0.403
Urbanization Level (Urb)	4.200	0.238
Degree of Openness (Tra)	2.380	0.420
Infrastructure Level (Infra)	1.330	0.752
Financial Development (Fin)	4.070	0.246
Mean VIF	2.820	—

Note: All VIF values are below the critical threshold of 10, confirming the absence of severe multicollinearity.

reports the VIF values, all of which were below the threshold of 10 (maximum: 4.20, mean: 2.82), indicating low collinearity. The DNQPF index recorded a VIF of 4.070, indicating a moderate correlation with variables such as financial development and urbanization, while maintaining orthogonality. These results validate the inclusion of all covariates in fixed-effects and spatial econometric models, ensuring a robust analysis of DNQPF’s contribution of DNQPF to sustainable transformation.

4.3. Baseline Regression Analysis

To assess the direct impact of Digital New-Quality Productive Forces (DNQPF) on sustainable industrial structural upgrading (SISU), this study estimates nested panel regression models using balanced data from 30 Chinese provinces (2011–2023), supporting green and inclusive transformation. Three models were employed: (1) a baseline model with DNQPF only, (2) an extended model with controls (government intervention, urbanization, openness, infrastructure, and financial development), and (3) a full model with two-way fixed effects.

Table 4. Baseline regression analysis results.

Variables	(1) Baseline	(2) + Controls	(3) + Fixed Effects
DNQPF	0.521 *** (0.177)	0.156 *** (0.035)	0.291 *** (0.062)
Gov	—	−0.185 *** (0.031)	0.324 *** (0.075)
Urb	—	0.301 *** (0.040)	0.031 (0.101)
Tra	—	−0.038 * (0.020)	0.016 (0.032)
Infra	—	−0.014 *** (0.004)	0.011 *** (0.003)
Fin	—	0.059 *** (0.004)	0.024 *** (0.007)
Constant	2.285 *** (0.034)	2.066 *** (0.023)	2.123 *** (0.062)
Control Variables	No	Yes	Yes
Province Fixed Effects	No	No	Yes
Time Fixed Effects	No	No	Yes
Observations	390	390	390
R2	0.595	0.836	0.975

Note: Standard errors are in parentheses. * p < 0.10, *** p < 0.01.

presents the results. Model (1) shows that DNQPF significantly enhanced SISU (β = 0.521, p < 0.01), indicating its role in green industrial transformation. Model (2) yielded a reduced but significant DNQPF coefficient (β = 0.156, p < 0.01), confirming its robustness. Model (3), controlling for province- and time-fixed effects, reports a DNQPF coefficient of 0.291 (p < 0.01), thus supporting H1. This result suggests that DNQPF fosters resource-efficient industrial upgrading by integrating data-driven technologies into production processes, aligning with endogenous growth theory’s emphasis on innovation-driven growth [24]. This highlights DNQPF’s role of the DNQPF in advancing sustainable economic structures, particularly in digitally advanced provinces. Diagnostics validated the fixed-effects model; the Hausman test (χ² = 23.86, p < 0.01) favored fixed effects, and likelihood-ratio tests (LR = 59.21, 599.15, p < 0.01) confirmed a superior fit. The R² increased from 0.595 to 0.975, reflecting enhanced explanatory power.

4.4. Robustness and Endogeneity Tests

To validate the reliability of Digital New-Quality Productive Forces (DNQPF) impacts on sustainable industrial structural upgrading (SISU), this study conducts comprehensive robustness and endogeneity tests, ensuring robust evidence for green and inclusive policy inferences. The results reported in

Table 5. Robustness and endogeneity test results.

