How Information Infrastructure Construction Drives Rural Inclusive Green Growth in China: The Mediating Role of Labor Mobility

Abstract

Against the backdrop of China’s digital rural revitalization, this study investigates the impact of information infrastructure construction (IIC) on rural inclusive green growth (RIGG) using panel data from 29 Chinese provinces (2011–2022). The findings demonstrate that IIC significantly promotes RIGG, a conclusion robust to two-way fixed effects, SYS-GMM estimations, and various robustness checks. This positive effect extends to the core dimensions of inclusive growth and green growth. Quantile regression reveals that the marginal effect of IIC is stronger in provinces with initially lower RIGG levels, suggesting its potential to narrow regional development gaps. Heterogeneity analysis further indicates a more pronounced impact in the central region compared to eastern and western regions. Mechanism tests confirm that rural labor mobility (RLM) serves as a critical transmission channel, where IIC fosters RIGG by accelerating labor movement, with this effect exhibiting persistence over time. These findings underscore the importance of regionally tailored IIC strategies to effectively promote sustainable and resilient rural development.

1. Introduction

Since the launch of reform and opening-up, China has achieved remarkable socio-economic progress. Yet, its development trajectory lacks inclusiveness and environmental sustainability, reflected in pronounced regional and urban–rural disparities—evidenced by a Gini coefficient consistently above the international alert level.

Since the start of reform and opening up, China has made great socioeconomic growth. However, its development trajectory lacks inclusivity and environmental sustainability, as indicated by significant regional and urban-rural disparities—a Gini coefficient persistently over the international warning threshold [1]. Extensive expansion patterns have also resulted in significant environmental damage. In recent years, the rapid expansion of information infrastructure, fueled by technologies like cloud computing, has gained major traction [2,3]. This infrastructure not only offers the hardware underpinning for emergent economic forms like the digital economy, but it also works as a crucial driver of inclusive green growth (IGG), creating crowding-in and multiplier effects through public investment [4,5,6].

Following the World Bank’s introduction of the “inclusive green growth” (IGG) concept in 2012, states throughout the world have aggressively pursued relevant policy initiatives and practices. Developed economies, notably EU members, frequently combine IGG with high environmental regulations, substantial social welfare, and technical innovation, emphasizing both green transition and social fairness [7,8]. Conversely, emerging nations prioritize balanced resource consumption, environmental conservation, and social inclusion throughout economic development., emerging nations prioritize balanced resource consumption, environmental conservation, and social inclusion throughout economic development [9]. China, the world’s biggest developing nation, is experiencing rural expansion and structural upheaval while dealing with resource restrictions, environmental concerns, and urban-rural differences [10]. Traditional development strategies are insufficiently inclusive and environmentally responsible, impeding long-term rural rehabilitation and agricultural modernization [11]. IGG constitutes a multidimensional concept integrating economic, social, and ecological aspects of sustainability [12], and is a key indicator of high-quality development in China [13]. The ongoing push for new infrastructure—exemplified by Sichuan Union Communications’ deployment of over 6000 rural base stations, boosting mobile coverage from below 70% to 98% and expanding gigabit fiber to townships [14]—has significantly enhanced rural livelihoods and regional development capacity [15].

Against this backdrop, this study aims to empirically investigate the role of information infrastructure construction (IIC) in promoting rural inclusive green growth (RIGG) in China. To achieve this central objective, we seek to address the following specific research questions:

(1) What is the overall impact of IIC on the level of RIGG?

(2) What is the underlying mechanism through which IIC affects RIGG? Specifically, does rural labor mobility (RLM) serve as a significant channel?

(3) Does the influence of IIC on RIGG exhibit regional heterogeneity?

By rigorously answering these questions, this study aims to provide nuanced insights and robust empirical evidence for formulating regionally tailored digitalization strategies to foster sustainable and resilient rural development.

2. Literature Reviews

IIG emerges from the integration of two foundational concepts: “inclusive growth” and “green growth.” Inclusive growth (IG) underscores that economic expansion should foster social equity and poverty reduction, emphasizing coordinated socioeconomic progress [16,17]. Green growth (GG), by contrast, prioritizes ecological conservation and the efficient use of natural resources within economic development, highlighting the sustainability of both environmental and economic systems [18,19]. IGG is not merely a combination of these two notions but represents an organic synthesis. This developmental paradigm promotes the efficient and clean utilization of natural resources, ensures that pollution remains within ecological carrying capacities, and embeds social inclusiveness across the growth process [20]. Consequently, it facilitates synergistic advancements across economic, social, and environmental dimensions [21,22].

In recent years, digital technology has introduced new impetus into rural development. As a foundational component of the digital era, IIC has achieved significant progress in rural regions [23,24]. IIC encompasses a range of technologies, including the internet, data centers, artificial intelligence, and the Internet of Things [25]. By 2024, 5G network coverage had been extended to remote rural areas. Initiatives such as the “Digital Village” program have improved rural economic conditions and quality of life through e-commerce and smart agriculture, while the deployment of smart grids has markedly enhanced energy efficiency and system stability in the countryside. Existing research demonstrates that IIC helps overcome spatio-temporal constraints, facilitates the flow of information and resources, improves transactional efficiency, and reduces costs in production and daily life—thereby exerting profound impacts on economic growth, industrial structure, regional innovation, environmental pollution, and energy consumption [26,27].

Despite the significant potential of IIC, it is crucial to critically acknowledge the challenges and risks accompanying its rapid deployment in rural areas [28]. If left unaddressed, these issues could undermine its positive contributions to RIGG. Firstly, the construction and operation of digital infrastructure, including data centers and communication networks, are inherently energy-intensive processes that generate substantial carbon emissions, creating a potential tension between digital transformation and environmental objectives [29]. Secondly, while IIC facilitates labor mobility, it may also accelerate “brain drain” from rural areas by attracting younger, more skilled workers, thereby potentially exacerbating rural hollowing-out and the erosion of local human capital [30].

While the literature on IGG is growing, existing research has paid relatively limited attention to its specific manifestation in rural contexts, particularly concerning its driving mechanisms. This study aims to provide new empirical evidence on the following key questions: First, although scholarly discussion on IGG is considerable, research focused specifically on the rural dimension RIGG remains insufficient. Given the substantial disparities between urban and rural areas in production modes, consumption structures, and resource endowments, research focusing on RIGG is not only necessary but also constitutes a primary focus of this paper. Second, in the current era of rapid digital technology advancement, whether and how the proliferation of digital infrastructure, specifically IIC, influences IGG in rural settings requires further in-depth investigation. Existing studies often examine the direct links between digitalization and economic or environmental outcomes separately, but systematic empirical tests of its relationship with the integrated concept of RIGG are still lacking. Third, and most importantly, the specific pathways and mechanisms through which these effects operate remain largely unexplored. Although labor mobility is a critical feature of rural transformation, its role as a mediating channel in the relationship between IIC and RIGG has not been empirically tested in prior research.

Therefore, this study contributes to the literature in the following ways: (1) by explicitly focusing the investigative lens on the rural context; (2) by providing robust empirical evidence on the impact of IIC on RIGG using comprehensive provincial-level rural panel data; and (3) by uniquely identifying and empirically testing rural labor mobility as a critical transmission mechanism, thereby offering deeper insights into the “how” behind this relationship.

3. Theoretical Basis and Research Hypothesis

3.1. IIC and RIGG

Despite comprehensive economic progress in China, imbalanced allocation of development resources and an underdeveloped energy infrastructure system have resulted in relative lag in sustainable development in rural areas [23]. As the physical foundation supporting digital technologies such as big data analytics and artificial intelligence, IIC enhances data processing and resource optimization capabilities, providing a new pathway for achieving RIGG [7,31].

