Digital Public Infrastructure and Agricultural Modernization: Causal Evidence from the Broadband China Policy

Abstract

Digital public infrastructure is increasingly viewed as a catalyst for rural and agricultural transformation, yet its structural impact on agricultural modernization remains insufficiently examined. Using the Broadband China policy as a quasi-natural experiment, this study applies a difference-in-differences framework to city-level panel data to identify the causal effect of digital public infrastructure on agricultural modernization. The estimates indicate that digital infrastructure significantly advances agricultural modernization. Further analysis shows that the effect operates through enhanced technological innovation and strengthened economic agglomeration, suggesting that digital connectivity reshapes both productivity and spatial organization within the agricultural sector. The impact is more pronounced in regions with developed logistics systems and lower information frictions, underscoring the importance of complementary infrastructure and institutional conditions. By linking digital public infrastructure to structural agricultural transformation, this study extends the literature on digital development and provides policy insights for developing economies pursuing infrastructure-driven modernization strategies.

1. Introduction

Agricultural modernization remains a fundamental challenge for developing economies, where structural constraints in infrastructure, technological innovation, and resource allocation continue to limit productivity and resilience. As the world’s largest developing country, China still faces substantial gaps in agricultural efficiency and technological upgrading compared with advanced agricultural economies [1,2]. At the same time, rapid advancements in next-generation information technologies—such as broadband networks, cloud computing, and big data—are reshaping production organization, market integration, and factor allocation. Recognizing this potential, China’s 2025 Central Document No. 1emphasizes extending digital infrastructure to rural areas and enhancing universal telecommunications services. Nevertheless, persistent disparities in digital connectivity, uneven infrastructure deployment, and information frictions continue to hinder the transformation of traditional agriculture into a modern, technology-intensive sector [3].

Existing research has primarily explored the impact of digital technologies on agriculture through various lenses, including national modern agricultural demonstration zones [4,5], industrial integration [6], digital inclusive finance [7,8,9,10], total factor productivity in agriculture [11], and digital rural development [12]. Additionally, some studies have examined the role of traditional infrastructure, such as rural infrastructure [13] or transportation infrastructure [14,15,16], in promoting agricultural development. As China’s innovation capabilities and R&D levels in big data, cloud computing, and related fields continue to advance, digital technologies have permeated agricultural production, consumption, and distribution. However, existing research has primarily focused on the application layer of digital technologies, such as the role of rural e-commerce in boosting farmer incomes [17], the impact of digital finance on corporate technological innovation [18], and the contribution of digital technologies to enhancing total factor productivity in manufacturing [19]. While these studies demonstrate significant economic spillover effects from digital technologies, they generally overlook the functional mechanisms of digital public infrastructure itself within agricultural development. In contrast, international research has paid greater attention to the impacts of digital public infrastructure on economic growth [20,21,22,23,24], urban entrepreneurship [25], and labor markets [26], with relatively limited exploration in the agricultural sector. Overall, despite digital public infrastructure’s broader coverage and superior spatio-temporal information transmission capabilities compared to traditional infrastructure, systematic empirical research remains scarce on its precise role in agricultural modernization and the mechanisms through which it influences agricultural development. The agricultural economics literature commonly employs panel fixed-effects models to control for regional heterogeneity [27,28,29,30], providing the methodological foundation for this study’s empirical design. Within the field of digital infrastructure research, existing studies have utilized difference-in-differences methods to examine the impact of telecommunications and broadband infrastructure on urban resilience and rural digital financial inclusion [31,32,33]. However, the current literature primarily focuses on macroeconomic dimensions, lacking a systematic examination of whether and how digital public infrastructure promotes agricultural modernization.

Against this backdrop, this study leverages China’s Broadband China policy as a quasi-natural experiment to identify the causal effects of digital public infrastructure on agricultural modernization. Using panel data from 300 prefecture-level and above cities from 2007 to 2023, this study estimates the policy’s overall impact, explores two key mechanisms, and assesses heterogeneous effects based on regional logistics development and information search costs. This study contributes to the literature by examining digital public infrastructure as a driver of agricultural modernization. While prior research has focused on specific outcomes such as productivity or income, or on constructing indicators of agricultural modernization and their spatiotemporal evolution [34,35,36], less attention has been paid to the broader process of structural transformation in agriculture. By shifting the focus to this dimension, the paper provides new evidence on the role of digital infrastructure in agricultural development. Second, the study also provides causal evidence by exploiting the Broadband China policy as a quasi-natural experiment. The difference-in-differences approach helps address endogeneity concerns arising from non-random infrastructure deployment and strengthens the identification of policy effects. Finally, the analysis explores the mechanisms underlying the observed effects. The results indicate that technological innovation and economic agglomeration are important channels, and that the impact of digital infrastructure depends on complementary conditions such as logistics systems and information costs. This highlights the role of digital infrastructure within a broader development framework.

2. Policy Context, Theoretical Analysis, and Research Hypotheses

2.1. The Broadband China Policy

Since 1994, China has achieved full connectivity with the global internet. After years of development, China’s digital public infrastructure has made tremendous breakthroughs. However, this development has been accompanied by issues such as regional imbalance in deployment, incomplete integration of applications, and low service efficiency [37]. Early manifestations included a triple digital divide: First, an urban-rural divide in China’s digital public infrastructure development. By 2012, while fiber to the home (FTTH) had begun rolling out in urban areas, vast rural regions still relied primarily on low-speed ADSL, with some administrative villages lacking broadband access altogether. Second, a regional divide in digital public infrastructure development. Network density and quality in eastern coastal regions far exceeded those in central and western regions [38]. Third, China faced an international digital public infrastructure gap, with overall broadband performance lagging behind developed nations. By the end of 2012, China’s average broadband penetration rate was approximately 13%, far below the 26% penetration rate in developed countries. Nearly 40% of Chinese users still relied on access speeds below 4 Mbps, significantly lower than the mainstream 18 Mbps speeds in developed countries [38]. This has become a critical bottleneck constraining the digital transformation and upgrading of agriculture.

