数字与智能技术的采用如何影响农户土地流转行为？来自中国粮食种植户的证据

• Article

• 

21 October 2026

32 Pages

,

,

and

1

College of Rural Revitalization, Jiangsu Open University, Nanjing 210036, China

2

Institute of Agricultural Economics and Development, Chinese Academy of Agricultural Sciences, Beijing 100081, China

3

School of Business, Jiangnan University, Wuxi 214122, China

4

Center for Rural Development Policy, Nanjing Forestry University, Nanjing 210037, China

Abstract

Farmland rental is an important means of optimizing agricultural land allocation and promoting appropriately scaled farming. Against the backdrop of accelerating digital and intelligent transformation in agricultural, how digital and intelligent technology (DIT) adoption relates to farmers’ farmland rental-in behavior require clear theoretical explanation and empirical validation. Using 2024 survey data from 385 grain-growing farmers in Jiangsu Province, China, this study examines the relationship between DIT adoption and farmland rental-in behavior using Probit, endogenous switching regression, IV-Probit, and conditional mixed-process models. The results show that DIT adoption is significantly and positively associated with both farmland rental-in decisions and rented-in scale. These findings remain robust across alternative samples and estimation specifications. Heterogeneity analyses indicate that this positive association is more pronounced for production-oriented than information-oriented DIT applications, for farms with larger baseline operational scales than smallholders, and for households with lower aging ratios. Mechanism analyses suggest that DIT adoption corresponds to farmland rental-in primarily through labor substitution effects and transaction cost reductions in rural land markets. This study extends farmland rental research to the context of agricultural digitalization, offering empirical insights for advancing digital agriculture, improving farmland rental markets, and fostering appropriately scaled farming operations.

1. Introduction

Efficient farmland allocation is fundamental to agricultural productivity growth and structural transformation [1,2]. As rural households diversify their livelihoods, differences in labor availability and farming capacity increase the need to reallocate land among operators. Where land rental markets suffer from frictions, skilled operators with the potential to expand encounter land constraints, whereas resource-deficient or labor-constrained households may underutilize their holdings [3]. The consequences of this mismatch depend in part on how operating scale interacts with labor availability and production technology to shape machinery use and the organization of farm operations [4]. Farmland rental allows households to adjust the area they operate without changing land ownership. By facilitating transfers toward more productive operators, well-functioning rental markets can improve agricultural efficiency [5,6]. Identifying the factors that shape farmers’ farmland rental-in behavior is therefore crucial for understanding how agricultural production scales adjust to shifting household resource endowments and broader structural transformations.

China provides a relevant setting for examining these issues, given its widespread household farming and ongoing efforts to facilitate farmland rental. Land transfer service centers, for example, support more formal and market-oriented rental transactions [7]. Yet improvements in market institutions do not eliminate negotiation costs, contractual uncertainty, or the constraints imposed by farmland’s fixed location [8]. For grain farmers, renting in additional land also entails greater demands on labor, capital, and management and may increase exposure to production and market risks. Whether farmers can take up rental opportunities therefore depends in part on their capacity to operate additional land. Examining rental-in participation alongside rented-in area helps assess how production and management conditions relate to entry into rental markets and the scale of current rental holdings.

Existing research has examined farmland rental from institutional, household, and production perspectives. Studies of land documentation and titling show that tenure arrangements play an important role in rental market development, although their implications vary with the institutional context [9,10]. Household endowments, perceived tenure security, and trust also influence land transfers between farmers with different production opportunities [11,12]. Other studies highlight the roles of social matching, farmers’ behavioral characteristics, and public intervention in rental transactions [13,14,15]. Beyond market participation, the literature links farmland rental to agricultural labor productivity and examines how machinery ownership and outsourced production services interact with land leasing [16,17,18]. Taken together, these studies suggest that farmland rental depends on the conditions under which land is exchanged and on farms’ technological and organizational capacity.

In recent years, the rapid development of digital and intelligent technologies (DIT) has created new opportunities to improve the allocation of agricultural production factors. In agriculture, the adoption of digital and intelligent technologies is commonly understood as the use of digital hardware, including smartphones, computers, sensors, drones, GPS devices, and automated and unmanned machinery, together with the Internet, big data, the Internet of Things, cloud computing, artificial intelligence, blockchain, and 5G communications to enable digital and intelligent agricultural production and management. Previous studies have found that digital information services, sensing technologies, intelligent equipment, and farm management applications have changed how farmers obtain information, organize field operations, and make production decisions [19,20,21,22]. On the one hand, digital extension has expanded access to agricultural market information and advisory services [23,24]. Digital advice can improve agricultural knowledge and encourage technology adoption, although its effectiveness varies with service content, delivery methods, and farmers’ circumstances [25,26,27]. Evidence on the outcomes of digital agricultural services is also mixed [28]. On the other hand, precision agriculture based on digital and intelligent technologies can improve the targeting of agricultural inputs and enhance production performance [29]. However, individual technologies differ in their functions, adoption patterns, and requirements for equipment use and management decisions [30,31,32]. Improvements in information access primarily affect the market side, whereas technological advances in field operations directly affect the production side. These two forms of application may therefore shape farmers’ willingness and ability to rent in additional farmland in distinct yet complementary ways. Their relationship with farmland rental-in may also vary with the conditions under which farmers adopt these technologies. Expected benefits, compatibility with existing farming practices, and farm resource endowments are all associated with technology adoption [33,34]. Studies of European farms show that digital technology adoption is related to farm characteristics, farmers’ experience, and their existing technological capabilities [35,36,37]. Access to digital infrastructure and the distribution of technological benefits also differ across farms [38,39].

Against this backdrop, scholars have begun to examine the role of digital technologies in farmland rental. One study found that Internet use promotes farmers’ decisions to rent out farmland [40], whereas another reported a negative association between Internet use and farmland rental-in [41]. Zeng et al. [42] found that digital technology use and its intensity were positively associated with participation in farmland rental-in, and identified stronger social networks and improved access to credit as potential pathways. More recent evidence links rural e-commerce to changes in rental market conditions, including rental rates and the participation of agricultural operators [43]. Taken together, these studies provide initial evidence of a close relationship between digital technologies and farmland transactions. At the same time, they highlight the need for further research. First, existing studies have not reached a consistent conclusion about the relationship between digital and intelligent technology adoption and farmers’ farmland rental-in behavior. Second, general Internet connectivity and use must be distinguished from the use of digital and intelligent technologies embedded in agricultural production and management. Third, although the relationship between digitalization and farmland rental has received attention, it remains necessary to distinguish technology adoption for agricultural information acquisition from adoption in production operations and to examine how each relates to farmers’ farmland rental-in behavior.

In light of this, using survey data from 385 grain-growing households in Jiangsu Province, China, collected in 2024, this study examines how DIT adoption relates to two observed dimensions of farmland rental-in behavior: participation decisions and rented-in scale. We characterize adoption by both status and intensity and distinguish between DIT use in agricultural information acquisition and production operations. Baseline Probit and OLS estimates are supplemented with endogenous switching, instrumental-variable, and conditional mixed-process estimates to assess the sensitivity of the findings to selection and potential endogeneity. The distinction between the two functions also informs our theoretical framework. Information-oriented applications may lower the cost of finding rental opportunities, whereas production-oriented applications may ease the labor demands of operating additional land. We examine evidence consistent with these pathways using agricultural labor input and access to farmland rental information. We further explore heterogeneity by current operated area and household labor-force aging.

Compared with existing literature, this study offers three contributions. First, in terms of research perspective and variable construction, this study advances beyond broad or single-dimensional metrics—such as macro-level digital economy indices, general Internet access, or rural e-commerce—by explicitly examining DIT embedded directly in farm-level production and management. Rather than treating digital technology as a single, undifferentiated practice, we decompose DIT into two core functional domains: information acquisition and production operations. Empirically, we operationalize this distinction through both a binary adoption status indicator (recording use in either domain) and a multi-item intensity index spanning both domains, supplemented by domain-specific estimations. This fine-grained variable design allows us to evaluate not only whether overall DIT adoption promotes farmland rental-in, but also which functional mechanism exerts a stronger influence, yielding clear and nuanced evidence on the micro-level drivers of appropriately scaled farming. Second, in its analytical framework and examination of mechanisms, the study links DIT adoption to two constraints on farmland rental-in: the labor required to operate additional land and the cost of finding rental opportunities. It considers both rental-in participation and rented-in area, using agricultural labor input and farmers’ perceived ease of obtaining rental information to assess evidence consistent with these pathways. This focus on production capacity and information search complements existing explanations that emphasize factors such as social networks and access to credit, and provides a more specific account of how DIT adoption may relate to farmers’ rental-in behavior. Third, in empirical identification and policy relevance, the study combines several econometric approaches to assess the robustness of its findings in the presence of potential self-selection and endogeneity in cross-sectional data. It also examines how responses vary by farm size and household aging, providing evidence on possible scale bias and age-related barriers in the benefits of DIT use. These findings have implications for tailoring agricultural technology extension policies and limiting the digital divide between larger farms and smallholders.