Variable	(1) Lagged Dependent	(2) Alternative DV	(3) Alternative Metric	(4) Exclude Municipalities	(5) Winsorized Sample	(6) 2SLS First Stage	(6) 2SLS Second Stage
DNQPF	0.203 *** (0.073)	1.271 *** (0.311)	0.012 *** (0.003)	0.362 *** (0.080)	0.367 *** (0.084)	—	0.431 ** (0.178)
IV	—	—	—	—	—	0.695 *** (0.062)	—
Constant	2.177 *** (0.078)	3.126 *** (0.548)	2.089 *** (0.073)	2.246 *** (0.112)	1.962 *** (0.056)	0.179 *** (0.067)	2.222 *** (0.204)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes
FE (Province/Time)	Yes	Yes	Yes	Yes	Yes	Yes	Yes
KP LM statistic	—	—	—	—	—	9.752 [0.002]	—
KP Wald F	—	—	—	—	—	124.431 {16.38}	—
Observations	360	390	390	338	390	360	360
R2	0.976	0.976	0.974	0.922	0.961	0.991	0.977

Notes: Robust standard errors are in parentheses. *** p < 0.01, ** p < 0.05. [ ] denotes p-values and { } indicates the 10% critical value from the Stock–Yogo weak identification test. The KP LM statistic tests for under-identification, whereas the KP Wald F statistic evaluates weak instrument validity.

consistently show that DNQPF has a significant positive effect on SISU.

Robustness checks included (1) a lagged dependent variable model, yielding a DNQPF coefficient of 0.203 (p < 0.01), indicating persistent impact; (2) an alternative dependent variable (tertiary-to-secondary industry output ratio), with a coefficient of 1.271 (p < 0.01); (3) a Principal Component Analysis (PCA)-based DNQPF index, with a coefficient of 0.012 (p < 0.01); (4) exclusion of municipalities (Beijing, Shanghai, Tianjin, Chongqing), with a coefficient of 0.362 (p < 0.01); and (5) 5% Winsorization, with a coefficient of 0.367 (p < 0.01). These tests confirmed DNQPF’s robust role of DNQPF in sustainable SISU.

Endogeneity is addressed through a two-stage least-squares approach, using lagged DNQPF as an instrumental variable. The first-stage results show strong identification (Kleibergen–Paap LM = 9.752, p = 0.002; Wald F = 124.431, above Stock–Yogo’s 16.38 threshold). The second-stage coefficient (0.431, p < 0.05) reinforced causal validity. These findings validate H1, demonstrating DNQPF’s consistent contribution of DNQPF to green and inclusive industrial transformation, supporting subsequent analyses.

4.5. Mediation Effect Analysis

Building on robust evidence that DNQPF significantly enhances SISU, this section examines whether DNQPF indirectly promotes SISU through sustainable consumption patterns (H2a). A two-step panel regression model assessed the mediating role of social consumption (consu).

Table 6. Results of mediation effect test.

Variable	(1) Consu	(2) SISU
DNQPF	0.765 *** (0.118)	0.219 *** (0.064)
Consu	—	0.095 *** (0.027)
Constant	−0.543 *** (0.131)	2.174 *** (0.065)
Controls	Yes	Yes
Region/Year FE	Yes	Yes
Observations	390	390
R2	0.681	0.976

Notes: Robust standard errors are in parentheses. *** p < 0.01.

shows that the DNQPF significantly enhanced Consu (β = 0.765, p < 0.01). In the second stage, DNQPF retained a direct effect on SISU (β = 0.219, p < 0.01), and Consu also positively influenced SISU (β = 0.095, p < 0.01), indicating partial mediation. This suggests that DNQPF drives sustainable consumption by enhancing digital platforms’ supply demand matching, stimulating demand for green products, and incentivizing firms to innovate for low-carbon, high-value output, which is consistent with neoclassical growth theory’s demand-side dynamics.

Sobel and Goodman tests (

Table 7. Sobel and Goodman tests mediation results.

Test	Coefficient	Std. Error	z-Value	p-Value
Sobel Test	0.072	0.024	3.064	0.002
Goodman-1 (Aroian)	0.072	0.024	3.037	0.002
Goodman-2	0.072	0.023	3.091	0.002
Coefficient a ($\mathrm{DNQPF}$ → Consu)	0.765	0.114	6.701	0.000
Coefficient b (Consu → SISU)	0.095	0.027	3.445	0.001
Indirect effect (a × b)	0.072	0.024	3.064	0.002
Direct effect (c′)	0.219	0.062	3.548	0.000
Total effect (c)	0.291	0.059	4.944	0.000
Proportion of intermediary effect	0.249
The ratio of indirect effects to direct effects	0.331
The ratio of total effect to direct effect	1.331

Note: Sobel and Goodman tests confirm the mediation significance. z-values > 2.58 indicate p < 0.01.

