1. Introduction
Following the official eradication of extreme poverty, the restructuring of China’s urban–rural relationship continues to face deep-seated institutional and structural challenges. While rural living standards have steadily improved, substantial disparities in income levels, infrastructure development, and access to public services persist between urban and rural areas. These disparities pose a substantial barrier to the pursuit of common prosperity. According to official statistics, in 2024, the per capita disposable income of urban residents in China was RMB 54,188, compared to RMB 23,119 for rural residents, yielding an urban–rural income ratio of 2.34:1. In Beijing, the ratio stood at 2.32:1, underscoring the continued significance and structural nature of the income divide.
One of the central barriers to integrated urban–rural development lies in the disconnection between the dominant smallholder-based agricultural production model and the institutional requirements of a modern agricultural system. Smallholder agriculture in China is commonly characterized by fragmented production, weak organizational capacity, and limited access to technology, all of which constrain agricultural productivity and rural income growth. In response, since 2013, China has adopted a centrally coordinated, provincially piloted approach to gradually establish a comprehensive agricultural socialized service system. This system seeks to restructure rural production through standardized operations, mechanization, and platform-based coordination. These services span the entire agricultural process—from plowing and planting to plant protection, harvesting, processing, and distribution—and aim to enhance production efficiency, reduce transaction costs, and strengthen the connection between smallholders, markets, and technologies.
This study conceptualizes agricultural socialization services as an institutional mechanism for the restructuring of socio-technical systems. Rather than being merely an efficiency-enhancing tool on the production side, these services function as a strategic pathway through which institutional intervention, organizational innovation, and resource reallocation are mobilized to restructure urban–rural relations and optimize the patterns of income distribution. As such, the agricultural service system not only reshapes the technological infrastructure of rural areas but also fundamentally transforms the institutional boundaries and factor mobility between urban and rural domains.
It is important to emphasize that the reform of agricultural socialization services is not an isolated policy intervention, but rather a core component of China’s broader state-led strategy for spatial reorganization and the optimization of population mobility. In recent years, the Chinese government has actively promoted the reform of the hukou (household registration) system, the construction of a national territorial spatial planning framework, and the advancement of a new urbanization strategy. These efforts form part of an integrated set of institutional instruments—including agricultural policies, land systems, and migration mechanisms—through which the state seeks to guide the orderly restructuring of socio-spatial configurations and establish a more balanced regional development pattern with optimized factor allocation [
1]. Against this backdrop, the institutionalization and policy expansion of agricultural socialization services should be understood as a critical element within China’s overarching system of state-led socio-spatial restructuring.
Against this background, this study makes three key contributions to the fields of rural development and income inequality research. Theoretically, it reconceptualizes agricultural socialization services as an institutionalized mechanism for restructuring socio-technical systems in rural China, rather than merely an instrumental tool for enhancing production efficiency. By introducing the perspectives of platform governance and institutional embeddedness, the study demonstrates how the organic integration of state intervention, technological deployment, and organizational coordination reshapes rural production relations and agricultural governance structures, effectively bridging institutional divides between urban and rural areas. This framework expands the institutional boundaries of agricultural modernization research and enriches the theoretical understanding of state capacity in driving rural transformation. Methodologically, this study applies the Double Machine Learning (DML) approach to identify the causal effects of agricultural service expansion under staggered policy pilot implementation. Compared with traditional methods such as Propensity Score Matching-Difference-in-Differences (PSM-DID) or linear fixed effects models, the DML framework offers significant advantages in addressing endogeneity, high-dimensional covariates, and nonlinear confounding variables. It enhances the robustness and credibility of causal inference in complex policy environments and provides a replicable empirical strategy for evaluating spatially targeted interventions. Empirically, the paper investigates three mediating mechanisms—technological diffusion, agricultural scale consolidation, and human capital mobility—through which agricultural socialization services affect urban–rural income disparities. In addition, it explores heterogeneity across regions with differing functional roles, institutional foundations, and service capacities. Rather than limiting the analysis to average treatment effects, the study emphasizes how, why, and under what conditions these services reduce inequality. This approach yields policy-relevant insights into the poverty-reduction and income-enhancing potential of agricultural service reform across diverse regional and institutional contexts.
2. Literature Review
2.1. Advancements in Research on Agricultural Socialization Services
Agricultural socialization services refer to agricultural-related social organizations providing comprehensive production services, including planting, breeding, processing, and transportation, to agricultural producers based on market demand. The emergence and development of such services have become a focal point for scholars both domestically and internationally, with research primarily focusing on their impact on farmers’ income, production efficiency, and ecological benefits. Existing studies indicate that agricultural socialization services have significantly increased farmers’ income. Based on national micro-survey data, Several scholars found that mechanization services notably increased per capita income among farm households, with the most pronounced benefits observed among low- and middle-income groups. These services also contributed to a reduction in intra-rural income inequality [
2]. In addition, other studies have confirmed that mechanization services improve farmers’ income and production incentives by substituting labor and reducing production costs [
3,
4,
5,
6,
7]. Research conducted by scholars from the Chinese Academy of Sciences further indicates that with the expansion of the socialized service market and the implementation of agricultural machinery purchase subsidies, the number and coverage of such service organizations have significantly increased nationwide. This expansion has facilitated the transfer of rural labor to non-agricultural sectors, thereby boosting farm household incomes [
2].