2. Theoretical Analysis and Research Hypotheses

Farmland rental enables households to adjust operated area in response to differences in productive capacity and economic opportunities. Because land is spatially immobile, its allocation across farmers changes primarily through transfers of ownership or operating rights. Rental markets allow such adjustments without transferring ownership and can improve the alignment between land, labor, and managerial resources [11].

Renting in and renting out farmland represent related but distinct household choices. Renting out may be constrained by subjective ownership and the endowment effect, even when continued cultivation offers limited returns [44]. It may also accompany household labor reallocation toward off-farm employment [45]. Renting in, meanwhile, requires farmers to evaluate the returns and risks of operating additional land, including the security of rental arrangements [12]. This study therefore focuses on rental-in participation and scale as two dimensions of grain farmers’ expansion decisions.

2.1. Direct Effect of DIT Adoption on Farmland Rental-In Behavior

For grain-growing households, farmland rental-in involves two related decisions. A household will rent in land when the expected gain from cultivating additional farmland, after accounting for rent, production costs, risk, and the costs of finding and securing a rental agreement, is positive. Once it enters the rental market, the household will continue to rent in land as long as the expected marginal return exceeds the marginal costs of rent, production, and management [4,8]. Limited information, labor constraints, and the difficulty of coordinating operations across plots can reduce the returns to additional land or increase these costs. DIT adoption may therefore influence both the decision to participate in the rental market and the area rented in.

One way DIT may affect these decisions is by improving the information available to farmers. Digital information services and decision-support systems can combine data on weather, soils, crop conditions, and markets, helping farmers assess the likely costs, returns, and risks of operating additional land [20,23,46]. Timely monitoring can also support adjustments to production practices when conditions change [19,46]. If these applications improve decisions and reduce avoidable losses, they can raise the expected net returns to rental-in and make market participation more attractive.

DIT may also expand the area a household is able to manage. Intelligent machinery, sensing technologies, and digital management systems can improve the precision of field operations and help farmers coordinate production across plots [19,29]. Automation may reduce the labor required for certain tasks [47]. Where these applications prevent labor and management costs from rising as quickly as the operated area, renting in further land becomes economically feasible at a larger scale [4]. These gains must be weighed against equipment costs, learning requirements, and access to complementary services [22]. When the expected gains exceed these costs and suitable rental land is available, DIT adoption may increase both farmers’ willingness to rent in land and their capacity to operate a larger rented-in area. We therefore propose the following hypothesis:

H1.

DIT adoption increases grain-growing farmers’ likelihood of renting in farmland and the area they rent in.

2.2. Mechanisms Linking DIT Adoption to Farmland Rental-In Behavior

Combining the functional attributes of DIT with its practical applications in agricultural production and management, this paper contends that DIT adoption promotes grain-growing farmers’ farmland rental-in primarily through two key pathways: alleviating agricultural labor constraints and lowering transaction costs in the farmland rental market.

2.2.1. Alleviating Agricultural Labor Constraints

Renting in farmland increases the work required for planting, crop management, harvesting, and supervision. These tasks are often concentrated within short agricultural seasons. Off-farm employment reduces the labor available within the household [45], while an aging labor force may make physically demanding operations more difficult to complete. When family labor is insufficient and hired workers are costly to obtain or supervise, the labor required for additional land raises the marginal cost of expansion [4].

DIT adoption may reduce this labor requirement in two ways. Production-oriented applications, including crop-protection drones and intelligent machinery, can automate selected field operations and reduce manual work [47]. Sensors, digital monitoring, and decision-support systems can also help farmers identify problems and organize interventions across plots, reducing some of the time spent on routine inspection and supervision [19,46]. These technologies do not replace all farm labor, but their use may lower labor input per unit of cultivated land. Evidence that machinery ownership and agricultural machinery services are related to land-leasing decisions further supports the relevance of operational capacity to rental-in [17,18].

When less labor is required to cultivate and supervise additional land, households can expand production with their existing workforce and rely less heavily on hired labor during peak periods. DIT adoption may therefore ease labor constraints that would otherwise prevent farmers from renting in land or limit the area they can manage. We propose:

H2.

DIT adoption promotes grain-growing farmers’ farmland rental-in by alleviating agricultural labor constraints.

2.2.2. Reducing Transaction Costs in the Farmland Rental Market

Renting in land also requires farmers to find available plots, assess their suitability, identify lessors, and agree on rental terms. Search and bargaining generate costs before a transaction, while monitoring agreements and resolving disputes may generate costs afterward [8,48]. Because plots differ in location and quality and rental information is dispersed, these costs can reduce the expected net returns to acquiring additional land.

DIT adoption may reduce the costs of finding and comparing rental opportunities. Reliance on acquaintances and local intermediaries can limit the plots and counterparties known to a farmer [13]. Where reliable rental information is available through digital communication or information platforms, farmers can compare plot location, quality, rent, and lease duration before contacting potential lessors. This may reduce repeated inquiries and improve matching, making rental-in more attractive and helping existing tenants identify further plots. Digital records may also make agreed terms easier to trace, although their contribution to contract compliance depends on verification and institutional support [7,12,49].

The information-search component is particularly relevant to this study. If DIT adoption makes suitable rental opportunities less costly to identify, it can lower the cost of entering the rental market and of acquiring additional plots. We therefore propose the following hypothesis:

H3.

DIT adoption promotes farmland rental-in among grain-growing farmers by reducing rental transaction costs, particularly information-search costs.

In summary, this paper proposes an overarching conceptual framework linking DIT adoption to farmland rental-in behavior (Figure 1).

Figure 1. Conceptual framework.

3. Materials and Methods

3.1. Sample and Data

The primary data used in this study were collected through a field survey of grain farmers conducted by the research team in Jiangsu Province, China, between August and October 2024. Jiangsu is one of China’s major grain-producing provinces and is also at the forefront of agricultural modernization and digital agriculture development. Owing to substantial regional differences in natural conditions, economic development, and agricultural production systems, Jiangsu exhibits considerable heterogeneity across northern, central, and southern regions. Northern Jiangsu shares many characteristics with China’s traditional grain-producing areas, whereas central and southern Jiangsu are more representative of agriculturally advanced regions undergoing rapid structural transformation. This regional diversity makes Jiangsu an appropriate setting for examining how DIT adoption affects farmers’ farmland rental-in behavior.

The survey used regional stratification and multistage sampling. Six prefecture-level cities were selected with attention to grain-sown area and regional agricultural conditions: Wuxi (south); Taizhou and Yangzhou (center); and Yancheng, Huai’an, and Suqian (north). Eight county-level units were selected: Yixing, Jingjiang, Baoying, Gaoyou, Funing, Jianhu, Jinhu, and Shuyang. One township was selected in each county-level unit, except Shuyang, where two were selected. Two administrative villages per township were selected, followed by a target of approximately 20–25 grain-farming households per village. The realized analytical sample includes 49 households in southern, 102 in central, and 234 in northern Jiangsu, across nine townships and 18 villages. Fieldwork planned 450 questionnaires and collected 396 completed questionnaires. Eleven questionnaires were excluded because of missing key fields, inconsistent answers, or implausible values, leaving a final analytical sample of 385 households, or 97.22% of the completed questionnaires.

To evaluate the external validity and robustness of our findings across broader geographic regions, we supplement our empirical analysis with data from the 2020 China Rural Revitalization Survey (CRRS), conducted by the Institute of Rural Development at the Chinese Academy of Social Sciences. The CRRS spans 50 counties, 156 townships, and 300 administrative villages across 10 provinces and autonomous regions (Guangdong, Zhejiang, Shandong, Anhui, Henan, Heilongjiang, Guizhou, Sichuan, Shaanxi, and Ningxia), offering broad national representativeness in terms of geographic coverage and socioeconomic development.

To ensure comparability with our baseline regression sample, we filtered the raw CRRS dataset comprising over 3800 households using a strict protocol: we retained only households actively engaged in grain cultivation during the survey year and excluded observations with severe missing values or logical inconsistencies regarding key variables (such as digital and intelligent technology adoption, farmland transfer, household labor force, and operational characteristics), yielding a final effective sample of 1478 grain-growing households. Additionally, to reconcile phrasing differences between the two survey instruments, we established rigorous variable harmonization rules. First, farmland rental-in behavior is consistently defined across both datasets as whether the household rented in cultivated land during the survey year and the scale of rented farmland. Second, regarding DIT adoption, while the Jiangsu baseline survey captured adoption through items reflecting the application of digital and intelligent technologies across production and management processes (including information acquisition and field operations across plowing, planting, field management, and harvesting), the 2020 CRRS questionnaire captured these dynamics through mapped items such as whether intelligent farm machinery or drone services were utilized and whether digital platforms or APPs were used to acquire agricultural information, allowing us to construct standardized, comparable indicators for adoption behavior. Furthermore, considering the rapid diffusion and technical iteration of digital technologies between 2020 and 2024, the 2020 dataset serves as a “baseline test” reflecting an earlier phase of technological penetration. To account for cross-provincial variations in agricultural endowments, marketization levels, and digital infrastructure, we incorporated Province Fixed Effects into the re-estimated empirical models, alongside county-level control variables including per capita GDP, topographical characteristics, and total agricultural machinery power. Demonstrating that the positive and statistically significant effect of DIT embedding on farmland rental-in persists even within the 2020 national sample—when technology adoption rates were relatively lower—provides compelling evidence for the robustness, durability, and generalizability of our core theoretical mechanisms.