) estimated an indirect effect of 0.072 (p < 0.01, z > 2.58), rejecting the null hypothesis of no mediation. A bootstrap test (1000 resamples,

Table 8. Bootstrap mediation test results (1000 resamples).

Effect Type	Coefficient	Std. Error	z-Value	p-Value	95% Confidence Interval
Indirect	0.072	0.028	2.54	0.011	[0.017, 0.128]
Direct	0.219	0.073	3.00	0.003	—

Note: Bootstrap test with 1000 replications. The 95% confidence interval for the indirect effect excluded zero, confirming the robustness of the mediation.

) confirmed the indirect effect (0.072, 95% CI: [0.017–0.128]), validating Consu’s role. These findings highlight DNQPF’s indirect contribution of the DNQPF to sustainable transformation by fostering green consumption patterns and informing policies to promote digitally driven demand for eco-friendly products.

4.6. Moderation Effect Analysis

To explore heterogeneous mechanisms, we examine three moderating variables—capital misallocation (MISK), regional innovation and entrepreneurship capacity (IRIEC), and urban–rural income inequality (Inequality)—capturing structural, institutional, and social factors conditioning the DNQPF–SISU relationship (H2b–d). The interaction terms (Equation (4)) were tested, and the results are listed in

Table 9. Regression results of moderation effects.

Variable	(1) Capital Misallocation	(2) Innovation & Entrepreneurship	(3) Income Inequality
DNQPF	0.287 *** (0.059)	0.292 *** (0.063)	0.288 *** (0.061)
MISK	−0.043 ** (0.018)	—	—
MISK × DNQPF	−0.171 ** (0.078)	—	—
IRIEC	—	0.001 (0.001)	—
IRIEC × DNQPF	—	0.006 * (0.003)	—
Inequality	—	—	0.384 * (0.204)
Inequality × DNQPF	—	—	−1.533 * (0.893)
Constant	2.191 *** (0.066)	2.018 *** (0.088)	2.019 *** (0.074)
Controls	Yes	Yes	Yes
Region/Year FE	Yes	Yes	Yes
Observations	390	390	390
R2	0.976	0.975	0.976

Notes: Robust standard errors are in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.10.

.

The interaction between DNQPF and capital misallocation exhibited a negative effect (β = −0.171, p < 0.05), IRIEC had a positive effect (β = 0.006, p < 0.10), and income inequality had a negative effect (β = −1.533, p < 0.10). Capital misallocation hinders DNQPF’s green efficiency of the DNQPF by constraining resource reallocation, aligning with new institutional economics. Innovation amplifies DNQPF’s impact of DNQPF by enhancing absorptive capacity, while income inequality fragments digital access and reduces inclusive growth. These findings can inform policies to optimize resource allocation, strengthen R&D ecosystems, and reduce disparities for sustainable transformation.

4.7. Heterogeneity Analysis: Regional and Foundational Differences

To explore the contextual variability of Digital New Quality Productive Forces (DNQPF) in driving sustainable industrial structural upgrading (SISU), this study examines regional and digital infrastructure disparities, revealing the conditions for green and inclusive transformation. These analyses can inform the development of tailored policies for sustainable digital development.

• Regional Heterogeneity

Using China’s National Bureau of Statistics classification, the sample was segmented into four regions: Eastern, Central, Western, and Northeastern China.

Table 10. Heterogeneity test regression results.

