Does New Infrastructure Promote the Development of Rural Industries? A Nonlinear Analysis Based on Provincial Panel Data from China

Abstract

Whether and how new infrastructure (NI) can drive the development of rural industries (DRI) is crucial to promoting urban–rural balance and ensuring national food security. Based on panel data from Chinese provincial-level regions (2013–2022), this study constructs a comprehensive DRI evaluation system encompassing production systems, economic benefits, industrial integration, and sustainable development. Using the entropy method to measure NI development levels, we examine its heterogeneous impacts on DRI through multiple analytical dimensions. The results reveal three key findings: First, a robust “inverted U-shaped” relationship exists between NI and DRI. Second, heterogeneity analysis demonstrates that (1) both eastern and western regions show this pattern while central and northeastern regions exhibit infrastructure saturation; (2) intelligent transportation infrastructure critically moderates this relationship—advanced regions achieve greater NI efficiency through digital infrastructure–transportation–industrial synergies, whereas underdeveloped regions face bottlenecks; and (3) provinces with lower population density but higher human capital show enhanced NI absorption capacity. Third, mechanism analysis confirms industrial structure upgrading and market expansion as key transmission channels. These findings suggest that implementing tiered intelligent transportation strategies, differentiated NI policies, human capital investment, and rural market expansion optimize NI’s rural development impacts.

1. Introduction

Within the overarching framework of global economic integration and digital transformation, digital technologies represented by artificial intelligence (AI) have fundamentally transformed the spatial production organization model based on physical space, while the virtual–physical space integration process driven by digital models has reshaped the logical system of rural spatial governance [1]. Against this backdrop, the long-standing impediments to urban–rural factor flow and the urban–rural value divide that have hindered rural development and integrated urban–rural development will gradually be mitigated, with digital rural spatial governance emerging as a critical breakthrough point for advancing the modernization of rural spatial governance and rural industrial development. On one hand, comprehensive urban infrastructure planning, improved accessibility and transportation potential of rural public service facilities [2], the application of Information and Communication Technologies (ICT), and the digital economy [2,3] collectively enhance rural resource efficiency and facilitate urban–rural factor mobility, while rapid transportation infrastructure development and significant progress in information technology promote the rescaling of urban–rural spaces and the flattening of spatial governance [4]. On the other hand, excessive transportation infrastructure and mobility may impose certain pressures on rural activities [5], which compounded by factors such as rural geographical remoteness, sparse population, and technological deficiencies [6], result in inherent weaknesses including overall low competitiveness, limited industrial chain extension, imperfect value chains, and insufficient breadth and depth of agricultural and rural service sector development [7]. These challenges in turn constrain the advancement of information technologies in rural areas, creating a “digital divide” [2] that further exacerbates urban–rural disparities. Consequently, leveraging spatial optimization and industrial upgrading based on information technologies has become a pivotal approach to addressing the issues of inadequate and uneven rural development.

Distinct from traditional infrastructure, new infrastructure (NI) is characterized by technology-driven innovation, information-network foundations, and features of novelty, interconnectivity, and sharing. Its unique capability to transcend physical boundaries may fundamentally transform conventional regional and rural development paradigms [8,9,10]. Stratan (2024) provides empirical evidence from Moldova showing that both transport and telecommunications infrastructure are critical enablers of business resilience, trade expansion, and regional inclusion [11]. However, Maciulyte and Butkus (2022), through their analysis of data from 28 EU countries, found that not all infrastructure types contribute significantly to economic growth, with only transport, information and communication technologies (ICT), and energy-related infrastructure demonstrating statistically positive impacts [12]. Notably, converged infrastructure (e.g., transportation and utilities) as physical networks and information infrastructure (e.g., the Internet) as relational networks collectively connect regions into an integrated system [13,14], exerting positive influences on regional productivity [15], interregional trade [16], and resource allocation in response to extreme weather events [17]. However, as traditional infrastructure such as roads and railways approaches saturation, diminishing marginal returns on investment have become evident [18,19]. Against this backdrop, a pivotal question emerges: Can NI continue to serve as a catalyst for rural industrial development while overcoming the limitations faced by traditional transportation infrastructure?

Within the context of rural industrial development, there remains ongoing academic debate regarding whether NI continues to drive growth and innovation or, similar to traditional infrastructure, eventually encounters diminishing returns [18,20]? On one hand, NI has the potential to overcome the constraints of traditional infrastructure by enhancing the efficiency and quality of economic operations [21]. Investments in NI such as high-speed rail and postal systems introduce novel products and markets to rural areas, stimulating innovation among agricultural household enterprises and enhancing economic value in rural regions [22]. The development of information infrastructure effectively mitigates rural income inequality [23,24], while enabling farmers to bridge both access and utilization divides [25]. On the other hand, as NI becomes more widespread and mature, the initial burst of growth effects may subside, leading to questions about the sustainability of its impact on rural industries [26,27]. The siting and operation of converged infrastructure projects such as high-speed rail have been shown to exacerbate intra-regional inequalities [28]. Furthermore, NI development alone cannot overcome agricultural specialization constraints imposed by legacy infrastructure systems [29], nor can it serve as an effective short-term stimulus for economic growth [30].

Building upon these considerations, this paper posits that the relationship between NI and the development of rural industries (DRI) may not be a straightforward linear one. Instead, it may exhibit a nonlinear pattern, where the benefits of NI investment in rural areas could follow an inverted U-shaped curve. The theoretical framework and hypotheses of this study are outlined in the third section, which seeks to unravel this puzzle. The subsequent chapters are arranged as follows: The second section provides a literature review. The third section outlines the theoretical framework and hypotheses of this study. The fourth section provides the model construction and variable explanations, as well as detailing the data sources and providing descriptive statistics. The fifth section presents the empirical analysis, including model regression results, robustness tests, heterogeneity analysis, and mediation effects. The sixth section provides a discussion. The seventh section concludes the study and offers policy implications. By addressing this theoretical conundrum, this paper aims to contribute to the understanding of how NI can be leveraged to promote sustainable DRI, providing insights for policy formulation and rural development strategies both in China and globally.

2. Literature Review

The literature pertinent to this study can be categorized into two main branches: research on the measurement of DRI indices and on the impact effects of NI on rural development.

Regarding the first branch, there is a dearth of research on measuring DRI indices. In 2017, China introduced the Rural Revitalization Strategy, which marked the first formal introduction of the concept of rural industry. The existing literature predominantly features qualitative studies on DRI, with a notable lack of quantitative research. An index for the revitalization of rural industries was developed focusing on four key dimensions: the degree of integration within rural industries, overall production capacity, the progress of specialized industries, and the establishment of interest linkage mechanisms [31,32]. Subsequent research evaluated the influence of this index on the income of those residing in rural areas [32]. It was discovered that the integration of rural industries notably boosts the income of farmers and helps to narrow the disparity between urban and rural regions [33]. Further analysis was conducted to understand how the rural digital economy propels the rejuvenation of rural industries, creating a measurement tool that assesses the level of rural industrial revitalization through three dimensions: the rise in agricultural production, the expansion of agricultural value, and the improvement of farmers’ earnings [34]. DRI indices commonly employ the comprehensive index approach, choosing indicators from areas such as rural production, industrial integration, industrial roles, and farmers’ incomes. Due to the benefits of objectivity, methodological rigor, and precise weighting, the entropy method or entropy weight method is frequently adopted in specific studies for quantification [35,36].

Regarding the second branch of the literature, existing research posits that NI plays a significant role in promoting agricultural development, increasing farmers’ incomes, and enhancing the efficiency of rural governance [37,38,39]. The primary research conclusions are divided into two types. The first is that NI has a significant positive impact on rural development [30]. NI, incorporating technologies and models like the Internet of Things, big data, and artificial intelligence, has propelled the upgrading and innovation of rural industries [40]. It has facilitated the digital transformation from traditional agriculture to handicrafts [41], enhancing production efficiency [42]. Additionally, NI has expanded the channels and markets for commodity sales, enabling rural industries to enter the global market through e-commerce platforms, attracting more consumers and investors. Furthermore, through services like online education and remote training, NI has improved the knowledge and skills of rural residents, fostering the integration of small-scale farming economies with modern agriculture. This has had a positive impact on food security and the increase in farmers’ incomes. The second perspective is the opposite, suggesting that NI may also have negative impacts on rural industries. Obstacles such as underdeveloped information infrastructure, a shortage of professional digital talent [43], and significant regional development disparities have led to an overall lag in the digital and informational development of rural areas. These factors, however, constrain the penetration and diffusion of the digital economy in rural regions [44], making it difficult for the inclusive effects of NI to benefit farmers widely [28], which could potentially widen the urban–rural digital divide further [45]. Moreover, it is necessary and urgent to investigate this. We need to determine whether NI, like traditional infrastructure, faces the risk of saturation and diminishing marginal returns on investment.