To alleviate the constraints imposed by deficiencies in digital public infrastructure on the digital development of China’s agriculture and rural areas, in 2013, the State Council released the implementation plan for the Broadband China policy and set phased objectives. First, China planned to achieve fiber to the home in urban areas and broadband access in rural villages by 2015. China aimed to achieve 95% broadband coverage in administrative villages by 2015, with urban and rural household broadband access speeds reaching 20 Mbps and 4 Mbps, respectively. Second, China planned to establish a technologically advanced broadband network infrastructure covering both urban and rural areas by 2020, targeting over 98% broadband coverage in administrative villages and boosting access speeds to 50 Mbps in urban areas and 12 Mbps in rural areas. To implement this top-level design, the Ministry of Industry and Information Technology (MIIT) and the National Development and Reform Commission (NDRC) selected a total of 117 demonstration cities in three batches between 2014 and 2016. Applicant cities were required to meet at least four of several key indicators, including “20 Mbps or higher broadband access capacity for urban households” and “fixed broadband household penetration rate.” Since the policy’s implementation, China’s broadband network infrastructure has achieved leapfrog development. By 2023, broadband coverage in administrative villages nationwide had far exceeded initial targets, with rural broadband users reaching nearly 200 million. In rural areas, users with access speeds of 100 Mbps or higher accounted for 96.5% of the total. Concurrently, 4G coverage in administrative villages exceeded 99%, while 5G networks achieved extensive coverage in key regions.

2.2. Policy Implementation and Agricultural Production

Some scholars have pointed out that the Broadband China policy not only impacts the construction of infrastructure such as the internet itself, but also extends the application of such infrastructure to both the production and consumption ends of economic and social development [39]. Beyond the expansion of network coverage, the actual implementation of the Broadband China policy has generated multi-dimensional impacts across different stages of agricultural production. As a nationwide strategic initiative, the policy not only improved rural broadband penetration but also integrated digital infrastructure into agricultural production, circulation, and market participation processes.

At the production stage, improved digital connectivity facilitates access to technical information, digital financial services, and precision agricultural technologies. Empirical evidence suggests that digital infrastructure enhances regional human capital accumulation, technological development capacity, and digital inclusive finance, thereby strengthening agricultural economic resilience [40]. It also significantly improves agricultural green total factor productivity [41] and promotes technological progress and farmland scale expansion in grain production [42], indicating that digital infrastructure contributes to productivity enhancement and sustainable agricultural development. At the income and circulation stage, digital infrastructure reduces transaction frictions and broadens market access. It has been found that digital infrastructure construction significantly increases farmers’ income by expanding market participation channels [43]. Meanwhile, improved digital connectivity enhances agricultural participation in global value chains and raises export value-added [41], suggesting that broadband expansion strengthens agricultural competitiveness in both domestic and international markets.

Theoretically, digital public infrastructure forms the foundation for economic digitization, yet its efficacy ultimately relies on broadband networks for transmission and coordination. Functionally, the Broadband China policy transcends a mere telecommunications network initiative. Based on this, this study argues that the Broadband China policy serves as an effective proxy variable for digital public infrastructure development levels.

2.3. Theoretical Analysis and Research Hypotheses

Agricultural modernization differs fundamentally from industrial upgrading due to the inherent structural characteristics of agricultural production. Unlike manufacturing sectors, agriculture is typically characterized by highly dispersed production units, small-scale household operations, strong dependence on timely market information, pronounced seasonality, and high transportation and preservation requirements. These structural features generate persistent information asymmetry, limited technological diffusion, and constrained economies of scale, thereby impeding productivity growth and structural transformation. Consequently, improvements in digital connectivity may exert significant impacts on agricultural modernization by alleviating these structural bottlenecks.

The widespread adoption of broadband networks has directly alleviated the long-standing information barriers in rural areas. According to information economics theory, the free flow of information is a prerequisite for optimizing resource allocation and reducing transaction costs. In traditional agriculture, the dispersion of economic entities resulted in high costs for farmers to obtain market information, leading to inefficient allocation of production factors and poor market connectivity for agricultural products [44]. The judicious application of information technology can nearly double the operational efficiency of the agro-industrial complex [45].

Through the staged expansion of high-speed broadband networks and universal telecommunications services, the Broadband China policy substantially improved rural information accessibility. New infrastructure, such as 5G communications and satellite internet, not only enables small-scale farmers in remote areas to remotely access broader markets, expanding the sales radius for agricultural products, but also allows demand-side entities to access richer supply information [46]. This improvement in information efficiency enhances the allocation of land, labor, and capital, thereby laying the foundation for agricultural modernization. The theoretical framework of the relationship between digital public infrastructure and agricultural modernization is illustrated in Figure 1. Based on this, this study proposes Hypothesis 1:

H1: Digital public infrastructure can promote agricultural modernization.

Figure 1. Theoretical Framework of Digital Public Infrastructure and Agricultural Modernization.

Figure 1. Theoretical Framework of Digital Public Infrastructure and Agricultural Modernization.

