Digital Rural Development Policy, Labor Employment Stickiness and Land Use Efficiency

Abstract

Maintaining stability in rural labor markets and enhancing labor employment stickiness (RLFS) are essential for alleviating the persistent outflow of rural labor. Based on data from the 2014–2022 China Family Panel Studies and Treating whether digital rural development plans were issued as a quasi-natural experiment, we employ a staggered difference-in-differences model to evaluate the impact of digital rural development policy implementation on RLFS. Meanwhile, we also explore the potential mechanisms through which the policy affects RLFS by combining the analysis with Order Logit model. The results show that the implementation of the digital rural development policy significantly increases RLFS, and these findings remain robust after a series of checks. Mechanism analysis indicates that the policy improves RLFS by strengthening rural workers’ embeddedness in local social networks, enhancing digital literacy and physical health, reducing speculative motives, and expanding local labor demand. Heterogeneity analyses reveal that the policy has stronger positive effects on RLFS among younger and middle-aged individuals, those with lower levels of human capital, and those engaged in agricultural work and that it is more effective in regions with diminishing demographic dividends and weaker land resource endowments. Further analysis suggests that although the policy increases the employment stickiness of younger rural workers and agricultural laborers, it does not improve the efficiency of rural land use. Therefore, the government should continue expanding the coverage of the digital rural development policy to fully leverage its positive effects on rural labor markets while also adjusting existing policy instruments to identify the key channels through which digital technologies can enhance land use efficiency.

1. Introduction

Addressing the constraints that hinder high-quality and sufficient employment among rural labor, ensuring stable employment for rural workers, and assessing the actual effects of new rural development policies in the digital era are critical for advancing rural revitalization and sustaining long-term national development. However, against the backdrop of rapid urbanization and industrialization, rural China is experiencing a substantial outflow of labor, particularly among young workers. The resulting economic consequences include not only an aging rural population, abandoned farmland, and the hollowing out of rural industries [1,2,3], but also, more fundamentally, a systemic weakening of RLFS. Conceptually, RLFS refers to the employment pattern of working-age individuals whose household registration or primary workplace is located in rural areas, characterized by low mobility and high employment stability driven by subjective preferences or the specificity of human capital. At the individual level, RLFS manifests as sustained and stable employment intentions or long, uninterrupted employment spells. In China’s dual urban–rural structure, RLFS is a key factor in maintaining stable employment relationships and is an important indicator of rural employment quality and the degree of integration between rural and urban labor markets. Higher RLFS implies that rural regions can effectively retain labor and sustain a stable local labor market; conversely, weakened RLFS signals persistent labor outflow and declining vitality in rural areas. Traditionally, the rural labor supply has relied on small-scale social networks rooted in local attachment or has been constrained by limited employment options [4,5]. These conditions generate substantial labor market mismatches, making it difficult for rural workers to form stable and long-term employment intentions. On the one hand, such distortions prevent rural labor from fully realizing its factor value; on the other hand, the rising frequency and risk of job switching undermine the healthy functioning of the labor market. Therefore, alleviating labor market distortions faced by rural workers, fostering stable employment intentions, and enhancing RLFS become one of the key issues in promoting rural development. Furthermore, as the stability of rural labor improves, whether digital rural development can effectively raise land resource utilization efficiency and activate regional factor vitality also constitutes urgent tasks for advancing rural revitalization and ensuring employment stability and sustained economic development.

The implementation of the digital rural development strategy offers a new opportunity to address these challenges. It is noteworthy that digital rural development is not merely the application of information tools but represents a profound transformation involving the restructuring of production relations and factor allocation mechanisms. Through the enhancement of digital infrastructure, the introduction of data as a factor, and the comprehensive empowerment provided by digital technologies, digital rural development profoundly affects rural economic structures, governance systems, and public service provisions [6,7], thereby creating new external conditions for the rural labor market. Moreover, digital rural development can not only generate high-quality local employment opportunities by fostering new business models [8] but also enhance human capital and community identity through innovative services such as remote education and online cultural services, thereby optimizing public service experiences via information equity. Collectively, these changes constitute a comprehensive ecosystem that enhances the “attractiveness” of rural areas, making it possible to address the weakening of RLFS.

As a key measure for achieving agricultural and rural modernization, the core of digital rural development policies lies in integrating new digital tools—such as digital infrastructure, digital governance, and digital public services—into the traditional rural economic system, thereby shaping a new external environment for rural labor employment and ultimately influencing individual employment decisions, particularly RLFS. Overall, digital rural development policies promote the further penetration of digital information technologies into rural areas. The establishment of digital information service platforms and remote online training platforms not only injects digital knowledge into the rural workforce but also provides residents with new digital services and tools. While enhancing the match between labor skills and job requirements, these initiatives also foster new rural business formats and transform traditional industries, thereby strengthening the long-term employment willingness of the labor force. Specifically, the direct effects of digital rural development policies on RLFS manifest in the following aspects:

First, the continuous improvement of digital infrastructure under policy promotion reduces both the spatial friction costs of cross-regional employment and the temporal friction costs of local employment for rural labor [9,10], enhancing the matching between labor demand and supply and thereby increasing individual RLFS. On the one hand, regarding spatial friction costs, geographic disparities in traditional labor markets often impose “market segmentation barriers” on rural labor. When local employment opportunities are scarce and cross-regional employment involves high risks, rural workers are compelled to engage in fragmented, temporary employment, which limits both their subjective RLFS and objective employment continuity. Digital rural development policies, through initiatives such as constructing digital base stations and promoting 5G deployment in rural areas, enhance digital infrastructure, reduce spatial friction costs in labor search, and mitigate market segmentation barriers caused by geographic differences. This alleviates labor misallocation due to information asymmetry, further strengthening the long-term employment willingness of labor individuals. For example, in Guiying Village, Longshan County, Hunan Province, the construction of a digital platform facilitated the active publication of convenience services and labor demand information, creating an online communication channel for the rural labor market. The platform also includes an “incident reporting” section, providing an accessible online feedback mechanism that enhances risk-avoidance capacity during cross-regional mobility, thereby supporting the improvement of RLFS. On the other hand, with respect to temporal friction costs, traditional agricultural labor relies heavily on past experience for planting decisions, resulting in path dependency, and the seasonal nature of farming leads to extended land idleness and seasonal unemployment [11]. Digital rural development policies enhance the application of agricultural sensors and big data platforms in crop cultivation, shortening average growth cycles, increasing annual planting batches per plot, and continuously improving agricultural production efficiency and stability. Consequently, labor willingness to engage in agriculture increases. Additionally, the development of remote online training enables laborers to improve their skills without reducing their planting efficiency, increasing their human capital stock and converting it into knowledge for subsequent planting decisions, thereby enhancing long-term agricultural RLFS.

Second, from the perspective of income utility, RLFS fundamentally reflects laborers’ employment decisions and work duration on the basis of the realized utility of different job-related incomes. When the actual utility derived from a given employment option exceeds that of alternatives, RLFS is reinforced. Traditional forms of rural employment, such as crop and livestock farming or casual labor, are characterized by low income levels, high vulnerability to natural risks and market price fluctuations, and limited opportunities for skill enhancement, resulting in low expected income utility and weak long-term employment intentions [12,13]. Digital rural development policies can enhance RLFS by both increasing the marginal returns of traditional employment activities and providing new avenues for career development. On the one hand, these policies promote the adoption of smart agriculture, precision agriculture, and agricultural big data management systems, digitally transforming the entire agricultural value chain and mitigating production-related natural risks and market risks in sales, thereby increasing income stability and average marginal returns. This strengthens laborers’ long-term employment intentions and RLFS within the agricultural sector. On the other hand, the integration of digital technology with traditional agriculture generates “agricultural new business models”, offering rural laborers new career development opportunities. Traditional agricultural work often suffers from low added value, labor intensity, and constrained career prospects, limiting skill and income growth and prompting labor outflow [14]. In contrast, digital integration fosters new agriculture-related positions such as live-streaming agriculture, e-commerce operations, and digital governance specialists, providing rural workers with alternative employment options. Participation in these new agricultural sectors enables laborers to obtain higher skill premiums and compensation, clarifying career trajectories and continuously enhancing RLFS. On the basis of the analysis above, we propose the following hypothesis:

H1. 