Variable	(1) East	(2) Central	(3) West	(4) Northeast	(5) High Digital Infrastructure	(6) Low Digital Infrastructure
DNQPF	0.295 *** (0.076)	0.467 * (0.243)	0.201 (0.136)	0.444 (0.323)	0.536 *** (0.138)	0.183 *** (0.066)
Gov	0.176 ** (0.080)	0.377 (0.314)	0.380 *** (0.096)	−0.396 * (0.200)	0.363 *** (0.097)	0.035 (0.109)
Urb	0.204 ** (0.100)	−0.611 (0.433)	−0.280 (0.278)	0.805 (1.415)	0.163 (0.240)	0.038 (0.099)
Tra	0.012 (0.032)	0.005 (0.216)	0.217 * (0.115)	−1.051 ** (0.401)	0.336 ** (0.144)	0.047 (0.037)
Infra	0.013 ** (0.005)	−0.013 *** (0.004)	0.009 * (0.005)	0.049 *** (0.012)	0.006 (0.005)	0.023 *** (0.005)
Fin	−0.000 (0.006)	0.105 *** (0.026)	0.022 ** (0.010)	0.022 (0.022)	0.030 *** (0.010)	0.015 ** (0.007)
Constant	2.183 *** (0.064)	2.298 *** (0.213)	2.250 *** (0.160)	1.713 * (0.838)	1.940 *** (0.131)	2.261 *** (0.061)
Region Fixed Effects	Yes	Yes	Yes	Yes	Yes	Yes
Year Fixed Effects	Yes	Yes	Yes	Yes	Yes	Yes
Observations	130	78	143	39	193	193
R2	0.995	0.971	0.890	0.986	0.914	0.988

Note: *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

(Columns 1–4) reports these results. In the Eastern region, DNQPF significantly enhances SISU (β = 0.295, p < 0.01), reflecting market maturity and innovation capacity that support green, low-carbon upgrading. The Central region shows a stronger effect (β = 0.467, p < 0.10), suggesting catch-up potential driven by demographic and infrastructural advantages. By contrast, the western (β = 0.201) and northeastern (β = 0.444) regions show insignificant effects, constrained by limited fiscal capacity and digital readiness, underscoring the need for region-specific green digital strategies.

• Heterogeneity in Digital Infrastructure Construction

Stratifying the sample by per capita Internet access ports (Table 10, Columns 5–6), high-infrastructure regions exhibit a robust DNQPF effect (β = 0.536, p < 0.01), driven by strong ICT systems that enhance sustainable digital integration into value chains. Low-infrastructure regions exhibited a weaker but significant effect (β = 0.183, p < 0.01), which was limited by connectivity and platform fragmentation. These findings highlight the pivotal role of the digital infrastructure in sustainable SISU.

These results reveal the context-sensitive impact of the DNQPF on green, inclusive industrial transformation. Regional and infrastructural disparities necessitate precise governance and tailoring digital policies to local economic, institutional, and green development capacities to maximize sustainability outcomes.

4.8. Threshold Effect Analysis: Testing Nonlinear Impacts and Conditional Mechanisms

To explore nonlinear dynamics in DNQPF’s influence on SISU, Hansen’s (1999) [26] panel threshold regression tests three variables—DNQPF, marketization (mark), and technological innovation (tec)—and reveals stage-dependent green transformation dynamics.

• Testing for Threshold Effects

Table 11. Threshold effect test results.

Threshold Variable	Inspection Type	F-Statistic	p-Value	Critical Value
				10%	5%	1%
DNQPF	Single threshold	41.25	0.012	25.840	30.051	41.137
	Double threshold	16.07	0.362	23.910	27.788	38.016
mark	Single threshold	49.93	0.004	28.762	34.528	42.802
	Double threshold	27.28	0.100	27.253	31.175	45.337
tec	Single threshold	41.01	0.050	34.512	40.992	60.954
	Double threshold	12.00	0.700	28.340	33.048	46.630

shows significant single thresholds (p < 0.05), with estimates in

Table 12. Estimated threshold values.

Threshold Variable	Threshold Type	Estimated Value	95% Confidence Interval
DNQPF	Single threshold	0.239	(0.235, 0.240)
mark	Single threshold	6.181	(6.109, 6.251)
tec	Single threshold	9.520	(9.515, 9.534)

showing DNQPF (0.239), mark (6.181), and tec (9.520). Figure 2 illustrates these thresholds, with likelihood ratios and 95% CIs confirming their stability.

• Estimation of Nonlinear Effects

Table 13. Regression results of threshold model.