Beyond improving information efficiency, digital public infrastructure also facilitates technological upgrading in agriculture. Under the new growth theory, technological progress serves as the core driver of endogenous economic growth [47]. Agricultural technological innovation, however, has long been constrained by limited R&D investment, slow diffusion of technical knowledge, and weak linkages between research institutions and farmers. The popularization of the Broadband China policy has facilitated the inflow of innovation factors into the digital agriculture sector. Digital public infrastructure not only provides technical support for building e-commerce platforms for agricultural products, smart agriculture cloud platforms, and agricultural socialized service platforms by matching supply and demand information online [25].

Moreover, digital connectivity generates a data productivity effect [48,49], whereby information generated through digital platforms reduces uncertainty in investment decisions and improves the efficiency of resource allocation toward agricultural R&D activities. The positive feedback mechanism between digital adoption and technological upgrading reinforces endogenous innovation capacity within the agricultural sector [50,51]. Through enhanced knowledge spillovers and accelerated technology adoption, digital public infrastructure contributes to productivity enhancement and structural upgrading in agriculture.

On the other hand, digital public infrastructure promotes structural reorganization and agglomeration within the agricultural economy. Krugman incorporated spatial factors into economic growth models [52], after which agglomeration economies became a core driver of endogenous economic growth, significantly boosting urban economic development [53]. Broadband expansion lowers coordination costs across regions, improves logistics information systems, and strengthens supply chain integration. In the agricultural context, digital connectivity supports the formation of leading agribusiness enterprises, specialized cooperatives, and modern agricultural industrial parks. These clusters enable intensive resource allocation, scale expansion, and specialization [54], improve agricultural production efficiency [55,56], and promote knowledge sharing and technology spillovers within the clusters [57]. For micro-entities in agricultural production, farmers can fully benefit from the infrastructure, specialized supporting services, and knowledge-technology spillover effects generated by industrial clusters, thereby reducing average agricultural production costs [58].

Taken together, digital public infrastructure promotes agricultural modernization through a multi-layered mechanism: by reducing information frictions, accelerating technological diffusion, and facilitating factor reallocation and agglomeration effects. Accordingly, this study proposes:

H2: Digital public infrastructure can promote agricultural modernization by driving technological innovation and economic agglomeration.

However, many rural areas in China suffer from insufficient logistics network coverage [59] and regional development imbalances [60]. The effectiveness of digital infrastructure is not uniform across regions. The agricultural sector exhibits strong dependence on complementary physical infrastructure and local information environments. Agricultural products are regionally concentrated, seasonal, and highly sensitive to logistics efficiency. In areas with underdeveloped logistics systems, improvements in digital connectivity may not fully translate into effective market access due to high transportation costs and limited cold chain coverage. Therefore, the interaction between digital and physical infrastructure conditions influences policy outcomes.

Furthermore, agricultural markets involve numerous farmers and dispersed buyers, where information asymmetry is prevalent during transactions [61]. As carriers of culture, the diversity of dialects not only hinders effective communication between groups [62] but also forces market participants to incur higher information search costs. Even in areas with broadband coverage, regions with high information search costs due to language diversity and cultural fragmentation face greater obstacles in effective information exchange. In contrast, regions with lower information search costs are more conducive to internalizing the benefits brought by improved digital connectivity. Therefore, the impact of digital public infrastructure on agricultural modernization depends on the level of complementary logistics development and the prevailing information search environment. Accordingly, this study proposes:

H3: The driving effect of digital public infrastructure is more pronounced in regions with higher logistics development levels and lower information search costs.

3. Research Design

3.1. Identification Strategy and Diagnostic Tests

Given the use of panel data in this paper, several diagnostic tests were conducted to determine the final model. The results of these tests are summarized in

Table 1. Diagnostic Results.

Test	Statistic	p-Value	Conclusion
Hausman Test	χ2(7) = 30.57	0.0001	Fixed effects preferred
Mean VIF	1.89	/	No multicollinearity
Wooldridge Test	F(1, 299) = 0.153	0.6961	No serial correlation

.

The Hausman test strongly favors the fixed-effects model over the random-effects model, indicating that fixed effects are more appropriate for this analysis. Multicollinearity indicates no significant multicollinearity issues among the explanatory variables, which supports the reliability of the coefficient estimates. The modified Wald test indicates the presence of heteroskedasticity across panels. Therefore, cluster-robust standard errors at the city level are employed to effectively mitigate heteroscedasticity. The Wooldridge test for autocorrelation indicates no significant first-order serial correlation in the residuals, suggesting that the model does not suffer from autocorrelation problems. These diagnostic results confirm the appropriateness of the two-way fixed-effects specification with cluster-robust standard errors.

3.2. Empirical Model

Based on the aforementioned literature review and theoretical analysis, to verify the driving effect of the Broadband China policy on agricultural modernization, this study adopts existing research methodologies by treating the Broadband China policy as a quasi-natural experiment for digital public infrastructure [63] and constructs a two-way fixed-effects difference-in-differences model.

$$S{\mathrm{core}}_{it}=\alpha +\beta DI{D}_{\mathrm{it}}+\gamma Control{s}_{it}+v_i+{\mu }_i+{\epsilon }_{it}$$

(1)

where the i denotes the prefecture-level cities and above, and t denotes the year. The dependent variable Score represents the composite score for the agricultural modernization level of the i-th prefecture-level or higher city in year t. DID serves as the core explanatory variable, indicating the policy implementation status of city i in year t. If city i has been selected as a Broadband China demonstration city, DID takes the value 1; otherwise, it takes the value 0. Controls represent relevant control variables affecting agricultural modernization. v denotes individual fixed effects, μ denotes time fixed effects, and $\epsilon$ denotes the random disturbance term.