While existing research provides a solid foundation for understanding the impact of NI on DRI, it is not without its limitations. Firstly, the literature has yet to clearly delineate whether the impact of NI on DRI is positive or negative. Second, when studying the influence of NI, existing studies often use the implementation of a specific policy as a proxy variable without fully considering the actual levels and differences in NI across various regions. Third, the mechanisms of action require further validation. In response, this paper constructs an evaluation system for NI and DRI based on provincial panel data and measures both constructs. Utilizing methods like double fixed effects and mediation analysis, the study examines the impact of NI on DRI from the perspectives of heterogeneity and mechanisms of influence. Additionally, the paper employs instrumental variable methods to address endogeneity issues.

The marginal contributions of our findings are as follows: Firstly, this research pioneers in systematically investigating the “inverted U-shaped” relationship between NI and DRI, filling a gap in the existing literature. The empirical validation of this nonlinear relationship offers a novel theoretical lens through which the complex dynamics of NI’s impact on DRI can be understood. Secondly, this study identifies and confirms that the upgrading of industrial structures and the expansion of transaction markets serve as pivotal mechanisms mediating the influence of NI on DRI. These findings provide actionable insights for policymakers, underpinning more informed and strategic planning for NI initiatives. Thirdly, this research delves into the heterogeneity of NI’s impact on DRI across various regions, considering factors such as geographical location, population density, and human capital. This nuanced analysis enhances the precision and applicability of the study’s recommendations for tailored policy interventions.

3. Theoretical Analysis and Hypothesis

3.1. The Impact of NI on DRI

The promotion and application of NI have significantly influenced DRI. The services of NI are continuously expanding into rural areas, with its technology and concepts permeating all aspects of agricultural production [46], rural economic development, and the lives of farmers. This has facilitated the optimization and upgrading of the agricultural industrial structure. It has also enhanced the level of rural public services and rural governance. The construction of NI, such as the internet, has bridged the digital divide between urban and rural areas, reduced the costs associated with information and management for farmers in their production and business activities, innovated agricultural production methods, and markedly enhanced agricultural production efficiency. The widespread availability of these infrastructures has introduced new ideas, knowledge, and technologies to rural areas. Through online platforms, it has promoted the flow of high-quality educational and learning resources to rural communities, which has played a positive role in realizing scientific agricultural production, improving rural human capital, inspiring entrepreneurial spirit among farmers [47], and increasing awareness of ecological conservation. NI, which relies on modern technologies like the internet and cloud computing, helps farmers quickly understand changes in market demand, adjusting production plans and product structures, alleviating the issue of unsold agricultural products, optimizing the allocation of agricultural resources, and achieving multi-level matching between supply and demand. However, research indicates that there is not a simple linear relationship between NI and regional development, a finding verified in the contexts of economic growth, total factor productivity [48], and the efficiency of financial resource allocation [49]. Once regional investment in NI reaches saturation, any further investment may result in resource wastage and diminished operational efficiency, thereby deviating from the optimal level of innovation performance. The digital transformation of agricultural production has not yet eliminated the dependence on traditional production factors. The relative scarcity of these factors determines that the level of agricultural development will not increase without bounds with the improvement of digitalization levels. When the development of NI meets the needs of agricultural growth, it may still face challenges. This is especially true when other supporting systems like policy support, technological research and development, and innovation transformation lag behind. It may lead to information redundancy, intensify intra-industry competition, increase the difficulty of service governance, affect the rational use of digital resources by farmers, and thereby constrain the further improvement of DRI. Thereby, this study posits a tri-phased theoretical framework to explain the “inverted U-shaped” relationship between NI and the DRI. The blue line in Figure 1 illustrates the hypothesized “inverted U-shaped” relationship between NI and DRI. In the initial phase of NI development, the infusion of new technologies and infrastructure, such as the internet and cloud computing, into rural areas will markedly boost DRI by enhancing agricultural production efficiency, lowering information and management costs, and improving access to educational and learning resources. As NI development progresses to the optimal phase, NI will reach a level where its benefits are maximized, and DRI will peak due to the efficient allocation of resources, the alignment of supply and demand across multiple levels, and the adoption of innovative production methods. However, in the saturation and diminishing returns phase, additional NI investment may yield diminishing returns, with the marginal benefits of NI declining because of issues such as resource wastage, operational inefficiencies, information redundancy, heightened competition, and governance challenges, ultimately constraining further advancements in DRI.

Hypothesis 1:

There exists an “inverted U-shaped” relationship between NI and DRI.

The impact of NI on DRI exhibits heterogeneous effects contingent upon regional characteristics including economic base, intelligent transportation infrastructure, population density, and human capital [50]. In areas with higher levels of economic development, the initial investment in NI can quickly integrate with the existing industrial base, generating positive synergistic effects that drive the rapid DRI increase [51]. However, as the infrastructure continues to improve, the marginal benefits gradually diminish [52]. Excessive investment may lead to resource wastage and reduced efficiency, thereby forming the latter half of the “inverted U-shaped” relationship [53]. In contrast, in regions with lower levels of economic development, where infrastructure is underdeveloped due to a late start, the initial investment in NI may not yield the anticipated effects. However, as infrastructure gradually improves, its positive effects will become apparent, though they may not reach the peak seen in regions with a higher economic base. Regions with high-level intelligent transportation infrastructure can enhance the efficiency of element allocation of NI. Firstly, the combination of digital infrastructure and intelligent transportation achieves real-time optimization of agricultural logistics and the expansion of the rural tourism market; secondly, the level of transportation digitization determines the marginal returns of NI, with underdeveloped transportation networks creating a “last mile” bottleneck that constrains the DRI; thirdly, high transportation connectivity expands the service radius of NI by reducing information asymmetry and transaction costs across regions. In densely populated areas, NI can better leverage economies of scale [54], reduce the cost of service provision, and promote the agglomeration and upgrading of rural industries. Nonetheless, an overly high population density can result in the excessive utilization and congestion of infrastructure, which in turn may diminish its effectiveness. In regions with low population density, this might not easily lead to economies of scale due to the lack of demand. As the population density increases moderately, the efficacy of the infrastructure is enhanced over time. In areas abundant in human capital, both farmers and technical experts are able to utilize the NI more efficiently, prompting innovation in agricultural technology and the upgrading of industries [55], thereby speeding up the progress of DRI. However, once human capital reaches a certain level, further investment in infrastructure may fail to be effectively utilized due to a shortage of sufficient high-skilled labor, leading to a decline in benefits. In areas with lower human capital, the potential benefits of NI may not be fully realized due to the insufficient skills of the workforce.

Hypothesis 2:

The NI-DRI relationship is moderated by regional characteristics (economic base, population density, human capital) and intelligent transportation infrastructure levels, with advanced transportation systems enabling higher NI efficiency through digital–logistical synergies.