3.2. Variable Construction

Dependent variables. Farmland rental-in decision (Rental-in decision) and farmland rented-in area (Rental-in scale). Focusing on the impact of DIT adoption on grain-growing farmers’ farmland rental-in behavior, and following existing literature [12,17], we decompose farmland rental-in behavior into two dimensions as our dependent variables: (1) Farmland rental-in decision, a binary variable assigned a value of 1 if the household rented in farmland during the survey year, and 0 otherwise; and (2) Farmland rental-in scale, a continuous variable captured by the actual area of farmland rented in by the household during the survey year. In the empirical regressions, the positive farmland rental-in scale is transformed into its natural logarithm to reduce right-skewness and improve distributional normality.

Key explanatory variables. DIT Adoption and DIT Adoption Intensity. Farm-level digitalization increasingly involves the joint adoption of complementary technologies rather than simple internet connectivity. Measurements solely based on internet access, online information acquisition, or the use of a specific smart device capture only fragmented aspects of digital adoption, failing to fully reflect how these technologies are embedded into agricultural production and farm management. Drawing on existing literature and pilot surveys, DIT application in farming encompasses two major dimensions: agricultural production digitalization and agricultural information acquisition digitalization [31,50]. Hence, this study examines both DIT adoption status and DIT adoption intensity to capture both the presence and depth of technological embedding.

(1) DIT Adoption (Binary Variable). To capture whether a household adopts DIT, farmers were surveyed on whether they used DIT during agricultural production (e.g., plant protection drones, remote sensing, agricultural IoT, and smart fertilization/irrigation systems) and whether they utilized internet-based technologies for acquiring agricultural information (e.g., production techniques, input procurement, farmland transfer, and agricultural credit). If a farmer reported using DIT in either dimension, the DIT adoption variable was assigned a value of 1, and 0 otherwise. (2) DIT Adoption Intensity (Continuous Variable). To capture internal variations in the degree of digitalization, we constructed a composite DIT Adoption Intensity Index as a continuous measure of adoption breadth and depth. This index comprises two sub-components: the first captures information-oriented DIT adoption intensity, measured on a 5-point Likert scale; the second reflects production-oriented DIT adoption intensity, measured by the number of production links adopting DIT and rescaled onto a 5-point scale (1 = non-adoption; 2 = adoption in 1 link; 3 = 2 links; 4 = 3 links; 5 = 4 or more links). Summing the scores of these two sub-components yields an overall DIT Adoption Intensity Index ranging from 2 to 10, where higher values indicate a higher degree of DIT embedding.

Control variables. Aside from DIT embedding, a household’s farmland rental-in behavior may be confounded by decision-maker characteristics, household and operational endowments, and village-level environments. To mitigate omitted variable bias, following relevant literature, we incorporate control variables across three tiers: (1) Decision-maker characteristics, including gender, age, and education level [12,50]; (2) Household and operational endowments, including household size, cadre status of family members, household aging rate, proportion of off-farm employed members, and per capita contracted farmland area [45,51,52]; (3) Village-level characteristics, including average village land rental rate, distance to the nearest county seat, and village per capita disposable income [15,49]. Additionally, given the substantial regional economic disparities and agricultural environment variations within Jiangsu Province, we further incorporate region fixed effects (categorized into Southern, Central, and Northern Jiangsu, i.e., Southern Jiangsu, Central Jiangsu, and Northern Jiangsu) to control for unobserved regional heterogeneity. To reduce right-skewness and mitigate the potential influence of extreme values, continuous variables—including farmland rental-in scale, per capita contracted farmland area, average village land rental rate, distance to the county seat, and village per capita disposable income—were log-transformed in empirical regressions. Notably, Table 1 presents descriptive statistics based on the raw, untransformed data for the full sample. Detailed variable definitions, operationalizations, and descriptive statistics are presented in Table 1.

Table 1. Variable definitions and descriptive statistics.

3.3. Econometric Models and Identification Strategies

3.3.1. Baseline Models

This study examines the impacts of DIT adoption on grain-growing farmers’ farmland rental-in decisions and farmland rental-in scale, respectively. Given that the farmland rental-in decision is a binary variable, we employ the Probit model for estimation. Conversely, because the farmland rental-in scale is a continuous variable, the OLS model is utilized. The specific model specifications are formulated as follows:

(1)

Area _ rent i = β 0 + β 1′D i + β 2′X i + δ r + ε i

(2)

In Equations (1) and (2), R i equals 1 if household i rented in farmland during the survey reference period and 0 otherwise. Area_rent i denotes the farmland rental-in scale of household i. D i is the focal explanatory variable and is measured alternately by the binary indicator of DIT adoption and the DIT adoption-intensity score. X i is a vector of decision-maker, household and farm-endowment, and village-level controls, while δᵣ denotes regional fixed effects. Φ(·) is the standard normal cumulative distribution function, and ε i is an idiosyncratic error term. The parameters of primary interest are α 1 and β 1. Because the estimated coefficients of the Probit model cannot be directly interpreted as changes in probability, we also report the average marginal effects.

3.3.2. Mechanism Tests Model

Based on the theoretical analysis presented above, we examine two potential pathways through which DIT adoption affects grain-growing farmers’ farmland rental-in behavior: alleviating labor constraints and reducing transaction costs. First, to test whether DIT embedding significantly impacts these mechanism variables, the model is specified as follows:

(3)

In Equation (3), M ki represents the mechanism variable k for household i. Specifically, k = L denotes labor input per mu (measured in person-days per mu), which captures the channel of labor constraint alleviation; k = T denotes a proxy for farmland rental information search costs, capturing a key component of farmland rental transaction costs. The coefficient θ 1k measures the association between DIT adoption and the corresponding mechanism variable. Subsequently, we further incorporate the mechanism variable M ki into the farmland rental-in decision and scale models to examine the relationship between these mechanism variables and farmland rental-in behavior.

(4)

Area _ rent i = v 0 + v 1 D i + v 2 M ki + v 3 X ki + δ r + τ i

(5)

The coefficients α 1 and β 1 in Equations (1) and (2) represent the total effects of DIT adoption on farmers’ farmland rental-in decisions and rented-in area, respectively. Equation (3) estimates the effect of DIT adoption on the mechanism variables representing agricultural labor constraints and rental transaction costs; the coefficient θ 1k on DIT adoption captures this effect. Equations (4) and (5) include both DIT adoption and the mechanism variables to estimate their effects on rental-in decisions and rented-in area. The coefficients η 1 and v 1 represent the direct effects of DIT adoption on these two outcomes after controlling for the mechanism variables, while the products θ 1k × η 2 and θ 1k × v 2 represent the corresponding indirect effects, and τ i is an idiosyncratic error term.

3.3.3. Endogeneity and Identification Strategy

The baseline estimates may suffer from endogeneity bias stemming from three primary sources. First, unobservable characteristics—such as managerial ability, risk preferences, digital literacy, and social networks—may jointly confound DIT adoption and farmland rental-in behavior. Second, reverse causality may arise if households operating larger farms or possessing stronger expansion intentions have greater incentives to adopt DITs. Third, DIT adoption is not randomly assigned but represents a voluntary choice driven by household endowments and expected returns, thereby introducing potential self-selection bias. To address these concerns, we employ a suite of complementary estimators [53]. Specifically, the IV-Probit model is utilized for the binary rental-in decision, while the 2SLS approach is applied to the continuous rental-in scale. Furthermore, recognizing that the binary DIT indicator constitutes a potentially endogenous treatment, we estimate an endogenous switching regression model to rigorously correct for selection bias in the continuous rental-in scale outcome. Finally, we employ the CMP framework [54] to jointly model DIT adoption and farmland rental-in, explicitly allowing the unobserved disturbances across equations to be correlated. Although these estimators rely on different identifying assumptions, consistent results across these alternative approaches provide robust evidence validating the baseline findings.

• Endogenous switching regression

The ESR model jointly estimates a DIT-selection equation and regime-specific outcome equations, thereby correcting for selection on both observed and unobserved characteristics and enabling counterfactual treatment-effect calculations [55]. Let W i* denote the latent net utility that household i derives from adopting DIT. The observed adoption decision is generated by:

(6)

In Equation (6), D i equals 1 when W i* > 0 and 0 otherwise. Z i contains the variables determining DIT adoption, including the covariates in X i and the excluded instrument defined below; γ is a vector of parameters, and μ i is the selection-equation disturbance. The regime-specific rental-in scale equations are:

(7)

In Equation (7), Y 1i and Y 0i represent the potential farmland rental-in outcomes for household i under DIT adoption and non-adoption, respectively. Crucially, the ESR model allows the unobserved error term of the selection equation (μ i) to be correlated with the disturbance terms of the regime outcome equations ε 1i and ε 0i, where the corresponding correlation coefficients are defined as ρ 1 = corr(μ i, ε 1i) and ρ 0 = corr(μ i, ε 0i). A statistically significant ρ 1 or ρ 0 confirms the presence of selection bias driven by unobservable characteristics. The selection and outcome equations are jointly estimated using the full-information maximum likelihood method.