3.3. Variable Selection and Data Description

3.3.1. Dependent Variable

This study uses the comprehensive score for agricultural modernization as the dependent variable. The Third Plenary Session of the 20th CPC Central Committee’s “Decision of the Central Committee of the Communist Party of China on Further Comprehensively Deepening Reforms and Advancing Modernization with Chinese Characteristics” proposed to “improve the support system for strengthening agriculture, benefiting farmers, and enriching rural communities.” China’s Plan for Accelerating the Construction of an Agricultural Power for the period 2024–2035 emphasizes “shared development and prosperity for farmers.” As one of the core objectives of agricultural modernization, the degree of achieving common prosperity directly reflects the quality and sustainability of this endeavor and should be incorporated into the evaluation framework. This serves both as an extension and supplement to the indicator system and as a response to the policy requirement that “the achievements of agricultural power construction benefit farmers more extensively and equitably.” Accordingly, this study references existing research methodologies to select 16 specific indicators [25,36,64]. Building upon the five pillars of “strong supply assurance, strong technological equipment, strong operational systems, strong industrial resilience, and strong competitive capacity,” it incorporates “common prosperity for farmers and rural areas” to construct an indicator system. The specific indicator system is presented in

Table 2. Indicator System for Agricultural Modernization.

Primary Indicators	Secondary Indicators	Explanatory Notes	Expected Signs
Supply Security	Comprehensive Grain Production Capacity	Comprehensive Grain Production Capacity (10,000 tons)	+
	Per Capita Vegetable Yield	Vegetable Yield/Total Population	+
	Per Capita Availability of Major Agricultural Products	Major Agricultural Products/Total Population (10,000 kcal/person)	+
Technology and Equipment	Intensity of Scientific Research Funding	Scientific and Technological Expenditure × Primary Industry Value Added/Regional GDP/Total Population	+
	Total Mechanical Power per Capita	Total Agricultural Machinery Power/Total Population	+
Competitiveness	Primary Industry Value Added per Capita	Primary Industry Value Added/Total Population	+
	Labor Productivity	Total Output Value of Agriculture, Forestry, Animal Husbandry, and Fisheries/Rural Population	+
	Land Productivity	Total Output Value of Agriculture, Forestry, Animal Husbandry, and Fisheries/Cultivated Area	+
Operating System	National-Level Modern Agricultural Industrial Parks	Number of National-Level Modern Agricultural Industrial Parks/Rural Population	+
	Family Farms	Number of Family Farms (Enterprises)/Rural Population	+
	Farmers’ Professional Cooperatives	Number of Farmers’ Professional Cooperatives/Rural Population	+
Industrial Resilience	Production Resilience	Pesticide and Fertilizer Application Volume/Total Population	−
	Agricultural Patents	Total Number of Authorized Agricultural Patent Applications	+
	Facility Agriculture	Facility Agriculture Land Area	+
Common Prosperity for Farmers and Rural Areas	Per Capita Net Income of Rural Residents	Per Capita Net Income of Rural Residents	+
	Urban-Rural Income Ratio	Urban Residents’ Income/Rural Residents’ Income	−

Note: “+” denotes a positive indicator and “−” denotes a negative indicator.

. Using the entropy method to calculate the weight of each indicator, it assesses the level of agricultural modernization at the prefecture-level city and above in China.

3.3.2. Explanatory Variable

This study employs the Broadband China policy as the core explanatory variable, specifically indicating whether a city was included in the pilot policy list. Following existing research [65], the Broadband China policy is coded as 1 when the sample city became a Broadband China demonstration city in year t, and as 0 when it was not selected as a Broadband China demonstration city.

3.3.3. Mechanism Variables

This study references existing research, using the ratio of invention patent applications to registered population in each region to measure technological innovation, and employing average nighttime light intensity to gauge economic agglomeration [65,66]. Specifically, agricultural invention patent applications reflect substantive breakthroughs in regional agricultural production and indicate the intensity of agricultural innovation investment. Therefore, this study uses the ratio of agricultural invention patent applications to registered population in cities at and above the prefecture level from 2007 to 2023 to measure technological innovation. Nighttime lights directly reflect local industrialization and urbanization levels. This study employs DMSP and VIIRS grid data on nighttime lights provided by NOAA from 2007 to 2023 to measure economic agglomeration.

3.3.4. Control Variables

This study adopts the following control variables based on existing research [67,68]. First, the degree of government intervention. Government intervention refers to local governments regulating resource allocation through administrative and legal means, reflecting their protective capacity. This study measures it using the ratio of local general public budget expenditures to total government fiscal expenditures. Second, economic density. Regions with higher economic development levels possess stronger consumer purchasing power, which is more conducive to agricultural modernization. This study represents this indicator using the ratio of regional GDP to administrative area size. Third is the human capital level. An increase in human capital accelerates the concentration of labor in high-tech industries, causing skilled labor to move away from agricultural production and thereby impacting local agricultural development. This study measures it using the ratio of full-time students in regular higher education institutions to the region’s year-end total population. Fourth is the intensity of fiscal investment. The infrastructure investment capacity of local governments is a crucial safeguard for advancing digital public infrastructure and agricultural modernization. This study measures it using the proportion of fixed-asset investment relative to local general public budget expenditures. Fifth is industrial structure. The development of non-agricultural industries may either compete with agriculture for resources or provide technological, capital, and market support. Its structural characteristics influence the external environment for agricultural development. This study measures it using the proportion of tertiary industry value-added relative to regional GDP.