3.2. Mechanism of the Impact of NI on DRI

NI, such as the internet, big data, and the Internet of Things, provides new technical support and production methods for rural industries, promoting the optimization and upgrading of the industrial structure [56]. Firstly, by enhancing agricultural production efficiency through the adoption of intelligent and precision agricultural technologies, the aim is to improve the quality and yield of agricultural products [57]. This approach drives the transition from traditional agriculture to modern agriculture by reducing resource wastage and increasing resource utilization efficiency [58]. Secondly, NI facilitates the extension of the industrial chain, contributing to the growth of sectors such as agricultural product processing, logistics, and sales, thereby creating a comprehensive industrial chain and increasing the industry’s added value [59]. Thirdly, by relying on NI, rural areas can incubate new industries such as e-commerce, rural tourism, and cultural and creative industries, achieving industrial diversification [60]. Therefore, the upgrading of the industrial structure becomes an important intrinsic mechanism by which NI promotes DRI. The construction and improvement of NI help break through geographical limitations and expand the market space for rural industries [61]. Firstly, the reduction in transaction costs: NI such as high-speed transportation networks and information communication networks can reduce the loss and expenses associated with product transportation and information transmission, thereby improving transaction efficiency. Secondly, the expansion of market scope: the NI allows rural industries to access national and even global markets more conveniently, expanding sales channels and increasing market demand. Finally, the enhancement of market competitiveness: with the NI, rural enterprises can better access market information, enhance product competitiveness, and increase market share. Therefore, the expansion of the transaction market has become an important external mechanism by which NI promotes DRI.

Hypothesis 3a:

NI drives DRI through the upgrading of industrial structures (UISs).

Hypothesis 3b:

NI promotes DRI by expanding trade markets (ETMs).

4. Materials and Methods

4.1. Model Specification

This paper employs a two-way fixed effects model and constructs three regression models to test the hypotheses proposed in the previous section. The specific settings are as follows:

$${DRI}_{i,t}={\alpha }_0+{\alpha }_1{NI}_{i,t}+{\alpha }_2{{NI}^2}_{i,t}+\mathsf{\omega }{Control}_{i,t}+{\mu }_i+{\phi }_t+{\epsilon }_{it}$$

(1)

$${Med}_{i,t}={\beta }_0+{\beta }_1{NI}_{i,t}+{\beta }_2{{NI}^2}_{i,t}+\mathsf{\lambda }{Control}_{i,t}+{\mu }_i+{\phi }_t+{\epsilon }_{it}$$

(2)

$${DRI}_{i,t}={\gamma }_0+{\gamma }_1{NI}_{i,t}+{\gamma }_2{{NI}^2}_{i,t}+{{\gamma }_{3}Med}_{i,t}+\mathsf{\pi }{Control}_{i,t}+{\mu }_i+{\phi }_t+{\epsilon }_{it}$$

(3)

In the aforementioned equations, i and t represent the provincial administrative regions and years, respectively; DRI denotes the development of rural industries; NI represents new infrastructure; ${NI}^2$ denotes the squared term of new infrastructure; Med is the mechanism variable, which includes upgrading industrial structure (UIS) and expanding transaction markets (ETM); Control stands for the control variables; and ${\mu }_i$, ${\phi }_t$, and ${\epsilon }_{it}$ represent the province fixed effects, time fixed effects, and error term, respectively.

In Equation (1), ${\alpha }_1$ and ${\alpha }_2$ are significantly positive and negative, respectively, and pass the “inverted U-shaped” test, indicating that the “inverted U-shaped” relationship between NI and DRI is established, thus validating Hypothesis 1. Based on Equation (1), mediation effect models are constructed (Equations (2) and (3)). Firstly, analyze the relationship between NI and the mediating variable. If ${\beta }_1$ and ${\beta }_2$ meet the expected signs, proceed to the next step of the test. Secondly, after including the mediating variable, analyze the relationship between NI and DRI. If ${\gamma }_3$ is significant and the signs of ${\gamma }_1$ and ${\gamma }_2$ remain unchanged, it indicates that the mediating variable plays a mediating role in the process of NI promoting DRI.

4.2. Variables

4.2.1. Dependent Variable

In order to scientifically evaluate DRI, this study, based on a comprehensive review of relevant research [31,62], fully considers the importance of sustainable development and recognizes that DRI is not only reflected in economic growth but also involves multiple dimensions such as production methods, economic structure, and industrial integration. Therefore, this paper constructs an indicator system to measure the development level of rural industries in China from the following four aspects, to comprehensively and systematically evaluate the status and potential of DRI. The rural industrial production system is key to evaluating the foundation and potential of rural industries, and the perfection of the production system directly affects the industry’s sustainable development capacity. The economic benefits of rural industries are the core indicators for measuring the achievements of rural industry development, reflecting the industry’s contribution to economic growth and the efficiency of resource allocation. The integration of the three sectors of rural industries represents a new trend in DRI. Through the integration of agriculture with secondary and tertiary industries, it can promote the extension of the industrial chain and industrial upgrading, enhancing the comprehensive competitiveness of rural industries. The sustainable DRI considers the long-term impact of DRI on the environment, society, and economy, and serves as an important basis for evaluating whether rural industries have the capacity for sustainable development.

Based on these four dimensions, this paper selects 11 specific indicators to construct the indicator system. In terms of measurement, the entropy method is used to determine the weight coefficients of each indicator, which can objectively reflect the relative importance of the indicators. Specifically, this study employs MATLAB R2019a software to perform extreme difference standardization on the raw data, thereby eliminating the impact of different indicator dimensions, and calculates the entropy values of each indicator to ultimately ascertain the weight coefficients [63]. The application of this method ensures the scientificity and accuracy of the evaluation results. The detailed calculation results of the weight coefficients are presented in

Table 1. Indicators and weights for the level of DRI.

Primary Indicator	Secondary Indicator	Tertiary Indicators	Indicator Type	Weight
Development of rural industries (DRI)	Production system	Cultivated land area grain unit output value	+	0.087
		Agricultural product producer price index	−	0.002
		Agricultural industry structure index	+	0.138
	Economic benefits	Total output value of agriculture, forestry, animal husbandry, and fishery	+	0.079
		Per capita output value of agriculture, forestry, animal husbandry, and fishery	+	0.164
		Disposable income of rural residents	+	0.076
	Integration of three industries	Large-scale agricultural product processing enterprises	+	0.175
		Proportion of agricultural employment in rural population	+	0.067
	Sustainable development	Number of certified green food enterprises	+	0.077
		National agricultural product geographical indications	+	0.121
		Fertilizer use intensity	−	0.015

Note: "+" indicates a positive indicator, "−" indicates a negative indicator.

.

Considering the differences in units and dimensions among the indicators, it is necessary to standardize the original data into dimensionless indicator evaluation values for comprehensive analysis. The data are standardized using the range standardization method, which scales the indicator values to a range between zero and one, as shown in Formulas (4)–(7). Within the indicator layer, indicators are categorized into positive and negative indicators, indicating the relationship between the selected indicators and the level of DRI. For instance, a higher per capita output value in agriculture, forestry, animal husbandry, and fishery signifies a better level of DRI; similarly, an increased number of certified green food enterprises indicates a higher level of DRI. The method is as follows:

$$\mathrm{Positive}\mathrm{indicators}:\mathrm{x}{’}_{ij}=(x_{ij}-x_{min})/(x_{max}-x_{min})$$

(4)

$$\mathrm{Negative}\mathrm{indicators}:\mathrm{x}{’}_{ij}=(x_{max}-x_{ij})/(x_{max}-x_{min})$$

(5)

Considering the need to avoid human interference in the data, this paper opts to use the entropy method to reduce the subjectivity imparted by external factors. The methodology is as follows:

Calculate the weight of the ith subject indicator value under the jth indicator:

$$p_{ij}=x{’}_{ij}/\sum _{i=1}^{m}x{’}_{ij}$$

(6)

Calculate the entropy of the jth evaluation index:

$$e_j=-1/lnm{\sum }_{i=1}^m{p}_{ij}ln{p}_{ij},$$

(7)

Calculate the weight of the evaluation index j:

$$w_j=(1-e_j)/\sum _{j=1}^n(1-e_j)$$

(8)

Utilizing the entropy method to measure DRI can assist government departments and policymakers in gaining a more comprehensive and objective understanding of the level of DRI. This, in turn, aids in the formulation of targeted policies and measures, promoting DRI towards a more sustainable and balanced direction.

4.2.2. Core Explanatory Variables

Based on the interpretation of the connotation of NI provided by the National Development and Reform Commission [64], this paper divides the evaluation index system of the NI level into three categories: informational infrastructure, integrated infrastructure, and innovative infrastructure [23]. Ultimately, 17 indicators were selected to construct the index system for the development level of China’s NI, as shown in

Table 2. Indicators and weights of the NI.