The Average Treatment Effect on the Treated (ATT) evaluates the difference between the expected outcome observed for DIT-adopting households and the counterfactual outcome those same households would have achieved had they not adopted DITs:

ATT = E(Y 1 i|D i = 1) − E(Y 0 i|D i = 1) = E(Y 1 i − Y 0 i|D i = 1)

(8)

In Equation (8), E(Y 1i|D i = 1) represents the expected actual farmland rental-in for DIT-adopting households, which is directly estimable; (Y 0i|D i = 1) denotes the counterfactual expected rental-in for these same households in the absence of DIT adoption. The difference between the two captures the average treatment effect on the treated of DIT adoption on farmland rental-in scale for actual adopters.

• Instrumental-variable Method

To address potential endogeneity concerns, the identification strategy requires at least one valid instrumental variable. Following the standard literature, we construct the instrument as the leave-one-out village mean of DIT adoption (or adoption intensity) among other grain-growing households within the same village. A valid instrument must satisfy both the relevance and exclusion restrictions. Regarding relevance, based on social network theory and peer effects, the adoption behavior of local peers provides learning opportunities and demonstration effects, which strongly increase household i’s propensity or intensity of adopting DITs. Regarding the exclusion restriction, the DIT adoption decisions of fellow villagers do not directly dictate households’ land allocation or farmland rental-in choices; rather, they affect households’ farmland rental outcomes solely through the channel of households’ own DIT adoption. Thus, the constructed leave-one-out mean serves as a plausible instrument. Instrument relevance is evaluated using the first-stage estimates and the applicable weak-instrument diagnostics.

To evaluate the sensitivity of our estimates to potential departures from strict exogeneity, this study incorporates the “plausibly exogenous” framework proposed by Conley et al. [56]. The core mechanics of this approach lie in relaxing the strict, knife-edge assumption that the instrument has zero direct effect on the dependent variable (i.e., the exclusion restriction coefficient γ = 0). Instead, it permits the instrument to exert a subtle direct effect on farmers’ farmland rental-in behavior, bounding this potential violation within a pre-specified plausible interval (γ ∈ [

,

]). Utilizing the Union of Confidence Intervals (UCI) method and Local-To-Zero (LTZ) approximations, this framework constructs recalibrated confidence intervals for the primary estimates under varying degrees of exclusion restriction relaxation. Crucially, this approach allows us to identify the critical breakdown value—defined as the minimum threshold of violation required to nullify the statistical significance of DIT adoption—thereby systematically evaluating the tolerance boundary of our identification strategy against deviations from perfect exogeneity.

4. Results

4.1. Baseline Results

Before estimating the baseline models, pairwise correlations and variance inflation factors (VIFs) were examined. All absolute pairwise correlations were below 0.40, and all VIFs were below 1.20. These diagnostics provide no indication of consequential multicollinearity among the regressors.

4.1.1. Farmland Rental-In Decision

Table 2 reports the baseline Probit estimates for grain farmers’ farmland rental-in participation. Because coefficients from a nonlinear Probit model cannot be directly interpreted as changes in probability, average marginal effects are reported alongside the coefficient estimates. Columns (1) through (4) present the estimated coefficients and average marginal effects of the binary DIT adoption indicator, both without and with controls. The estimates are positive and statistically significant across all specifications, indicating that DIT adoption is associated with a higher probability of renting in farmland. This finding is broadly consistent with previous evidence reported by Zeng et al. [42]. Columns (5) through (8) report the corresponding results for DIT adoption intensity. Both the coefficients and average marginal effects are positive and statistically significant at the 1% level, regardless of whether controls are included. These results indicate that broader and more intensive use of DIT in agricultural production and management is associated with a higher probability of farmland rental-in.

Table 2. Baseline Probit estimates for the farmland rental-in decision.

Taken together, the estimates based on both DIT measures reveal a significant positive relationship between DIT adoption and grain farmers’ farmland rental-in decision, providing initial support for H1.

Regarding the control variables, the age of the household decision-maker, the household aging ratio, the share of household laborers engaged in off-farm employment, and village economic development are significantly negatively associated with farmland rental-in participation. These relationships suggest that agricultural labor aging, the reallocation of household labor toward nonfarm employment, and the greater availability of nonfarm opportunities may weaken farmers’ willingness to expand their agricultural operations. Because the controls are included primarily to account for observable household and village characteristics and reduce potential omitted-variable bias, their estimates are not discussed further.

4.1.2. Farmland Rental-In Scale

Table 3 reports the regression estimates for the relationship between DIT adoption and grain farmers’ farmland rental-in scale. Columns (1) and (2) present the estimates without and with control variables, respectively. The coefficient on DIT adoption is positive and statistically significant at the 1% level in both specifications. This relationship remains significant after controlling for individual characteristics, household and farm endowments, village characteristics, and regional fixed effects, indicating a consistent positive association between DIT adoption and rental-in scale. Columns (3) and (4) report the corresponding estimates for DIT adoption intensity. Its coefficient is also positive and statistically significant at the 1% level regardless of whether controls are included. These findings suggest that broader and more intensive use of DIT in agricultural production and management may strengthen farmers’ capacity to expand their operations through farmland rental-in.

Table 3. Baseline OLS estimates for farmland rental-in scale.

Together with the estimates for the rental-in decision, these results indicate that DIT adoption is positively associated with both grain farmers’ likelihood of renting in farmland and their rental-in scale, providing further support for H1.

4.2. Addressing Endogeneity

4.2.1. Correcting for Self-Selection

Because a household’s decision to adopt DIT in agricultural production is non-random—representing a self-selection choice driven by household resource endowments, policy support, local infrastructure, technological accessibility, and relative farming returns—we employ Endogenous Switching Probit (ESP) and Endogenous Switching Regression (ESR) models to estimate the treatment effects of DIT adoption on farmland rental-in behavior while correcting for selection bias. The detailed estimation results are presented in Table 4. Specifically, Columns (1) and (4) report the selection equation estimates for DIT adoption under the ESP and ESR frameworks, respectively. Columns (2)–(3) and Columns (5)–(6) display the regime outcome equation estimates for farmland rental-in decision (under ESP) and rental-in scale (under ESR) across DIT adopters and non-adopters, respectively. The Wald test for independence of equations statistically rejects the null hypothesis of mutual independence between the selection and outcome equations at the 1% significance level. Let ρ 1 and ρ 0 denote the correlation coefficients between the error terms of the selection equation and the regime outcome equations for DIT adopters and non-adopters, respectively. The estimated ρ 1 is statistically significant at the 1% level, confirming the presence of unobserved confounders that jointly influence DIT adoption decisions and farmland rental-in behavior (i.e., selection on unobservable). Ignoring this self-selection would yield biased estimates, thereby validating the necessity and appropriateness of the ESP/ESR modeling strategy. Furthermore, the goodness-of-fit Wald chi 2 statistics for the ESP and ESR models are statistically significant at the 1% and 5% levels, respectively, demonstrating strong overall model fit and highly stable empirical findings.

Table 4. ESP and ESR estimates correcting for self-selection.

Table 5 further presents the Average Treatment Effects on the Treated derived from the ESP and ESR models, capturing the difference in expected farmland rental-in outcomes for DIT-adopting households between their actual adoption state and their counterfactual non-adoption state. Regarding the farmland rental-in decision (ESP model), the ATT estimate is 0.131, statistically significant at the 1% level. This implies that, relative to their counterfactual scenario of non-adoption, DIT adoption elevates the probability of actual adopters renting in farmland by an average of 13.1 percentage points. Regarding the farmland rental-in scale (ESR model), the ATT estimate is 1.918, which is likewise statistically significant at the 1% level, demonstrating that DIT adoption exerts a substantial positive effect on expanding farmland rental-in scale. Consequently, after rigorously controlling for selection bias originating from both observable and unobservable confounders, DIT adoption consistently increases the probability and expands the scale of farmland rental-in among grain-growing households, providing robust empirical support for Hypothesis H1.

Table 5. Estimation Results for Average Treatment Effects on the Treated.