The implementation of a digital rural development policy can significantly promote RLFS.

From the perspective of transmission mechanisms, the implementation of a digital rural development policy may influence RLFS through five pathways. First, the policy can enhance RLFS by increasing the degree of rural laborers’ integration into local social networks. Within a given region, digital rural initiatives leverage advances in digital technology and communication infrastructure to transfer social networks traditionally based on kinship and geographic proximity onto digital platforms, overcoming the small-group limitations inherent in conventional rural social ties [15,16]. As digital governance tools continue to improve, new social networks emerge, reshaping local network integration patterns and stabilizing employment activities. Across regions, community platforms such as WeChat and TikTok facilitate the maintenance and expansion of external social networks for rural residents, reduce barriers to cross-regional labor mobility, enhance belonging and identification for migrant workers, and strengthen their long-term employment intentions. Second, the policy can enhance RLFS by improving rural laborers’ digital literacy. Targeted training programs in smart agriculture management and data analytics for large-scale farmers and family farms increase their understanding of new agricultural models, bolster confidence in continued agricultural engagement, and lay the groundwork for further digital transformation. For migrant workers, remote skills training and management or operations courses provided under digital rural initiatives improve labor market skill matching, supporting access to stable employment positions.

Third, the implementation of a digital rural development policy can enhance RLFS by improving laborers’ health conditions. Imbalances in the supply and demand of medical resources have long constrained traditional rural labor, significantly affecting both work capacity and employment stability. With the ongoing development of digital rural infrastructure, integrated online-offline healthcare systems have gradually emerged. The rise in teleconsultations and online medical services reduces the costs and delays associated with seeking medical care for rural laborers [17]. The establishment of electronic health records increases transparency regarding workers’ health status, effectively preventing major illnesses while avoiding redundant medical expenses. These improvements mitigate sudden employment interruptions and strengthen RLFS.

Furthermore, the implementation of a digital rural development policy can enhance RLFS by reducing laborers’ speculative motives. For rural laborers engaged in agricultural production, the integration of the IoT and big data under the policy enables precision farming and better alignment of production with market demand, mitigating supply–demand fluctuations and price risks and thereby stabilizing operational returns. This ensures more predictable labor income expectations for agricultural workers [18,19]. The development of digital inclusive finance also provides smallholders with more accessible financing channels, alleviating liquidity constraints in agricultural operations, fostering long-term business confidence, and reducing speculative employment choices. For rural laborers engaged in off-farm work, the establishment of online data platforms increases transparency regarding job requirements, wage standards, and firm credibility, reducing information asymmetries that drive short-term, speculative employment behaviors, and thereby strengthening the RLFS between laborers and their current positions.

Finally, on the labor demand side, the implementation of a digital rural development policy can enhance RLFS by expanding the scale of labor demand. By continuously improving rural digital infrastructure, fostering digital industrial clusters, and promoting the digital transformation of traditional agricultural enterprises, the policy substantially broadens the development space for rural firms. As rural industry chains are strengthened and new industries and business models emerge, the labor demand structure shifts from predominantly manual work to a diversified mix of manual and technical tasks, increasing overall labor demand. This expansion not only generates sufficient local employment opportunities for rural laborers but also provides a stable job supply. Moreover, the availability of diverse positions allows rural laborers to match roles to their capabilities, reducing labor interruptions caused by poor job fit and thereby enhancing RLFS. Consequently, we propose the following hypothesis:

H2. 

The implementation of a digital rural development policy influences RLFS through five channels: enhancing social network integration, improving laborers’ digital literacy, increasing health status, reducing speculative labor motives, and expanding the scale of enterprise labor demand.

2. Literature Review

The existing literature indicates that digital technologies and the digital economy have multidimensional and complex effects on labor markets. With respect to employment opportunities, some studies have reported that participation in the digital economy can significantly increase protective farming behaviors among rural households, thereby contributing to labor retention in the agricultural sector [20]. The impacts of digital skills vary. Some scholars have argued that digital skills, combined with the development of telecommunications infrastructure, primarily promote nonagricultural employment and entrepreneurial activities, with limited effects on agricultural employment [21,22]. Conversely, other studies have suggested that digital skills can simultaneously increase employment opportunities in both agricultural and nonagricultural sectors while reducing informal employment and enhancing overall employment security [23]. In terms of employment quality, rural digital infrastructure development also significantly improves the quality of nonagricultural employment [24].

In terms of labor mobility and stability, digital technologies have a dual effect. On the one hand, access to digital technologies significantly facilitates rural labor mobility [25]; however, the technologies themselves may introduce governance challenges. For instance, a comparative study in the German agricultural sector indicated that digital technologies can be used to intensify labor supervision and control, which may undermine worker autonomy and reduce work motivation [26]. Moreover, improvements in farmers’ digital literacy also drive employment transitions, manifested in increased local part-time work and extended periods of migrant labor [27]. On the other hand, digital technologies strengthen intravillage social networks and a sense of belonging, helping to mitigate labor outflow and promote collective action [28]. The development of inclusive digital finance further reinforces the promotion of nonagricultural employment [29,30].

In terms of the employment structure, the development of agricultural digital technologies has driven a profound transformation. This effect manifests as a coexistence of “substitution” and “creation”, with smart agriculture replacing certain low-skilled laborers while new business models such as e-commerce live streaming and online training generate additional employment opportunities [31]. Overall, the advancement of rural digital technologies facilitates the shift in rural labor from production-oriented positions to service-oriented and technical roles [32]. However, this process is accompanied by employment polarization and risk, as low-skilled workers may face marginalization [33], and the uncertainties associated with the development of rural e-commerce platforms introduce new risks for laborers [34].

Another key concept relevant to this study is labor flexibility, defined as the ability to reallocate labor across different sectors [35]. In contrast, labor stickiness, as a core measure of employment stability, embodies a multilayered connotation in the literature. At the firm level, labor stickiness is considered an organizational characteristic significantly associated with firm survival and growth probabilities [36,37]. At the individual employment level, labor stickiness is directly reflected in employment stability. Factors affecting this stability span both the micro and macro dimensions: at the micro level, household migration satisfaction and individual human capital have been identified as critical determinants [38]; from a macro perspective, labor migration patterns and the level of the rural credit supply significantly influence employment stability in different contexts [39,40].

In summary, the existing studies have examined the impact of digital technology or digital economy penetration into rural areas on the labor market from multiple perspectives. However, several gaps remain. First, most studies have focused on the effects of the digital economy on labor “mobility” or “employment choices”, neglecting its systematic impact on “employment stability”. At the policy level, the intrinsic relationship between digital rural development policies and labor stickiness has not been adequately analyzed theoretically or empirically. Most existing studies treat identity-based and health-protection policies, such as hukou system reforms or social insurance reforms, as quasi-natural experiments to examine their effects on rural labor mobility. For example, Ning (2016) and Shen et al. (2017) find that rural insurance system reforms increase rural labor working hours and slow labor market exit [41,42], while hukou system reforms promote rural labor out-migration but reduce return migration to rural areas [43]. Other studies take the implementation of digital rural development policy as a quasi-natural experiment to investigate its economic effects on rural household income and the development of new agricultural business entities; for instance, Wang et al. (2026) show that the digital rural strategy significantly narrows the urban–rural income gap by increasing complementary resource allocation [44], and digital rural development policy strengthens the cultivation of new agricultural business entities, thereby promoting overall agricultural economic development [45,46]. Overall, existing policy evaluation studies have not sufficiently explored the economic linkages between digital rural development policy and the rural labor market. Second, although some studies have investigated the stabilizing effects of digital technology, a systematic examination of the mechanisms through which digital rural development policies influence RLFS is lacking. In particular, previous research has not integrated multiple dimensions, such as social networks, health security, occupational expectations, and industry demand, into a comprehensive analytical framework. Third, discussions on the heterogeneity of policy effects remain insufficient, with limited identification of which groups or regions benefit most from digital rural development policies, constraining targeted policy implementation. Finally, the existing studies have largely not analyzed labor allocation effects and have failed to assess whether enhanced labor stickiness translates into improved efficiency in the use of core rural resources, such as land, thereby limiting the comprehensive evaluation of policy effectiveness.