Threshold Variable	Threshold Interval	$\mathbf{Coefficient}\mathbf{of}\mathbf{DNQPF}\mathbf{on}\mathbf{Industrial}\mathbf{Upgrading}$	Significance
DNQPF	DNQPF < 0.239	0.518 ***	p < 0.01
	DNQPF ≥ 0.239	0.323 ***	p < 0.01
mark	mark < 6.181	0.412 ***	p < 0.01
	mark ≥ 6.181	0.210 *	p < 0.10
tec	tec < 9.520	0.371 ***	p < 0.01
	tec ≥ 9.520	0.213 *	p < 0.10

Note: *, and *** denote significance at the 10%, and 1% levels, respectively.

reports conditional effects: DNQPF’s coefficient drops from 0.518 (p < 0.01) to 0.323 (p < 0.01), mark from 0.412 to 0.210 (p < 0.10), and tec from 0.371 to 0.213 (p < 0.10). These nonlinearities indicate diminishing returns as digitalization, marketization, or innovation saturates, reflecting resource constraints or institutional bottlenecks in advanced provinces, consistent with the nonlinear dynamics in economic transitions. Policies must target stage-specific interventions to balance digital investment with institutional reforms to sustain green upgrades.

4.9. Spatial Spillover Effect Analysis: Mechanisms of Digital New-Quality Productive Forces Under Regional Linkage

Amid China’s uneven development and integrated regional economies, this study examines whether Digital New-Quality Productive Forces (DNQPF) generate spatial spillovers that enhance sustainable industrial structural upgrading (SISU) across provinces. Spatial autocorrelation tests, using a nested geographic-economic weight matrix, revealed significantly positive Moran’s I values (0.076–0.109, p < 0.05,

Table 14. Moran’s I index for digital new-quality productive forces (2011–2023).

Year	Moran’s I	z-Value	p-Value
2011	0.094	2.475	0.013
2012	0.089	2.390	0.017
2013	0.083	2.260	0.024
2014	0.095	2.470	0.014
2015	0.076	2.105	0.035
2016	0.086	2.327	0.020
2017	0.078	2.164	0.030
2018	0.085	2.277	0.023
2019	0.088	2.340	0.019
2020	0.097	2.508	0.012
2021	0.098	2.527	0.012
2022	0.103	2.611	0.009
2023	0.109	2.734	0.006

) from 2011 to 2023, indicating the clustering of provinces with similar DNQPF levels. The Moran scatter plot in Figure 3 highlights high-high clusters (e.g., Jiangsu, Zhejiang, and Guangdong) forming green innovation hubs and low-low clusters in inland regions, signaling digital divides. These geographic patterns demonstrate that eastern coastal provinces, with advanced digital infrastructure, drive green innovation spillovers, amplifying SISU in the neighboring areas, while inland Western and Northeastern provinces face connectivity constraints, limiting diffusion and exacerbating digital divides.

Specification tests (

Table 15. Model specification tests.

Test Name	Statistic	p-Value
LM-Error	26.103	0.000
Robust LM-Error	25.409	0.000
LM-Lag	4.710	0.030
Robust LM-Lag	4.016	0.045
Wald (SDM vs. SEM/SLM)	24.82–31.00	0.000
Hausman Test	23.86	0.000
LR: Region FE	59.21	0.000
LR: Time FE	599.15	0.000

) confirmed spatial dependence (LM-Error: 26.103, p < 0.001; Robust LM-Lag: 4.016, p < 0.05), supporting the Spatial Durbin Model SDM (Equation (6)) with two-way fixed effects (Wald: 24.82–31.00, p < 0.001).

Table 16. Empirical results of spatial spillover effect estimation.

Variable	(1) Econ-Geo	(2) Adjacency	(3) Econ-Only	(4) Add Controls	(5) Winsorized
DNQPF	0.299 ** (0.122)	0.179 * (0.103)	0.284 ** (0.125)	0.320 *** (0.107)	0.314 ** (0.130)
W × DNQPF	1.192 *** (0.435)	0.238 * (0.130)	0.303 *** (0.091)	1.128 ** (0.483)	1.317 *** (0.459)
Direct effect	0.282 ** (0.124)	0.182 * (0.104)	0.285 ** (0.125)	0.303 *** (0.106)	0.296 ** (0.133)
Indirect effect	0.947 ** (0.438)	0.260 * (0.150)	0.303 *** (0.078)	0.794 ** (0.384)	1.077 ** (0.461)
Total effect	1.229 *** (0.390)	0.442 *** (0.137)	0.587 *** (0.188)	1.097 *** (0.378)	1.373 *** (0.411)
R2	0.5665	0.540	0.6795	0.366	0.540

Note: *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

reports the significant direct (0.282–0.320, p < 0.05) and indirect (0.260–1.317, p < 0.05) effects, validating H4.