4.2.2. Instrumental-Variable Estimates

To address potential endogeneity stemming from reverse causality and omitted variables, we employ the leave-one-out average DIT adoption status of other grain farmers in the same village as an instrumental variable, estimating two-step IV-Probit and 2SLS models (Table 6). Columns (1) and (3) present the first-stage estimates. The instrument coefficient is positive and statistically significant (at the 10% level), and the reported first-stage F-statistic, alongside the Anderson–Rubin and Wald test statistics, provides empirical support for instrument relevance while mitigating weak-identification concerns. Columns (2) and (4) report the second-stage results. After instrumenting for DIT adoption, the average marginal effect on farmland rental-in participation remains positive and statistically significant (Column 2). Likewise, the 2SLS estimate in Column (4) confirms a positive and statistically significant association between DIT adoption and rented-in area, consistent with our baseline estimates. Furthermore, recognizing that instrument relevance alone does not guarantee the exclusion restriction—which cannot be directly tested—we follow Conley et al. [56] to evaluate the sensitivity of our findings to limited departures from exact exogeneity. Under the UCI approach (Column 5), the baseline inference remains robust within the recalibrated confidence set. Similarly, the LTZ specification in Column (6) shows that the estimated effect on rental-in scale remains positive and statistically significant. Overall, these results demonstrate that our main findings are robust to potential endogeneity and minor deviations from strict exogeneity.

Table 6. Instrumental-variable estimates for DIT adoption.

Table 7 reports the instrumental variable estimates using DIT adoption intensity as the key explanatory variable. The instrument is the mean DIT adoption intensity of other grain farmers in the same village. The first-stage coefficients reported in columns (1) and (3) are statistically significant at the 1% level. The reported Anderson–Rubin and Wald statistics for the IV-Probit model and the first-stage F statistic for the 2SLS model are also statistically significant, supporting the relevance of the instrument. The average marginal effect reported in column (2) and the coefficient reported in column (4) are positive and statistically significant. Both estimates are larger than their corresponding baseline estimates in column (8) of Table 2 and column (4) of Table 3. This pattern suggests that endogeneity may also generate downward bias in the baseline estimates for DIT adoption intensity. However, accounting for potential endogeneity does not alter the positive relationship between DIT adoption intensity and either farmland rental-in participation or scale. Overall, the instrumental variable estimates based on both DIT adoption status and intensity are consistent with the baseline findings. The results provide further evidence that DIT adoption increases grain farmers’ probability and scale of farmland rental-in, thereby lending additional support to H1.

Table 7. Instrumental-variable estimates for DIT adoption intensity.

4.3. Robustness Checks

4.3.1. Alternative Estimator

As a complementary approach to the IV-Probit and 2SLS estimators, the CMP method jointly estimates the DIT adoption equation and the farmland rental-in outcome equation while allowing their disturbances to be correlated. Table 8 reports the results. Panel A presents the estimates using the binary DIT adoption indicator. Columns (1) and (2) report the equations for DIT adoption and farmland rental-in participation, respectively, while columns (3) and (4) report the corresponding equations for DIT adoption and rental-in scale. The instrument has a positive and statistically significant coefficient in both DIT adoption equations, supporting its relevance. The estimated Atanhrho_12 is statistically significant at the 1% level in both systems, indicating that unobserved factors affecting DIT adoption are correlated with those affecting farmland rental-in participation and scale. This result supports the use of joint estimation to account for potential endogeneity. The coefficients on DIT adoption in columns (2) and (4) are positive and statistically significant at the 1% level, consistent with the baseline estimates. Panel B reports the CMP estimates using DIT adoption intensity. The coefficients on DIT adoption intensity in columns (2) and (4) are also positive and statistically significant at the 1% level for farmland rental-in participation and scale, respectively. These estimates are consistent with the baseline results.

Table 8. Conditional mixed-process estimates.

Overall, the CMP estimates based on both DIT adoption status and intensity confirm that the positive relationships with farmland rental-in participation and scale remain after accounting for correlated unobserved factors across equations. The results provide further support for H1.

4.3.2. Excluding Villages Exposed to Digital-Agriculture Projects

Areas that have implemented digital agriculture development projects typically receive targeted funding, technical assistance, and infrastructure investment. These projects may improve local digital infrastructure, accelerate the diffusion of digital technologies, and strengthen complementary agricultural services. Farmers located in such areas may therefore exhibit greater willingness and capacity to integrate DIT into agricultural production and management. If exposure to these projects also affects farmland rental-in, the baseline estimates may partly capture the influence of the project environment and consequently overstate the independent relationship between household DIT adoption and farmland rental-in. To address this concern, we exclude 94 households located in villages that had implemented digital agriculture development projects and re-estimate the baseline models using the remaining 291 households. Table 9 reports the results. After excluding these households, both DIT adoption status and intensity remain positively and significantly associated with farmland rental-in participation and scale. These findings indicate that the baseline results are not driven solely by exposure to digital agriculture development projects and provide further evidence of their robustness.

Table 9. Robustness checks excluding villages exposed to digital-agriculture projects.

4.3.3. Alternative Data Source

To evaluate whether our baseline findings are sensitive to the specific Jiangsu sample, we conduct an independent sample robustness check using micro-data on 1478 grain-growing households across 10 provinces from the CRRS. Key variables are reconstructed using questionnaire items with comparable economic meaning. Given differences in survey waves, phrasing, and measurement operationalizations, this exercise evaluates out-of-sample consistency rather than serving as an exact technical replication. As reported in Table 10, both DIT adoption status and adoption intensity exhibit positive and statistically significant relationships with farmland rental-in participation and scale. These results demonstrate that the positive baseline relationship holds across a geographically broader national sample, providing further empirical support for the robustness and external validity of our primary findings.

Table 10. Independent-sample robustness check using the CRRS.

4.4. Mechanism Tests

Building upon our theoretical framework, this section empirically explores two potential behavioral channels through which DIT adoption relates to farmers’ farmland rental-in decisions: relaxing agricultural labor constraints on the production side and lowering transaction costs (specifically, information search costs) on the land market side.

Agricultural labor constraints are operationalized via labor input per unit area in grain cultivation (person-days per mu), capturing household operational labor intensity per unit of land. Transaction costs are operationalized specifically through perceived information-search difficulty, measured using a reverse-coded survey item assessing the difficulty of accessing timely and reliable farmland rental information. Conceptually, transaction costs in farmland leasing encompass both pre-contractual search frictions and post-search bargaining, contracting, monitoring, and enforcement costs. In this study, our choice to focus specifically on information search is grounded in the contextual reality of rural China, where information asymmetry regarding land availability, rent prices, and plot characteristics constitutes the primary pre-contractual entry barrier that digital platforms directly address. Conversely, post-search transaction costs are predominantly mediated through local social networks, informal norms, and village relational governance, where household DIT adoption plays a less direct role [49]. The empirical procedure first evaluates the association between DIT adoption and each mechanism variable, and subsequently examines how these mechanism variables correlate with farmland rental-in participation and scale.

Table 11 presents the empirical results for these behavioral channels. Column (1) shows that DIT adoption is negatively and statistically significantly associated with households’ labor input per mu. Columns (2) and (3) further demonstrate that labor input per mu is negatively correlated with both farmland rental-in participation and scale at the 1% significance level. These patterns are consistent with the theoretical expectation that DIT adoption—particularly through production-oriented applications—economizes on field labor input and relaxes peak-season operational labor constraints, thereby facilitating operational scale expansion. Overall, these results yield empirical evidence consistent with Hypothesis H2.

Table 11. Mechanism-test results.

Regarding the transaction cost channel, Column (4) of Table 11 indicates that DIT adoption is associated with a statistically significant reduction in perceived information-search costs. Columns (5) and (6) reveal that information-search costs are statistically significantly and negatively correlated with farmland rental-in participation and rented-in area, respectively. These results align with existing literature highlighting the role of digital access in mitigating information asymmetries in farmland rental markets [40]. By lowering pre-contractual search frictions, DIT adoption—especially information-oriented applications—appears to ease market entry barriers for households seeking to rent in land, providing supportive empirical patterns for Hypothesis H3.

4.5. Heterogeneity Analysis

4.5.1. Functional Dimensions of DIT Adoption

DIT adoption in agriculture spans two functional dimensions: agricultural production and information acquisition. Because these dimensions augment farmers’ operational capacity through distinct pathways, their impacts on farmland rental-in behavior may vary. We therefore examine production-oriented and information-oriented DIT adoption separately. Panels A and B of Table 12 report the corresponding estimation results. For farmland rental-in participation, the average marginal effects of both DIT dimensions—alongside their respective adoption intensity measures—are positive and statistically significant. For farmland rental-in scale, the estimated coefficients across both dimensions remain positive and statistically significant. These findings confirm that DIT adoption in either agricultural production or information acquisition significantly promotes grain farmers’ participation in farmland rental markets and expands their rented-in area.

Table 12. Heterogeneity by functional dimension of DIT adoption.

Notably, the estimated effects are consistently larger for production-oriented DIT adoption across both outcomes, indicating that DIT adoption in production exerts a stronger driving force on farmland rental-in behavior. Mechanistically, production-oriented DIT directly restructures farming operations—utilizing intelligent machinery, crop-protection drones, and digital management systems to economize on labor inputs per unit of land and lower operational oversight costs. By directly expanding farmers’ effective management capacity boundary, production-oriented DIT creates a more immediate incentive to rent in additional farmland. In contrast, information-oriented DIT primarily alleviates information asymmetries and search costs in farmland transactions. While it enhances decision-making efficiency, translating improved information access into actual farmland rental-in expansion remains contingent on a household’s baseline production capacity and resource endowments. Consequently, production-oriented DIT adoption serves as a more direct and potent catalyst for driving farmers’ farmland rental-in participation and scale expansion.