On this basis, the present study aims to systematically examine the effects of a digital rural development policy on the allocation of rural labor resources from the core dimension of RLFS. The potential marginal contributions of this study are as follows: First, it shifts the research focus from the broad concept of “employment effects” to the deeper dimension of “employment stickiness”, offering a novel perspective for evaluating the economic impact of digital rural development policies; second, it constructs and tests the microlevel mechanisms through which digital rural development policies enhance RLFS across five dimensions: social network embedding, labor skill improvement, health security enhancement, speculation suppression, and labor demand stabilization; third, it provides an in-depth analysis of policy heterogeneity across groups with different employment choices and educational levels, as well as regions with varying population and land resource conditions, offering empirical evidence for the formulation of targeted and differentiated digital rural development policies; and fourth, by moving beyond the limitations of the existing research, it further investigates the impact of digital rural development policies on land use efficiency, extending the analytical perspective from “employment stability” to “resource productivity”, thereby providing a deeper theoretical basis for optimizing digital rural policy.

3. Materials and Methods

This study treats the issuance of the “Digital Rural Development Action Plan” by provincial governments as a quasinatural experiment to examine the impact of digital rural development policy implementation on the RLFS and its underlying mechanisms, aiming to better understand how to mitigate factor misallocation in the rural labor market and enhance the RLFS. Tracing the evolution of Digital Rural, the 2018 No. 1 Central Document in China first proposed the “digital rural development strategy”. To promote its nationwide rollout, the Ministry of Agriculture and Rural Affairs and other authorities subsequently issued guiding policies such as the Digital Rural Development Strategy Outline and the Digital Rural Construction Guide 1.0. Building on these directives, provincial governments formulated regional digital rural development plans and released Digital Rural Development Action Plans to advance policy implementation. In terms of core measures, the digital rural development policy encompasses upgrades to rural digital infrastructure, the digital transformation of agricultural production, the cultivation of new rural digital industries, and improvements in digital governance capacity. Specific initiatives include the supplementary construction of 4G base stations in rural areas, the integration and sharing of rural information service stations, the development of mobile internet applications, the establishment of foundational agricultural and rural databases and rapid data-processing platforms, and demonstration projects that integrate agricultural machinery with information technologies.

3.1. Data Source and Description

The data used in this study are drawn from the China Family Panel Studies (CFPS) conducted by the Institute of Social Science Survey (ISSS) at Peking University across 25 provinces, municipalities, and autonomous regions in China. The CFPS employs a multistage probability sampling method and encompasses both urban and rural residents, as well as individual- and household-level information, forming a large-scale public survey dataset. The baseline for sampling and survey implementation was 2010, with a biennial survey interval. To date, the dataset comprises seven waves: 2010, 2012, 2014, 2016, 2018, 2020, and 2022. Since the 2012 and earlier waves did not provide temporally ordered information on the most recent labor start and end dates, this study ultimately utilizes an unbalanced panel dataset covering five waves from 2014 to 2022.

Given that this study focuses primarily on rural labor, the data were preprocessed as follows. First, individual-level survey data were used to extract information on the respondent’s province, household registration type, starting year and month of the most recent main job, and ending year and month of the most recent main job, as well as other individual work and characteristic information. Second, household-level survey data were used to obtain information on household size, total household loans, and other family characteristics, which were then matched with individual data on the basis of household and individual identifiers. Third, rural labor samples were identified by restricting individuals to ages 16–65 and to those not currently in school [47] and by selecting respondents whose household registration was agricultural or whose surveyed work location was rural; individuals meeting either criterion were considered valid rural labor samples. In addition to individual labor, personal characteristics, and household characteristics, data were also collected from other sources. The regional economic and industrial characteristic data were obtained from provincial statistical yearbooks, the National Bureau of Statistics online database, and the Anjuke (https://www.anjuke.com/, accessed on 14 January 2025) website. Information on the implementation timing of digital rural development policies and competing policies was collected from provincial government websites. The data related to provincial labor turnover rates were obtained from the Zhilian Recruitment (https://www.zhaopin.com, accessed on 20 August 2025) website. To address missing or anomalous values, missing data were filled via linear interpolation or annual averages, and outliers were winsorized at the 1% level on both ends. The final dataset comprises 13,067 rural individuals over an eight-year span, forming an unbalanced panel with 25,841 observations.

3.2. Variable Selection

3.2.1. Explained Variable: Rural Labor Force Stickiness (RLFS)

RLFS measures the stability of an individual’s employment in the same job. The longer an individual continuously engages in the same job, the greater their RLFS. On the basis of this concept, this study calculates RLFS using the start and end dates of the individual’s most recent main job as recorded in the CFPS individual survey questionnaire. The specific calculation method is as follows:

$$RLFS=\ln [(endyear-startyear)\times 12+(endmonth-startmonth)+1]$$

(1)

$startyear$ and $endyear$ denote the start and end years, respectively, of the individual’s most recent main job as recorded in the survey year, while $startmonth$ and $endmonth$ indicate the corresponding start and end months, respectively. Additionally, the minimum value for job duration is set to 1.

3.2.2. Explanatory Variable: Digital Rural Development Policy (DID)

The issuance of provincial Digital Rural Development Action Plans or Digital Agriculture and Rural Construction Plans provides an ideal quasinatural experimental setting to examine the impact of digital rural development policy implementation on RLFS. Specifically, an individual is considered affected by the policy if their region had issued either of the above plans in that year, with the explanatory variable coded as 1; otherwise, it is coded as 0. Additionally, we collected the month of policy issuance for each province. In subsequent robustness checks, when the survey year coincides with the policy year, the identification is refined by comparing the survey month with the policy issuance month, allowing for a more accurate determination of whether the individual was exposed to the digital rural development policy.

3.2.3. Control Variables

To ensure that other individual characteristics, family personal characteristics and regional conditions did not affect the research findings, and to avoid potential endogeneity arising from omitted variables correlated with the explanatory variables, thereby ensuring that the estimated coefficients of the explanatory variables are accurately identified. in accordance with the existing research, we selected some potential confounding factors that may affect RLFS at the individual, family and geographical levels.

First, at the individual level, the following personal characteristic control variables were included: individual age (AGE) and the square of the individual’s actual age in the survey year (SAGE), represented by the sample’s actual age and its squared term [48]. Given that individuals meet the basic labor age requirements, the opportunity cost associated with job changes first decreases but then increases with age, thereby affecting actual labor stability. The inclusion of age and age squared as control variables helps isolate factors that may directly affect RLFS. Marital status (MTS), recorded as the individual’s marital status in the survey year [49], takes five integer values ranging from 1 to 5, representing unmarried, married, cohabiting, divorced, and widowed, respectively. Marital status significantly influences employment stability among rural residents, thereby constraining labor mobility in terms of both frequency and scope. Controlling for marital status allows a clearer identification of the policy’s pure effect. Level of education (LOE), measured as the highest education attained by the individual in the survey year, ranges from 0 (illiterate or semiliterate) to 9 (doctoral degree). Educational attainment, representing human capital accumulation, plays a crucial role in employment and labor stability. For rural residents whose primary income derives from wages, higher education improves labor market competitiveness and increases the likelihood of stable employment. Conversely, for households that rely mainly on agricultural income, lower education reduces the opportunity cost of continuing agricultural work and decreases the willingness to leave rural labor, thereby prolonging sustained agricultural employment. Controlling for education level contributes to the accurate estimation of the effect of the digital rural development policy. Family affairs (FAF) is a binary variable indicating whether the individual frequently performs household chores or cares for elderly family members during the survey year; it takes a value of 1 if the individual engages in such activities and 0 otherwise. Time devoted to nonproductive activities such as household chores or elder care limits individuals’ flexibility in their employment decisions, reducing the range of potential employment options. Pressure of parenting children (PPC) is measured as the number of children under 16 that the individual has in the survey year. The existing research has shown that child-rearing significantly hinders labor market participation. In particular, for rural women, greater childcare pressure reduces labor market participation and increases the probability of employment interruption.