Figure 3 further visualizes these spillovers, showing more substantial DNQPF impacts in eastern hubs (β = 0.295) and weaker effects in inland areas (β = 0.150–0.201), underscoring regional disparities in green transformation. Policies should prioritize inter-provincial data sharing, cloud computing for low-carbon integration, and investments in lagging regions to bridge digital divides, thereby enhancing green regional transformation.

4.10. Discussion

This study’s spatial spillover findings (Table 16, direct 0.282–0.320, indirect 0.260–1.317) deepen the understanding of DNQPF’s role of DNQPF in sustainable industrial transformation by highlighting regional interdependencies, aligning with spatial economics. Figure 3’s Moran scatter plot reveals High-High clusters in Eastern provinces (e.g., Jiangsu, Zhejiang, β = 0.295, Table 10), where advanced digital infrastructure amplifies green SISU through knowledge and technology diffusion, and low-low clusters in inland Western and Northeastern regions (β = 0.150–0.201), where connectivity constraints limit spillover benefits and underscore digital divides. This extends digital-real economy integration by demonstrating DNQPF’s role in fostering coordinated green development across provinces. Similarly, Xu et al. (2025) [35] confirmed the digital economy’s role in green innovation, but lacked multidimensional DNQPF indices and spatial analysis, which our study addresses. These findings suggest policies to strengthen cross-provincial R&D alliances and digital infrastructure in lagging regions, align with China’s carbon-neutrality goals, and advance regional green transformation. Limitations such as provincial data aggregation suggest the need for future micro-level studies to refine spatial mechanisms (Section 5.3).

5. Conclusions and Policy Implications

5.1. Key Findings

Using a balanced panel of 30 Chinese provinces from 2011 to 2023, this study constructs a Digital New-Quality Productive Forces (DNQPF) index to evaluate its influence on Sustainable Industrial Structural Upgrading (SISU) and to advance green, low-carbon, and inclusive growth. By incorporating data-driven production factors into endogenous growth theory, the analysis demonstrates how digitalization drives sustainable industrial transformation through regional heterogeneity and nonlinear dynamics, thereby supporting SDG 9 (Industry, Innovation and Infrastructure) and SDG 12 (Responsible Consumption and Production).

Five key findings emerged: (1) DNQPF drives sustainable SISU (β = 0.291, p < 0.01), enhancing resource-efficient economic growth; (2) household consumption mediates this effect, fostering sustainable consumption patterns; (3) capital misallocation and income inequality hinder, while innovation amplifies DNQPF’s green impact; (4) nonlinear thresholds (DNQPF: 0.239, coefficient drops from 0.518 to 0.323, p < 0.01; marketization: 6.181; innovation: 9.520) indicate stage-dependent dynamics; and (5) spatial spillovers promote regional green SISU (direct effects 0.282–0.320, indirect 0.260–1.317, p < 0.05) through knowledge and infrastructure diffusion (H4).

Regional heterogeneity is evident: Eastern provinces show strong effects (β = 0.295, p < 0.01), Central provinces exhibit catch-up potential (β = 0.467, p < 0.10), while Western and Northeastern provinces show insignificant effects, reflecting digital infrastructure disparities. These results contribute to sustainable industrial transformation by highlighting DNQPF’s role in balancing economic and environmental efficiency across diverse regional contexts, extending digital-real economy integration.

5.2. Strategic Policy Implications

To fully realize the sustainable potential of Digital New-Quality Productive Forces (DNQPF), three strategic policy pillars are recommended. These align with the Sustainable Development Goals (SDGs), particularly SDG 9 (Industry, Innovation and Infrastructure) and SDG 12 (Responsible Consumption and Production), and provide a roadmap for integrating digital transformation with inclusive and low-carbon industrial upgrading.