Primary Indicator	Secondary Indicator	Tertiary Indicators	Quaternary Indicators	Indicator Type	Weight
New infrastructure development (NI)	Information infrastructure	Internet infrastructure	IPv4 address density	+	0.095
			Internet domain name density	+	0.083
			Internet broadband access port density	+	0.048
		Communications infrastructure	Fiber optic cable line density	+	0.042
			Mobile phone base station density	+	0.055
		New technology infrastructure	Number of computers used per 100 people	+	0.051
			Proportion of enterprises with e-commerce transaction activities	+	0.036
	Converged infrastructure	Intelligent transportation infrastructure	Railway operating mileage density	+	0.025
			Mileage density of high-speed highways	+	0.022
		Smart weather infrastructure	Automatic weather station density	+	0.035
		Smart home infrastructure	Proportion of cable radio and television subscribers to total households	+	0.049
			Proportion of digital TV subscribers to total households	+	0.044
		E-commerce infrastructure	Number of companies with R&D activities	+	0.127
			Number of enterprises with R&D institutions	+	0.108
	Innovative infrastructure	Scientific and educational infrastructure	Density of R&D institutions	+	0.071
			Density of regular institutions of higher learning	+	0.026
		Innovation infrastructure for high-tech industries	Density of high-tech enterprises	+	0.084

Note: "+" indicates a positive indicator.

. The normalization processing consistent with the previous section was adopted, and the entropy method was used to determine the indicator weights. Finally, a comprehensive score was obtained using linear weighting.

4.2.3. Mediating Variables

The upgrading of industrial structure (UIS) is represented by the proxy variable “Advanced Industrial Structure Index (AIS)”. The Advanced Industrial Structure (AIS) refers to the process or trend of the industrial structure transitioning from a lower level to a higher level, which is a change in the situation of economic construction and the transformation of economic development patterns. With the integration and development of information technology, intelligent manufacturing, modern agriculture, manufacturing, and services in the industrial chain, this paper selects the upgrade of industrial structure as a mediating variable. According to the characteristics of AIS, this study uses the index of advanced industrial structure to measure the upgrade of China’s industrial structure, measuring the upgrade of the industrial structure from the proportion of the three types of industries in the total output value. The specific calculation formula is as follows:

$${AIS}_{it}={\displaystyle \frac{Y_{1it}}{Y_{it}}}\times 1+{\displaystyle \frac{Y_{2it}}{Y_{it}}}\times 2+{\displaystyle \frac{Y_{3it}}{Y_{it}}}\times 3$$

(9)

Among them, $Y_{it}$ represents the regional GDP of province i in year t, and ${\displaystyle \frac{Y_{1it}}{Y_{it}}}$, ${\displaystyle \frac{Y_{2it}}{Y_{it}}}$, and ${\displaystyle \frac{Y_{3it}}{Y_{it}}}$ represent the ratio of the added value of the primary, secondary, and tertiary industries to the regional GDP of province i in year t, respectively.

The expansion of the trading market (ETM) is represented by the proxy variable “Log of Express Delivery Revenue (lnER)”. Express delivery revenue can serve as a basis for the development of e-commerce and the expansion of the trading market in rural areas. By analyzing the relationship between express delivery revenue and the growth of rural industrial sales, as well as the expansion of market coverage, we can examine the impact of trading market expansion on DRI.

4.2.4. Control Variables

To avoid interference from other variables on the regression results, this study designs four control variables. First is the level of economic development (lngdp), measured by the regional per capita GDP. Second is the degree of population density (Dpd). Population density, to a certain extent, mitigates the negative impact of an aging society, improves the quality structure of the labor force, upgrades consumption supply, and promotes the quality of economic development. It is measured by the natural logarithm of the number of people per square kilometer. Third is the intensity of investment in technological innovation (Iii). Investment in technological innovation can promote social production, enhance economic efficiency, and narrow regional disparities. It is measured by the intensity of R&D expenditure. Fourth is human capital (Hc), measured by the proportion of college students in the total population.

4.3. Data Sources and Descriptive Statistics

The sample data for this paper consist of panel data from part of provinces, municipalities, and autonomous regions in China from 2013 to 2022, including Beijing, Tianjin, Hebei Province, Shanxi Province, Inner Mongolia Autonomous Region, Liaoning Province, Jilin Province, Heilongjiang Province, Shanghai, Jiangsu Province, Zhejiang Province, Anhui Province, Fujian Province, Jiangxi Province, Shandong Province, Henan Province, Hubei Province, Hunan Province, Guangdong Province, Guangxi Zhuang Autonomous Region, Hainan Province, Chongqing, Sichuan Province, Guizhou Province, Yunnan Province, Shaanxi Province, Gansu Province, Qinghai Province, Ningxia Hui Autonomous Region, and Xinjiang Uyghur Autonomous Region (data of some provincial-level administrative regions of China including Xizang, Hong Kong, Macau, and Taiwan is partially unavailable). The primary data sources include the “China Statistical Yearbook” (2014–2023), the “China Urban and Rural Statistical Yearbook” (2014–2023), the “China Science and Technology Statistical Yearbook” (2014–2023), and the statistical yearbooks of various provinces (regions, cities) (2014–2023) and the EPS data platform at https://www.epsnet.com.cn/index.html#/Index (accessed on 14 December 2024). Missing data are interpolated using the mean substitution method and the regression substitution method. Descriptive statistics for each variable are presented in

Table 3. Descriptive statistics.

Variable	Obs	Mean	Std. Dev.	Min	Max
DRI	300	0.154	0.091	0.014	0.503
NI	300	0.092	0.083	0.021	0.532
lngdp	300	9.974	0.868	7.66	11.768
Dpd	300	7.894	0.381	6.965	8.62
Iii	300	1.843	1.165	0.45	6.83
Hc	300	0.024	0.021	0.001	0.246
AIS	300	2.399	0.123	2.177	2.836
lnER	300	13.408	1.495	9.385	17.038

.

Figure 2 visually depicts the spatial evolutionary trend of China’s DRI in 2013, 2016, 2019, and 2022 through maps. The analytical results indicate a significant and steady growth pattern in China’s DRI index. In 2013, the lowest recorded DRI index was 0.014, which subsequently increased annually, reaching a minimum value of 0.028 in 2022, signifying a 100% growth over the past decade. This achievement is closely related to China’s recent vigorous implementation of the Rural Revitalization Strategy, for example, Guizhou Province has effectively enhanced its local DRI level by developing a characteristic tea industry and expanding sales channels through e-commerce platforms. However, it is noteworthy that despite the overall DRI index in China maintaining a high standard in 2022, there was a declining trend in its maximum value. This phenomenon may foreshadow that some advanced regions are facing pressures such as resource and environmental constraints, market saturation, and the need for economic restructuring, all of which have impacted the further enhancement of the DRI index. Moreover, Figure 3 reveals the dynamic changes in China’s NI over the same time series. Observing the data, it is evident that the growth trend of China’s NI index is particularly pronounced. Between 2013 and 2022, the minimum NI index value increased from 0.021 to 0.038, while the maximum value surged from 0.199 to 0.532, reflecting a growth of 167.34% over the decade. This significant growth magnitude prominently demonstrates the rapid development of China’s NI. For instance, in terms of 5G base station construction, China has built the world’s largest 5G network, providing a solid foundation for the digital transformation of various industries. Combined with Figure 2, we can infer that the advancement of new infrastructure has provided new momentum for DRI, for example, the application of 5G technology in the agricultural sector has promoted the development of smart agriculture and enhanced agricultural production efficiency and product quality, thereby promoting the growth of DRI.

5. Results

5.1. Baseline Regression Results

The baseline regression results of NI on DRI are shown in

Table 4. OLS Overall regression results.