3.3.5. Data Sources and Processing

This study utilizes panel data from 300 cities at and above the prefecture level in China, spanning 2007–2023 as its sample. The number of national-level modern agricultural industrial parks, family farms, and farmer cooperatives was manually collected and compiled, while other data sources include the China Urban Statistical Yearbook, provincial and municipal statistical yearbooks, DMSP stable light data, and VIIRS data. Individual missing values were imputed using the mean. Descriptive statistics for each variable are presented in

Table 3. Summary statistics.

Variables	Definition	Obs.	Mean	Std. Dev	Min	Max
Score	Comprehensive Score for Agricultural Modernization	5100	0.007	0.013	0.000	0.318
DID	Broadband China Pilot Policy Virtual Variable	5100	0.193	0.394	0.000	1
Human Capital Level	Number of Students Enrolled in Regular Higher Education Institutions/Total Population	5100	0.393	7.491	0.000	202.608
Economic Density	Ln (Regional GDP/Administrative Land Area)	5100	7.037	1.546	0.089	14.925
Government Intervention Level	General Revenue of Local Governments/General Expenditures of Government	5100	0.216	0.203	0.044	2.898
Fiscal Investment Intensity	Fixed Asset Investment/Local General Public Budget Expenditure	5100	5.710	4.020	0.010	41.677
Industrial Structure	Tertiary Industry Value-Added Share	5100	0.420	0.115	0.086	1.978
Technological Innovation	Patent Applications/Registered Population	5100	11.837	43.645	0.000	857.406
Economic Agglomeration	Nighttime Illumination Average	5100	7.643	9.300	0.000	60.053

.

4. Empirical Results

4.1. Baseline Regression

Based on the panel fixed-effects model specified above, this study conducts a baseline regression analysis on the impact of the Broadband China policy on agricultural modernization status. Column (1) in

Table 4. Baseline Regression Results.

Variables	(1)	(2)	(3)	(4)
	Score	Score	Score	Score
DID	0.009 ***	0.011 ***	0.003 ***	0.003 ***
	(0.000)	(0.001)	(0.001)	(0.001)
City FE	N	Y	Y	Y
Year FE	N	N	Y	Y
Control variables	N	N	N	Y
R2	0.077	0.121	0.321	0.325
Observations	5100	5100	5100	5100

Notes: *** denotes the levels of significance of 1%. Y indicates “Yes” and N indicates “No”.

presents the estimation results without controlling for time fixed effects or including control variables. Columns (2) to (4) progressively incorporate time fixed effects and control variables. The results indicate that the regression coefficient for the Broadband China policy remains significantly positive at the 1% level. This demonstrates that the implementation of the pilot policy significantly enhances the agricultural modernization level.

4.2. Robustness Test

4.2.1. Parallel Trend Test

The core premise for the validity of the two-way fixed-effects difference-in-differences model is that the outcome variables of the treatment and control groups exhibit parallel trends before policy implementation. Specifically, before the implementation of the Broadband China policy, there should be no significant trend differences in the level of agricultural modernization between pilot cities and non-pilot cities. This study uses 2014, the earliest pilot year, as the base period. The dummy variable for the year preceding the policy is omitted and treated as the baseline group, constructing the following model.

$$S{\mathrm{core}}_{it}={\alpha }_0+{\displaystyle {\sum }_{m=-7}^{-1}{\theta }_{\mathrm{m}}\mathrm{Pr}{\mathrm{e}}_{i,t+m}+{\theta }_{0}Curren{t}_{it}+{\displaystyle {\sum }_{m=1}^9{\theta }_{m}P{\mathrm{ost}}_{i,t+m}+{\lambda }_i+{\mu }_t+{\epsilon }_{it}}}$$

(2)

Among these, $Scor{e}_{it}$ represents the comprehensive score for agricultural modernization in city i during year t; $\mathrm{Pr}{e}_{i,t+m}$ is the pre-policy dummy variable. If city i was a pilot city during the m years preceding the Broadband China pilot program, it is 1; otherwise, it is 0; $Curren{t}_{it}$ is the policy-period dummy variable. It equals 1 if city i was a pilot city in 2014, and 0 otherwise; $\mathrm{P}os{t}_{i,t+m}$ is the policy-post dummy variable. It equals 1 if city i was a pilot city in the m years following the Broadband China pilot program, and 0 otherwise; ${\lambda }_i$ and ${\mu }_t$ are the region and year fixed effects, respectively, while ${\epsilon }_{it}$ is the random disturbance term. The results of the parallel trends test are shown in Figure 2.

As shown in Figure 2, before the implementation of the Broadband China policy, the estimated coefficients for each year were not significantly different from zero. This indicates that there was no significant difference in the trend of comprehensive score changes between pilot cities and non-pilot cities, satisfying the prerequisite for parallel trend testing. Beginning with the post-policy implementation period, particularly from 2017 to 2022, the estimated coefficients were significantly greater than zero. This indicates that the Broadband China policy substantially enhanced agricultural modernization, preliminarily supporting the baseline regression results presented in this study. Furthermore, the estimated coefficient declined and approached zero in 2023. This may be attributed to two factors: first, over time, non-pilot cities narrowed the development gap with pilot cities by addressing digital public infrastructure deficiencies through subsequent investments; second, the development of agricultural modernization is a long-term process. The policy effects of the Broadband China policy may unfold through dynamic adjustments driven by technology diffusion and industrial upgrading, leading to a diminishing marginal impact of certain short-term policy dividends.