(1)

Institutional and Green Talent Development: Governments should strengthen digital–green ecosystems by investing in digital literacy, green skills training, and institutional capacity-building. This includes improving innovation governance, enhancing regulatory transparency, and supporting cross-sector coordination mechanisms. These actions directly support SDG 9.5, which emphasizes enhancing scientific research and upgrading technological capabilities, and SDG 12.a, which calls for strengthening institutional capacity in developing countries to shift toward sustainable consumption and production systems.

(2)

Low-Carbon Infrastructure Investment: Targeted investments in intelligent infrastructure—such as 5G networks, cloud computing facilities, and the industrial Internet—are essential for enabling low-carbon manufacturing, platform-based productivity, and circular economy models. Integrating digital infrastructure with renewable energy grids and intelligent logistics systems can reduce resource intensity and emissions, thereby advancing SDG 9.1 (quality, sustainable infrastructure) and SDG 9.c (universal ICT access), while reinforcing SDG 12.2, which promotes the efficient use of natural resources.

(3)

Regional Coordination and Inclusive Digitalization: To address spatial inequality and maximize spillover benefits, policymakers should foster interprovincial cooperation through data-sharing platforms, R&D consortia, and inclusive digital access in rural and underdeveloped regions. Initiatives such as computing power alliances, cross-regional innovation corridors, and broadband extension projects can reduce digital divides and enhance system-wide resilience. These measures are consistent with SDG 9.b (supporting domestic technology development and innovation) and SDG 12.1 (developing and implementing national frameworks for sustainable production and consumption).

These strategies apply not only to China but also to other emerging economies that aim to achieve sustainable industrial upgrading within the framework of digital-green integration.

5.3. Limitations and Future Prospects

This study has several limitations that present valuable directions for future research, particularly in advancing the empirical understanding of digital–green transformation aligned with the Sustainable Development Goals (SDGs).

First, the analysis relies on provincial-level panel data, which may mask firm-level heterogeneity in digital adoption, green innovation behavior, and resource allocation efficiency. Future research should incorporate micro-level data, such as enterprise surveys on the adoption of AI, IoT, and big data, to explore intra-industry variation in Digital New-Quality Productive Forces (DNQPF) and their differential effects on sustainability. This would yield more granular policy recommendations and better align with SDG 9.5, which emphasizes upgrading research and innovation systems. Second, due to the unavailability of complete environmental datasets for 2023, key green efficiency indicators such as carbon emissions and Green Total Factor Productivity (GTFP) could not be included. This limitation constrains the empirical assessment of environmental mediation pathways and the analysis of digitalization’s contribution to ecological outcomes. Future studies should integrate real-time environmental metrics as data availability improves, enabling direct evaluation of digital drivers of decarbonization. This would support not only SDG 9 and SDG 12 but also SDG 13 (Climate Action) by quantifying the roles of digital technologies in achieving carbon neutrality. Third, methodological extensions remain necessary. Dynamic panel models (e.g., system GMM) can better capture potential feedback loops between digital capacity and sustainable industrial transformation. Additionally, spatial structural equation modeling (SEM) could uncover latent spillover channels, such as interregional talent flows, innovation diffusion, and institutional coordination. These refinements would deepen the theoretical foundations of endogenous growth theory in the digital era and enhance empirical modeling of spatially uneven SDG progress. Finally, there is substantial scope for comparative and cross-national studies. Applying the DNQPF framework to other emerging economies, such as India, Brazil, or Indonesia, would enable testing of the model’s external validity and its contextual adaptation. Such research could reveal institutional, infrastructural, and cultural differences in digital–green integration and facilitate benchmarking progress across regions. These efforts are essential to evaluating global alignment with the 2030 Agenda and supporting knowledge-sharing among countries pursuing inclusive industrial upgrading.

By addressing these directions, future research can strengthen the theoretical and empirical foundations of the DNQPF framework, enhance its applicability to diverse development contexts, and accelerate global sustainability transitions consistent with SDG 9, SDG 12, SDG 13, and the broader objectives of the 2030 Agenda for Sustainable Development.