	(1)	(2)	(3)	(4)	(5)
NI	1.958 ***	1.225 ***	0.416 *	0.816 ***	0.246 *
	(0.191)	(0.199)	(0.223)	(0.181)	(0.134)
NI^2	−4.137 ***	−2.590 ***	−1.224 ***	−1.652 ***	−0.920 ***
	(0.459)	(0.338)	(0.331)	(0.364)	(0.205)
lngdp		0.070 ***	0.037 ***	0.019 **	0.025 ***
		(0.005)	(0.008)	(0.008)	(0.008)
Dpd		0.023 ***	0.006	0.008	0.003
		(0.009)	(0.011)	(0.012)	(0.012)
Iii		−0.016 **	0.010	0.023 **	0.011
		(0.007)	(0.008)	(0.011)	(0.011)
Hc		−1.132 ***	−0.780 ***	−0.235	−0.551 **
		(0.205)	(0.205)	(0.158)	(0.203)
_cons	0.037 ***	−0.739 ***	−0.299 ***	−0.375 **	−0.158
	(0.010)	(0.073)	(0.107)	(0.151)	(0.131)
id			NO	YES	YES
year			YES	NO	YES
N	300	300	300	300	300
r2_a	0.366	0.686	0.643	0.917	0.582

Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

. Models (1) and (2) are OLS regressions of DRI on NI. The first and second terms of NI are significant at the 1% level, verifying the nonlinear impact of NI on DRI. Meanwhile, the F-test rejects the OLS pooled effect, indicating the presence of provincial individual heterogeneity in the model. The Hausman test rejects the random effects model at a 1% significance level. To eliminate provincial individual heterogeneity, this paper establishes a two-way fixed effects model (5). In models (3)–(5), the coefficients of the first and second terms of NI are significant, indicating a clear nonlinear relationship between NI and DRI. The regression results from Model (5) demonstrate that the coefficient for the linear term of NI is 0.264, while the quadratic term shows a coefficient of −0.920 with a 95% confidence interval of [−1.338, −0.502], indicating an “inverted U-shaped” relationship between these variables. When NI = 0, a one-unit increase in NI raises DRI by 0.264 units; when NI reaches the inflection point value of 0.134, the marginal effect declines to zero; and when NI exceeds this threshold, the effect turns negative. This pattern of diminishing returns aligns with the infrastructure productivity threshold hypothesis proposed by Aschauer (1989) [65]. This implies that in the initial stage, a unit increase in NI leads to significant improvements in the annual revenue of rural enterprises, penetration rate of rural e-commerce, and agricultural labor productivity. When NI surpasses the inflection point, the marginal benefit of additional investment becomes negative, typically manifested by near-saturation of digital financial inclusion coverage and relatively high vacancy rates in digital parks in some regions. Furthermore, the “Utest” command in the Stata 16.0 is used to further test the “inverted U-shaped” relationship. The results show that the p-value of the Utest is much less than 0.01, rejecting the null hypothesis of no “inverted U-shaped” relationship at the 1% level. Specifically, the “inverted U-shaped” curve between NI and DRI shows a characteristic of an initially positive average slope (1.784, p < 0.01) followed by a negative slope (−2.442, p < 0.01), and the inflection point value of NI (0.237) falls within its range of values (0.221, 0.257). This indicates an “inverted U-shaped” relationship between NI and DRI, validating Hypothesis 1.

5.2. Robustness Checks

5.2.1. Endogeneity Test

The relationship between NI and DRI is likely to be influenced by political and administrative drivers of investment patterns. This is because, in many countries, especially China, large-scale infrastructure projects are often initiated, funded, and managed by the government. This includes NI projects such as 5G networks, data centers, and intelligent transportation systems. The government’s priorities, policy objectives, and administrative decisions play a significant role in determining the location and manner of these investments. Additionally, governments often implement regional development policies aimed at narrowing the urban–rural gap or promoting growth in specific areas. These policies can lead to targeted infrastructure investments in certain rural areas, thereby influencing the observed relationship between NI and DRI. Therefore, one can attempt to use instrumental variables (IVs) that are related to NI but not directly influenced by political or administrative decisions, which can help identify the causal effect of NI on DRI. Meanwhile, the instrumental variable method can ensure that estimates are not biased by omitted variable problems or two-way causality.

The Hausman test’s p-value of 0.007 indicates that the baseline regression model has endogeneity issues. Considering the fact that digital infrastructure is importantly carried by information networks, we selected the number of post offices per million people in 1984 as the instrumental variable for NI. The rationale behind this choice is as follows: On one hand, as an important part of traditional communication, post offices to some extent foreshadow a region’s communication capabilities and information dissemination efficiency, reflecting the level of early information transmission infrastructure. Similarly, the development of NI cannot be separated from the evolution of traditional communication technologies. Areas with a high number of post offices in history are likely to also have better-developed telecommunications infrastructure. On the other hand, the impact of historical post office numbers on the economies of urban and rural areas is gradually diminishing and so currently struggles to affect the efficiency of DRI, which satisfies the relevance and exogeneity principles of instrumental variable selection. Logically, the current DRI will not affect the number of local post offices. Therefore, there is no reverse causality. The instrumental variable method can be used to alleviate potential omitted variable bias and reverse causality issues between NI and DRI. Additionally, because the sample in this study is made up of balanced panel data, using only the number of post offices per million people in 1984 at the city level as an instrumental variable would pose measurement challenges due to the application of the fixed effects model. To address this, we drew on the setup method by Nunn and Qian [66] and constructed an interaction term between the number of post offices per million people in 1984 (cross-sectional data) and the national IT service revenue of the previous year (time-series data) as the IV for NI. Then, we employed the Two-Stage Least Squares (2SLS) method to identify the causal relationship between NI and DRI. The results, as shown in

Table 5. Endogeneity and robustness tests.

	(1)	(2)	(3)	(4)
NI	0.158		0.315 **
	(1.054)		(0.140)
NI^2	−1.689 *		−1.158 ***
	(0.990)		(0.238)
NI_p		0.003 ***
		(0.001)
NI_p^2		−0.000 **
		(0.000)
NI_lag				0.149
				(0.314)
NI_lag^2				−0.883 ***
				(0.315)
lngdp	0.0316 **	0.021 *	0.028 ***	0.023 ***
	(0.0135)	(0.010)	(0.008)	(0.008)
Dpd	0.0106	0.014	0.007	0.004
	(0.0120)	(0.012)	(0.011)	(0.012)
Iii	0.0324	0.012	0.008	0.013
	(0.0200)	(0.015)	(0.010)	(0.011)
Hc	−0.805 *	−0.421 *	−0.887 ***	−0.486 **
	(0.454)	(0.205)	(0.266)	(0.182)
_cons	−0.302 *	−0.209	−0.211	−0.135
	(0.168)	(0.142)	(0.127)	(0.134)
id	YES	YES	YES	YES
year	YES	YES	YES	YES
N	300	300	300	270
r2_a	0.442	0.330	0.391	0.311

Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

Column (1), indicate that the regression coefficient of the instrumental variable is positive and passes the 1% significance test, suggesting that NI still has a significant “inverted U-shaped” relationship with DRI, validating the stability of the baseline results. The chosen instrumental variable passed the weak instrument test (F-statistic 22.68) and the under-identification test (Anderson LM statistic 12.666), rejecting the hypotheses of weak instruments and non-identification, indicating that the instrumental variable selection is reasonable and effective. Therefore, the instrumental variable regression results also confirm that the impact of NI on DRI is robust.

5.2.2. Replacing the Core Explanatory Variable

Mobile base stations are an important component of NI, particularly in terms of communication infrastructure. Therefore, the distribution and development of mobile base stations can reflect the penetration and progress of NI to a certain extent. To avoid research errors caused by variable setting bias, we use the number of mobile phone base stations (NI_p) to replace the original core explanatory variable and re-estimate the baseline model. The results are shown in Table 5, Column (2). It is not difficult to see that there have been no significant changes in the significance and direction of the coefficient of the core explanatory variable, demonstrating that the estimation results of the baseline model are robust.

5.2.3. Winsorization Treatment

To avoid the potential bias that extreme values may introduce into the research results, the relevant data for the variables used in the article were trimmed at the top and bottom 1%, and the regression analysis was conducted again. The results are shown in Column (3) of Table 5. Observing the data, there is no significant difference in the sign and significance of the regression coefficients for NI compared to the baseline regression results, indicating that the baseline regression results are reliable.

5.2.4. Reverse Causality Test

To further test for reverse causality, we also use the lagged values of the independent variable to regress the current values of the dependent variable, that is, we used the one-period lag of the explanatory variable NI to regress DRI. The results, shown in Column (4) of Table 5, indicate a significant nonlinear relationship between NI and DRI, indicating that the baseline regression results are reliable.