4.5.2. Heterogeneity by Farm Size

To examine potential heterogeneity across operational scales, we conduct subgroup regressions by classifying households according to their historical operated farmland area in the previous year. Using pre-adoption farm size directly addresses potential endogeneity concerns, as current operated area may already incorporate current rented-in land and create a “bad control” problem. In light of the smallholder-dominated context of Chinese agriculture and our sample distribution, we apply a threshold of 20 mu (approximately 1.33 ha) to split the sample into a smaller baseline-scale group (<20 mu) and a larger baseline-scale group (≥20 mu).

Columns (1)–(4) of Table 13 report the subgroup regression estimates alongside formal tests for coefficient differences between groups. In the larger baseline-scale group, DIT adoption status and adoption intensity are positively and statistically significantly associated with both farmland rental-in participation and rented-in area. By contrast, the corresponding estimates in the smaller baseline-scale group are smaller in magnitude and lack statistical significance. The tests for coefficient differences confirm that these between-group variations are statistically significant at the 10% level.

Table 13. Results of heterogeneity analysis by historical operating scale.

These subgroup results demonstrate that the positive association between DIT adoption and farmland scale expansion is primarily driven by households with larger baseline farm sizes. From an economic standpoint, households with larger baseline operations possess stronger financial, equipment, and managerial capacities, enabling them to better absorb technology adoption fixed costs and achieve scale economies when combining DIT with other productive assets. Conversely, resource and technical constraints may limit smaller-scale households from fully exploiting digital efficiency gains. This heterogeneous pattern aligns with growing literature on digital divide concerns, suggesting that agricultural digital tools may unintentionally widen scale disparities if smallholders lack the necessary complementary resources [38,39].

4.5.3. Heterogeneity by Household Aging

Demographic aging within the agricultural labor force may impose a human capital bottleneck that conditions a household’s capability to translate DIT adoption into land expansion. To examine this structural heterogeneity, we divide the sample at the median share of household agricultural laborers aged 65 or older, distinguishing between lower-aging and higher-aging households.

Columns (1)–(4) of Table 14 present the subgroup estimation results alongside inter-group coefficient difference tests. For farmland rental-in participation, DIT adoption status exerts a positive and statistically significant effect among lower-aging households, whereas this effect loses statistical significance in the higher-aging subsample. While DIT adoption intensity significantly increases rental-in participation across both cohorts, the average marginal effect is markedly larger in the lower-aging group. Regarding farmland rental-in scale, both DIT adoption status and intensity yield positive and statistically significant coefficients across both subgroups, with the estimated magnitudes consistently larger for lower-aging households. Tests of coefficient differences statistically confirm that the land expansionary responsiveness to DIT adoption varies across demographic aging gradients.

Table 14. Results of heterogeneity analysis by household labor-force aging.

These empirical patterns demonstrate that labor force aging serves as a demographic drag that attenuates the land-expansionary impetus generated by DIT adoption. Economically, a younger agricultural labor force possesses greater human capital flexibility and lower cognitive friction, enabling households to master digital tools rapidly, streamline information search in farmland rental markets, and operationalize labor-saving machinery. By effectively alleviating labor constraints and operational supervision costs, DIT adoption empowers younger households to aggressively participate in farmland rental markets and enlarge their rented-in scale. Conversely, highly aged agricultural households face steeper learning curves, digital literacy barriers, and shorter operational investment horizons. These demographic constraints diminish their ability to leverage digital technologies for effective labor substitution, rendering them more risk-averse toward the long-term commitments required for farmland expansion. Consequently, labor force aging acts as a key structural boundary that weakens the positive impact of DIT adoption on farmers’ farmland rental-in participation and scale expansion.

5. Discussion

5.1. Theoretical Implications

This study systematically confirms the positive and statistically significant impact of digital and intelligent technology (DIT) adoption on both the probability and scale of farmland rental-in among grain farmers. This finding aligns with growing evidence that Internet use facilitates participation in farmland rental markets through improved information acquisition and market matching [40]. Moving beyond prior studies focused primarily on general Internet connectivity, this study disaggregates agricultural digitalization into two functional dimensions: information acquisition and production management. We demonstrate that DIT adoption serves not merely as an information search tool, but fundamentally expands farmers’ operational capacity boundaries, thereby acting as a potent driver of land expansion.

The mechanism analysis yields empirical evidence consistent with dual underlying pathways. Regarding the labor constraint channel, DIT adoption corresponds to capital-labor substitution by economizing on labor input per mu, complementing studies that emphasize the labor-saving and efficiency-enhancing roles of precision agriculture and digital tools [20,29]. By relaxing household labor constraints, DIT adoption appears to expand the manageable operational scale achievable under a given family labor endowment, with production-oriented DIT exhibiting a particularly pronounced relationship. Regarding the transaction cost channel, DIT adoption is associated with a statistically significant reduction in perceived information search costs, supporting prior literature on how information frictions restrict rural land market liquidity [8,15]. While this proxy specifically highlights pre-contractual search frictions, it targets the primary market-entry barrier where digital technologies exert direct, friction-reducing effects in decentralized land markets. Looking ahead, future research could further extend this empirical scope by incorporating post-search transaction-cost metrics as dynamic, contract-level tracking data become available.

The heterogeneity analysis reveals a strong complementarity between DIT adoption and baseline resource endowments. For farms with larger initial operational scales and demographically younger labor forces, technology adoption better dilutes fixed overheads and unlocks scale economies, consistent with evidence that adoption costs, digital skills, and farm characteristics shape the benefits of digital agriculture [37,57]. However, this finding also indicates that DIT adoption may exacerbate operational disparities between commercial farms and resource-constrained smallholders, reflecting the broader digital divide identified in agriculture [38,39]. Moreover, digital farming platforms may shape farmers’ production choices and induce technological dependence, further influencing how the benefits of digitalization are distributed across rural economies [58].

5.2. Practical Implications

Agricultural digitalization policies should pivot from expanding general network connectivity toward promoting deep, production-oriented technological applications, prioritizing intelligent machinery, crop-protection drones, precision-input systems, and digital farm management software. Meanwhile, to prevent high capital barriers from excluding smallholders, shared machinery platforms, digital service networks, and custom-hiring agricultural service organizations should be fostered to improve technology access without requiring individual ownership.

Digital infrastructure should be deeply integrated with institutional governance of farmland rental markets. Upgrading rural property rights exchanges to provide standardized, verifiable information—such as plot telemetry, land quality, rental prices, contract durations, and legal encumbrances—while coupling digital registries with plot verification, standardized contracts, credit mechanisms, and dispute-resolution services can substantially minimize contractual uncertainty and market frictions.

Policy support must account for differences in farm size, demographic age structure, and digital capability by providing targeted training, simplified interfaces, local technical assistance, and demonstration programs for smallholders and older farmers. The overarching objective should not be farm expansion per se, but rather facilitating economically viable and environmentally sustainable moderate-scale operations through the coordinated alignment of technology, socialized services, and land market institutions.

5.3. Limitations and Future Research

First, while this study provides robust empirical evidence using primary survey data from Jiangsu Province alongside out-of-sample validation with national CRRS data, the findings should be interpreted within the context of cross-sectional identification. Although extensive village-level controls, cluster-robust standard errors, and Conley et al. [56] sensitivity analyses are incorporated to mitigate potential endogeneity and instrument validity concerns, cross-sectional specifications naturally face limitations in capturing dynamic behavioral adjustments over time. Future studies could build upon these findings by leveraging multi-period panel datasets or quasi-experimental policy rollouts to further evaluate the long-term, dynamic effects of DIT adoption.

Key adoption and mechanism variables are currently operationalized via self-reported survey questionnaire items. Future research could combine household surveys with platform records, machinery IoT telemetry, satellite remote sensing, and multidimensional measures of transaction costs (e.g., bargaining, contracting, monitoring, and enforcement). Furthermore, while our mechanism analysis establishes empirical relationships aligned with theoretical channels, fully identified causal mediation analysis remains a priority for future work.

Cross-sample testing across the Jiangsu and CRRS datasets should be interpreted as an independent-sample robustness check rather than an exact technical replication due to differences in survey wave and variable construction. Future longitudinal studies using harmonized surveys across regions should systematically evaluate whether DIT-driven farmland expansion enhances total factor productivity, farm profitability, environmental performance, income distribution, and overall household welfare over the long term.