First, IIC plays a crucial role in rural revitalization, economic diversification, and energy transition [2,32]. It transforms traditionally extensive and inefficient production modes, broadens channels for knowledge dissemination and acquisition, and enables innovation entities to optimize resource allocation, thereby facilitating efficient output of green technology innovations [33,34]. Traditionally, rural energy consumption has heavily relied on highly polluting and energy-intensive resources such as coal, which not only exacerbates environmental pressure but also hinders sustainable economic and social transformation [35]. With the spread of IIC, particularly e-commerce services in rural areas, farmers’ income levels have increased, and the structure of rural energy consumption has gradually improved [36].

Second, IIC promotes the application of advanced energy systems, significantly enhancing the capabilities of rural communities in production, operation, and monitoring. It facilitates the integration of rural areas with external technological advancements, including smart grids, distributed generation, and solar energy utilization, thereby improving the reliability and efficiency of energy systems [26,27]. Smart grids enable real-time monitoring, automatic voltage regulation, and reduction in transmission and distribution losses; big data and cloud platforms assist authorities in optimizing energy dispatch, reducing supply costs, and promoting efficient energy use in rural areas [37,38]. The deep integration of information technologies makes energy production and consumption smarter and more efficient, helping to reduce energy waste.

Accordingly, the following hypothesis is proposed:

Hypothesis 1.

IIC can significantly promote RIGG.

3.2. Intermediation Mechanism

IIC can indirectly promote RIGG by alleviating factor misallocation. On one hand, it accelerates the dissemination of information technology and the flow of information elements, breaking down information barriers, reducing transaction costs, and improving capital allocation efficiency. On the other hand, it broadens employment channels, enhances the inclusiveness of development outcome distribution, and improves information access for disadvantaged groups, thereby mitigating labor misallocation.

Consistent with the Harris-Todaro model, rural-urban migration exhibits significant age selectivity, with young laborers attracted by higher expected urban incomes [39]. IIC deepens socioeconomic actors’ reliance on information technologies, eliminates information barriers between labor market demand and supply, promotes information sharing, and reduces information asymmetry faced by groups such as migrant workers and university graduates [40]. It overcomes geographical constraints, improves rural access to urban employment information, and accelerates the rural labor migration process [41]. This transformation directly affects household income, energy demand, and agricultural management models [42]. Increased income enhances households’ capacity to adopt clean energy, thereby improving living standards and reducing dependence on traditional polluting energy sources [43]. Meanwhile, localized labor shortages triggered by outmigration promote the application of automation technologies, such as mechanized equipment and smart irrigation systems, enhancing agricultural productivity, reducing pollution, and thus contributing to RIGG [39]. Based on the above mechanisms, the following hypothesis is proposed:

Hypothesis 2.

RLM plays a moderating role in the process through which IIC promotes RIGG.

Based on the above analysis, it can be theorized that the impact of IIC on RIGG encompasses both direct effects and indirect effects achieved through the promotion of RLM. Accordingly, the analytical framework of this study is presented in Figure 1.

4. Research Method and Empirical Data

4.1. Model Building

Building on prior research, we assert that RIGG is multifaceted. To examine the relationship between IIC and RIGG, we use a panel data model with double fixed effects for regression analysis. The model is specified as follows:

$$\begin{array}{c}{lnRIGG}_{i,t}={\alpha }_0+{\alpha }_1{lnIIC}_{i,t}+\rho ln{Control}_{i,t}+{\lambda }_t+{\sigma }_i+{\epsilon }_{i,t}\end{array}$$

(1)

In the specific analyzes, where i and t represent provinces and years respectively.

Control denotes control variables. ${\lambda }_t$ and ${\sigma }_i$ denote individual fixed effects and time fixed effects, respectively. ${\epsilon }_{i,t}$ denotes a random perturbation term.

4.2. Variable Measurement

We analyze the relationship between IIC and RIGG using balanced panel data (2011–2022) from 29 Chinese provinces, excluding Shanghai due to its high level of urbanization and the absence of rural data. Detailed variable descriptions follow.

4.2.1. Independent Variable: IIC

IIC primarily refers to facilities evolved from a new generation of information technologies. These include communication network infrastructures such as 5G, the Internet of Things (IoT), industrial internet, and satellite internet; technology infrastructures represented by artificial intelligence, cloud computing, and blockchain; as well as computing infrastructures such as data centers and intelligent computing centers. Regarding the measurement of IIC, many scholars adopt single indicators, such as broadband penetration, fixed-line subscriptions, and network coverage [44,45]. Other researchers employ the entropy method to evaluate IIC in a more comprehensive and objective manner [32,46]. By selecting multiple appropriate indicators tailored to the features of IIC, this approach reduces the subjective bias associated with single-indicator measurements.

Accordingly, we selected 13 indicators across five dimensions (

Table 1. Comprehensive Evaluation Index System for IIC.

System	Primary Indicators	Secondary Indicators	Weights	Attribute
Information Infrastructure Construction (IIC)	Telecom Major Communication Capacity	Mobile Phone Base Stations (10,000 units)	0.048	+
		Mobile Phone Switch Capacity (10,000 units)	0.034	+
		Fiber Optic Cable Line Length (km)	0.047	+
	Main Internet Indicators Development	Number of Domain Names (10,000 units)	0.104	+
		Number of Websites (10,000 units)	0.169	+
		Internet Broadband Access Ports (10,000 units)	0.045	+
		Internet Broadband Access Users (10,000 units)	0.049	+
	Enterprise Informationization Level	Number of Enterprises (units)	0.074	+
		Number of Websites per 100 Enterprises (unit)	0.009	+
	Enterprise E-Commerce Situation	Proportion of Enterprises Engaged in E-Commerce Transactions (%)	0.026	+
		E-Commerce Sales Revenue (10,000 yuan)	0.111	+
	Software and Information Services Industry Development Level	Software Business Revenue (10,000 yuan)	0.146	+
		Software Product Revenue (10,000 yuan)	0.136	+

): Telecom Major Communication Capacity, Main Internet Indicators Development, Enterprise Informatization Level, Enterprise E-Commerce Situation, and Software and Information Services Industry Development Level. Using the entropy method, we conducted a comprehensive evaluation of the IIC level of each province. The selection of the 13 secondary indicators across these five dimensions is grounded in a comprehensive understanding of IIC as the physical and technological foundation of the digital economy [44]. This indicator system is designed to concurrently capture both the supply-side capacity and the application-level penetration of IIC, thereby ensuring a holistic measurement.

Specifically, telecom major communication capacity comprises indicators such as the number of mobile phone base stations and the length of fiber optic cable lines. These metrics directly reflect the coverage scale and transmission capability of fundamental network infrastructure, serving as a prerequisite for supporting all digital applications [43,44]. Main internet indicators development encompasses the number of websites and internet broadband access subscribers. These indicators characterize the scale of internet utilization and the breadth of access penetration, forming a core basis for evaluating the development level of a digital society [41]. Enterprise informatization level and e-commerce situation is utilized to assess the adoption of IIC by key economic entities. The number of enterprises with websites and their e-commerce sales revenue signify the depth to which firms have integrated digital tools into their operational processes, thereby driving productivity enhancement and market expansion [23,24]. Software and information services industry development level represents the high-value-added and innovation-driven segment of the digital economy. Software business revenue acts as a proxy for the scale and technological capability of the regional software industry, providing essential support for the in-depth deployment and application of IIC across various sectors [25].