5.3. Heterogeneity Analysis

5.3.1. Regional Heterogeneity

Due to China’s vast territory, there are significant gaps among various regions in terms of the foundation for the development of the digital industry, digital talent, and investment in capital elements. This leads to the possibility of differences in the impact of NI on DRI. Therefore, according to the classification standards of the National Bureau of Statistics, the total sample of the study is divided into four regions: east, central, west, and northeast. We conducted an analysis on the heterogeneity of the impact of NI on DRI in different regions, with the regression results shown in

Table 6. Regional heterogeneity analysis.

	(1)	(2)	(3)	(4)
NI	0.353 **	−0.562	2.044 *	3.267
	(0.176)	(2.424)	(1.194)	(13.238)
NI^2	−0.605 ***	−3.300	−5.662 *	−19.540
	(0.225)	(9.572)	(2.903)	(105.728)
_cons	0.627 **	−0.479	0.000	0.028
	(0.266)	(0.726)	(0.000)	(1.103)
id	YES	YES	YES	YES
year	YES	YES	YES	YES
N	100	60	110	30
r2_a	0.956	0.926	0.878	0.841

Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

, Columns (1)–(4). The regression results of Model (1) demonstrate a statistically significant inverted U-shaped relationship between NI and DRI in eastern regions, with the linear term coefficient of NI being 0.353 (p < 0.01) and the quadratic term coefficient being −0.605 (95% CI [−1.047, −0.163]). Calculations show that when NI levels are below the inflection point of 0.292, each additional unit of NI investment increases DRI by 0.353 units, whereas when NI exceeds this threshold, the marginal returns of additional investment approach zero. The estimation results of Model (3) reveal that although western regions exhibit a similar inverted U-shaped pattern (linear term coefficient 2.044, p < 0.05; quadratic term coefficient −5.662, 95% CI [−11.353, 0.029]), the quadratic term’s confidence interval including zero indicates weaker statistical significance of the nonlinear relationship. Notably, the inflection point value in western regions (0.181) is significantly lower than that in eastern regions (0.292). Compared to the western region, the inflection point in the eastern region is further to the right, which may be because the level of economic development is higher, the infrastructure is more comprehensive, and the industrial base is stronger, allowing for more effective utilization of the advantages brought by NI. Therefore, it can withstand a higher level of NI investment before negative effects appear at a higher investment level. Although the western region has a relatively lower level of economic development, the marginal benefit of NI may be higher because improvements in infrastructure can bring about greater development space. The industrial structures in the central and northeast regions of China are currently in a period of transition and may be approaching the point of infrastructure saturation, with NI failing to effectively translate into a driving force for DRI. This is primarily attributed to the distinctive nature of their economic structures, insufficient innovation-driven capabilities, deficiencies in policy guidance and financial investment, as well as limitations in the level of marketization and openness. Specifically, the heavy industry and agriculture-oriented economic structures in these two regions result in a differentiated dependency and demand for NI compared to other areas, and the integration and adaptation of NI require time. Moreover, the relatively weak innovation-driven capacity and the lagging development of high-tech industries constrain the technological integration and application of NI, thereby affecting its promotional role in DRI. In addition, insufficient policy support, an imbalanced investment structure, and the inadequate role of market mechanisms in resource allocation collectively contribute to the failure of NI resources to effectively transform into a driving force for DRI.

5.3.2. Heterogeneity of Intelligent Transportation Infrastructure

Intelligent transportation infrastructure serves as a core carrier of NI, and its development directly impacts the penetration efficiency of technologies such as 5G and the Internet of Things in rural areas. Some studies suggest that NI exhibits inclusive characteristics [67], while others point out that it may exacerbate regional imbalances [22]. Intelligent transportation infrastructure may act as a key variable in modulating the inflection point of the “inverted U-shaped” relationship between NI and DRI by altering technology application scenarios, logistics efficiency, or market accessibility. To test the heterogeneity of the impact of NI on DRI under varying levels of intelligent transportation infrastructure, this study uses “Railway operating mileage density” and “Mileage density of high-speed highways” as indicators to measure the level of regional intelligent transportation infrastructure. The entropy method is employed to assign weights and calculate the level of intelligent transportation infrastructure for each province. The median value of 32.723 is selected as the threshold, with provinces below this median classified as having a low level of intelligent transportation infrastructure and those at or above the median as having a high level. The regression results are presented in

Table 7. Analysis of heterogeneity in intelligent transportation infrastructure, population density, and human capital.

	(1)	(2)	(3)	(4)	(5)	(6)
NI	0.469	0.262	1.195 **	0.246	1.053 **	1.312 **
	(1.006)	(0.185)	(0.466)	(0.240)	(0.460)	(0.648)
NI^2	−2.755	−0.864 ***	−2.763 ***	−0.871 ***	−2.699 ***	−2.158 ***
	(1.917)	(0.249)	(0.994)	(0.334)	(1.032)	(0.791)
_cons	−0.178	−0.198	0.275	−0.394 *	0.000	−0.546 *
	(0.228)	(0.207)	(0.176)	(0.211)	(.0.000)	(0.287)
id	YES	YES	YES	YES	YES	YES
year	YES	YES	YES	YES	YES	YES
N	150.000	150.000	147	153	214	86
r2_a	0.906	0.941	0.967	0.920	0.948	0.938

Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

, Columns (1) and (2). The results indicate that in provinces with high levels of intelligent transportation infrastructure, there exists an “inverted U-shaped” relationship between NI and DRI, with a linear term coefficient of 0.262 and a quadratic term coefficient of −0.864 (95% CI [−1.047, −0.163]). This implies that when intelligent transportation levels are high, each unit increase in NI investment raises DRI by 0.262 units during the phase when NI is below 0.152, while the marginal effect turns negative when NI exceeds 0.152. In provinces with low levels of intelligent transportation infrastructure, the regression coefficients are statistically insignificant. This may occur because when intelligent transportation infrastructure levels are low, NI investments fail to form complete application scenarios, face constrained service radii, and quickly reach investment saturation, leading to an earlier emergence of the inflection point (0.085). In regions with higher levels of intelligent transportation infrastructure, the factor allocation efficiency of NI is significantly enhanced, industrial integration space expands, and the inflection point appears later (0.152). These findings align with the conclusions of Banerjee et al. (2020) regarding the threshold effects of transportation infrastructure [68]. It follows that intelligent transportation infrastructure serves as a key moderating variable determining the inflection point position in the “inverted U-shaped” relationship between NI and DRI.

5.3.3. Heterogeneity of Population Density

To test the heterogeneity of the impact of NI on DRI in provinces with different population densities, the logarithm of the number of people per square kilometer is used as an indicator to measure regional population density. The median population density of 7.894 is selected, with provinces below the median being classified as low-population-density provinces and those equal to or above the median as high-population-density provinces. The regression results are shown in Table 7, Columns (3) and (4). The empirical results demonstrate significant population density heterogeneity in the relationship between NI and DRI. The low population density subsample exhibits a statistically significant inverted U-shaped relationship, with a quadratic term coefficient of −2.763 (95% CI [−4.711, −0.814]) and an inflection point at 0.216. Similarly, the high population density subsample shows an inverted U-shaped pattern, but with a smaller absolute value of the quadratic term coefficient (−0.871, 95% CI [−1.526, −0.215]) and an earlier inflection point at 0.141. These findings robustly support Hypothesis 2. The regional differences in inflection points primarily stem from the following mechanisms. First, there are regional differences in the marginal diminishing pattern of resource utilization efficiency. In provinces with low population density, although the initial construction of NI faces higher unit costs, the subsequent marginal utility diminishes at a slower rate due to the relatively higher per capita infrastructure capital stock. In contrast, in densely populated regions, while digital infrastructure initially yields higher returns, the marginal benefits turn negative more rapidly due to faster saturation of service demand. Second, the moderating effect of economic development stages is significant. Less developed regions (mostly provinces with low population density) exhibit higher “demand elasticity” for digital infrastructure, allowing the positive effects of NI to persist for a longer duration. Finally, there are distinct heterogeneous responses across industrial structures. Provinces with low population density are predominantly characterized by agriculture-dominated industries, where digital technologies have greater transformative potential for traditional agriculture. Conversely, in provinces with high population density, where the service sector accounts for over 45% of the economy, the competitive saturation effect in digital services leads to a more rapid decline in the marginal returns of NI.