6. Conclusions

6.1. Main Findings

Based on primary micro-survey data from 385 grain-producing households in Jiangsu Province, supplemented by out-of-sample validation using data from the China Rural Revitalization Survey, this study empirically examines the relationship between digital and intelligent technology adoption and farmland rental-in behavior. Three main conclusions emerge from the analysis:

First, both DIT adoption status and intensity exhibit a statistically significant positive association with farmers’ propensity to rent in farmland and their rented-in area. This positive relationship remains robust across alternative model specifications and when tested on the independent national CRRS sample. These findings indicate that DIT adoption correlates not only with within-farm operational efficiency improvements, but also strongly aligns with factor reallocation patterns and farmers’ land market participation decisions.

Second, labor substitution and transaction cost reduction represent two complementary behavioral pathways corresponding to farmland rental-in decisions. On the one hand, by economizing on labor input per unit of land, DIT adoption aligns with the relaxation of family labor constraints and the expansion of farmers’ manageable operational scale boundaries. On the other hand, DIT adoption is significantly associated with reduced information search frictions in rural land markets, facilitating timely access to land rental opportunities. Together, these dual channels demonstrate that digitally enabled farmland rental-in behavior encompasses both physical operational capacity enhancement and transaction cost reduction.

Third, the association between DIT adoption and farmland rental-in behavior exhibits marked functional and structural heterogeneity. In the functional dimension, production-oriented DIT adoption displays a noticeably stronger positive association with farmland rental-in than information-oriented DIT adoption. In the structural dimension, this positive association is significantly more pronounced among households with larger baseline operational scales and demographically younger labor forces. These divergent patterns highlight a fundamental complementarity between digital technologies and farmers’ baseline resource endowments, implying that without targeted policy support for smallholders and aging farm households, agricultural digitalization may promote land consolidation unevenly and potentially reinforce the rural “digital divide.”

Finally, regarding external validity and empirical boundaries, although validation using broader CRRS data confirms the overall stability of these empirical patterns, caution is warranted when generalizing these findings to regions with vastly different agrarian structures, digital infrastructure, or topographical conditions. In particular, in economically lagging or mountainous areas where land fragmentation remains high and digital access barriers persist, the positive associations observed in this study may be attenuated.

6.2. Policy Implications

Based on our empirical findings, we propose the following actionable policy recommendations:

First, policy support should prioritize the practical application of production-oriented DIT over general digital connectivity. Agricultural extension agencies and local advisory services should organize field demonstrations and technology trials tailored to regional cropping systems, baseline farm sizes, and local labor constraints. To overcome capital barriers for small-scale farmers, local governments should foster agricultural socialized service organizations and rural cooperatives. Providing affordable service contracts for smart machinery operations, plant-protection drone applications, and digital farm management enables smallholders to access advanced technology without incurring prohibitive equipment purchasing costs.

Second, digital infrastructure should be integrated with institutional governance of farmland rental markets to lower transaction costs. Given the evidence on information search frictions, regional property rights exchanges should be upgraded into standardized digital land market platforms. These platforms should provide regularly updated, verifiable information regarding plot locations, soil quality, rental prices, lease durations, and land-use restrictions. Coupled with plot verification, standardized legal contracts, and dispute-resolution mechanisms, such digital services can substantially minimize information asymmetries and lower search effort for farmers seeking to expand operations.

Third, DIT promotion strategies must implement targeted interventions that account for heterogeneous structural constraints across farm sizes and labor age profiles. Because the land-expansionary effects of DIT are weaker among smallholders and aging households, policy measures must actively mitigate the risk of a widening rural “digital divide.” For small-scale operators, service-based adoption models should be expanded to deliver technology benefits without requiring individual capital ownership. For households with aging labor forces, extension providers should offer simplified mobile interfaces, practical field demonstrations, assisted digital services, and local technical guidance, ensuring that farmers across diverse resource endowments can leverage DIT according to their operational needs.

Fourth, digital agriculture initiatives should be coordinated with land market institutions and agricultural service systems to promote sustainable moderate-scale operations. Local agricultural services should assist farmers in evaluating their realistic operational capacity boundaries—aligning land expansion decisions with available labor, financial liquidity, and service accessibility. Land rental arrangements must strictly adhere to voluntary principles and accommodate local tenure conditions. Furthermore, program evaluation should extend beyond changes in operated area to incorporate multidimensional metrics, including farm profitability, labor savings, service accessibility, and resource-use efficiency, thereby supporting economically viable and environmentally appropriate scale expansion suited to local conditions.

Author Contributions

Conceptualization, G.H. and Q.G.; literature review, G.H.; methodology, G.H. and G.W.; formal analysis, G.H.; data curation, G.W.; investigation, G.H. and G.W.; validation, G.H. and J.W.; visualization, G.H.; writing—original draft preparation, G.H.; writing—review and editing, G.H., J.W. and Q.G.; supervision, Q.G.; funding acquisition, G.H. and J.W. All authors have read and agreed to the published version of the manuscript.

Funding

This research was supported by the National Social Science Fund of China Major Project (Grant No. 25&ZD282); Anhui Provincial Outstanding Young Teachers Cultivation Program (YQYB2026056); Chuzhou University Research Launch Fund Project (2026qd69).

Institutional Review Board Statement

According to the Civil Code of the People’s Republic of China, Article 1035, the collection and use of private information in this study were verbally consented to by the participants, and confidentiality measures were implemented. Therefore, ethical approval is not required for this type of study, and it is exempt from ethical review.

Data Availability Statement

Restrictions apply to the availability of the data used in this study. The primary survey data were collected as part of a government-commissioned public livelihood survey and are not publicly available because they contain confidential information. These data may be obtained from the corresponding author upon reasonable request, subject to applicable confidentiality and data-use requirements. The CRRS data were obtained from a third party, the Rural Development Institute of the Chinese Academy of Social Sciences, and are available at http://rdi.cass.cn/dcsj/202306/t20230607_5643271.shtml (accessed on 15 March 2026) with the permission of the data provider.

Acknowledgments

We thank the field investigators for their contributions to the household survey. We also thank the Rural Development Institute of the Chinese Academy of Social Sciences for providing the China Rural Revitalization Survey (CRRS) data used in this study. We are grateful to the editor and anonymous reviewers for their careful review and constructive suggestions.

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:

DIT Digital and Intelligent Technologies
CRRS China Rural Revitalization Survey
OLS Ordinary Least Squares
ESR Endogenous Switching Regression
ESP Endogenous Switching Probit
IV Instrumental Variable
2SLS Two-Stage Least Squares
IV-Probit Instrumental-Variable Probit Model
CMP Conditional Mixed Process
ATT Average Treatment Effect on the Treated
VIF Variance Inflation Factor
AR Anderson-Rubin Test
LTZ Local-to-Zero Approach
UCI Union of Confidence Intervals Approach