The calculation formula of the entropy method is as follows:

Due to differences in dimensions and scales in the selected data, it is necessary to standardize the data. In this study, the min-max normalization method is applied for standardization, as shown below:

Positive indicator:

$$\begin{array}{c}{x}_{ij}^t={\displaystyle \frac{x_{ij}^t-min\left\{x_{ij}^t\right\}}{max\left\{x_{ij}^t\right\}-min\left\{x_{ij}^t\right\}}},\left(i=1, 2, \cdots , m; j=1, 2, \cdots , n\right)\end{array}$$

(2)

Negative indicator:

$$\begin{array}{c}{x}_{ij}^t={\displaystyle \frac{max\left\{x_{ij}^t\right\}-x_{ij}^t}{max\left\{x_{ij}^t\right\}-min\left\{x_{ij}^t\right\}}},\left(i=1, 2, \cdots , m; j=1, 2, \cdots , n\right)\end{array}$$

(3)

$max\left\{x_{ij}^t\right\}$ refers to the maximum value of the indicator in year t, $min\left\{x_{ij}^t\right\}$ refers to the minimum value of the indicator in year t, and $x_{ij}^t$ represents the dimensionless result.

After standardizing the relevant indicator data, the entropy method is employed to assign weights to the indicators. The specific calculation steps of the entropy method are as follows:

First, calculate the comprehensive development level of the subsystem, ${\omega }_{ij}$ represents the proportion of j-th indicator in the i-th year.

$$\begin{array}{c}{\omega }_{ij}={\displaystyle \frac{x_{ij}}{\sum _{i=1}^m x_{ij}}}\end{array}$$

(4)

Next, define the indicator information entropy $e_j$ and redundancy $d_j$:

$$\begin{array}{c}{e}_j=-{\displaystyle \frac{1}{\mathrm{l}\mathrm{n}m}}{\displaystyle \sum _{i=1}^m} \left({\omega }_{ij}\times \mathrm{l}\mathrm{n}{\omega }_{ij}\right)\end{array}$$

(5)

$$\begin{array}{c}{d}_j=1-e_j\end{array}$$

(6)

where $0\le e_j\le 1$, m represents the number of years to be evaluated.

Then, calculate the weight of each indicator based on the defined information entropy and redundancy:

$$\begin{array}{c}{\phi }_j={\displaystyle \frac{d_j}{\sum _{j=1}^m d_j}}\end{array}$$

(7)

Finally, the comprehensive score for the subsystems of IIC is determined by calculating the weights of each indicator:

$$\begin{array}{c}{U}_i={\displaystyle \sum _{j=1}^m} {\phi }_j\times {\omega }_{ij}\left(i=1, 2\right)\end{array}$$

(8)

In this context, $U_i$ represents the comprehensive score of the i-th system, with the index value ranging from 0 to 1.

4.2.2. Dependent Variable: RIGG

The dependent variable in this paper is RIGG. Existing literature often employs either Data Envelopment Analysis (DEA) [47,48] or indicator system construction for measurement [49,50]. DEA essentially evaluates input-output efficiency, which does not align with the concept of RIGG examined in this study. In contrast, the indicator system approach assesses RIGG from a scale-oriented perspective, utilizing multiple relevant indicators to comprehensively reflect the features of IGG and provide a multi-dimensional evaluation.

Aligned with the World Bank’s definition in 2012, we treat RIGG as an organic integration of IG and GG. Socially, IIC aims to enhance human welfare, mitigate social inequality, and ensure the equitable distribution of essential resources like labor, livelihood opportunities, and energy. Economically, this entails that development is not defined solely by GDP expansion, but by a green economy characterized by sustained technological innovation, continuous environmental improvement, and declining economic inequality [47]. Accordingly, specific indicators were chosen to measure each component.

Table 2. Comprehensive Evaluation Index System for RIGG.

Systems		Primary Indicators	Secondary Indicators	Weights	Attribute
Rural Inclusive Green Growth (RIGG)	Inclusive Growth (IG)	Rural economic development	Primary Industry/GDP (%)	0.218	+
			Rural Per Capita Disposable Income (yuan)	0.074	+
			Rural Express Delivery Volume (km)	0.213	+
		Rural inclusive development	Urban Residents’ Average Income/Rural Residents’ Average Income (%)	0.150	−
			Number of Rural Students (people)	0.163	+
			Number of Rural Administrative Institutions (unit)	0.181	+
	Green Growth (GG)	Rural energy consumption level	Rural total energy consumption/GDP (tec per 10,000 yuan)	0.010	−
			Rural total carbon emissions/GDP (tons per 10,000 yuan)	0.001	−
			Rural total energy consumption/Rural total population (tec per person)	0.009	−
			Total carbon emissions/Total population (tons per person)	0.001	−
		Rural green technology level	Urban employment in information transmission, software, and IT services (10,000 people)	0.337	+
			Rural per capita solar energy and solar house area (10,000 square meters)	0.197	+
			Rural water-saving irrigation machinery (unit)	0.204	+
			Rural electricity generation (10,000 kwh)	0.241	+

summarizes the details of these indicators.

Specifically, the IG subsystem is designed to measure the extent to which economic opportunities and benefits are broadly shared across the rural population. This study selects two dimensions for its evaluation: rural economic development and rural inclusive development. In terms of rural economic development, the focus is on capturing the creation of opportunities and wealth. The primary industry/GDP ratio reflects the economic significance and structural foundation of the agricultural sector. Rural per capita disposable income serves as a direct indicator of the material prosperity of rural residents, representing a fundamental objective of inclusive development. Rural express delivery volume acts as a proxy for the integration of rural areas into the modern digital economy, facilitating market access and consumption [15]. The dimension of rural inclusive development primarily assesses the distribution of opportunities and the fairness of development. The urban-to-rural income ratio is a classical measure of intersectoral income inequality, directly quantifying the developmental gap between urban and rural areas [1]. The number of rural students represents investment in rural human capital, which is crucial for long-term social mobility and breaking the cycle of poverty. The number of rural administrative institutions indicates the presence and accessibility of basic public services and governance in rural areas; these institutions are key entities for policy implementation and ensuring that developmental outcomes are effectively supported.

The GG subsystem is designed to assess the resource use efficiency, environmental pressure, and green transition capacity of rural areas. This study employs two first-level indicators for its measurement: rural energy consumption level and rural green technology level. The rural energy consumption level dimension focuses on evaluating the environmental pressure and resource use efficiency of rural economic activities. The four selected indicators collectively capture the carbon and energy intensity of the rural economy from both production and consumption perspectives, which are central to the concept of green growth [2]. The rural green technology level dimension aims to gauge the capacity and ongoing progress of rural areas in transitioning towards a green economy. Among these, the indicator urban employment in Information transmission, software, and IT Services, while statistically categorized as an urban metric, effectively proxies the overall technological development at the provincial level. Given that rural areas are not closed systems, their development is inevitably influenced by the broader technological environment of the province; this indicator thus reflects the potential technological capabilities and knowledge spillovers that can enable smart agriculture, resource management, and green production methods in rural settings [27,36]. The indicators rural per capita solar energy and solar house area and rural electricity generation directly measure the actual adoption and output scale of renewable energy within rural areas [33]. Meanwhile, the number of rural water-saving irrigation machinery serves as a direct manifestation of the application of green technology in agricultural production, signaling the advancement of resource-conserving farming practices [31].