Second, at the family level, the following control variables were included: Families receiving government subsidies (FRG), a binary variable indicating whether the household received any form of government subsidy during the survey year; it takes a value of 1 if the household received subsidies and 0 otherwise. Labor income and government subsidies interact complementarily in rural household income generation, influencing the subjective willingness to interrupt employment. Family scale (FCA), measured as the logarithm of total household members, reflects household size, which is closely related to labor endowment conditions. Outstanding loans to households (OLH), measured by the total outstanding household debt recorded in the family survey, captures the debt burden faced by the individual’s household. A higher debt burden induces household members to increase labor supply and reduce employment interruptions, thereby enhancing subjective labor stability.

Third, we considered regional economic variables as follows: regional economic development (RED), measured by the logarithm of per capita regional GDP, reflecting the local economic condition and its impact on RLFS; industry structure (RIS), measured by the ratio of tertiary industry value added to secondary industry value added [50], capturing the effect of industrial structure on rural labor demand; and level of social insurance (LSI), measured as the logarithm of the actual number of beneficiaries of urban and rural social pension insurance, representing the degree of social protection. The existing studies have indicated that insufficient social security in rural areas compels laborers to allocate additional working hours or seek higher wages to mitigate personal survival risks arising from social protection gaps. Level of housing price (LHP) is measured as the logarithm of the annual average regional housing transaction price, with prices deflated to 2013 using the consumer price index. Housing prices are typically negatively correlated with labor participation and stability; when prices are excessively high, both labor participation and employment stability tend to decline.

3.2.4. Mechanistic Variables

To examine the primary channels through which the digital rural development policy affects RLFS, this study constructs five mechanism variables as follows:

Degree of social network embedding (DSNE). The closeness of laborers to local social networks influences their long-term employment willingness. We use the CFPS individual survey item “trust in strangers” to represent the degree of social network embedding [51]. This variable is an integer ranging from 0 to 10, with higher values indicating greater trust in local strangers and, thus, greater embedding in local social networks. It is worth noting that this variable is inherently subjective and may generate upward or downward bias in the estimation results due to potential discrepancies between individuals’ subjective responses and objective facts. Given that respondents in the CFPS are randomly sampled, we argue that the expected value of these two types of bias approximately converges to zero.

Digital literacy (IDT). Digital literacy, an important component of human capital, affects employment outcomes. We measure this using the CFPS item “importance of the internet for work”, an integer ranging from 1 to 5, with higher values indicating that the individual perceives the internet as more important for work, reflecting higher subjective digital literacy in their employment.

Labor health status (LHS). Health status is fundamental for maintaining stable long-term employment. We use the CFPS survey item “health condition”, an integer ranging from 1 to 5, with higher values representing better health.

Labor speculative motivation (LSM). Speculative motivation during labor supply is a key factor influencing long-term employment willingness. This is measured by the individual’s response to the statement “hard work will be rewarded”, with values of 1–4 indicating strongly disagree, disagree, agree, and strongly agree, respectively. Due to limitations in data availability, the measurement of labor speculative motivation in this study is subject to certain constraints. Individuals who report a high level of agreement with the statement that “hard work is rewarded” may still exhibit speculative behavior in actual labor supply, which implies that the estimated effects of the digital rural development policy may be subject to some degree of upward bias.

Labor demand scale (LDS). From the labor demand perspective, larger firms are more likely to provide stable employment. We measure this by the logarithm of the number of employees in the laborer’s employing unit.

Table 1. Descriptive statistics.

	(1)	(2)	(3)	(4)	(5)
Variable	Obs	Mean	Sd	Min	Max
RLFS	25,841	2.977	1.537	0.000	7.299
DID	25,841	0.218	0.413	0.000	1.000
AGE	25,841	38.993	13.911	14.000	93.000
SAGE	25,841	1713.983	1218.938	196.000	8649.000
MTS	25,841	1.945	0.730	0.000	5.000
LOE	25,841	7.924	5.468	0.000	8.000
FAF	25,841	0.338	0.473	0.000	1.000
PPC	25,841	0.658	1.043	0.000	10.000
FRG	25,841	0.351	0.477	0.000	1.000
FCA	25,841	4.343	2.116	1.000	21.000
OLH	25,841	1.075	8.832	0.000	500.000
RED	25,841	10.858	0.387	10.150	12.049
RIS	25,841	1.275	0.461	0.704	5.244
LSI	25,841	6.389	0.736	1.281	7.393
LHP	25,841	8.884	0.463	0.000	10.889

presents descriptive statistics for the main variables in this research.

3.3. Model Design

3.3.1. Benchmark Model

To investigate the impact of the digital rural development policy on rural labor force stickiness, we constructed a staggered difference-in-differences design as follows:

$$RLF{S}_{\mathrm{it}}={\alpha }_0+{\alpha }_{1}DIDit+{\alpha }_{2}Contro{l}_{it}+{\gamma }_i+{\mu }_t+{\epsilon }_{it}$$

(2)

In Equation (2), the explained variable $RLF{S}_{it}$ represents the rural labor force stickiness of individual $i$ in year $t$. $DI{D}_{it}$ indicates whether individual $i$ in year $t$ is affected by the digital rural development policy. $Contro{l}_{it}$ denotes a set of multidimensional control variables, including individual-level, household-level, and regional-level characteristics. ${\gamma }_i$ and ${\mu }_t$ represent individual fixed effects and time fixed effects, respectively; ${ϵ}_{it}$ is the random disturbance term; and ${\alpha }_1$ is the core coefficient of interest and captures the impact of the digital rural development policy on $RLFS$. If ${\alpha }_1>0$, the digital rural development policy clearly promoted rural labor force stickiness.

3.3.2. Mechanism Analysis Model

To intuitively identify the main pathways through which the digital rural development policy affects RLFS, this study incorporates the numerical characteristics of the mechanism variables by combining a two-step mechanism test with an ordered logit model. The mechanism analysis model is specified as follows:

$$Channe{l}_{it}={\beta }_{0}DI{D}_{it}+{\beta }_{1}Contro{l}_{it}+{\lambda }_i+{\mu }_t+{\epsilon }_{it}$$

(3)

In Equation (3), $Channe{l}_{it}$ denotes the mechanism variables, including social network embeddedness, digital literacy, health status, labor demand scale, and speculative labor motivation. ${\beta }_1$ is the core coefficient of interest, and ${\beta }_0$ is the constant term. Given that most mechanism variables are ordinal categorical variables, an ordered logit model is employed to estimate Equation (3).

4. Results

4.1. Benchmark Regression

Table 2. Benchmark regression results.

	(1)	(2)	(3)	(4)
Variable	RLFS	RLFS	RLFS	RLFS
DID	0.143 ***	0.135 ***	0.136 ***	0.133 ***
	(0.041)	(0.041)	(0.042)	(0.042)
AGE		0.115 ***	0.115 ***	0.115 ***
		(0.021)	(0.021)	(0.021)
SAGE		−0.001 ***	−0.001 ***	−0.001 ***
		(0.000)	(0.000)	(0.000)
MTS		0.049 *	0.048 *	0.045 *
		(0.026)	(0.026)	(0.026)
LOE		−0.021 ***	−0.021 ***	−0.021 ***
		(0.004)	(0.004)	(0.004)
FAF		0.041	0.040	0.043
		(0.040)	(0.040)	(0.040)
PPC		−0.087 ***	−0.089 ***	−0.087 ***
		(0.020)	(0.020)	(0.020)
FRG			0.013	0.013
			(0.031)	(0.031)
FCA			−0.016 *	−0.016 *
			(0.008)	(0.008)
OLH			−0.000	−0.000
			(0.000)	(0.000)
RED				−0.146
				(0.089)
RIS				0.070
				(0.050)
LSI				0.003
				(0.044)
LHP				0.049
				(0.054)
Individual Fe	YES	YES	YES	YES
Year Fe	YES	YES	YES	YES
Cons	2.945 ***	−0.280	−0.213	0.825
	(0.012)	(0.744)	(0.745)	(1.167)
N	25,841	25,841	25,841	25,841
R2	0.575	0.577	0.577	0.578

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

reports the benchmark regression results for the impact of the digital rural development policy on RLFS. Columns (1)–(4) present the results obtained by gradually adding individual-level, family-level, and regional-level control variables. Across all the model specifications, the estimated coefficients of the core explanatory variable DID remain positive and statistically significant at the 1% level. This pattern indicates that the implementation of the digital rural development policy significantly extends the duration of individuals’ main jobs and enhances RLFS. Taking Column (4), which simultaneously incorporates individual, family, and regional controls, as an example, the average marginal effect of the digital rural development policy on RLFS is 0.133, implying that the policy increases RLFS by 13.3%. These findings provide preliminary support for Hypothesis 1.