5.3.4. Heterogeneity of Human Capital

Considering that human capital is also an important factor influencing the role of NI in DRI, this paper uses the proportion of college students in the total population as an indicator to measure regional human capital. The median level of human capital is set at 0.236. Provinces with a proportion below the median are classified as having low human capital, while those at or above the median are classified as having high human capital. The regression results are shown in Table 7, Columns (5) and (6). The results indicate an “inverted U-shaped” relationship between NI and DRI, further validating Hypothesis 1. In the coefficient of the impact of digital NI on DRI, the inflection point for provinces with high human capital (0.304) occurs later than that for provinces with low human capital (0.195). This is mainly because provinces with high human capital have a larger labor force with professional skills and innovative capabilities, who are more adept at learning new knowledge and skills, thus being able to maintain the marginal benefits of NI for a longer period of time. Additionally, provinces with high human capital may have a larger market demand and potential, more reasonable industrial structures, and better resource allocation and investment efficiency, allowing these regions to maintain the marginal benefits of NI for a longer time. This part of the empirical results validates Hypothesis 2.

5.4. Mechanism Analysis

Table 8. Mediation analysis.

	(1)	(2)	(3)	(4)
	AIS	DRI	lnER	DRI
NI	1.627 **	0.048	4.581 ***	0.110
	(0.652)	(0.158)	(0.704)	(0.162)
NI^2	−2.096 **	−0.665 ***	−10.878 ***	−0.597 **
	(0.766)	(0.191)	(1.093)	(0.261)
AIS		0.122 **
		(0.053)
lnER				0.030 *
				(0.017)
_cons	2.174 ***	−0.422 **	11.488 ***	−0.499 **
	(0.256)	(0.169)	(0.368)	(0.239)
id	YES	YES	YES	YES
year	YES	YES	YES	YES
N	300	300	300	300
r2_a	0.642	0.394	0.980	0.381

Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

shows the analysis of the transmission pathways of NI on DRI. The text previously discussed that NI may affect DRI through channels such as industrial structure upgrading and the expansion of transaction markets. To test whether these transmission pathways exist, the study explores the ways in which NI impacts DRI based on Equations (1)–(4). The regression results are presented in Table 8. Table 8 Column (1) shows that the coefficient of NI is significantly positive, and the coefficient of the quadratic term is negative, indicating that NI can promote the upgrading of the industrial structure. There is an “inverted U-shaped” relationship between NI and industrial structure upgrading. In Column (2), both the quadratic term of NI and the coefficient of the industrial structure upgrading index are significant, but compared to Column (4) in Table 1, the coefficient of NI has decreased. This may be because industrial structure upgrading absorbs part of the positive impact of NI on DRI, thereby refining and adjusting the impact of NI. This suggests that industrial structure upgrading can explain part of the impact of NI, making its effect on DRI more moderate. Additionally, the study uses Sobel tests and Bootstrap tests to demonstrate the robustness of the above mechanism tests. After 1000 Bootstrap samples, the mediating effect of industrial structure upgrading is confirmed with a p-value less than 0.01. It is noteworthy that the indirect effect is negative, indicating a substitution effect between NI and industrial structure upgrading. This could be due to the construction of NI attracting resources away from rural areas to cities or more favorable regions, leading to hindered industrial structure upgrading or resource scarcity in rural areas, and consequently negatively affecting DRI. Columns (3)–(4) of Table 8 test the mechanism of transaction market expansion. The results indicate that NI can promote the expansion of transaction markets, and there is an “inverted U-shaped” relationship between NI and transaction market expansion. In Column (2), both the quadratic term of NI and the coefficient of the transaction market expansion index are significant, but compared to Column (5) in Table 1, the coefficient of NI has decreased, suggesting that transaction market expansion plays a partial mediating role between NI and DRI. This part of the test validates Hypotheses 3a and 3b.

6. Discussion

The findings of this study contribute significantly to the ongoing discourse on the relationship between NI and DRI. The discussion below aims to elaborate on the theoretical insights, policy implications, and the limitations of the study, while also suggesting directions for future research.

6.1. Discussion on Theoretical Implications

The discovery of an “inverted U-shaped” relationship between NI and DRI is a particularly intriguing theoretical contribution. This finding challenges the monolithic view that NI is either entirely beneficial or detrimental to rural revitalization [69,70]. Instead, it suggests a nuanced perspective that acknowledges the complexity of NI’s impact. The identification of an inflection point where NI’s effects shift from positive to negative is a critical insight that can guide future theoretical development. This revelation suggests that the relationship between NI and the DRI is not static but dynamic, and it depends on specific contextual factors, a perspective that resonates with the multifaceted and complex nature of rural development. In this sense, the current study builds upon the work of previous scholars who have emphasized the environmental variability of infrastructure impacts [71] while also providing a more precise empirical basis for the concept of optimal levels of NI investment. Exploring the mechanisms through which NI affects non-agricultural industries, particularly via industrial structure upgrading and transaction market expansion, fills a significant gap in the literature. While prior research has touched upon the role of infrastructure in promoting economic growth [72], this study delves deeper into the specific channels through which NI can facilitate rural revitalization. The empirical validation of these mechanisms not only offers a clearer understanding of how NI influences rural economies but also provides a more granular analysis that can guide more targeted and effective policy interventions.

6.2. Discussion on Heterogeneity

The heterogeneity of geographical location is a particularly interesting aspect of this study. The variation in the “inverted U-shaped” relationship between NI and DRI across different regions highlights the importance of context in infrastructure development. Both the eastern and western regions exhibit this relationship, but with different inflection points, indicating that the optimal level of NI investment varies by region. Compared to the western region’s inflection point of 0.181, the higher inflection point in the eastern region (0.292) suggests that more developed areas may have a higher capacity to absorb and utilize NI before reaching saturation. In contrast, the central and northeastern regions are closer to infrastructure saturation, emphasizing the need for a more nuanced approach to NI investment in these areas where the traditional infrastructure-led growth model may no longer be as effective.

This study reveals the critical moderating role of intelligent transportation infrastructure in the relationship between NI and DRI. In regions with advanced intelligent transportation systems, NI establishes a virtuous cycle of “digital infrastructure—intelligent transportation—industrial upgrading” by enhancing factor allocation efficiency and industrial synergy, demonstrating a significant “inverted U-shaped relationship” with DRI. Conversely, in areas with underdeveloped intelligent transportation, new NI investments fail to overcome developmental thresholds due to the “bucket effect” of infrastructure deficiencies and factor outflows, resulting in statistically insignificant policy effects. These findings not only validate the applicability of Krugman’s core–periphery theory in the digital economy era but also provide a novel explanatory perspective for reconciling the academic debate between the “inclusive nature” and “regional disparities” of NI and the developmental level of intelligent transportation essentially constitutes a threshold condition for the effective implementation of NI policies.

The study’s examination of the heterogeneity of population density and human capital provides further theoretical depth. Provinces with low population density and high human capital can utilize NI more effectively, indicating certain types of NI with a high concentration of human capital can be better utilized in areas with a skilled labor force [73]. This finding supports the theory that the interaction between infrastructure and human capital is a key determinant of development outcomes.