The RIGG indicator system is designed to reflect its dual nature as the coupling coordination between IG and GG. Therefore, diverging from approaches that directly compute composite indicators using the entropy method, this paper first employs the entropy method to measure the levels of IG and GG separately, then computes their degree of coupling coordination. The resulting coupling coordination degree value is used to represent the RIGG level, which is defined as the coordinated and coupled outcome of IG and GG. Unlike previous studies, this paper measures rural IG and GG levels separately and then calculates their degree of coupling coordination. The coupling coordination degree value obtained is used to represent the level of RIGG, which is defined as the coordinated and coupled outcome of IG and GG. Based on the standardized indicators, the entropy method is used to assign weights to the indicators and calculate the system’s coupling coordination degree, which is derived from both coupling degree and coordination degree. The calculation method for the coupling coordination degree is as follows:

$$\begin{array}{c}C={\left\{U_1\times U_2)/{\left(\left[U_1+U_2\right]/2\right)}^2\right\}}^{1/2}\end{array}$$

(9)

$$\begin{array}{c}T=\alpha \times U_1+\beta \times U_2\end{array}$$

(10)

$$\begin{array}{c}D=\sqrt{C\times T}\end{array}$$

(11)

where, C is the coupling degree between IG and GG systems; T is the coordination degree between the two systems; D is the coupling coordination degree, indicating the level of RIGG; $U_1$ is the comprehensive score for the IG subsystem; $U_2$ is the comprehensive score for the GG subsystem; $\alpha$ and $\beta$ are undetermined weights assigned to the IG and GG subsystems, respectively. In this evaluation, it is assumed that both systems are of equal importance, and therefore, both weights are set to 0.5. In subsequent analyses, we will also assign different ratios to calculate the coupling coordination degree between IG and GG, in order to test the robustness of the results.

The synergy theory explains the dynamic coordination between two systems, progressing from low to high levels. High coordination is achieved when the coupling between the systems is strong, with minimal performance gaps [51]. The coordination degree between IG and GG, the level of RIGG, is shown in

Table 3. Classification criteria for the level of RIGG.

Development Stages	Value Range	Classification
Unacceptable range	$[0, 0.1)$	extreme imbalance
	$[0.1, 0.2)$	serious imbalance
	$[0.2, 0.3)$	moderate imbalance
	$[0.3, 0.4)$	mild imbalance
Transition zone	$[0.4, 0.5)$	almost imbalance
	$[0.5, 0.6)$	almost coordination
Acceptable range	$[0.6, 0.7)$	primary coordination
	$[0.7, 0.8)$	intermediate coordination
	$[0.8, 0.9)$	good coordination
	$[0.9, 1.0)$	perfect coordination

.

Figure 2 illustrates the changes in the provincial distribution of China’s RIGG index between 2011 (a) and 2022 (b). In 2011, eight provinces recorded a RIGG index exceeding 0.4, thereby entering the transition zone. Among these, Beijing was the only region classified at the almost coordination level, while the other seven provinces remained in the almost imbalance category. The remaining 21 provinces fell within the unacceptable range. By 2022, a notable improvement was observed in the spatial pattern of RIGG. The number of provinces in the unacceptable range dropped to three, while 21 provinces advanced to the transition zone, and five provinces reached the acceptable range. Beijing progressed further to the intermediate coordination level.

Overall, China’s RIGG exhibited a marked upward trend from 2011 to 2022, indicating enhanced regional coordination in rural green and inclusive development. This progress can be attributed to a combination of factors. First, the ongoing implementation of the national Rural Revitalization Strategy and agricultural modernization policies has reinforced institutional support for green and inclusive growth in rural areas. Second, the implementation of local measures in areas such as rural industrial ecology, equalization of public services, and improvements to the living environment has progressively yielded results, contributing to the synchronized advancement of the rural economy, society, and environment. Additionally, the application of digital and green technologies in rural areas has played a crucial role in enhancing resource efficiency and fostering inclusive development. These advancements reflect the significant progress China has made in promoting a higher-quality, more equitable, and more sustainable development path for rural areas.

4.2.3. Other Variables

Regarding control variables, drawing on previous studies, IGG is influenced by a multitude of factors. This study selects a set of macro-level control variables at the provincial level to account for key confounding influences. These include:

Industrial structure (ind), measured as the ratio of the value-added of the tertiary industry to that of the secondary industry, which is expected to positively affect RIGG by signaling a shift towards a less polluting, more service-oriented economy.

Rural education level (edu), represented by the number of students enrolled in rural schools, as a proxy for human capital stock, which is anticipated to foster RIGG by enhancing the capacity for adopting green technologies and securing higher-income employment.

Rural internet penetration (internet), indicated by the number of rural broadband subscribers, which is hypothesized to complement IIC and directly promote RIGG by facilitating information sharing, e-commerce, and access to green production techniques.

Rural circulation facilities (road), measured by the length of rural delivery routes in kilometers, which is expected to support RIGG by improving market accessibility for rural products (boosting income) and enhancing the efficiency of logistics networks (reducing energy waste).

As for the mediating variable, this study examines how the development of IIC facilitates the movement of rural labor to urban areas, thereby affecting RIGG. Accordingly, the mediating variable, rural labor mobility (rlm), is measured by the ratio of the number of rural laborers migrating out annually to the total rural labor force.

4.3. Data Source

Data for all control variables were compiled from the China Rural Statistical Yearbook and China Statistical Yearbook. To mitigate the impact of multicollinearity on the model, all variables were log-transformed.

Following rigorous validation and cleaning procedures,

Table 4. Descriptive statistics for variables.

Variable	Obs	Mean	Std. Dev.	Min	Max
lnrigg	348	0.363	0.067	0.206	0.586
lnig	348	0.278	0.103	0.073	0.675
lngg	348	0.126	0.068	0.018	0.350
lniic	348	0.115	0.105	0.003	0.558
lnind	348	0.770	0.238	0.418	1.842
lnedu	348	5.864	0.818	2.935	7.239
lninternet	348	5.128	1.339	0.405	7.354
lnroad	348	11.509	0.863	8.515	12.681
lnrlm	348	0.330	0.110	0.050	0.669

presents the descriptive statistics for all variables in this study, and the correlation analysis of the core variables is presented in

Table A1. Correlation matrix.

Variables	lnrigg	lniic	lnind	lnedu	lninternet	lnroad	lnrlm
lnrigg	1
lniic	0.739 ***	1
lnind	0.606 ***	0.408 ***	1
lnedu	0.187 ***	0.345 ***	−0.327 ***	1
lninternet	0.503 ***	0.590 ***	0.0600	0.706 ***	1
lnroad	0.186 ***	0.342 ***	−0.297 ***	0.785 ***	0.761 ***	1
lnrlm	0.167 ***	0.223 ***	−0.269 ***	0.452 ***	0.528 ***	0.368 ***	1

Note: * p < 0.1, ** p < 0.05, *** p < 0.01.

.

5. Results and Discussion

5.1. Impact of IIC on RIGG

5.1.1. Benchmark Estimation

To examine the relationship between IIC and RIGG, a two-way fixed effects panel model incorporating both province and year effects was employed (

Table 5. Benchmark regression result.