In terms of enhancing agricultural production efficiency, agricultural socialization services have optimized resource utilization and improved output quality. Based on empirical econometric modeling, Several scholars found that agricultural socialization services effectively improved the ecological efficiency of farmland by reducing pollution emissions and input consumption, while maintaining stable output levels. The impact of hired labor services was particularly significant in major grain-producing regions, whereas the effect of machinery rental services in non-grain areas appeared to be relatively limited. Mechanism analysis further revealed that these services promoted a more specialized division of labor within the agricultural sector, which contributed approximately 11.4% to improvements in ecological efficiency. Accordingly, the expansion of large-scale service provision and the promotion of green mechanization are conducive to advancing sustainable agricultural development [
8].
Moreover, agricultural socialization services play a critical role in promoting the transition to green production. Studies have shown that such services encourage farmers to adopt green technologies and inputs by replacing manual labor with mechanization and standardizing production processes. This transformation facilitates a shift in agricultural production toward energy conservation, emission reduction, and environmentally friendly practices [
8]. Tao and Zhao (2024) noted that the pilot program of full-process socialization services contributed to ensuring food security by reducing production costs, promoting the development of new agricultural business entities, and increasing farmers’ incomes [
9]. Related research based on provincial panel data in China confirmed that socialization services significantly enhance the stability of grain yields, with more pronounced effects observed in major grain-producing and agriculturally developed regions. Taken together, existing studies have strongly affirmed the role of agricultural socialization services in improving agricultural efficiency, output, and farm household income, thereby creating favorable conditions for urban–rural integration and common prosperity.
Recent studies emphasize that agricultural socialization services also serve as an important institutional vehicle for integrating smallholder farmers into organized production networks and transforming both production relations and resource allocation. Service-oriented agricultural reforms have not only advanced agricultural specialization, scale, and digitalization but have also effectively facilitated the linkage between “small-scale production” and the “large-scale market” [
10]. Furthermore, empirical research on service-driven technology diffusion suggests that the “service–technology–benefit” cycle not only enhances agricultural ecological efficiency but also constitutes a critical mechanism through which agricultural socialization services indirectly narrow the urban–rural income gap via technological progress [
11].
2.2. Advancements in Research on the Urban-Rural Income Gap
Research on the urban–rural income gap has primarily focused on measurement methods, influencing factors, and dynamic evolution. Since the launch of the reform and opening-up policy, income levels for both urban and rural residents in China have increased, though the income gap initially widened before gradually narrowing. Thanks to targeted poverty alleviation and rural revitalization strategies, the urban–rural income ratio has declined steadily in recent years, with rural incomes experiencing continuous growth. According to reports from the relevant authorities, the overall urban–rural income gap has continued to narrow; however, regional disparities and intra-rural inequalities remain pronounced. Scholars generally agree that structural factors, institutional arrangements, and unequal resource allocation are the root causes of the urban–rural income gap. The persistence of the dual urban–rural system, industrial structural divergence, and imbalanced factor endowments all contribute to the existing income distribution pattern. Additional critical factors include low levels of rural human capital, weak education and infrastructure systems, limited industrial spillover effects, and disparities in financial and social security access between urban and rural areas. Labor migration to cities, land system reform, regional coordinated development, and the expansion of agricultural socialization services have all been identified as effective approaches to mitigating urban–rural inequality [
12,
13]. Utilizing a double machine learning approach, Lu et al. (2025) found that farmland transfer not only facilitated urban–rural integration but also significantly reduced income disparities [
14]. The structure of urban–rural income distribution is jointly shaped by multiple variables, including labor mobility, land transfer, government subsidies, and socialized service provision.
In terms of measurement methods, both domestic and international studies primarily employ indicators such as the Gini coefficient, the Theil index, and the urban–rural income ratio [
13,
15]. Among these, the Theil index and the urban–rural income ratio are particularly effective in capturing both absolute and relative structural disparities. The study by Wang and Ouyang integrated the income ratio with the Theil index to evaluate the income gap and further analyzed its implications for economic growth [
15]. These measurement tools facilitate a multi-dimensional understanding of distributional imbalances.