4.2.2. Placebo Test

To avoid the baseline regression results being influenced by unobservable omitted variables, this study adopts a placebo test by randomly replacing treatment group cities, drawing from existing research [69]. This study randomly selected an equal number of cities from the sample pool to re-estimate the baseline model. This process was repeated 1000 times, yielding 1000 regression coefficients and their corresponding p-values, with the specific distribution shown in Figure 3. The results indicate that all regression coefficients are insignificant and far from the true estimates in the baseline regression. This suggests that the coefficient estimates in the baseline regression represent low-probability events in the placebo test and are unlikely to have been obtained by chance. This finding rules out the possibility that the estimation results in the baseline regression were influenced by unobservable factors.

4.2.3. Elimination of Expected Effects

A potential requirement for using the double difference method is that individuals should not form expectations about the policy in advance. To eliminate this effect, this study sequentially incorporates dummy variables representing the year before policy implementation (Before_1) and the two years before policy implementation (Before_2). The results presented in

Table 5. Exclusion of Expected Effects.

Variables	(1)	(2)	(3)
	Current Period	Before_1	Before_2
DID	0.003 ***	0.003 ***	0.003 ***
	(0.001)	(0.001)	(0.001)
Before_1		0.0001
		(0.0003)
Before_2			0.0002
			(0.0002)
Control variables	Y	Y	Y
Observations	5100	5100	5100
R2	0.331	0.331	0.331

Notes: *** denotes the levels of significance of 1%. Y indicates “Yes” and N indicates “No”.

indicate that the estimated coefficients for both variables are insignificant, thereby ruling out the interference of the anticipation effect.

4.2.4. Heterogeneity Treatment Effects

When treatment effects exhibit significant heterogeneity, traditional two-way fixed-effects difference-in-differences models may produce estimation biases. To enhance the robustness of baseline regression results, this study employs the multi-period double robust estimator (CSDID) proposed by scholars to test for heterogeneous treatment effects [70]. Three types of average treatment effects were calculated: (1) Simple Average Treatment Effect (Simple ATT); (2) Group Average Treatment Effect (Group ATT); and (3) Dynamic Average Treatment Effect (Dynamic ATT).

Table 6. Heterogeneity in Treatment Effects.

Variables	(1)	(2)
	Score	Score
Simple ATT	0.021 **
	(0.002)
Group ATT		0.023 **
		(0.002)
Control variables	Y	Y
Observations	5100	5100

Notes: ** denotes the levels of significance of 5%. Y indicates “Yes” and N indicates “No”.

reports the results for Simple ATT and Group ATT, both of which are significantly positive. This indicates that estimation biases arising from heterogeneous treatment effects do not pose a serious issue in this study, and the baseline results are robust.

The results of the dynamic average treatment effect are shown in Figure 4. Before the implementation of the Broadband China Pilot Program, the treatment effects across different periods fluctuated around zero and were statistically insignificant, satisfying the prerequisite for the parallel trends assumption. Following the program’s implementation, a significant positive impact emerged, further validating the policy’s sustained contribution to agricultural modernization.

4.2.5. Additional Robustness Tests

This study employs two strategies for robustness testing: first, trimming. Considering that extreme values in the sample may distort regression results, the dependent variable undergoes 1% trimming to mitigate estimation bias caused by outliers. Columns (1) and (2) in

Table 7. Additional Robustness Test Results.

Variables	(1)	(2)	(3)	(4)
	Score	Score	Score	Score
DID	0.018 ***	0.011 ***	0.004 ***	0.003 ***
	(0.003)	(0.003)	(0.001)	(0.001)
City FE	Y	Y	Y	Y
Year FE	N	Y	N	Y
Control variables	Y	Y	Y	Y
Observations	5100	5100	5032	5032
R2	0.375	0.462	0.250	0.322

Notes: *** denotes the levels of significance of 1%. Y indicates “Yes” and N indicates “No”.

report regression results after trimming, where Column (1) omits time fixed effects while Column (2) incorporates them. Results indicate that DID coefficients remain statistically significant at the 1% level with positive signs regardless of time fixed effects inclusion, consistent with baseline regressions. This confirms that after outlier adjustment, the Broadband China policy’s positive impact on agricultural modernization across cities remains robust. Second, this study excluded samples from municipalities directly under the central government. Given the significant differences in administrative level, resource endowments, and policy support between municipalities and other cities, this study follows existing research by excluding the four municipalities of Beijing, Tianjin, Shanghai, and Chongqing [71]. Columns (3) to (4) in Table 7 report the corresponding results. Column (3) does not control for time-fixed effects, while column (4) does. Both DID coefficients remain statistically significant at the 1% level, consistent with the baseline regression findings. This indicates that after controlling for the unique impact of municipalities directly under the central government, the policy’s promotional effect on agricultural modernization remains valid, further validating the robustness of the baseline regression.

4.3. Heterogeneity Analysis

4.3.1. Level of Logistics Development

This study draws upon existing research [61,72] by selecting the proportion of employees in the transportation, warehousing, and postal services sector relative to the total population as an indirect indicator of regional logistics development. This metric represents the scale of human capital investment in the logistics industry; a higher proportion of employees typically corresponds to greater logistics activity and industrial prosperity. Using a median cutoff method, the sample is divided into regions with higher and lower logistics development levels. Columns (1) to (2) in

Table 8. Results of the Heterogeneity Analysis.