Variables	Fixed Effects			SYS-GMM
	(1)	(2)	(3)	(4)	(5)	(6)
	lnrigg	lnig	lngg	lnrigg	lnig	lngg
lniic	0.2238 ***	0.2197 ***	0.4156 ***	0.0931 ***	0.0979 ***	0.1793 ***
	(9.46)	(8.20)	(12.61)	(11.28)	(4.80)	(20.13)
lnind	0.0660 ***	0.1672 ***	0.0322 **	0.0383 ***	0.1246 ***	−0.0230 ***
	(6.03)	(13.50)	(2.11)	(15.85)	(27.70)	(−4.80)
lnedu	0.0069 *	0.0040	0.0043	0.0027 ***	0.0026 **	0.0014
	(1.69)	(0.88)	(0.77)	(3.10)	(2.33)	(0.60)
lninternet	0.0088 ***	−0.0006	0.0095 ***	0.0022 *	−0.0019	0.0053 ***
	(4.13)	(−0.25)	(3.20)	(1.93)	(−1.57)	(5.71)
lnroad	0.0059 **	0.0068 **	0.0054	0.0062 ***	0.0131 ***	0.0001
	(1.99)	(2.02)	(1.31)	(3.98)	(11.02)	(0.08)
L.lnrigg				0.6511 ***
				(21.45)
L.lnig					0.6904 ***
					(33.50)
L.lngg						0.5901 ***
						(37.39)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Province FE	Yes	Yes	Yes	Yes	Yes	Yes
N	348	348	348	290	290	290
R2	0.9772	0.9877	0.9569
AR (1)				−2.98	−3.71	−2.22
				[0.003]	[0.000]	[0.027]
AR (2)				−1.22	−0.99	−1.27
				[0.223]	[0.324]	[0.205]
Hansen test of overid. restrictions				27.98	27.24	25.62
				[1.000]	[1.000]	[1.000]

Note: t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01, p-values of Arellano-Bond test for AR (1), AR (2) and Hansen test of overid. Restrictions are in [x].

). The results show a statistically significant positive coefficient for lniic at the 1% level, indicating that higher IIC levels significantly enhance RIGG. To address potential endogeneity and improve estimation accuracy, the SYS-GMM model was further applied to re-estimate this relationship. Before interpreting the dynamic panel estimates, the validity of two key assumptions was examined. The empirical results in Table 5 reveal that the p-values for AR (1) and AR (2) are less than 0.1 and greater than 0.1, respectively, supporting the appropriateness of the estimation strategy. Moreover, the sign of the estimated coefficient in the fixed effects model is consistent with that from the SYS-GMM model, confirming that IIC exerts a significant positive effect on RIGG, thus validating the rationale of Hypothesis 1.

Given that RIGG is derived from the coupling coordination between IG and GG, we further investigated the impact of IIC on IG and GG separately using both the two-way fixed effects and SYS-GMM models. The results demonstrate that IIC significantly promotes both IG and GG, aligning with its positive effect on RIGG. It is worth noting, however, that the magnitude of IIC’s impact on GG is stronger than that on IG and RIGG.

5.1.2. Asymmetric Check

To investigate the asymmetric effects of IIC on RIGG, this study adopts a panel quantile regression technique. The particular regression findings are shown in

Table A2. Results of Quantile Regression: lniic on lnrigg.

Variables	10	25	50	75	90
lniic	0.2362 ***	0.1859 ***	0.1810 ***	0.1754 ***	0.1964 ***
	(7.29)	(6.50)	(5.38)	(3.06)	(7.43)
lnind	0.0571 ***	0.0703 ***	0.0666 ***	0.0648 ***	0.0559 ***
	(2.69)	(3.25)	(3.33)	(3.40)	(5.20)
lnedu	−0.0004	0.0043	0.0064	0.0099	0.0021
	(−0.11)	(0.73)	(0.56)	(0.80)	(0.85)
lninternet	0.0069 ***	0.0065 ***	0.0065 ***	0.0053 *	0.0072 ***
	(3.12)	(2.81)	(2.78)	(1.78)	(3.71)
lnroad	0.0074 ***	0.0086 ***	0.0073	0.0061	0.0057
	(2.73)	(4.35)	(0.60)	(0.65)	(1.20)
Constant	0.1933 ***	0.1581 ***	0.1721	0.1807 *	0.2347 ***
	(4.44)	(4.39)	(1.17)	(1.84)	(4.36)
N	348	348	348	348	348
R2	0.9668	0.9729	0.9748	0.9696	0.9653

Note: t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

. Figure 3 shows how the calculated coefficients of IIC vary between RIGG quantiles. Overall, IIC consistently has a beneficial influence on RIGG, however the intensity of this effect varies somewhat. Specifically, at lower quantiles of RIGG, such as the 10th quantile, the regression coefficient of IIC is much higher than at other quantile levels, demonstrating that the improving impact of IIC is more noticeable in places with originally poorer RIGG performance. Furthermore, this study looks at the asymmetric effects of IIC on IG and GG. The results are shown in

Table A3. Results of Quantile Regression: lniic on lnig.

Variables	10	25	50	75	90
lniic	0.1990 ***	0.1937 ***	0.1761 ***	0.1932 ***	0.1847 ***
	(4.28)	(4.14)	(3.82)	(6.60)	(8.94)
lnind	0.2021 ***	0.1884 ***	0.2024 ***	0.1787 ***	0.1953 ***
	(12.78)	(10.46)	(8.93)	(15.92)	(24.23)
lnedu	0.0001	−0.0006	−0.0010	0.0028	0.0042 **
	(0.01)	(−0.14)	(−0.21)	(0.87)	(2.19)
lninternet	0.0010	0.0004	0.0002	−0.0017	−0.0037 ***
	(0.20)	(0.06)	(0.07)	(−0.93)	(−2.74)
lnroad	0.0050	0.0016	0.0044	0.0083 **	0.0113 ***
	(0.89)	(0.47)	(1.12)	(2.04)	(6.13)
Constant	0.0181	0.0945 **	0.0668	0.0607	0.0138
	(0.29)	(2.27)	(1.28)	(1.49)	(0.66)
N	348	348	348	348	348
R2	0.9814	0.9834	0.9861	0.9854	0.9807

Note: t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

and

Table A4. Results of Quantile Regression: lniic on lngg.

Variables	10	25	50	75	90
lniic	0.3777 ***	0.4226 ***	0.3887 ***	0.4296 ***	0.4163 ***
	(2.79)	(5.86)	(4.17)	(11.09)	(8.24)
lnind	0.0264	0.0344	0.0293	0.0288	0.0195 *
	(1.06)	(1.61)	(1.26)	(1.62)	(1.71)
lnedu	0.0054	0.0048	0.0074	0.0018	−0.0007
	(0.68)	(0.65)	(1.24)	(0.35)	(−0.30)
lninternet	0.0078 **	0.0076 **	0.0078 ***	0.0058	0.0057 ***
	(2.34)	(2.04)	(2.86)	(1.64)	(3.93)
lnroad	0.0060	0.0081 *	0.0140 ***	−0.0002	−0.0032
	(0.99)	(1.67)	(3.66)	(−0.01)	(−0.46)
Constant	−0.0809	−0.1020	−0.1406 **	0.0335	0.1052
	(−0.85)	(−1.38)	(−2.29)	(0.26)	(1.35)
N	348	348	348	348	348
R2	0.9358	0.9470	0.9514	0.9474	0.9378

Note: t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

. Figure 4 and Figure 5 present the changes in the coefficients of IIC on IG and GG, respectively. Although the influence of IIC on both IG and GG varies somewhat across quantiles, the differences are relatively small, suggesting that the effect of IIC on these two indicators remains stable.

In summary, IIC generally plays a positive role in promoting RIGG. However, the strength of its impact exhibits an asymmetric pattern depending on the initial level of RIGG, with a more substantial promoting effect observed particularly in regions with lower RIGG levels.

5.2. Robustness Test

5.2.1. Endogenous Problems

Although our analysis incorporates fixed effects model and SYS-GMM model to account for key confounding factors in RIGG estimation, potential endogeneity concerns may persist, potentially limiting the robustness of our findings. To address this methodological challenge, we implement the instrumental variable method.