4.2. Robustness Tests

4.2.1. Parallel Trend Test

The validity of identifying the policy’s marginal effects using the difference-in-differences approach relies on the premise that, for the specific research objective, the treatment and control groups do not exhibit systematic differences prior to the policy shock. If the two groups fail to share similar prepolicy trends, the observed increase in RLFS after the policy may be driven by inherent group differences rather than by the policy intervention itself, thereby undermining the credibility of the baseline results. To address this concern, this study uses the three periods prior to policy implementation as the reference window and the event study method to perform a parallel trend test [52], with the results presented in Figure 1. The estimates show that before the implementation of the digital rural development policy, there is no significant difference in RLFS between the treatment and control groups; all prepolicy coefficients are statistically insignificant. Thus, the parallel trend assumption is supported at the 95% confidence level, confirming the validity of the identification strategy. Moreover, the dynamic effects indicate that the impact of the digital rural development policy on RLFS continues to expand over time, suggesting that the policy has a persistent and long-term influence on rural labor.

4.2.2. Eliminating the Interference of Sample-Specific Characteristics on the Estimation Results

The sample used in this study may be influenced by regional, individual, and temporal characteristics, which could interfere with the baseline regression results. First, regarding regional characteristics, labor mobility across different positions varies significantly across provinces, with some provinces exhibiting relatively high employment stability and faster responses to external shocks. Ignoring these regional differences could lead to an overestimation of the effect of digital rural development policy implementation. To address this, we first identified the top 30 cities with the highest job-hopping rates on the basis of the “White-Collar Job-Hopping Index Report” published by China’s major recruitment platform Zhilian and defined provinces whose cities were all outside this range as having relatively stable labor markets. We then excluded these provinces and reran the regressions, with the results reported in Column (1) of

Table 3. Robustness test with respect to inherent sample heterogeneity and special subsamples.

	(1)	(2)	(3)	(4)
Variable	RLFS	RLFS	RLFS	RLFS
DID	0.126 **	0.133 **	0.150 ***	0.183 ***
	(0.055)	(0.053)	(0.043)	(0.065)
Individual Control	YES	YES	YES	YES
Family Control	YES	YES	YES	YES
Regional Control	YES	YES	YES	YES
Individual Fe	YES	YES	YES	YES
Year Fe	YES	YES	YES	YES
Cons	2.547	0.825	0.708	2.330
	(1.872)	(1.746)	(1.153)	(1.422)
N	19,917	25,841	24,454	19,373
R2	0.579	0.578	0.570	0.591

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

. Second, with respect to individual characteristics, specific health conditions or employment stages may bias the policy effect estimation, and individual employment behavior often violates the homoscedasticity assumption. To mitigate this, (i) we employed robust standard errors clustered at the provincial level and (ii) excluded retired individuals whose preretirement employment changes are generally infrequent. The results controlling for individual characteristics are reported in Columns (2) and (3) of Table 3. Third, regarding temporal characteristics, the outbreak of COVID-19 distorted the normal allocation of labor. To address this, we excluded observations for 2020, with the results reported in Column (4) of Table 3 [53]. Overall, after separately accounting for regional, individual, and temporal characteristics, the DID coefficient remains significantly positive at the 5% level or higher, confirming the robustness of the baseline results.

4.2.3. Variable Substitution and Alternative Research Methods

Using a single-variable measurement method and research model may expose the baseline regression results to incidental risks inherent to the chosen approach. Therefore, we reconstructed both the explanatory and dependent variables and re-estimated Equation (1) using a double machine learning model. Specifically, RLFS was remeasured using the duration of the individual’s previous main job in months and years, with the results reported in Columns (1) and (2) of

Table 4. Results of variable substitution and alternative research methods.

	(1)	(2)	(3)	(4)	(5)
Variable	RLFSO	RLFSY	RLFS	RLFS	RLFS
DID	9.243 ***	0.076 **	0.098 **	0.112 ***	0.139 ***
	(2.869)	(0.030)	(0.041)	(0.022)	(0.022)
Individual Control	YES	YES	YES	YES	YES
Family Control	YES	YES	YES	YES	YES
Regional Control	YES	YES	YES	YES	YES
Individual Fe	YES	YES	YES	YES	YES
Year Fe	YES	YES	YES	YES	YES
Cons	93.152	−0.200	0.556	0.019 **	0.027 ***
	(79.441)	(0.827)	(1.166)	(0.010)	(0.010)
N	25,841	25,841	25,841	25,841	25,841
R2	0.600	0.599	0.578	-	-

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

. The DID variable was reconstructed on the basis of whether the individual’s region had issued the Digital Rural Development Action Plan during the month and year of the survey, with the results shown in Column (3) of Table 4. Furthermore, using 3:2 and 5:4 splits for the training and validation sets, the double machine learning model was applied to re-estimate Equation (1), with the results reported in Columns (4) and (5), respectively. Across all specifications, the baseline regression results remain substantively unchanged.

4.2.4. Counterfactual Estimation and Heterogeneous Treatment Effect

To further verify the robustness of our findings, we conducted a counterfactual test. Specifically, we removed the postimplementation samples of the digital rural development policy and, within the remaining sample period, advanced the policy implementation year by 2 or 3 years for provinces that had originally issued the Digital Rural Development Action Plan, constructing a “placebo digital rural development policy shock” (DID_pre2 or DID_pre3) as the key explanatory variable in the regression. If the regression coefficient of the placebo shock is significant, it would indicate that unobserved shocks could also cause significant changes in RLFS between the treatment and control groups, suggesting potential interference with the policy effect estimation. Conversely, if the coefficient is not significant, it supports the robustness of the baseline results. Columns (1) and (2) of

Table 5. Results of counterfactual estimation and heterogeneous treatment effects.

	(1)	(2)	(3)	(4)
Variable	RLFS	RLFS	RLFS	RLFS
DID			0.178 ***	0.144 ***
			(0.041)	(0.044)
DID_pre2	0.055
	(0.070)
DID_pre3		0.038
		(0.071)
Individual Control	YES	YES	YES	YES
Family Control	YES	YES	YES	YES
Regional Control	YES	YES	YES	YES
Individual Fe	YES	YES	YES	YES
Year Fe	YES	YES	YES	YES
Exclusive policy	No	No	No	YES
Cons	−1.310	−1.309	−2.780	0.746
	(2.787)	(2.790)	(2.191)	(1.278)
N	10,646	10,646	25,841	25,841
R2	0.576	0.576	0.586	0.578

Standard errors are in parentheses. *** p < 0.01.

show that the coefficients for DID_pre2 and DID_pre3 are not significant, indicating that the main conclusions of this study are not driven by other prepolicy shocks.

Simultaneously, considering the potential “negative weight” problem in multiperiod difference-in-differences (DID) due to heterogeneous treatment effects, we followed the approach of Callaway and Sant’Anna (2021) [54] to recalculate the average treatment effect for each period, aggregating them using the proportion of each group’s sample size as weights to obtain a CSDID estimator that accounts for heterogeneous treatment effects. The CSDID estimates reported in Column (3) of Table 5 indicate that the coefficient of the core explanatory variable remains significant at the 1% level, confirming that the baseline conclusions of this study remain robust even after accounting for heterogeneous treatment effects.