Variables	Level of Logistics Development		Information Gathering Costs
	High Level	Low Level	High Costs	Low Costs
	(1)	(2)	(3)	(4)
DID	0.003 **	0.002	0.003 **	0.002
	(0.002)	(0.001)	(0.001)	(0.001)
City FE	Y	Y	Y	Y
Year FE	Y	Y	Y	Y
Control variables	Y	Y	Y	Y
Observations	2337	2319	2023	2040
R2	0.308	0.349	0.406	0.402
p-value	0.05 **		0.05 **

Notes: ** denotes the levels of significance of 5%. Y indicates “Yes” and N indicates “No”.

show that in the higher-level logistics sub-sample, the Broadband China policy exhibits a statistically significant positive effect at the 5% level. However, in the lower-level logistics sub-sample, this effect fails to pass the significance test. This study employs a Bootstrap group-level coefficient difference test, yielding a p-value of 0.07, which significantly rejects the null hypothesis, indicating a significant difference in regression coefficients between the two groups. Columns (1) to (2) in Table 8 show that in the group with higher logistics development, the Broadband China policy significantly promotes agricultural modernization at the 5% significance level. In contrast, the policy effect fails to pass the significance test in the group with lower logistics development. Hypothesis 3 is thus validated.

4.3.2. Information Gathering Costs

This study references existing research and utilizes the Chinese dialects spoken in 2113 counties and higher-level observation units listed in the Comprehensive Dictionary of Chinese Dialects [73]. After completing matching and value assignment based on the 2000 administrative divisions, a dialect differentiation index was derived to indirectly reflect information search costs [74,75]. Using the median division method, the sample was categorized into regions with higher and lower information search costs. The results of the grouped regression analysis are presented in columns (3) to (4) of Table 8. Findings indicate that in the subsample with lower information search costs, the Broadband China policy exhibits a statistically significant positive effect at the 5% level. Conversely, in the subsample with higher information search costs, this effect fails to pass statistical testing. This study employs a Bootstrap group-level coefficient difference test with a p-value of 0.05, significantly rejecting the null hypothesis. This confirms that the coefficients from the group-level regression exhibit significant differences between the two groups. Table 8 reports the estimation results. Columns (3) to (4) show that in the group with lower information search costs, the Broadband China policy significantly promotes agricultural modernization at the 5% significance level. In contrast, the policy effect fails to pass the significance test in the group with higher costs. Hypothesis 3 is thus validated.

5. Further Analysis

The theoretical analysis above demonstrates that digital public infrastructure development stimulates the internal vitality of rural economies through two distinct channels: enhancing agricultural technological innovation and strengthening economic agglomeration. This section will conduct specific mechanism testing and empirical analysis centered on technological innovation and economic agglomeration.

5.1. Enhancing Technological Innovation

In the agricultural sector, digital public infrastructure facilitates online matching of agricultural product information, provides technical support for relevant cloud platforms, channels innovative resources toward digital agriculture, and reinforces the positive feedback effect of transforming and applying digital agricultural technological achievements. To validate this mechanism, this study measures technological innovation levels using the ratio of patent applications to registered population and examines the impact of the Broadband China policy on technological innovation. Column (1) of

Table 9. Mechanism Analysis.

Variables	Technological Innovation	Economic Agglomeration
	(1)	(2)
DID	20.683 ***	1.523 ***
	(5.312)	(0.420)
City FE	Y	Y
Year FE	Y	Y
Control variables	Y	Y
Observations	5100	4768
R2	0.231	0.618

Notes: *** denotes the levels of significance of 1%. Y indicates “Yes” and N indicates “No”.

shows that the Broadband China policy is significantly positive at the 1% statistical level, indicating that it significantly promotes technological innovation in agriculture.

5.2. Strengthening Economic Agglomeration

Digital public infrastructure has overcome geographical constraints, enabling the clustering of new economic entities such as leading enterprises and specialized cooperatives. This not only optimizes resource allocation efficiency but also vigorously advances the scaling-up of agricultural operations. To validate this mechanism, this study uses nighttime light intensity as a proxy for economic agglomeration to examine the impact of the Broadband China policy on economic clustering. The results are presented in Column (2) of Table 9. The regression results indicate that the Broadband China policy is significantly positive at the 1% statistical level, suggesting that this policy significantly promotes the agglomeration of agriculture-related economic activities. This finding validates Hypothesis 2.

6. Discussion

The empirical results based on the Broadband China policy provide consistent evidence that digital public infrastructure plays a significant role in promoting agricultural modernization. This finding aligns with the core proposition of information economics that reductions in information asymmetry can improve resource allocation efficiency. In the agricultural sector, where production is decentralized and market information is often fragmented, improvements in digital connectivity help reduce transaction costs, enhance market matching, and facilitate the integration of production and circulation. Existing studies have primarily examined the effects of digital technologies on specific outcomes such as farmers’ income or agricultural productivity [11,76], whereas this study shows that digital infrastructure exerts a broader structural influence by advancing overall agricultural modernization. This suggests that digital public infrastructure should be understood not merely as a supporting condition, but as a fundamental driver of structural transformation in agriculture.

The mechanism analysis further clarifies the channels through which digital public infrastructure affects agricultural modernization. The results indicate that technological innovation and economic agglomeration are two key transmission mechanisms. On the one hand, consistent with new growth theory [47], digital infrastructure facilitates knowledge diffusion and reduces the cost of accessing information, thereby strengthening incentives for technological innovation in agriculture. Improved access to digital information supports the adoption of new production technologies, enhances learning effects, and promotes productivity growth [18,48]. On the other hand, digital infrastructure reduces coordination costs across regions and improves factor mobility, which contributes to the spatial concentration of agricultural activities. Such agglomeration effects can generate economies of scale and knowledge spillovers [52,53], ultimately improving production efficiency. Compared with studies that focus on a single mechanism, the evidence here supports a more integrated perspective, indicating that digital infrastructure simultaneously operates through multiple channels to reshape the organization of agricultural production.