Following the identification strategy developed by Ivus et al. (2015), we employ terrain relief as an instrumental variable for IIC [52]. The instrument variable meets the relevance criterion because terrain relief affects IIC expenses. Rugged terrain increases installation expenses and technical barriers, negatively correlating with IIC feasibility. It also meets the exclusion restriction, being a geologically fixed factor independent of socioeconomic conditions. Given its constancy, the paper uses the product of terrain relief and yearly Internet ports per province as the instrumental variable (iv1) for 2SLS regression.

Furthermore, following the technique of Gao et al. (2024), this analysis employs the number of fixed-line telephones per 100 persons in 1984 as an instrumental variable for IIC [53]. The choice of this instrument is defended as follows: The early adoption of fixed-line telephones represented the nascent form of IIC, and places with higher historical telephone penetration rates provided a firmer foundation for subsequent IIC development, satisfying the relevance criterion. At the same time, as a historical and predefined variable, early fixed-line telephone penetration is unlikely to have a direct impact on current outcomes like RIGG. Its effect is focused entirely on shaping the emergence of current IIC, which meets the exclusion criteria.

Because the number of fixed-line telephones per 100 inhabitants at the province level in 1984 was cross-sectional, this study’s fundamental sample data is balanced panel data. To address this, we use the technique suggested by Nunn and Qian (2014) to generate the panel instrumental variable, which includes a time-varying instrumental variable [54]. Specifically, we generate the instrumental variable (iv2) for IIC by combining the number of fixed-line telephones per 100 persons in 1984 with the number of broadband internet users the previous year.

The regression results are presented in

Table 6. Endogeneity test results.

Variables	(1)	(2)	(3)	(4)	(5)	(6)
	The First Stage	The Second Stage	The First Stage	The Second Stage	The First Stage	The Second Stage
lniic		0.1523 ***		0.2513 ***		0.2051 ***
		(3.0884)		(5.6274)		(6.3429)
iv1	0.1966 ***				0.2178 ***
	(8.1823)				(8.0938)
iv2			0.0115 ***		0.0126 ***
			(5.8179)		(7.5206)
Constant	0.0816	0.1689 ***	−0.0476	0.1659 ***	0.0017	0.1673 ***
	(0.8836)	(4.3174)	(−0.4544)	(4.3582)	(0.0221)	(4.3828)
Kleibergen-Paap rk LM	22.581		35.65		61.806
	(0.0000)		(0.0000)		(0.0000)
Cragg-Donald Wald F	67.53		81.063		115.247
	[16.38]		[16.38]		[19.93]
Hansen J statistic					2.594
					(0.1073)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Province FE	Yes	Yes	Yes	Yes	Yes	Yes
Observations	348	348	348	348	348	348

Note: * p < 0.1, ** p < 0.05, *** p < 0.01. z-values and p-values of estimated coefficients are in (x), and critical values of Stock-Yoho weak instrumental variable identification F-test at the 10% significance level are in square brackets.

. Columns (1) and (2) report the two-stage least squares results using only instrumental variable iv1, while columns (3) and (4) report those using only iv2. In both settings, the Kleibergen-Paap rk LM statistic is significant at the 1% level, rejecting the null hypothesis of underidentification. Furthermore, the Cragg-Donald Wald F statistic exceeds the Stock-Yogo critical value at the 10% level, indicating no concern for weak instruments. Since only one instrument is used in each specification, over-identification is not an issue. These results suggest that both iv1 and iv2 are valid and reliable instrumental variables.

We also use both iv1 and iv2 as instruments for the main explanatory variable lniic. The results are presented in Table 6, columns (5) and (6). The Kleibergen-Paap rk LM statistic is still significant at the 1% level, rejecting underidentification. The Cragg-Donald Wald F statistic also exceeds the Stock-Yogo weak instrument test critical value at the 10% threshold, indicating that there are no weak instruments. Furthermore, the Hansen J statistic produces a p-value of 0.1073, which does not reject the null hypothesis that all instruments are exogenous, implying that there is no over-identification problem.

5.2.2. Robustness Check: Exclusion of Outliers

Outliers in the sample may affect the coefficient estimates. To mitigate potential distortion caused by outliers, all continuous variables were winsorized at the 1% level on both tails. Columns (1) to (3) in Table 6 present the results of robustness tests after removing outliers, demonstrating that the empirical findings remain reliable and stable. The results indicate that the estimated coefficients of IIC on RIGG, IG, and GG are all significantly positive at the 1% level, consistent with the baseline findings, thereby passing the robustness check.

5.2.3. Replacing the Independent Variable

Firstly, in using the entropy method, we applied the min-max method to standardize the original data. For the robustness test, we used the Z-score method to standardize the original data and then applied the entropy method to calculate IIC.

$$\begin{array}{c}Z={\displaystyle \frac{X-\mu }{\sigma }}\end{array}$$

(12)

In the formula, $X$ represents the raw observed values of the indicators used to construct IIC, $\mu$ denotes the sample mean, and $\sigma$ is the standard deviation. The resulting value $Z$ is the standardized measure of IIC obtained using the z-score method.

The results presented in columns (4) to (6) of

Table 7. Robustness results.

Variables	Exclusion of Outliers			Replacing the Independent Variable
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
	lnrigg	lnig	lngg	lnrigg	lnig	lngg	lnrigg	lnig	lngg
lniic	0.2238 ***	0.2197 ***	0.4156 ***
	(9.46)	(8.20)	(12.61)
lniic_zscore				0.0842 ***	0.0936 ***	0.1411 ***
				(9.30)	(9.44)	(10.67)
lniic_pca							0.0791 ***	0.0749 ***	0.1341 ***
							(8.45)	(7.07)	(9.80)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Province FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	348	348	348	348	348	348	348	348	348
R2	0.9772	0.9877	0.9569	0.9770	0.9884	0.9522	0.9761	0.9871	0.9500

Note: t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

show that the coefficients of lniic_zscore are positive and statistically significant at the 1% level for lnrigg, lnig, and lngg. This confirms that IIC significantly promotes RIGG, IG, and GG. The close alignment of these results with our baseline estimates further reinforces the robustness of the study’s main findings.

Secondly, the entropy method was initially employed to measure IIC. As a commonly employed alternative, we also applied Principal Component Analysis (PCA). Consequently, we recalculated IIC using the PCA method based on the original dataset and re-estimated its impact on RIGG, IG, and GG using a fixed effects model. As shown in columns (7) to (9) of Table 7, the coefficients of IIC on RIGG, IG, and GG are all significantly positive at the 1% level, indicating that IIC promotes the development of RIGG, IG, and GG. These results align with those obtained from the entropy method, further confirming the robustness of our findings.

5.2.4. Replacing the Dependent Variable (RIGG)

In the coupling coordination model, the parameters $\alpha$ and $\beta$ in the calculation of the coordination degree T are set to $\alpha =\beta =0.5$, assuming equal importance between IG and GG, following common practice. The corresponding results are reported in column (3) of

Table 8. Robustness results: Replacing the dependent variable.

Variables	(1)	(2)	(3)	(4)	(5)
lniic	0.2589 ***	0.2391 ***	0.2238 ***	0.2120 ***	0.2028 ***
	(9.32)	(9.45)	(9.46)	(9.30)	(8.99)
lnind	0.0480 ***	0.0575 ***	0.0660 ***	0.0738 ***	0.0810 ***
	(3.74)	(4.91)	(6.03)	(7.00)	(7.76)
lnedu	0.0083 *	0.0076 *	0.0069 *	0.0062	0.0055
	(1.75)	(1.75)	(1.69)	(1.58)	(1.43)
lninternet	0.0101 ***	0.0093 ***	0.0088 ***	0.0084 ***	0.0083 ***
	(4.04)	(4.09)	(4.13)	(4.13)	(4.08)
lnroad	0.0040	0.0049	0.0059 **	0.0071 **	0.0085 ***
	(1.16)	(1.54)	(1.99)	(2.49)	(2.98)
Constant	0.1230 ***	0.1306 ***	0.1335 ***	0.1326 ***	0.1288 ***
	(2.63)	(3.06)	(3.34)	(3.45)	(3.39)
Year FE	Yes	Yes	Yes	Yes	Yes
Province FE	Yes	Yes	Yes	Yes	Yes
N	348	348	348	348	348
R2	0.9704	0.9744	0.9772	0.9790	0.9798

Note: t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

. To examine the robustness of this assumption, we adjusted the weights of IG and GG.