In addition, other policies implemented concurrently with the digital rural development policy may confound the baseline regression results, and ignoring these policies could overstate the actual effect of the digital rural development policy. Overall, the impact of the digital rural development policy on RLFS may be influenced by the following four concurrent policies:

• Carbon emission trading pilot programs. The green transformation driven by carbon emission trading pilots can alter the composition of rural migrant labor, potentially undermining labor protection and ultimately affecting employment stability [55]. To control for this effect, we construct a dummy variable “carbon emission trading pilot”, which takes a value of 1 if the individual’s region had established a pilot program in the survey year and 0 otherwise. Information on carbon emission trading pilots is sourced from the official website of the National Development and Reform Commission of China;
• Establishment of local credit information platforms. The establishment of provincial credit platforms can impose regulatory and binding effects on enterprise labor employment, influencing the actual allocation of labor [56]. We introduce a dummy variable “local credit platform”, coded 1 if the individual resides in a province with an established platform in the survey year and 0 otherwise. Information on local credit platforms is obtained from the official website of the People’s Bank of China;
• Construction of public data open platforms. Data released through public data open platforms can enhance labor market transparency and optimize employment choices for rural labor [57]. We manually collect the establishment years of public data open platforms in each province and construct a dummy variable, “public data open platform”, which takes a value of 1 if the sample is located in a province with a platform in the survey year and 0 otherwise;
• Telecommunications digital service pilot projects aimed at enhancing network coverage in rural areas may provide favorable conditions for the comprehensive development of digital rural initiatives by improving digital infrastructure. We construct a dummy variable for the telecommunications digital service pilot, coded 1 if the individual’s province participated in the pilot in the survey year and 0 otherwise. Information on the pilot projects is sourced from the official website of the Ministry of Industry and Information Technology of China.

Column (4) of Table 5 reports the regression results after simultaneously including the aforementioned policy variables. After accounting for the influence of other concurrent policies, the implementation of the digital rural development policy still significantly enhances RLFS, with the core explanatory variable remaining positive and significant at the 1% level, which is consistent with the baseline regression results.

4.2.5. Placebo Test

In the preceding analysis, we verified that the effect of the digital rural development policy is not driven by omitted prepolicy shocks or confounded by other concurrent policies. However, the baseline regression results may still be subject to disturbances from other unobserved or random factors. To obtain more reliable conclusions while preserving the data distribution, we conducted a placebo test. Specifically, we randomly assigned policy implementation timing and randomly selected policy regions and performed 500 iterations of resampling regressions. Figure 2 presents the kernel density plot of the coefficients of the core explanatory variable from these 500 resampled regressions. It is evident that the estimated coefficients are mostly concentrated around zero, indicating that unobserved random factors are not the primary driver of the significant differences in RLFS between the treatment and control groups.

4.2.6. Endogeneity Analysis

In the preceding analysis, we conducted various robustness checks from multiple perspectives. However, potential endogeneity may still exist because of reverse causality. For provinces characterized by more severe distortions in the rural labor market and lower RLFS, governments are more likely to adopt proactive digital strategies to improve rural employment conditions through the widespread application of digital technologies in rural areas. To address the issues above, we construct a 2SLS model using an instrumental variable (IV) that is exogenous and correlated with the explanatory variable. Following Wen et al. (2024), the interaction between the reciprocal of provincial terrain ruggedness and a time trend is selected as the instrument for the DID [58]. The theoretical premise of selecting the instrumental variables above is that: first, terrain ruggedness affects the difficulty of digital technology application; areas with greater variation in terrain face higher costs in building digital government infrastructure. Therefore, the instrument satisfies the relevance criterion by being correlated with the implementation of digital rural development policy. Second, as a geographic characteristic of a region, terrain ruggedness is largely exogenous and is unlikely to be directly related to economic variables, thus meeting the exclusion restriction required for a valid instrument. Results are reported in columns (1) and (2) of

Table 6. Results of endogeneity analysis.

	(1)	(2)
Variable	ONE STAGE	TWO STAGE
DID		0.214 ***
		(0.051)
IV	0.001 ***
	(0.000)
Individual Control	YES	YES
Family Control	YES	YES
Regional Control	YES	YES
Individual Fe	YES	YES
Year Fe	YES	YES
KP Wald F	287.13	—
Under-identification test p Value	0.000	—
N	25,841	25,841
R2	—	0.125

Standard errors are in parentheses. *** p < 0.01.

.

The first-stage results show that the coefficient of the instrumental variable is positive and significant at the 1% level, indicating a clear positive effect of terrain ruggedness on DID. The KP Wald F statistics are 287.13, and exceed the critical value of 10% for weak instruments. Indicating that weak instrument bias is unlikely to substantially affect the first-stage regression results. The second-stage regression shows that, after the instrumental variable is employed, the DID continues to significantly enhance RLFS at the 1% significance level, with an increased estimated coefficient. We also conduct tests for under-identification, providing additional evidence supporting the validity of the chosen instrumental variable.

4.3. Mechanism Analysis

The degree of rural labor integration into local social networks, digital literacy, labor health status, intensity of labor speculative motivation, and labor demand scale are key channels affecting RLFS. On the labor supply side, deeper employment network integration strengthens laborers’ sense of belonging, fostering long-term employment intentions. Concurrently, improvements in digital literacy and personal health enable workers to meet the skill and health requirements of their jobs, enhancing job-person fit and promoting employment stability. Moreover, reducing speculative motives in the labor supply is another crucial pathway for enhancing RLFS. From a traditional perspective, blind labor speculation due to information gaps or incomplete information increases mobility across jobs, thereby lowering employment stability. As information transparency improves, career paths within the same job become clearer, and information flows across different positions are smoother, constraining speculative labor behaviors. On the labor demand side, enterprises and farms constitute the core sources of labor demand. As their production scales expand and operational returns increase, the overall labor demand scale increases, and their capacity to provide long-term, stable employment platforms strengthens, objectively creating favorable external conditions for enhancing RLFS. On the basis of these considerations, this study further examines the primary channels through which the digital rural development policy exerts its effects, specifically whether policy implementation can promote RLFS by increasing the integration of rural labor into local employment networks, enhancing digital literacy and health, reducing speculative motives, and expanding labor demand. The regression results of the mechanism analysis are presented in

Table 7. Mechanism tests.

	(1)	(2)	(3)	(4)	(5)
Variable	DSNE	IDT	LHS	LSM	LDS
DID	0.177 ***	0.203 **	0.218 ***	0.188 ***	0.169 ***
	(0.067)	(0.097)	(0.071)	(0.089)	(0.066)
Individual Control	YES	YES	YES	YES	YES
Family Control	YES	YES	YES	YES	YES
Regional Control	YES	YES	YES	YES	YES
Individual Fe	YES	YES	YES	YES	YES
Year Fe	YES	YES	YES	YES	YES
N	19,267	15,895	17,603	21,337	20,599
Pseudo R2	0.014	0.540	0.012	0.835	0.087

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

.

Column (1) of Table 7 reports the regression results of digital rural development policy implementation on the integration of labor into local social networks. The coefficient of the core explanatory variable is positive and significant at the 1% level, indicating that policy implementation significantly increases the likelihood that respondents report higher levels of trust in local strangers. This finding suggests that the digital rural development policy enhances RLFS by facilitating the integration of rural laborers into local social networks. Columns (2) through (4) present the effects of policy implementation on individuals’ digital literacy, health status, and speculative labor motivation. In all these columns, the coefficients of the core explanatory variable are positive and significant at the 5% level or higher, demonstrating that the policy promotes RLFS by improving digital literacy and health while reducing speculative behavior in labor supply. Additionally, Column (5) shows that on the labor demand side, policy implementation significantly increases the labor demand scale, providing a stable external environment for rural employment and further enhancing RLFS. Collectively, the results above support Hypothesis H2.

4.4. Heterogeneity Analysis

From the perspective of the full sample, the aforementioned analysis preliminarily confirms that digital rural development policy implementation positively influences RLFS. To further clarify the actual effects of policy implementation on individuals under different conditions and to fully leverage its positive impact on RLFS, we further examine the heterogeneous effects of the digital rural development policy. Specifically, we delved into the heterogeneity of the policy’s impact on RLFS on the basis of individual characteristics, family characteristics, and regional characteristics.