The urban–rural income gap is not merely a structural phenomenon; rather, it is closely tied to institutional arrangements and resource misallocation. As argued by Peng et al. (2022), the integration of agricultural socialization services helps dismantle the barriers imposed by traditional production relations and optimizes resource distribution, thereby supporting the theoretical view that institutional design lies at the root of urban–rural income disparities [
16]. Furthermore, a growing body of empirical research has identified technological progress, land scale operations, and human capital accumulation as key intermediary channels through which policy reforms influence income distribution [
2,
17,
18,
19,
20,
21,
22,
23,
24,
25]. For example, Lu et al. (2025) found that farmland transfer policies promote urban–rural integration by facilitating scale operations and labor mobility [
14], while Yang et al. (2023) demonstrated that human capital upgrading accelerates the transfer of rural labor to non-agricultural sectors, thereby indirectly narrowing the income gap [
26].
In summary, the existing literature has laid a solid foundation for uncovering the multifaceted causal mechanisms underlying the urban–rural income gap, while also highlighting the importance of an integrated analysis that considers institutional, technological, and organizational factors—an approach that aligns closely with the empirical framework adopted in this study.
2.3. Research on the Impact of Agricultural Socialization Services on the Urban–Rural Income Gap
In recent years, an increasing number of scholars have focused on how agricultural socialization services influence the urban–rural income gap through multiple mechanisms. Existing studies have not only highlighted the direct effects of such services but also systematically traced the causal pathways through which they operate—particularly via technological progress, scale operations, and labor reallocation. The following section reviews and evaluates the current literature from these three perspectives, with the aim of laying a solid foundation for the theoretical framework and empirical analysis presented in subsequent sections.
From the perspective of the technological progress mechanism, a large body of research has shown that agricultural socialization services significantly enhance rural production efficiency by accelerating the diffusion and application of agricultural technologies, thereby indirectly narrowing the urban–rural income gap. Several scholars found that socialization services serve as vital carriers for the dissemination of new agricultural technologies, helping smallholder farmers overcome barriers related to knowledge and financial constraints. These services effectively improve total factor productivity (TFP), which in turn raises rural household incomes [
8]. From a theoretical standpoint, Peng et al. (2022) further argued that agricultural socialization services can facilitate a transformation in agricultural production relations through technological advancement, leading to a more optimal allocation of production factors [
16]. Moreover, with the growing adoption of digital and smart farming technologies, the role of socialization services in driving technological progress has become even more pronounced. This transformation enables rural labor to be freed from repetitive manual tasks and redirected toward higher value-added activities, thus creating additional momentum for narrowing the income gap between urban and rural areas [
2]. Technological advancement has, therefore, emerged as one of the core pathways linking agricultural socialization services to improvements in the urban–rural income distribution.
With respect to the scale operation mechanism, agricultural socialization services play a critical role in narrowing the urban–rural income gap by promoting the large-scale consolidation of land and other production factors. Empirical evidence from Ma et al. (2023) indicates that the introduction of socialization services accelerates farmland transfer, enabling more concentrated land use, which in turn leads to economies of scale and improvements in both per-unit output and farmers’ incomes [
10]. In addition, service organizations—through entrusted management and cooperative arrangements—have enhanced farmers’ bargaining power in the market, thereby facilitating a deeper integration of smallholders into formal market systems [
16]. Using a double machine learning approach, Several scholars confirmed that scale operation serves as a key mediating variable through which agricultural socialization services affect the structure of urban–rural income distribution, with particularly pronounced effects in major grain-producing and highly mechanized regions [
14]. The scale mechanism is reflected not only in the consolidation of land and service functions but also in the intensification and specialization of the service organizations themselves, further improving production efficiency in rural areas.
At the labor force mechanism level, agricultural socialization services provide an important channel for narrowing the urban–rural income gap by releasing rural labor and facilitating its diversified mobility. The widespread adoption of socialization services has reduced the time and skill barriers associated with agricultural work, enabling a large number of rural laborers to shift into non-agricultural sectors or engage in entrepreneurial activities, thereby expanding income sources for rural households [
2]. Changes in variables such as labor mobility and land transfer indicate that socialization services not only enhance the skill levels of the rural labor force but also optimize the allocation of labor between urban and rural areas [
26]. In line with the mechanisms discussed above, the labor force pathway functions as a crucial intermediary link in the relationship between socialization services and income inequality, underscoring the practical importance of human capital improvement and structural optimization in promoting common prosperity.
In summary, the existing literature has provided a relatively comprehensive account of the causal pathways through which agricultural socialization services influence the urban–rural income gap, particularly through mechanisms of technological progress, scale operations, and labor mobility. However, research that systematically integrates these mechanisms and conducts rigorous empirical quantification remains limited. Building upon the theoretical framework and mechanism analysis outlined in this study, we aim to further investigate the specific impact channels of agricultural socialization services on the urban–rural income gap through multi-path mediation tests and machine learning techniques, thereby contributing to both theoretical advancement and empirical enrichment in the relevant research domain.