First, when assigning greater importance to GG, we considered two scenarios: (1) ($\alpha =0.3, \beta =0.7$) and (2) ($\alpha =0.4, \beta =0.6$). The results, shown in columns (1) and (2) of Table 8, indicate that the coefficient of lniic on lnrigg remains significantly positive, suggesting that IIC continues to promote RIGG.

Second, when giving greater weight to IG, we considered two scenarios: (1) ($\alpha =0.6, \beta =0.4$) and (2) ($\alpha =0.7, \beta =0.3$). The results, shown in columns (4) and (5) of Table 8, also show consistently positive and significant coefficients.

Therefore, varying the relative weights of IG and GG in the coordination degree calculation leads to minor differences in RIGG values but does not alter the positive effect of IIC on RIGG. The conclusion that IIC promotes RIGG remains robust.

5.3. Heterogeneity Test

Columns (1) and (2) of

Table 9. The heterogeneous effects of urban and rural areas.

Variables	Level of Urbanization		Different Regions of China
	(1)	(2)	(3)	(4)	(5)
	lnrigg	lnrigg	lnrigg	lnrigg	lnrigg
lniic	0.2931 ***	0.2980 ***	0.3067 ***	0.6178 ***	0.2131 ***
	(7.34)	(7.45)	(9.23)	(6.90)	(4.17)
lnind	0.0830 ***	0.0636 ***	0.0498 **	0.1448 ***	0.0288 *
	(2.94)	(5.32)	(2.10)	(9.02)	(1.78)
lnedu	0.0049	0.0096 *	0.0066	0.0526 ***	−0.0080
	(0.70)	(1.87)	(1.40)	(3.92)	(−1.19)
lninternet	0.0133 ***	0.0047	0.0125 ***	−0.0148 **	0.0085 ***
	(4.08)	(1.63)	(4.23)	(−2.63)	(2.63)
lnroad	−0.0549 ***	0.0105 ***	−0.0558 ***	0.0055	0.0055 *
	(−3.10)	(3.27)	(−3.83)	(0.24)	(1.95)
Constant	0.7917 ***	0.0789 *	0.8342 ***	−0.1171	0.2549 ***
	(4.29)	(1.74)	(5.25)	(−0.42)	(5.06)
Year FE	Yes	Yes	Yes	Yes	Yes
Province FE	Yes	Yes	Yes	Yes	Yes
N	96	252	120	96	132
R2	0.9836	0.9740	0.9808	0.9814	0.9850
The p-values for coefficient differences	[0.9384]		[0.0012]
				[0.0001]

Note: t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01. The p-values for coefficient differences are based on the test of group coefficient differences using the seemingly unrelated estimation (suest) method.

report the estimated effects of IIC on RIGG in highly urbanized and less urbanized areas, respectively. The regression results demonstrate that IIC significantly promotes RIGG in both types of regions. Although the estimated coefficients differ, tests based on Seemingly Unrelated Estimation (SUEST) indicate that this difference is statistically insignificant, suggesting that the effect of IIC on RIGG is not significantly moderated by the level of urbanization.

Columns (3) through (5) of Table 9 show the estimated impacts of IIC on RIGG in China’s eastern, central, and western regions, respectively. The empirical data show that IIC has a considerable favorable influence on RIGG in all areas, however the intensity of this effect varies greatly by region. IIC has a very strong promoting impact in the central area. Seemingly Unrelated Estimation (SUEST) tests demonstrate that coefficient differences across the central, eastern, and western areas are statistically significant. This basic conclusion is supported by independent findings on geographical variance in rural inclusive development [23] and the spatially varied impact of the digital economy, with the central area showing the largest reaction [18].

This heterogeneity can be insightfully interpreted through the lens of new urbanization. The central region, in its current developmental phase, often acts as a critical nexus for the inter-regional flow of factors spurred by new urbanization policies [23]. Unlike the more saturated eastern markets, the central region possesses significant growth potential and absorption capacity. Simultaneously, compared to the western region, it typically boasts more robust foundational conditions. Therefore, the synergistic effect between IIC—which provides the “digital arteries”—and new urbanization—which facilitates the “physical and economic integration”—is likely most potent in the central region. This synergy unlocks greater marginal benefits by optimally channeling urban capital, technology, and talent into rural revitalization, thereby producing the strongest observed promotion effect on RIGG.

This variation may be due to the combined effect of regional industrial growth phases and policy dividends. On the one hand, the central area is now in the process of relocating industrial facilities from the eastern region and modernizing its own industrial structure. As a general-purpose technology, IIC has the potential to significantly empower this process, increasing industrial efficiency and green transformation. In contrast, in the eastern area, where the service sector is dominant, the marginal effect of IIC may be quite small. On the other hand, the implementation of national policies such as the “Rise of Central China” program has offered increased policy support and resource allocation to the central region, hence increasing IIC’s efficacy.

5.4. Transmission Channel

The aforementioned empirical results confirm the positive impact of IIC on RIGG. To identify the underlying channels, this study focuses on the mediating role of RLM, as suggested by theoretical analysis, and employs the following mediation model to empirically examine this mechanism:

$$\begin{array}{c}{lnRLM}_{i,t}={\gamma }_0+{\gamma }_1{lnIIC}_{i,t}+\rho ln{Control}_{i,t}+{\lambda }_t+{\sigma }_i+{\epsilon }_{i,t}\end{array}$$

(13)

$$\begin{array}{c}{lnRLM}_{i,t-1}={\gamma }_0+{\gamma }_1{lnIIC}_{i,t}+\rho ln{Control}_{i,t}+{\lambda }_t+{\sigma }_i+{\epsilon }_{i,t}\end{array}$$

(14)

In our model specification, we examine the effects of IIC on both the current term of RLM and its one-period lagged value. This approach not only accounts for the potential time-lagged nature of IIC’s influence but also helps mitigate potential reverse causality between the variables. Besides, to address potential endogeneity concerns, we also employ a SYS-GMM model to further test the channel through which IIC influences RIGG.

$$\begin{array}{c}{lnRLM}_{i,t}={\eta }_0+{\eta }_1{lnRLM}_{i,t-1}+{\eta }_2{lnIIC}_{i,t}+\rho ln{Control}_{i,t}+{\lambda }_t+{\sigma }_i+{\epsilon }_{i,t}\end{array}$$

(15)

$$\begin{array}{c}{lnRLM}_{i,t-1}={\eta }_0+{\eta }_1{lnRLM}_{i,t-2}+{\eta }_2{lnIIC}_{i,t}+\rho ln{Control}_{i,t}+{\lambda }_t+{\sigma }_i+{\epsilon }_{i,t}\end{array}$$

(16)

${lnRIGG}_{i,t}$ denotes the dependent variable for province i in year t, and ${lnRIGG}_{i,t-1}$ represents its first-order lag term. ${lnRLM}_{i,t}$ and ${lnRLM}_{i,t-1}$ are mediating variable in this study, which is the rural labor mobility and its first-order lag term. $ln{Control}_{i,t}$ refers to the set of control variables. In the two-step mediation effect model, the focus lies on testing whether the effects of the core explanatory variable on both the dependent variable and the mediating variable are statistically significant [50].