4.4.1. Individual Heterogeneity

At the individual level, we primarily examine the heterogeneous effects of the digital rural development policy on individuals of different age groups, employment types, human capital levels, and engagement in agriculture-related work. First, young and middle-aged or older workers differ in their ability to absorb various digital rural initiatives. Young workers exhibit greater willingness and faster adaptation to digital technologies in rural areas, making the policy’s impact on their labor more pronounced. To assess the policy’s effect across age groups, we divide the sample into a youth group and a middle-aged and older group on the basis of whether the individual’s average age during the observation period exceeds 40 years and perform group regressions. The results are shown in Columns (1) and (2) of Appendix A 

Table A1. Individual heterogeneity results for age and work selection.

	(1)	(2)	(3)
Variable	RLFS	RLFS	RLFS
DID	0.144 ***	0.060	0.184 ***
	(0.046)	(0.090)	(0.045)
DID*LAE			−0.152 ***
			(0.049)
Individual Control	YES	YES	YES
Family Control	YES	YES	YES
Regional Control	YES	YES	YES
Individual Fe	YES	YES	YES
Year Fe	YES	YES	YES
N	17,648	8193	25,841
R2	0.542	0.562	0.578

Standard errors are in parentheses. *** p < 0.01.

. The results show that compared with middle-aged and older individuals, the implementation of the digital rural development policy more significantly enhances RLFS among rural youth, with the DID coefficient positive and significant at the 1% level.

Second, on the basis of the location of employment, we construct a “local employment” dummy variable (LAE) and employ a triple-difference approach to examine the heterogeneous effects of the digital rural development policy across different employment choices. Specifically, if an individual’s workplace is “the current village/community of residence” or “other villages/communities within the current township/town/subdistrict”, the dummy takes a value of 1, indicating local employment; if the workplace is “other townships within the residing county” or other provinces/cities or countries, the dummy takes a value of 0, indicating nonlocal employment. The triple-difference regression results are reported in Column (3) of Appendix A Table A1. The results show that, compared with locally employed individuals, the digital rural development policy has a more pronounced effect on RLFS for nonlocal workers, with the triple-difference coefficient significantly negative. This outcome arises because the policy not only broadens labor information channels for nonlocal workers, reducing the risk of labor market mismatch but also enhances labor protection through transparent digital supervision mechanisms, ultimately improving their RLFS.

Furthermore, the impact of the policy on individuals often varies according to the level of human capital represented by educational attainment [59]. Generally, workers with higher levels of education possess better information literacy and a stronger capacity to utilize modern digital technologies and are better able to benefit from the digital rural development policy, ultimately influencing their labor outcomes. Additionally, better-educated rural residents are more likely to recognize new rural models and emerging industries earlier, facilitating the faster integration of technological supply into individual decision-making and positioning them as policy beneficiaries. On this basis, we divide the sample into a “higher human capital group” and a “lower human capital group” according to whether the individual’s highest completed education level is a bachelor’s degree or above. Columns (1) and (2) of Appendix A 

Table A2. Individual heterogeneity results for education level and agricultural work.

	(1)	(2)	(3)	(4)
Variable	RLFS	RLFS	RLFS	RLFS
DID	−0.171	0.153 ***	0.224 **	0.117 **
	(0.153)	(0.045)	(0.104)	(0.049)
Individual Control	YES	YES	YES	YES
Family Control	YES	YES	YES	YES
Regional Control	YES	YES	YES	YES
Individual Fe	YES	YES	YES	YES
Year Fe	YES	YES	YES	YES
N	1024	23,485	6603	14,517
R2	0.597	0.572	0.590	0.564

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

report the subgroup regression results. The findings indicate that the digital rural development policy has a more pronounced effect on RLFS for the “lower human capital group”. It can be observed that the finding above is inconsistent with existing theoretical arguments. possible explanation is that, relative to workers with higher human capital, skill and educational requirements in the labor market impose higher employment barriers on rural workers with lower human capital. The agricultural digital transformation and the downward extension of rural digital services promoted by the digital rural development policy have generated employment positions with relatively low entry thresholds that are better suited to low-skilled workers, thereby more precisely matching their actual labor market needs and enhancing both their long-term employment intentions and job matching quality.

Finally, we examine the effect of the digital rural development policy on RLFS for individuals engaged in agricultural versus nonagricultural work. The policy has fostered new employment modes and positions in the agricultural sector, such as “agricultural product live streaming” and “drone operators”, which reduce the physical and time demands of traditional agricultural work. These developments provide rural residents with novel agricultural employment opportunities and increase their willingness to engage in agricultural work over the long term. On the basis of the CFPS survey question “whether engaged in agricultural work”, we divide the sample into “agricultural workers” and “nonagricultural workers”. The subgroup regression results are reported in Columns (3) and (4) of Appendix A Table A2. The underlying economic explanation is that the agricultural sector constitutes the core arena in the process of digital rural development, allowing policy measures to reach agricultural employers more rapidly. Following the implementation of the digital rural development policy in rural areas, labor development pathways within the agricultural sector are restructured; the emergence of new agricultural formats and models diversifies employment patterns and makes career progression paths more transparent, thereby accelerating the improvement of RLFS within the sector.

4.4.2. Location Heterogeneity

At the regional level, we further examine how variations in population dynamics and land resource endowments influence the effectiveness of the digital rural development policy. Clarifying these effects helps address how to better implement digital rural development and enhance its economic impact under China’s aging population and uneven land distribution. First, on the basis of the natural population growth rate of each province in the sample, we construct a dummy variable indicating whether a region has high population growth (HHNR). Specifically, if a province’s natural population growth rate exceeds the national average in the corresponding year, the dummy takes a value of 1; otherwise, it takes a value of 0. This high-population-growth dummy is interacted with the DID variable, and a triple-difference regression is conducted. The results are reported in Column (1) of Appendix A 

Table A3. Location heterogeneity results.

	(1)	(2)
Variable	RLFS	RLFS
DID	0.103 **	0.111 **
	(0.052)	(0.046)
$\mathrm{DID} \times$ HHNR	0.049
	(0.051)
$\mathrm{DID} \times$ HLR		0.063
		(0.053)
Individual Control	YES	YES
Family Control	YES	YES
Regional Control	YES	YES
Individual Fe	YES	YES
Year Fe	YES	YES
N	25,841	25,841
R2	0.578	0.578

Standard errors are in parentheses. ** p < 0.05.

. The results show that positive effect of the digital rural development policy on RLFS is concentrated mainly in regions with relatively low natural population growth, as the DID coefficient is significantly positive, while the triple-difference interaction term is not significant. Public service theory suggests that the natural population growth rate affects the pressure on government public service provision, with lower population growth generally implying relatively lower fiscal and service burdens. Accordingly, the above phenomenon can be economically interpreted as follows: in regions with lower natural population growth rates, reduced public service pressure enables governments to allocate more resources to the digital rural development policy, strengthens the complementarity between public services and digital rural initiatives, and thereby allows the effects of digital rural development on the labor market to be more fully realized.

Second, on the basis of per capita arable land and forestland areas for rural residents in each province, we construct a dummy variable for regions with relatively favorable land resource endowments (HLR). If a province’s per capita rural arable land and forestland exceeds the national average in that year, the dummy is assigned a value of 1; otherwise, it is assigned a value of 0. The triple-difference regression results are reported in Column (2) of Appendix A Table A3. The results show that the DID coefficient is significantly positive, whereas the triple-difference coefficient is positive but not significant, indicating that the policy’s effect on RLFS is concentrated mainly in regions with lower land resource endowments. This outcome arises because the new rural industries generated by digital rural development are not strictly dependent on land resource endowments. The integration of digital technology in regions with scarce land resources has created new positions, such as “smart agricultural technicians” and “leisure agriculture planners”, enhancing labor willingness and reinforcing RLFS.