2.4. Dual Machine Learning in Economic Research
Double Machine Learning (DML) is a recent approach in causal inference, introduced by Chernozhukov et al. [
27]. This approach integrates high-dimensional machine learning with traditional semiparametric estimation to robustly estimate causal parameters, including policy effects, by effectively managing the bias of high-dimensional covariates through orthogonalization and cross-sample partitioning techniques. Chernozhukov et al. (2018) were the pioneers in introducing this framework within the economics domain for evaluating treatment effects and structural parameters [
27]. Yang et al. (2020) utilized the gradient boosting method within a dual machine learning model, effectively addressing the modeling challenges associated with large datasets and attaining favorable outcomes in audit quality [
28]. Empirical economics researchers have employed DML to analyze policy effects and derive conclusions regarding causal relationships across various issues.
In empirical economic studies, Double Machine Learning (DML) has been increasingly utilized by researchers for causal inference and policy effect evaluation across a variety of contexts. For instance, Zhang and Li (2023) employ a dual machine learning methodology to analyze the impact of network infrastructure on inclusive green growth and regional development disparities, highlighting the significance of infrastructure investment in coordinating regional development [
29]. Existing studies have compared traditional policy evaluation methods such as the Double Machine Learning (DML) model and the Difference-in-Differences (DID) approach, highlighting the advantages of the DML model over previous policy evaluation methods. The DML model employs machine learning algorithms to model high-dimensional covariates and nonlinear relationships, thereby eliminating confounding biases and yielding more accurate estimates of treatment effects. In contrast, the DID model relies on linear assumptions and cannot adequately control for the nonlinear and interactive effects of high-dimensional covariates, potentially leading to bias [
27,
30]. Conventional linear regression may introduce specification bias and produce less robust estimators, whereas DML leverages the strengths of machine learning algorithms in handling nonlinear data, effectively avoiding model misspecification issues. The DML model utilizes machine learning algorithms to automatically fit high-dimensional and complex nonlinear relationships. The DID model, relying on traditional regression techniques, struggles to handle high-dimensional data and nonlinear terms, making it susceptible to model misspecification errors [
27,
28]. The machine learning algorithms in DML employ cross-validation and regularization techniques to mitigate the excessive influence of model misspecification on results. In contrast, DID’s heavy reliance on linear assumptions renders it more prone to model instability [
27]. The DML method does not depend on the parallel trends assumption because it adjusts for covariate effects through machine learning models and directly estimates treatment effects, thereby exhibiting lower reliance on parallel trends [
27]. Additionally, based on the concepts of instrumental variable functions, two-stage predictive residual regression, and sample-splitting fitting, DML can alleviate the “regularization bias” present in machine learning estimates, ensuring an unbiased estimation of treatment effect coefficients even in small samples [
29]. For instance, Lu et al. (2025) utilizing a DML-based causal identification method to examine the impact of China’s land transfer policy on urban-rural integration, conducted a comparative analysis of regression results identified by continuous DID and DML models. Their findings revealed that across six different identification strategies, the coefficients of explanatory variables in DML exhibited greater statistical significance with smaller standard errors in terms of robustness and significance, whereas DID exhibited model instability when controlling for complex relationships. When comparing complex settings involving fixed effects and high-dimensional control variables, DML utilized machine learning algorithms to automatically fit high-dimensional and complex nonlinear relationships, while DID demonstrated limited adaptability to complex models. In terms of model specification robustness, DML’s estimation results remained consistent across different strategies (e.g., incorporating fixed effects or high-dimensional covariates), whereas DID’s estimation results were highly sensitive to model specifications, with coefficients fluctuating substantially across strategies [
14].
These studies demonstrate that DML can adeptly manage high-dimensional covariates and nonlinear relationships, thereby mitigating the omitted variable bias often present in conventional regression methods, so significantly bolstering the trustworthiness of empirical conclusions in economics. The utilization of dual machine learning techniques in economics is advancing, offering a robust instrument for evaluating the effects of intricate policies, including agricultural social services.
In conclusion, previous research has explored the multifaceted effects of agricultural socialization services, the sources and evolution of the urban–rural income gap, and the application of double machine learning (DML) methods in causal inference. However, existing studies have rarely integrated these perspectives into a unified analytical framework.
This study advances the literature in several important ways. First, it reconceptualizes agricultural socialization services as a socio-technical system reform mechanism that connects institutional intervention with structural transformation in rural China. Second, it innovatively applies a DML framework to a staggered provincial pilot policy context, improving the credibility of causal identification by mitigating high-dimensional confounding and endogeneity. Third, it incorporates a mechanism-oriented approach by empirically testing three distinct pathways—technological advancement, agricultural scale expansion, and human capital mobility—through which services affect urban–rural income distribution patterns. These theoretical and methodological innovations collectively enable a more precise empirical examination of the redistributive effects of agricultural socialization services, providing new explanatory pathways for how agricultural modernization functions as an institutional lever for inclusive development.