6.3. Discussion on Mechanisms

The upgrading of industrial structure serves as a significant pathway through which NI promotes rural economic growth, involving multiple dimensions such as technological advancement, extension of industrial chains, and increased added value. Specifically, NI facilitates the transformation of agriculture towards high-value-added industries by providing advanced information technology and logistics support, thereby enhancing the economic benefits of rural areas [60]. Concurrently, the expansion of transaction markets creates favorable conditions for the development of non-agricultural industries in rural areas by improving market access and reducing transaction costs [61]. These two mechanisms are not isolated but rather mutually reinforcing and interdependent. For instance, market expansion may further stimulate the demand for industrial structure upgrading, while the optimization of the industrial structure provides more quality products and services for market expansion. These mechanisms interact with policy support, social capital, and the natural environment in rural settings, collectively shaping the complex landscape of rural development. Policy support, as a crucial external force driving industrial structure upgrading, encourages technological innovation and industrial transformation through measures such as tax incentives, fiscal subsidies, and credit support [74,75]. Investments in areas like smart agriculture promote the upgrading of the industrial structure towards higher-value-added service industries and high-tech industries, providing a dual guarantee of funds and markets for this upgrading. Social capital, including trust, norms, and networks accumulated in rural areas, facilitates information sharing, resource integration, and technology dissemination. Social networks like farmers’ cooperatives and industry associations promote the application of new technologies, accelerating the upgrading of the industrial structure [76]. The condition of the natural environment directly affects the choice and upgrading path of the industrial structure. NI investments in ecologically vulnerable areas focus on ecological protection, promoting the development of green industries, and a favorable natural environment is also the basis for developing specialized industries such as tourism. In terms of transaction market expansion, policy support reduces transaction costs and optimizes the market environment, such as the construction of agricultural product trading platforms and the improvement of logistics distribution systems. Social capital reduces transaction uncertainties, enhances market trust, and forms stable trading partner relationships. The natural environment influences the expansion of transaction markets, where transportation convenience and resource abundance determine the types and quantities of products, affecting market size and structure. In summary, these interacting elements constitute a dynamic system, and understanding them is essential for formulating effective rural revitalization strategies.

6.4. Discussion on Limitations and Future Research Prospects

This paper, from the perspective of NI and DRI, empirically tests the relationship between NI and DRI and its mechanism of action. There is a certain degree of innovation in the research perspective, but there are still some shortcomings. Firstly, the statistical data of rural areas in China are yet to be enriched, so the selected indicator dimensions in the process of indicator construction are relatively few, which can only capture the characteristics of DRI as accurately as possible, but it is difficult to achieve absolute accuracy. Second, the data used in this study are provincial-level panel data in China, without considering the urban–rural differences and NI differences between and within provinces. Meanwhile, since the concept of NI has not been proposed for a long time, the amount of time-series data on this is small. Although the conclusions of this study have been verified by many empirical processes, there is still room for further exploration of the data and methods. Therefore, future research can proceed from three angles: (1) Enrich research data. Collect more comprehensive and detailed statistical data on rural areas, refine the dimensions of DRI. Continuously collect relevant time-series data to facilitate long-term trend analysis. (2) Consider regional differences and hierarchy. Focus on the urban–rural differences between and within provinces in terms of NI, as well as regional differences. Collect as much diverse data as possible, such as city and county or even township-level data, to more accurately analyze the impact of NI on DRI. (3) Diverse methodological research. With the improvement of data availability in rural areas in the future and the increasingly mature mechanism testing techniques, follow-up studies can further diversify the dimensions of indicator construction and can also use machine learning algorithms for preliminary dimension screening to improve the accuracy of index construction and enrich the connotation of indicators. Use spatial econometric methods to analyze the spatial distribution of NI and its impact on DRI.

7. Conclusions and Policy Recommendations

7.1. Conclusions

This study examines the relationship between NI and DRI using panel data from Chinese provincial-level regions (2013–2022). The principal findings demonstrate three key insights. First, there is a robust “inverted U-shaped” relationship between NI and DRI, with an inflection point at 0.134, consistent across instrumental variable analysis, alternative variable specifications, and winsorization procedures. Second, heterogeneity analysis reveals significant variations: regional heterogeneity analysis shows eastern and western regions exhibit the “inverted U-shaped” relationship with rightward-shifted inflection points at 0.292 and 0.181, respectively, while central and northeastern regions display infrastructure saturation; intelligent transportation infrastructure analysis indicates advanced regions achieve greater NI efficiency with a 0.152 inflection point through digital infrastructure-intelligent transportation–industrial upgrading synergies, whereas underdeveloped regions face infrastructure bottlenecks with a 0.085 inflection point; population density and human capital analysis reveals provinces combining lower population density with higher human capital demonstrate superior NI absorption capacity. Third, mechanism analysis establishes industrial structure upgrading and market expansion as critical channels through which NI influences DRI development.

7.2. Policy Recommendations

Our findings offer several valuable insights for policymakers:

(1) Differentiated policy support and optimal resource allocation. Given the varying “inverted U-shaped” relationships between NI and DRI across different regions, policymakers should consider regional heterogeneity when formulating policies. Specifically, the eastern and western regions should pay attention to the inflection points of infrastructure construction. Considering the eastern region, it is recommended to establish an investment early warning mechanism triggered by an assessment when NI approaches saturation levels, and to set up an industrial upgrading fund, allocating a certain proportion of NI investment to support the transformation of traditional industries towards digitalization and intelligentization. Considering the western region’s larger space for infrastructure development, it is proposed to implement a “NI+” bundled investment strategy, integrating transportation infrastructure construction with digital infrastructure development, and advancing infrastructure upgrades in stages, from basic communication network coverage to industrial digital applications, and finally to intelligent agricultural production. In the central and northeastern regions, the focus should be on improving the utilization efficiency of existing infrastructure, conducting efficiency audits of existing facilities, and upgrading facilities with low utilization rates. Implement a “Digital Empowerment” initiative, supporting a certain number of enterprises in digital transformation annually, and innovate policy instruments, such as issuing special bonds and establishing regional talent centers, to promote industrial upgrading and innovative development.

(2) Implementation of a tiered development strategy for intelligent transportation infrastructure. First, establish an evaluation system for intelligent transportation infrastructure, including core indicators such as road network digitalization rate and vehicle networking penetration rate, and classify regions into different levels. For regions with a high level of intelligent transportation infrastructure, a leading development strategy should be formulated; for regions with a low level, a strategy of addressing weaknesses and strengthening foundations should be adopted. Second, for regions with a high level of intelligent transportation infrastructure, focus on promoting the construction of 5G-V2X deep integration infrastructure, including real-time high-precision map updating systems and intelligent logistics upgrade projects, thereby improving the efficiency of agricultural product cold-chain logistics and providing technical support for the intelligent management of rural tourism. Third, for regions with a low level of intelligent transportation infrastructure, focus on implementing basic digital transformations such as intelligent upgrades of rural roads, digital upgrades of village-level logistics outlets, and popularization of Beidou navigation terminals, driving the development of rural e-commerce clusters, and supporting collaborative layouts of e-commerce warehousing and low-cost IoT solutions. Fourth, all construction projects must be equipped with dynamic monitoring systems to achieve benefit evaluation and plan adjustment, ensuring maximized investment returns and realizing the coordinated DRI and NI.

(3) Implementation of a differential human capital development strategy. In view of the important role of human capital in promoting rural industry development through NI, precise policies should be implemented for regions with different levels of human capital. For regions with a high level of human capital, policies should focus on improving the investment benefits of NI. Specific measures include establishing university–enterprise joint R&D centers, focusing on applied research of NI in rural development; and setting up a special talent introduction plan for NI. For regions with a weaker human capital base, promote the “Internet + Education” model, increase investment in rural education, implement rural teacher training and improvement programs, improve the infrastructure of rural schools, and ensure the quality of education. At the same time, carry out vocational skills training programs, cultivate professional and technical personnel according to the needs of NI, such as professionals in information technology, new energy, and intelligent manufacturing. Finally, establish and improve the rural talent introduction and incentive mechanism, attract high-level talents to return to their hometowns for employment and entrepreneurship, and enhance the overall level of human capital in rural areas.

(4) Expanding transaction markets and strengthening the marginal benefits of NI. Firstly, efforts should be made to promote the establishment and widespread use of e-commerce platforms in rural areas, optimizing the logistics and distribution network to ensure the rapid circulation of agricultural products. Secondly, information infrastructure should be strengthened to improve network coverage and support rural e-commerce. In provinces with low population density, focus should be on investing in transportation and information infrastructure to enhance positive marginal benefits, promote effective resource flow, and integrate regional economies. Thirdly, measures should be taken to avoid overinvestment in NI. This includes establishing a pre-investment evaluation system to ensure investment efficiency, building logistics hubs in key areas to reduce the distance between urban and rural logistics, and promoting intelligent agricultural technologies to improve agricultural production efficiency, thereby maximizing the economic benefits of NI.