The heterogeneity results highlight that the effectiveness of digital public infrastructure depends on complementary conditions. In particular, the promotion effect is more pronounced in regions with better-developed logistics systems and lower information search costs. This finding is consistent with the view that digital technologies are subject to “complementarity constraints” [20,23]. Without adequate physical infrastructure, improvements in digital connectivity may not translate into actual reductions in transaction costs. For example, even if market information becomes more accessible, poor transportation conditions can still limit the circulation of agricultural products. Similarly, high information search costs—arising from linguistic, institutional, or market barriers—can weaken the effectiveness of digital technologies. These results suggest that digital public infrastructure alone is insufficient to drive agricultural modernization; rather, its effects depend on the broader institutional and infrastructural environment. This also helps explain the heterogeneous impacts of digitalization observed across regions and countries in the existing literature.

From a theoretical perspective, this study contributes by integrating information economics and new growth theory into a unified analytical framework that links digital public infrastructure, factor allocation, and agricultural modernization. By explicitly identifying information cost reduction, technological innovation, and economic agglomeration as key channels, the analysis extends existing theories to the context of agricultural development. In this sense, digital public infrastructure can be understood as a general-purpose technology that simultaneously influences multiple dimensions of economic activity, thereby generating systemic transformation effects. This integrated framework helps bridge the gap between micro-level technological adoption and macro-level structural change, which has been relatively underexplored in the literature on digital agriculture.

From a broader development perspective, the findings also have implications beyond China. Many developing countries face similar structural challenges, including fragmented production systems, limited market integration, and underdeveloped infrastructure. The experiences of India and other developing countries indicate that while digital technologies can enhance market efficiency and agricultural output, their effectiveness varies depending on local conditions [77,78,79,80]. The results of this study reinforce the view that digital infrastructure policies need to be coordinated with investments in logistics systems, human capital, and institutional capacity. Without such coordination, the potential benefits of digital transformation may not be fully realized. Therefore, policy design should emphasize the integration of digital and physical infrastructure, as well as the improvement of local absorptive capacity.

Several limitations of this study should be noted. First, the analysis is based on city-level panel data, which may not fully capture micro-level behavioral responses of farmers and agricultural enterprises. Second, some key variables, such as technological innovation and economic agglomeration, are measured using proxy indicators, which may introduce measurement error. Third, although the difference-in-differences approach, combined with robustness checks, helps mitigate potential endogeneity concerns, unobserved factors cannot be completely ruled out. Future research could address these issues by using micro-level data, adopting alternative identification strategies, and examining long-term dynamic effects. In addition, further studies could explore whether the mechanisms identified in this paper hold in different institutional contexts, thereby enhancing the external validity of the findings.

7. Conclusions and Policy Implications

7.1. Conclusions

Based on the data of Chinese cities at the prefecture level and above from 2007 to 2023, this paper constructs a DID model to study the impact of digital public infrastructure on agricultural modernization and regards the Broadband China policy as a policy impact. Based on the above research, it is concluded that digital public infrastructure promotes agricultural modernization, which can be achieved by promoting technological innovation and economic agglomeration. Further study found that this effect is more significant in areas with a higher level of logistics development and lower cost of information search. This study demonstrates that digital public infrastructure constitutes a systemic force reshaping agricultural development, providing empirical support for integrating digital strategies into agricultural modernization policies.

7.2. Policy Implications

The findings of this study offer several important policy recommendations:

First, promote the coordinated development of digital and logistics infrastructure. The heterogeneity analysis indicates that digital infrastructure yields stronger modernization effects in regions with well-developed logistics systems. Therefore, digital expansion strategies should be integrated with improvements in rural transportation networks, cold chain logistics, and agricultural product distribution systems. In regions with weak logistics foundations, investments in broadband infrastructure should be accompanied by parallel upgrades in physical supply-chain facilities to ensure that digital connectivity can effectively translate into market access and value-chain integration.

Second, reduce information search costs through digital literacy enhancement and localized information platforms. In areas characterized by high dialect diversity or limited digital capabilities, the benefits of digital infrastructure may not be fully realized. Policymakers should implement systematic digital skills training programs for farmers, promote standardized language communication in digital transactions, and develop localized agricultural information service platforms adapted to regional linguistic and cultural contexts. These measures can lower adoption barriers and enhance the inclusiveness of digital transformation.

Third, strengthen innovation-driven agricultural clusters. Since technological innovation and economic agglomeration are key transmission channels, policies should support the development of digital agriculture innovation hubs, smart agricultural industrial parks, and integrated platforms that combine research institutions, enterprises, and farmers. Financial incentives and targeted subsidies can be directed toward agricultural big data applications, intelligent equipment, and digital service providers. By guiding capital, technology, and talent toward agriculture-related clusters, governments can amplify the scale and spillover effects of digital public infrastructure.

7.3. Limitations and Future Research

This study has certain limitations that warrant further investigation. First, while city-level panel data allow for policy identification and long-term analysis, they may not fully capture micro-level behavioral adjustments of individual farmers or agricultural enterprises. Future research could incorporate household-level or firm-level data to examine how digital infrastructure affects production decisions, technology adoption, and income distribution at the micro level. Second, future research should explore the long-term dynamic effects of digital infrastructure, examine interactions between digital infrastructure and rural financial development, or integrate emerging data sources such as satellite data and platform-level transaction records. Such extensions would deepen understanding of how digital public infrastructure reshapes agricultural development in the digital era. In addition, future research could explore whether the benefits of digital infrastructure in advancing agricultural modernization can be replicated in other countries or regions—particularly those with significant differences in socioeconomic conditions, levels of agricultural technology development, and farmers’ digital literacy.