4. Research Methodology
4.1. Study Area
This study encompasses all provinces in mainland China, excluding Tibet, Hong Kong, Macao, and Taiwan, utilizing 30 of the 31 administrative units as its primary research region. Tibet is excluded from the analysis due to prolonged data deficiencies, inconsistent data quality, and the absence of a regularly executed agricultural socialization service system. This analysis similarly excludes Hong Kong, Macao, and Taiwan as they are not relevant to the policy context or statistical framework of agricultural socialization services.
The selection of provincial administrative units as the research unit is predicated on three primary considerations:
There is a significant correlation between the degree of policy implementation and regional disparities. The policy of agricultural socialization services, particularly the “full range of agricultural socialization services pilot”, is primarily administered by the central government and executed at the local level, with significant variation in implementation across provincial units. For example, while certain western provinces such as Gansu and Qinghai are in the nascent phases of service systems, major agricultural provinces like Hebei, Shandong, and Henan have pioneered the development of agricultural production service supply, financial assistance, and the establishment of digital agricultural platforms. The regional heterogeneity makes the provincial level a suitable analytical unit for illustrating the relationship between the expansion of agricultural services and the income discrepancy between urban and rural areas [
18,
26].
Analyzing information on services in conjunction with data regarding income inequality between urban and rural regions. Provincial-level statistics regarding China’s urban-rural income gap, encompassing rural residents’ net household income, the urban-rural income ratio, and per capita disposable income for both urban and rural populations, are often disseminated by the National Bureau of Statistics. The primary statistical indicators of agricultural socialization services, including service area, service type coverage, and number of service subjects, are predominantly aggregated at the provincial level and disseminated through the annual work report of the Ministry of Agriculture and Rural Development and local agricultural yearbooks. Utilizing provincial administrative units as the research unit enhances data accuracy, alignment, and facilitates a horizontal comparison analysis of critical variables.
Lastly, it facilitates the implementation of spatial identification and the investigation of regional heterogeneity. The provincial-level analysis facilitates a more accurate identification of regional disparities in agricultural service provision, farmer engagement capability, and urban-rural dynamics. This also facilitates the integration of a subregional heterogeneity test into the forthcoming extended study, aiding in the comprehension of the influence of agricultural socialization services across various regions and their respective magnitudes [
39]. Moreover, the scale exhibits robust academic comparability and practical applicability, evidenced by the recurrent utilization of provincial panel data in empirical studies of agricultural services conducted by researchers such as Li et al. (2024) and Sun et al. (2023) [
40,
41].
This study analyzes 30 provincial administrative units in mainland China, excluding Tibet, Hong Kong, Macao, and Taiwan. This selection aligns with the actual dynamics of policy implementation and service system development, while also satisfying the requirements for data on urban-rural income gap and agricultural service indicators. It establishes a comprehensive data repository and a basis for regional comparison, facilitating the application of advanced causal inference techniques, including dual machine learning.
4.2. Modelling
This study will employ the previously delineated theoretical analytical framework to experimentally investigate the impact of agricultural socialization service pilots on the urban-rural income disparity. Challenges related to sample selection bias, data structure limitations, and rigid premise assumptions hinder conventional policy evaluation techniques (such as propensity score matching, synthetic control, and difference-in-differences methods) from delivering an accurate representation of causality in complex policy environments. The Double Machine Learning (DML) model was introduced to capture nonlinear interactions among variables using nonparametric methods. This approach enables an impartial evaluation of the policy disposition effect under limited sample conditions by eliminating the influence of high-dimensional confounding variables in a quasi-natural experiment and utilizing the orthogonalization framework [
27]. The model concurrently uses the “instrumental variable method” to mitigate regularization bias in machine learning estimates and establish a robust statistical basis for causal inference. Current research in causal inference relies on these studies, which lend methodological validity to the DML model by illustrating its superiority in managing high-dimensional data, nonlinear relationships, and endogeneity challenges [
27,
28,
29].
Referring to previous research [
27], this paper constructs a partial linear model using double machine learning (for the specific model design, please refer to
Appendix A). To enhance the stability and reliability of the model estimates, this paper employs a 5-fold cross-validation method to process the regression samples and adjusts the results by taking the median of 101 repeated samples, with the final estimate being the median value.
4.3. Variable Selection
4.3.1. Explained Variable
The Urban-Rural Income Gap (URIG) functions as the principal explained variable. The urban-rural income ratio, the Gini coefficient, and Theil’s index are the three predominant methodologies for quantifying urban-rural income difference. This research uses Theil’s index as the principal explained variable, supplemented by the urban-rural income ratio (inc) and the Gini coefficient (gini) to perform a robustness test. This aims to ensure the reliability and robustness of the study’s conclusions, considering elements including data availability, decomposition capability, and measurement precision. Theil’s index quantifies income inequality by the entropy weight principle, offering great additivity, effective decomposability, and insensitivity to outliers, making it particularly apt for analyzing structural disparities among various locations. The Thiel index, utilizing China’s pronounced urban-rural split, may analyze the income gap between urban and rural areas and assess the impact of social service programs on income distribution [
15]. The Theil’s index is less reliant on sample size and does not necessitate quantile data, making it ideal for assessing and comparing cross-sections of provincial yearly panel data. Research conducted by scholars Zheng Jun and Yi Huanhuan (2023) presents a measurement paradigm that serves as a reference; it utilizes the Thiel index as the principal explained variable to investigate the impact of institutional instruments, such as agricultural insurance, on income inequality within the context of urban-rural integration and development [
13].