4.5. Further Analysis: The Impact of Digital Rural Development Policy on LLUE

Currently, agricultural operations in rural China are still dominated by smallholder households. In this context, improving land use efficiency and increasing the real returns of land assets remain crucial for achieving agricultural modernization and sustainable rural development [60]. As noted earlier, the implementation of the digital rural development policy can significantly enhance the RLFS of rural agricultural laborers and young workers, stabilizing their willingness to remain in the same job positions. These findings suggest that the digital rural development policy can, to some extent, alleviate rural population hollowing and labor outflow in China. However, this study further seeks to examine how the digital rural development policy affects land use efficiency. Specifically, it investigates whether the influx of young labor into agricultural and rural areas induced by policy implementation can ultimately translate into improvements in land use efficiency. To test this hypothesis, we use household questionnaire data from the CFPS dataset and measure household land resource utilization efficiency (LLUE) in two ways: the ratio of total household operational and property income to land asset value and the ratio of total household operational and property income to household-owned land area. The regression results are reported in Columns (1) and (2) of

Table 8. Results of the impact of DID on LLUE.

	(1)	(2)
Variable	LLUE	LLUE
DID	−0.007	−1.702
	(0.015)	(1.386)
Family Control	YES	YES
Regional Control	YES	YES
Individual Fe	YES	YES
Year Fe	YES	YES
N	12,549	12,549
R2	0.567	0.533

.

The results show that the estimated coefficient of the DID term is negative and not statistically significant. From a deeper perspective, the above phenomenon may be explained by two main factors. On the one hand, the increase in RLFS induced by the digital rural development policy primarily reflects the matching of labor with emerging digital agriculture–related positions whose job content does not fully concentrate on land-based production activities, implying that enhanced RLFS cannot be directly transformed into greater land-use vitality or further improvements in land-use efficiency. On the other hand, policy measures are mainly tilted toward digital infrastructure construction, the cultivation of rural digital industries, and digital skills training, resulting in insufficient digital empowerment of land factors during policy implementation; inadequate support for land-related digital technologies—such as IoT-based agricultural monitoring, intelligent agricultural machinery, and climate big-data early-warning platforms—constrains the ability of rising RLFS to translate into higher land productivity through deeper integration with digital technologies. Therefore, the finding indicates that the labor influx induced by the implementation of the digital rural development policy has not led to improvements in land resource utilization efficiency in rural areas. The current digital rural development policy does not effectively increase the economic value of land resources in rural China, suggesting that related policy measures still require further refinement.

5. Discussion

The rapid development of digital technologies provides new avenues for alleviating labor market distortions in rural areas and addressing rural labor hollowing; therefore, it is essential to rigorously assess the positive impacts of digital rural development policy, which promotes the deep integration of digital technologies with agriculture and rural development, on rural labor markets, and to examine how such policies affect RLFS. This study attempts to reveal the relationship between digital rural development policy implementation and RLFS from both direct effects and indirect effects. Compared with the existing literature, our findings extend and complement studies on digital rural development and rural labor mobility. Prior research generally argues that digital rural development accelerates the transfer of surplus rural labor and promotes labor mobility [27,61,62]. This study further demonstrates that digital rural development not only lengthens the duration of continuous employment for surplus rural labor—thereby strengthening RLFS—but also enhances the RLFS of out-migrants, namely rural workers engaged in off-farm employment, ultimately improving overall conditions in the rural labor market. From the perspective of policy evaluation, we also advance existing approaches to assessing digital rural development policy implementation: whereas prior studies typically proxy digital rural development by whether a region is selected as a digital rural pilot or a “Broadband China” pilot, or by the provision of rural digital services [63,64,65], this study relies on whether governments have formally issued concrete digital rural development plans, thereby more accurately capturing the actual effects of policy implementation.

However, several limitations remain. First, the measurement of RLFS is relatively unidimensional, and the study does not capture its dynamic aspects. Specifically, at the static level, RLFS is defined as the stability of a worker in the same position, i.e., the duration of continuous employment in the same job. At the dynamic level, RLFS should also reflect whether workers can maintain stability when facing unexpected shocks or whether, after temporary employment interruptions, they have the willingness and capacity to quickly restore their employment to the previous state. This dynamic dimension is not considered in the current study. Second, although the study revealed a significant increase in RLFS in regions implementing digital rural development policies, policy implementation did not further improve rural land resource utilization efficiency. Building on this finding, the study does not conduct deeper analyses to identify the key reasons why the digital rural development policy fails to enhance land use efficiency, nor does it propose targeted policy adjustments. Third, the study finds that the digital rural development policy has a more pronounced effect on RLFS for less-educated laborers. Existing research generally suggests that highly educated individuals are better able to adopt digital tools and leverage digital technology effectively. The present findings contradict this conventional view, yet the study does not offer a satisfactory comprehensive theoretical and empirical explanation for this phenomenon.

These limitations offer insights and directions for future research: on the one hand, future studies may more precisely assess the economic effects of the digital rural development policy by identifying whether concrete policy measures have been effectively implemented at the village level, such as upgrading local digital infrastructure and strengthening residents’ digital technology adoption; on the other hand, focusing on labor skills and productivity, further research should systematically examine why the current digital rural development policy has failed to further improve land-use efficiency in rural areas, thereby informing policy refinement and promoting deeper integration of digital technologies with agriculture and rural development.

6. Conclusions

Under the proactive implementation of the digital rural development strategy, the ongoing digitalization process has exerted a notable influence on industrial transformation and upgrading in rural China, as well as on adjustments in labor employment behavior. This study examines, from a policy support perspective, whether the introduction of digital rural development action plans—represented as the digital rural development policy—can effectively enhance RLFS. Specifically, we first theoretically elucidate both the direct and indirect effects of digital rural development policy implementation on RLFS. On this basis, using individual survey data from the 2014–2022 CFPS database, we employ a multiperiod difference-in-differences model to empirically examine the impact and mechanisms through which the digital rural development policy affects RLFS.

This study reveals the following: First, the implementation of the digital rural development policy can effectively enhance RLFS. Second, the digital rural development policy affects RLFS through five channels: increasing labor embeddedness in local social networks, enhancing digital literacy, improving physical health, reducing speculative motives, and expanding labor demand. Third, the policy has a stronger positive effect on RLFS among young and middle-aged individuals, laborers with lower human capital, and those engaged in agricultural work; moreover, its impact is more pronounced in regions with declining demographic dividends and poorer land resource endowments. Fourth, although the policy increased RLFS among young rural laborers and agricultural workers, it did not lead to a significant improvement in rural land resource utilization efficiency.

On the basis of the findings mentioned above, we present specific policy recommendations as follows:

First, governments at all levels should actively implement the Digital Rural Development Action Plan to clarify the positive impact of the digital rural development policy on rural labor employment behavior. On the one hand, the scope of the Action Plan should be further expanded, promoting the deep integration of digital technologies with rural industries, culture, and governance and thereby fully leveraging the spillover effects of digital technologies. On the other hand, the relevant measures and requirements of the Action Plan should be further refined. In advancing targeted digital rural development, supervision and evaluation mechanisms should be established to track the progress and status of digital rural development in real time, providing practical evidence and data for dynamically adjusting the policy direction. Second, considering the heterogeneity among rural residents and regional differences in resources and population, digital rural policies should be implemented in a categorized, multistage, and targeted manner to ensure context-specific application. Specifically, efforts should be made to actively promote and disseminate the policy, enhancing young and middle-aged rural laborers’ and agricultural workers’ understanding of the policy measures. In regions with relatively scarce land resources and declining demographic dividends, the stabilizing effect of the digital rural development policy on local labor should be fully utilized. During policy implementation, the phased outcomes can be compiled into best-practice cases or manuals, offering official guidance for laborers on how to better absorb policy benefits and align with policy requirements, thereby maximizing the economic value of the digital rural development policy. Third, current policy measures should be actively adjusted to promote the deep integration of digital technologies with rural land resources and identify the key pathways for improving land resource utilization efficiency. Against the backdrop of digitalization, the new economic value of rural land resources should be cultivated. Although the existing digital rural development policy has partially mitigated the outflow of young and middle-aged labor in the rural and agricultural sectors, it has not effectively improved rural land utilization efficiency, and new growth poles in agriculture remain undeveloped. Therefore, policy measures should be actively refined to expand the application scenarios for integrating digital technologies with land resources, establish new models of digital-land resource synergy, enhance land use efficiency, and strengthen the sustainable development capacity of the rural economy.