The mediating impact of IIC on RLM is reported in

Table 10. Mechanism verification.

Variables	Fixed Effects		SYS-GMM
	(1)	(2)	(3)	(4)
	lnrlm	L.lnrlm	lnrlm	L.lnrlm
lniic	0.1039 ***	0.1481 ***	0.0281 **	0.0852 ***
	(2.72)	(3.67)	(2.65)	(4.12)
L.lnrlm			0.8771 ***
			(104.25)
L2.lnrlm				0.9724 ***
				(46.52)
AR (1)			−3.13	−3.53
			[0.002]	[0.000]
AR (2)			1.21	1.17
			[0.225]	[0.241]
Hansen test of overid. restrictions			28.34	25.60
			[1.0000]	[1.0000]
Control variables	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Province FE	Yes	Yes	Yes	Yes
N	348	319	290	261

Note: t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01. p-values of Arellano-Bond test for AR (1), AR (2) and Hansen test of overid. Restrictions are in [x].

column (1). The findings demonstrate that, at the 1% level, the coefficient of lniic on lnrlm is significantly positive. Additionally, lniic considerably improves lnrlm under the SYS-GMM design (column (3)). These results imply that increased provincial IIC levels efficiently promote rural labor mobility, which in turn helps to improve RIGG in rural regions.

We add the lagged term of RLM in column (2) to investigate the dynamic nature of this connection. The findings demonstrate that the coefficient on lniic is still significantly positive, suggesting that IIC’s boosting influence on RLM is enduring. There are a number of reasons for this delayed impact. First, the improvement of IIC and its penetration into rural labor markets takes time, as its impact on the dissemination of job information, the enhancement of workers’ digital literacy, and the translation into remote employment opportunities typically materializes gradually. Second, labor migration decisions are inherently sticky; there is often an adaptation and adjustment period for workers as they become aware of, build trust in, and make arrangements utilizing new information channels. We also use the SYS-GMM estimator to address possible endogeneity issues, and the findings (column (4)) continue to be strong, demonstrating the dependability of this delayed impact.

The observed gradual manifestation of IIC’s influence over time is conceptually echoed and empirically substantiated by Ma et al. (2025) [39]. Their research illustrates how digital financial inclusion disrupts the entrenched “preference for proximity” in migration—a behavioral pattern whose transformation inherently occurs over an extended period. The process of overcoming information barriers and facilitating a structural shift of laborers into the tertiary sector [39] logically accounts for the persistent effects identified in our study. The substantial scale of this digital influence is further corroborated by Wang et al. [40], who report that overall digitization elevates labor mobility by 51.04%, thereby contextualizing the significant long-term coefficient observed in our model. This result aligns with the New Economics of Labor Migration theory, which posits that labor migration and subsequent remittances can enhance rural livelihoods by alleviating production constraints, diversifying income sources, and supplying investment capital [55]. The powerful, digitally driven reallocation of labor resonates with the perspective that mobility constitutes a development opportunity [56]. Our study specifically identifies IIC as a key infrastructure that renders this opportunity more accessible and sustainable for rural laborers.

6. Conclusions, Policy Implications and Limitations

Based on panel data from rural areas of 29 provinces in China between 2011 and 2022, this study empirically examines the impact of IIC on RIGG. The main findings can be summarized as follows:

(1) IIC significantly promotes RIGG. This conclusion remains robust after applying two-way fixed effects models, system GMM estimations, and a series of robustness checks, thereby validating Hypothesis 1. Moreover, IIC also exerts significantly positive effects on both inclusive growth and green growth.

(2) Quantile regression results indicate that the marginal promoting effect of IIC is stronger in provinces with initially lower levels of RIGG, suggesting that IIC plays a more substantial role in driving development in lagging regions.

(3) Heterogeneity analysis reveals regional disparities in the impact of IIC on RIGG, with a more pronounced effect in the central region compared to the eastern and western regions.

(4) Mechanism tests demonstrate that IIC indirectly fosters RIGG primarily by accelerating rural labor mobility, and this effect exhibits persistence over time, thus supporting Hypothesis 2.

Based on the above findings, the following policy recommendations are proposed:

(1) Enhance targeted investment in IIC to empower less developed regions. Given the quantile regression results indicating a stronger marginal effect of IIC in areas with initially lower levels of RIGG, it is recommended that under national initiatives such as the “East Data, West Computing” project and the “Digital Village” program, priority be given to counties ranked lower in the RIGG index. A “one-county-one-policy” approach should be adopted to strengthen IIC in these regions. Investment should focus on cost-effective mobile internet and 5G networks to rapidly bridge information gaps and narrow inter-regional digital disparities.

(2) Implement regionally differentiated IIC development strategies to improve overall resource allocation efficiency. Heterogeneity analysis reveals that IIC has the most pronounced promoting effect on RIGG in the central region. Therefore, at the national level, efforts to advance the “Digital Village” initiative and deploy new infrastructure should prioritize resource allocation to central, western, and less urbanized counties. This includes accelerating the deployment of 5G base stations, the Internet of Things, rural broadband, and data centers to fully leverage IIC’s role in driving green and inclusive growth. Meanwhile, in the eastern region, where infrastructure is relatively advanced, emphasis should be placed on the high-quality integration and application of IIC. Deepening its use in sectors such as smart agriculture, rural tourism, and e-commerce will facilitate the green transformation of modern agriculture and further enhance IIC’s dual support for both green growth and inclusive growth.

(3) Improve digital services for labor mobility to unleash the potential of rural human resources. Mechanism tests confirm that rural labor mobility plays a key mediating role in the process by which IIC promotes RIGG. It is recommended that the government leverage IIC to build an integrated urban-rural information platform, break down information barriers in the labor market, and promote orderly mobility of rural labor. Furthermore, a comprehensive digital service system covering recruitment, training, and rights protection should be established to reduce the costs of mobility and improve the efficiency and quality of labor allocation. In regions experiencing labor outflows, specialized digital skills training should be provided to remaining residents to enhance their employability in new sectors such as remote work, e-commerce entrepreneurship, and smart agriculture. This will help mitigate rural hollowing-out and achieve a positive interaction between population mobility and local development.

Although this study systematically examines the impact of IIC on RIGG, including its asymmetric and heterogeneous effects as well as underlying mechanisms, certain limitations still exist, offering opportunities for future research.

First, regarding indicator construction, although the entropy method and the coupling coordination degree model were employed to measure IIC and RIGG, respectively, which to some extent mitigates the subjectivity and bias associated with single-indicator measures, the evaluation process is not entirely free from subjective factors. Currently, there is no unified standard in academia for measuring IIC and RIGG, and the relevant indicator systems are still in the exploratory stage. As national-level statistical data continue to improve in the future, more comprehensive and authoritative indicators could be introduced to enhance the scientific rigor and comparability of the measurements. Second, the underlying mechanisms of regional heterogeneity have not been fully investigated. This study finds that the promoting effect of IIC on RIGG is more pronounced in provinces with initially lower development levels and is strongest in the central region. As the causes of this phenomenon fall outside the focus of our research, we have only provided a preliminary interpretation based on the contextual background, without constructing mechanism-testing models to thoroughly uncover the specific pathways behind these regional disparities. Future research could introduce moderation or mediation effect models to explore the intrinsic mechanisms of regional heterogeneity from multiple dimensions, such as regional foundations, policy environments, and industrial structure. Additionally, case studies or other qualitative methods could be applied for in-depth validation.