The urban-rural income ratio is the most definitive and clear indicator of inequality, as it compares the disposable income per capita of urban and rural residents, providing the most direct evidence of disparity. This indicator solely represents the disparity in average values, neglects alterations in the income distribution structure, is unresponsive to marginal enhancements and acute poverty, and exhibits a specific “mean trap”. Integrating data within the urban-rural dichotomy presents hurdles, despite the Gini coefficient’s prevalent application in assessing overall income inequality. The Gini coefficients for urban and rural areas in China have been reported using diverse units, standards of measurement, and calculation methodologies over the years. Nonetheless, a considerable volume of data is absent at the provincial level, complicating subregional attribution analysis and hindering its application in continuity studies. Consequently, URG serves as the primary explanatory variable in the principal regression of this paper. The robustness test employs Inc and Gini, along with the urban-rural income ratio and urban-rural income, to evaluate if the primary conclusions remain valid across various measurement methodologies.
4.3.2. Core Explanatory Variable
Agricultural socialization services (AGSS). Referring to Zhou et al. (2024) and Zhou et al. (2023) [
42,
43], a staggered difference-in-differences approach is employed to analyze the impact of policy pilot programs on regional rural income disparities among farmers. This approach was founded on two policy documents: the 2013 Guidance for Implementing the Pilot Work on the Socialization of the Entire Agricultural Production Process and the 2016 Notice of the Pilot Work on the Socialization of the Entire Agricultural Production Process. The policy is represented by the policy dummy variable (Treat × Time), which equals 1 if the sample province is a pilot province and 0 if it is not. Time is represented as a year dummy variable; however, due to the staggered implementation of the policy by provincial governments, it is assigned a value of 1 in the year the policy was implemented in the respective province and in the years thereafter, and 0 otherwise.
4.3.3. Control Variable
This article also considers supplementary variables that may influence rural and urban prosperity, enabling an accurate evaluation of policy outcomes with the currently available data. Due to its regularization methodology, double machine learning proficiently manages multidimensional control variables. Below is a compilation of factors that can be regulated to affect the extent of agricultural and regional development: (1) Industrialization Level (IDL) measured by the ratio of industrial added value to regional GDP; (2) Informationization Level (ILS), measured by the ratio of total postal and telecommunication services to regional GDP; (3)The Human Capital Level (HCL), measured by the ratio of students enrolled in general higher education institutions to the resident population at year-end. (4) Transportation Infrastructure Level (TIS), measured by the natural logarithm of total road mileage; (5) Soil Farmland quality (SFQ), represented by the area under soil erosion control measures; (6) Irrigation and water conservancy facilities (IWF), measured by the proportion of irrigated farmland to total cultivated land; (7) Agricultural Disaster Affected Degree (ADA), measured by the total area of farmland affected by disasters.
4.3.4. Mediator
This research develops a dual machine learning causal model to examine the effect of agricultural socialization services on the urban-rural income gap. The model considers the impacts of technological advancement, agricultural scale, and human capital, and incorporates three types of mediating variables to elucidate this link. Research underscores the bifunctional nature of agricultural socialization services. They improve the efficiency of agricultural production. Conversely, they indirectly influence the urban-rural income distribution by facilitating the optimal deployment of agricultural resources and enhancing farm family patterns [
44,
45].
The initial aspect is the influence of technological progress, measured by agricultural total factor productivity (ATFP) and the level of agricultural mechanization (AMS). Agricultural socialization services facilitate the spread and application of agricultural technology and represent a crucial aspect of modern agricultural factor organization. By standardizing operations, utilizing new agricultural machinery, and advocating for digital agricultural platforms, professional service organizations can enhance the technological intensity and labor productivity of agricultural output. On the other hand, socialization services can assist smallholder farmers in surmounting their deficiencies in knowledge, capital, and technology, hence enhancing their production technology standards. “Agricultural total factor productivity” (ATFP) denotes the extent to which technological progress and optimal resource allocation influence agricultural output. Utilizing the green total factor productivity metrics for regional agriculture established by Li et al. (2024) [
45]. The proportion of mechanized cultivated area to total cultivated area serves as an indicator of the level of agricultural mechanization (AMS), based on data from the China Agricultural Mechanization Yearbook and several provincial statistical yearbooks. Increased rates of mechanization boost the ability of service businesses to export technology and accelerate technological penetration.