References

1. Adamopoulos, T.; Brandt, L.; Leight, J.; Restuccia, D. Misallocation, selection, and productivity: A quantitative analysis with panel data from China. Econometrica2022, 90, 1261–1282. [Google Scholar] [CrossRef] [Scilit]
2. Chen, C.; Restuccia, D.; Santaeulàlia-Llopis, R. Land misallocation and productivity. Am. Econ. J. Macroecon.2023, 15, 441–465. [Google Scholar] [CrossRef] [Scilit]
3. Jin, S.; Deininger, K. Land rental markets in the process of rural structural transformation: Productivity and equity impacts from China. J. Comp. Econ.2009, 37, 629–646. [Google Scholar] [CrossRef] [Scilit]
4. Foster, A.D.; Rosenzweig, M.R. Are there too many farms in the world? Labor-market transaction costs, machine capacities, and optimal farm size. J. Political Econ.2022, 130, 636–680. [Google Scholar] [CrossRef] [Scilit]
5. Deininger, K.; Jin, S.; Liu, S.; Xia, F. Property rights reform to support China’s rural–urban integration: Household-level evidence from the Chengdu experiment. Aust. J. Agric. Resour. Econ.2020, 64, 30–54. [Google Scholar] [CrossRef] [Scilit]
6. Chen, C.; Restuccia, D.; Santaeulàlia-Llopis, R. The effects of land markets on resource allocation and agricultural productivity. Rev. Econ. Dyn.2022, 45, 41–54. [Google Scholar] [CrossRef] [Scilit]
7. Fan, P.; Mishra, A.K.; Feng, S.; Su, M. From informal farmland rental to market-oriented transactions: Do China’s Land Transfer Service Centers help? J. Agric. Econ.2025, 76, 74–97. [Google Scholar] [CrossRef] [Scilit]
8. Yang, Z. Measuring transaction costs in China’s rural land rental market: Ex-ante bargaining associated with ex-post contract-violation uncertainty and location specificity. J. Econ. Behav. Organ.2024, 224, 688–716. [Google Scholar] [CrossRef] [Scilit]
9. Wang, H.; Riedinger, J.; Jin, S. Land documents, tenure security and land rental development: Panel evidence from China. China Econ. Rev.2015, 36, 220–235. [Google Scholar] [CrossRef] [Scilit]
10. Cheng, W.; Xu, J.; Zhou, N.; He, Z.; Zhang, L. How did land titling affect China’s rural land rental market? Size, composition and efficiency. Land Use Policy2019, 82, 609–619. [Google Scholar] [CrossRef] [Scilit]
11. Deininger, K.; Jin, S. The potential of land rental markets in the process of economic development: Evidence from China. J. Dev. Econ.2005, 78, 241–270. [Google Scholar] [CrossRef] [Scilit]
12. Ma, X.; Heerink, N.; van Ierland, E.; Lang, H.; Shi, X. Decisions by Chinese households regarding renting in arable land—The impact of tenure security perceptions and trust. China Econ. Rev.2020, 60, 101328. [Google Scholar] [CrossRef] [Scilit]
13. Macours, K.; de Janvry, A.; Sadoulet, E. Insecurity of property rights and social matching in the tenancy market. Eur. Econ. Rev.2010, 54, 880–899. [Google Scholar] [CrossRef] [Scilit]
14. Qian, C.; Li, F.; Antonides, G.; Heerink, N.; Ma, X.; Li, X. Effect of personality traits on smallholders’ land renting behavior: Theory and evidence from the North China Plain. China Econ. Rev.2020, 62, 101510. [Google Scholar] [CrossRef] [Scilit]
15. Zhang, J.; Mishra, A.K.; Ma, X. Mechanism of Chinese farmers’ land rental participation: The role of invisible markets and public intervention. Food Policy2023, 117, 102453. [Google Scholar] [CrossRef] [Scilit]
16. Zhang, J.; Mishra, A.K.; Zhu, P.; Li, X. Land rental market and agricultural labor productivity in rural China: A mediation analysis. World Dev.2020, 135, 105089. [Google Scholar] [CrossRef] [Scilit]
17. Qian, L.; Lu, H.; Gao, Q.; Lu, H. Household-owned farm machinery vs. outsourced machinery services: The impact of agricultural mechanization on the land leasing behavior of relatively large-scale farmers in China. Land Use Policy2022, 115, 106008. [Google Scholar] [CrossRef] [Scilit]
18. Liu, Y.; Heerink, N.; Li, F.; Shi, X. Do agricultural machinery services promote village farmland rental markets? Theory and evidence from a case study in the North China plain. Land Use Policy2022, 122, 106388. [Google Scholar] [CrossRef] [Scilit]
19. Wolfert, S.; Ge, L.; Verdouw, C.; Bogaardt, M.-J. Big Data in Smart Farming—A review. Agric. Syst.2017, 153, 69–80. [Google Scholar] [CrossRef] [Scilit]
20. Khanna, M. Digital transformation of the agricultural sector: Pathways, drivers and policy implications. Appl. Econ. Perspect. Policy2021, 43, 1221–1242. [Google Scholar] [CrossRef] [Scilit]
21. Klerkx, L.; Jakku, E.; Labarthe, P. A review of social science on digital agriculture, smart farming and agriculture 4.0: New contributions and a future research agenda. NJAS Wagening. J. Life Sci.2019, 90–91, 100315. [Google Scholar] [CrossRef] [Scilit]
22. Birner, R.; Daum, T.; Pray, C. Who drives the digital revolution in agriculture? A review of supply-side trends, players and challenges. Appl. Econ. Perspect. Policy2021, 43, 1260–1285. [Google Scholar] [CrossRef] [Scilit]
23. Aker, J.C. Information from markets near and far: Mobile phones and agricultural markets in Niger. Am. Econ. J. Appl. Econ.2010, 2, 46–59. [Google Scholar] [CrossRef] [Scilit]
24. Aker, J.C. Dial “A” for agriculture: A review of information and communication technologies for agricultural extension in developing countries. Agric. Econ.2011, 42, 631–647. [Google Scholar] [CrossRef] [Scilit]
25. Fabregas, R.; Kremer, M.; Schilbach, F. Realizing the potential of digital development: The case of agricultural advice. Science2019, 366, eaay3038. [Google Scholar] [CrossRef] [Scilit]
26. Cole, S.A.; Fernando, A.N. ‘Mobile’izing agricultural advice: Technology adoption, diffusion and sustainability. Econ. J.2021, 131, 192–219. [Google Scholar] [CrossRef] [Scilit]
27. Van Campenhout, B.; Spielman, D.J.; Lecoutere, E. Information and communication technologies to provide agricultural advice to smallholder farmers: Experimental evidence from Uganda. Am. J. Agric. Econ.2021, 103, 317–337. [Google Scholar] [CrossRef] [Scilit]
28. Porciello, J.; Coggins, S.; Mabaya, E.; Otunba-Payne, G. Digital agriculture services in low- and middle-income countries: A systematic scoping review. Glob. Food Secur.2022, 34, 100640. [Google Scholar] [CrossRef] [Scilit]
29. Finger, R.; Swinton, S.M.; El Benni, N.; Walter, A. Precision farming at the nexus of agricultural production and the environment. Annu. Rev. Resour. Econ.2019, 11, 313–335. [Google Scholar] [CrossRef] [Scilit]
30. Lowenberg-DeBoer, J.; Erickson, B. Setting the record straight on precision agriculture adoption. Agron. J.2019, 111, 1552–1569. [Google Scholar] [CrossRef] [Scilit]
31. Miller, N.J.; Griffin, T.W.; Ciampitti, I.A.; Sharda, A. Farm adoption of embodied knowledge and information intensive precision agriculture technology bundles. Precis. Agric.2019, 20, 348–361. [Google Scholar] [CrossRef] [Scilit]
32. Groher, T.; Heitkämper, K.; Walter, A.; Liebisch, F.; Umstätter, C. Status quo of adoption of precision agriculture enabling technologies in Swiss plant production. Precis. Agric.2020, 21, 1327–1350. [Google Scholar] [CrossRef] [Scilit]
33. Pathak, H.S.; Brown, P.; Best, T. A systematic literature review of the factors affecting the precision agriculture adoption process. Precis. Agric.2019, 20, 1292–1316. [Google Scholar] [CrossRef] [Scilit]
34. Shang, L.; Heckelei, T.; Gerullis, M.K.; Börner, J.; Rasch, S. Adoption and diffusion of digital farming technologies—Integrating farm-level evidence and system interaction. Agric. Syst.2021, 190, 103074. [Google Scholar] [CrossRef] [Scilit]
35. Paustian, M.; Theuvsen, L. Adoption of precision agriculture technologies by German crop farmers. Precis. Agric.2017, 18, 701–716. [Google Scholar] [CrossRef] [Scilit]
36. Kernecker, M.; Knierim, A.; Wurbs, A.; Kraus, T.; Borges, F. Experience versus expectation: Farmers’ perceptions of smart farming technologies for cropping systems across Europe. Precis. Agric.2020, 21, 34–50. [Google Scholar] [CrossRef] [Scilit]
37. Gabriel, A.; Gandorfer, M. Adoption of digital technologies in agriculture—An inventory in a European small-scale farming region. Precis. Agric.2023, 24, 68–91. [Google Scholar] [CrossRef] [Scilit]
38. Rotz, S.; Gravely, E.; Mosby, I.; Duncan, E.; Finnis, E.; Horgan, M.; LeBlanc, J.; Martin, R.; Tait Neufeld, H.; Nixon, A.; et al. Automated pastures and the digital divide: How agricultural technologies are shaping labour and rural communities. J. Rural Stud.2019, 68, 112–122. [Google Scholar] [CrossRef] [Scilit]
39. Mehrabi, Z.; McDowell, M.J.; Ricciardi, V.; Levers, C.; Martinez, J.D.; Mehrabi, N.; Wittman, H.; Ramankutty, N.; Jarvis, A. The global divide in data-driven farming. Nat. Sustain.2021, 4, 154–160. [Google Scholar] [CrossRef] [Scilit]
40. Zou, B.; Mishra, A.K. How internet use affects the farmland rental market: An empirical study from rural China. Comput. Electron. Agric.2022, 198, 107075. [Google Scholar] [CrossRef] [Scilit]
41. Zhang, F.; Bao, X.; Deng, X.; Xu, D. Rural land transfer in the information age: Can Internet use affect farmers’ land transfer-in? Land2022, 11, 1761. [Google Scholar] [CrossRef] [Scilit]
42. Zeng, H.; Chen, J.; Gao, Q. The impact of digital technology use on farmers’ land transfer-in: Empirical evidence from Jiangsu, China. Agriculture2024, 14, 89. [Google Scholar] [CrossRef] [Scilit]
43. Zheng, L.; Li, L.; Jin, S. Opening the village door: China’s agricultural e-commerce facilitates the transformation of land rental markets. Agric. Econ.2025, 56, 622–636. [Google Scholar] [CrossRef] [Scilit]
44. Yan, J.; Yang, Y.; Xia, F. Subjective land ownership and the endowment effect in land markets: A case study of the farmland “three rights separation” reform in China. Land Use Policy2021, 101, 105137. [Google Scholar] [CrossRef] [Scilit]
45. Huang, J.; Gao, L.; Rozelle, S. The effect of off-farm employment on the decisions of households to rent out and rent in cultivated land in China. China Agric. Econ. Rev.2012, 4, 5–17. [Google Scholar] [CrossRef] [Scilit]
46. Zhai, Z.; Martín