The second is the agricultural scale effect, quantified through land-scale management (LSM) and service-scale management (SSM). Agricultural social services offer a feasible avenue for small farmers to integrate into contemporary agricultural systems via a “service scale” that progresses to “land scale”. Socialization services reduce barriers for farmers to access the market and engage in large-scale operations, facilitating the intensive management of fragmented arable land. On the other hand, the evolving nature of service entities indicates specialization and conglomeration, resulting in a multiplier effect of “organization + scale” [
10]. Large Scale Management (LSM): The ratio of total cultivated area to the total number of individuals employed in agriculture serves as one indicator. Monitoring of Land Transfers: The Ministry of Agriculture and Rural Development and the China Rural Statistics Yearbook are two data sources. Service Scale Management (SSM) is a methodology for evaluating the density and distribution of service organizations in agricultural output. The assessment is predicated on the production value of professional and ancillary activities in agriculture, forestry, animal husbandry, and fisheries, together with the workforce size in these sectors.
Third, the human capital effect is assessed through rural innovation and entrepreneurship (RIE) and rural labor transfer (RLT). Agricultural social services enhance rural human capital accumulation and mobility by liberating agricultural labor and enabling factor reallocation, thus ameliorating structural income disparities between urban and rural regions. A beneficial outcome of service intervention is that it reduces the time and skill demands of farm households’ productive activities, hence creating greater options for non-farm work or entrepreneurship. Conversely, service-oriented innovation in rural regions is propelled by the unrestricted exchange of knowledge, technology, and capital, which enhances the framework of rural human capital [
13,
26]. A secondary indicator of the amount of rural innovation and entrepreneurship (RIE) is the number of farmers’ cooperatives and people engaged in agriculture, forestry, animal husbandry, and fisheries. This indicator relies on data obtained from statistical bulletins of local agriculture and rural offices. By utilizing data from the National Bureau of Statistics’ Labor Force Survey Yearbook and provincial statistical yearbooks, we may ascertain the amount of rural labor transfer (RLT) as the ratio of those employed in the primary industry to those engaged in the secondary and tertiary industries.
4.4. Source of Data
This analysis examines the period from 2005 to 2022, excluding Tibet, Hong Kong, Macao, and Taiwan, utilizing currently available data. The information presented in this article primarily derives from official Chinese sources, such as the China Statistical Yearbook, China Rural Statistical Yearbook, and China Rural Management Statistical Yearbook, among others. The data were interpolated to address any gaps. The descriptive statistics of the main variables are shown in
Table 1.
6. Conclusions and Discussion
In the context of urban-rural integration and development, the academic and policy communities are currently profoundly concerned with the issue of utilizing agricultural policy instruments to reduce the income gap between urban and rural areas and foster shared prosperity. By employing China’s pilot program of comprehensive socialization services for agricultural production, this paper empirically determines the influence of agricultural socialization services on the urban-rural income divide. In order to conduct a more comprehensive analysis of the mechanism of action and the characteristics of heterogeneity, it then implements a dual machine learning (DML) approach with provincial panel data from 2005 to 2022.
6.1. Agricultural Socialization Services Significantly Reduce the Urban-Rural Income Gap
Primary discoveries of the investigation include the following:
The income gap between urban and rural areas is substantially diminished by agricultural socialization services. The benchmark regression results indicate that the pilot agricultural socialization services (AGSS) have a significant negative impact on the urban-rural income gap, even after accounting for a variety of control variables. Additionally, the core explanatory variable, the urban-rural Theil index (Theil), consistently exhibits a negative coefficient for AGSS. Variable shrinkage treatment, counterfactual tests, and a variety of machine learning models that incorporate sample division ratios, as well as sensitivity analyses of alternative explanatory variables, such as the urban-rural income ratio and the urban-rural Gini coefficient, substantiate the results. The reliability and validity of the findings are illustrated by these analyses.
6.2. Technological Progress, Scale Operations, and Human Capital Accumulation Are Important Pathways to Narrowing the Disparity
Technological advancements, the accumulation of human capital, and the implementation of large-scale operations are critical methods for reducing the income gap. In terms of mechanism analysis, this paper incorporates a mediation effect identification approach into the double machine learning model, examining the mechanisms from three dimensions: technological progress, scale effects, and human capital. The findings indicate that agricultural socialization services contribute to the reduction of the urban-rural income gap by promoting technological advancement—through improvements in agricultural total factor productivity (ATFP) and mechanization levels (AMS); enhancing scale efficiency—via the expansion of land scale management (LSM) and service scale management (SSM); and facilitating human capital accumulation—by promoting the transfer of rural labor to non-agricultural sectors (RLT) and stimulating rural innovation and entrepreneurship (RIE). This discovery offers additional evidence that agricultural socialization services are a critical instrument for increasing the efficacy of agricultural output and are also a fundamental component of the broader pattern of integrated urban-rural development.
6.3. The Impact on Urban—Rural Income Gap Exhibits Significant Regional Heterogeneity. According to Heterogeneity Studies, the Impact of Agricultural Socialization Services on the Urban-Rural Income Gap Is Significantly Influenced by Regional, Policy, and Resource—Endowed Factors
Initially, agricultural socialization services have a more significant effect on the reduction of poverty and the increase in income in the primary food-producing regions, as indicated by regional functional allocation. This may be due to the fact that the urban-rural income divide is more significantly influenced by social services, the service sector is generally well-developed, and a greater proportion of the workforce is employed in agriculture. In contrast, policies are enforced with greater vigor in the primary producing regions.
Secondly, the severity of the policy reaction differs between the East, the central, and the West regions when it is deconstructed by geography. The socialization services system is flawless in the eastern region, where agriculture is not as prevalent, illustrating the policy’s high-level marginal development. In the western and central regions, where there is a substantial agricultural population and limited opportunities for non-farm employment, socialization services are more critical in adjusting the income structure and reducing the urban-rural income divide.
Third, the urban-rural income gap is more significantly influenced by agricultural socialization services in regions with a high concentration of cooperatives, as evidenced by the number of farmers’ professional cooperatives. New agricultural management organizations, such as cooperatives, are well-suited to ensure the effective release of policy dividends, provide socialization services, and integrate services due to their strong organizational and coordination skills.
Lastly, the socialization services’ capacity to reduce the urban-rural income gap is most effective in highly mechanized regions, as indicated by the agricultural mechanization level subgroups. This suggests that the service system is more efficient at allocating production factors and has stronger intergenerational income spillovers when integrated into the technology system.
6.4. Policy Implications and Future Research
Based on the data presented above, the report concludes with the following policy recommendations:
The scope and availability of agricultural socialization services should be further expanded. The accessibility and accuracy of services may be enhanced through the implementation of financial inputs, service standardization, and public platform construction.
Agricultural socialization services continue to encounter challenges, such as fragmented supply and imbalanced regional growth. Encourage the integration of urban and rural areas by utilizing agricultural socialization services to the fullest. Institutional integration and factor circulation should be achieved through agricultural modernization, and socialization services should be integrated into rural industrial, ecological, and governance systems. This will facilitate the convergence of rural and urban production and lifestyles.
Synergy in the development of organizational systems and technological advancement is a priority. Foster agricultural science and technology social service organizations, enhance the agricultural machinery purchase subsidy policy, and encourage the construction of high-quality farmland. These are merely a few of the essential measures that must be taken to fortify the technical foundation of services and facilitate the systematic development of new, high-quality agricultural productivity.
Encourage the development of utility networks in the less developed western and central regions of China. By directing investments, enhancing property rights protection, and attracting talented individuals to rural areas, we can encourage interregional synergistic poverty reduction and direct agricultural socialization services to the less fortunate.
Given the pronounced regional disparities in the impact of agricultural socialization services on the urban-rural income gap, policy formulation should be tailored to local conditions and adopt differentiated strategies: In major grain-producing areas, where the agricultural service system is relatively well-established, continuous investments in technology and mechanical equipment should be made to advance the modernization of agricultural production, thereby sustaining their contribution to narrowing the urban-rural income gap; in major grain-consuming areas, efforts should be accelerated to build the service system, increase financial and technical support, and address the shortcomings of insufficient demand, so as to fully leverage the role of agricultural services in bridging the urban-rural income divide; in areas with balanced production and marketing, efficient service models can be cultivated based on actual conditions to promote the integrated production and circulation of agricultural products. For the eastern, central, and western regions, differentiated measures should be implemented according to their respective agricultural foundations and development levels: the eastern region can prioritize enhancing the quality and added value of agricultural services, encouraging efficient operations and branded development; the central and western regions need to strengthen infrastructure and institutional capacity building, reinforce the organization of cooperatives and the construction of mechanized service teams, so as to narrow the gap with the eastern region. Additionally, it is essential to enhance economies of scale in regions with a large number of cooperatives and provide policy support and technical training in areas with fewer cooperatives or low levels of mechanization. Through these differentiated policies that more precisely leverage agricultural socialization services, it is anticipated that the urban-rural income gap will be effectively narrowed, thereby facilitating the realization of the goal of common prosperity.
Finally, the findings presented in this study are subject to certain limitations. Initially, provincial panels are the primary data source at the data level, but they do not accurately represent individual variations or institutional nesting processes at a more micro level. Secondly, the DML framework is effective in recognizing the mediating impact; however, the mechanism’s complexity may still be deficient. Future research can explore the mechanism of policy effects and spatial spillover paths in agricultural socialization services for urban-rural integration by utilizing micro data from farm households, prefectures, and municipal levels, as well as spatial econometric or structural modeling methods. This will enhance the theoretical framework and policy system.