The Agricultural Ecological Effects of Rural Labor Migration: A Perspective Based on Green Total Factor Productivity

Abstract

In the context of promoting sustainable and low-carbon agricultural development, this study investigates the effects of rural labor migration (RLM) on agricultural ecological efficiency from the perspective of green total factor productivity (GTFP). Using panel data from 30 Chinese provinces (autonomous regions, municipalities) over 2011–2022, agricultural GTFP is calculated via the SBM–Global Malmquist–Luenberger (SBM–GML) index. Baseline regressions and the spatial Durbin model (SDM) are employed to examine the impacts of labor migration. The research results show that: (1) Agricultural ecological efficiency exhibits significant spatial clustering, demonstrating “high–high” and “low–low” aggregation patterns. (2) RLM significantly enhances local agricultural ecological efficiency while also generating a positive spatial spillover effect. (3) The effects are heterogeneous: northern regions and highly urbanized areas experience stronger positive impacts, whereas southern regions and less urbanized areas show weaker effects. The findings highlight the pivotal role of RLM in promoting agricultural modernization and provide insights for enhancing regional coordination and ecological efficiency.

1. Introduction

1.1. Research Background

In recent years, global climate change and ecological constraints have become increasingly severe, making carbon reduction a central priority for promoting sustainable economic and social development [1,2,3]. Agriculture, as a resource-intensive and environmentally sensitive sector, is not only fundamental to ensuring food security [4] but also constitutes a significant source of greenhouse gas emissions. Greenhouse gases generated solely from agricultural production account for approximately 10% of global anthropogenic emissions [1,5], primarily including CH₄ from crop production, N₂O from soil nitrogen fertilizer application, and CO₂ released during the processing and transportation of fertilizers, pesticides, and other agricultural inputs. Against this backdrop, improving ecological efficiency has emerged as a key indicator for assessing the green and sustainable development of agriculture. Its essence lies in maximizing output growth while minimizing resource consumption and environmental damage (e.g., carbon emissions, pollutant residues) while achieving agricultural output growth, essentially representing a “output-resource-environment” coordinated optimization [6,7]. Traditional productivity metrics focus on economic ratios, analyzing the relationship between output and factor inputs, but fail to incorporate environmental dimensions [8,9]. Agricultural GTFP fundamentally evolves and enriches this core, integrating environmental impacts into consideration, thereby establishing a more comprehensive conceptual basis suitable for evaluating low-carbon agricultural ecological efficiency [10,11]. As the world’s largest agricultural producer and greenhouse gas emitter, China has explicitly proposed the “dual carbon” strategy. However, improving agricultural ecological efficiency faces multiple challenges, including the high difficulty of coordinated reduction across diverse emission sources (e.g., methane from paddy fields and nitrous oxide from fertilizers), reliance on imbalanced input structures leading to fertilizer and pesticide dependence and insufficient technological investment, and the constraints imposed by weak agricultural infrastructure on green practices [6,11]. Consequently, optimizing factor allocation to enhance agricultural ecological efficiency and achieving a transition from “high-input, high-output” to “high-quality, low-emission” production has become an urgent practical issue.

Among various paths for optimizing factor allocation, the large-scale RLM to urban areas, as a key factor mobility phenomenon in China’s economic and social development, has profoundly reshaped the “labor-land-capital” allocation structure in agricultural production [12], thereby exerting a significant impact on agricultural ecological efficiency. In terms of scale, data from the National Bureau of Statistics show that China’s rural resident population was approximately 934 million (over 60% of the total population) in 2000, and had declined to about 478 million (less than 35%) by the end of 2023. This unprecedented scale of labor mobility has a dual effect on agricultural ecological efficiency: on one hand, labor shortages may restrict labor-intensive green practices such as intensive cultivation and ecological pest management, temporarily inhibiting improvements in ecological efficiency [13]. On the other hand, labor outflow forces “mechanization to replace labor” and “capital to compensate for labor shortages,” promoting land transfer and large-scale farming. Meanwhile, urban technologies and management experiences brought back by migrating labor contribute to the dissemination of green production practices such as precision fertilization and low-carbon agricultural machinery, which is conducive to long-term improvements in ecological efficiency [14,15].

More importantly, both theory and empirical evidence suggest that such impacts are not confined to local areas. According to knowledge spillover theory and spatial econometrics, production technology changes and governance innovations induced by labor migration in one region can be transmitted to neighboring areas through cross-regional demonstration, industrial chain linkages, and factor flows, thereby generating spatial spillover effects on ecological efficiency [16]. Ignoring such regional interconnections can lead to a one-sided understanding of the ecological effects of labor migration; therefore, it is particularly necessary to systematically examine its mechanisms from a spatial perspective. Additionally, differences in agricultural base conditions and development stages across regions may also result in heterogeneous impacts of labor migration.

1.2. Research Questions and Initial Hypotheses

In existing research, scholars have extensively investigated agricultural production efficiency, Agricultural GTFP, and rural sustainable development [17,18]. However, three major limitations remain in the current literature: First, the ecological implications of RLM have received insufficient attention. Most studies focus on the relationship between labor mobility and economic growth, income structure, or production efficiency [19,20], while systematic analysis of ecological and environmental dimensions remains relatively scarce. Second, spatial interdependencies are frequently overlooked. The majority of research assumes interprovincial agricultural eco-efficiency to be mutually independent [21,22], failing to adequately account for the cross-regional ecological linkages and efficiency spillovers triggered by labor migration. Third, research on regional heterogeneity tailored to the characteristics of Chinese agriculture remains insufficient. Chinese agriculture exhibits significant north–south disparities and urbanization gradient distributions. Variations in natural endowments, industrial structures, and urbanization levels may lead to divergent directions and intensities of labor migration’s impact on ecological efficiency across different regions.

Based on the aforementioned research background and the limitations in existing literature, this paper develops a spatial effects analysis framework of labor migration-agricultural eco-efficiency, focusing on addressing the following three key questions: (1) Can RLM significantly enhance agricultural eco-efficiency? This question focuses on the local effects of labor migration. (2) Does RLM generate spatial spillover effects? Specifically, can migration behaviors positively influence agricultural eco-efficiency in neighboring regions through mechanisms such as knowledge diffusion, industrial linkages, and policy demonstration? (3) Does the impact of RLM on agricultural eco-efficiency exhibit significant regional heterogeneity? In particular, do divergences in north–south agricultural disparities and urbanization levels lead to differentiated characteristics in the ecological effects of labor migration?

Theoretically, the impact of labor migration on agricultural eco-efficiency can be explained through three fundamental pathways. First, based on factor substitution theory and agricultural technological change theory [23,24], labor outflow may promote the upgrading of production modes through “labor-capital” substitution [25], thereby improving agricultural eco-efficiency. Second, drawing on knowledge spillover and spatial correlation theory [26,27], labor migration not only facilitates local factor restructuring but may also generate spatial spillover effects on eco-efficiency through cross-regional diffusion and demonstration mechanisms. Finally, from the perspective of regional differentiation and new economic geography [28], differences in natural endowments, agricultural structures, and urbanization levels across regions will lead to significant heterogeneity in the ecological effects of labor migration [29].

Therefore, this paper proposes three preliminary hypotheses: first, RLM can significantly positively influence local agricultural eco-efficiency (H1); second, this effect can generate spatial spillover effects in neighboring regions through spatial linkage mechanisms (H2); third, the impact of labor migration on agricultural eco-efficiency shows significant variations across different geographical regions and urbanization stages (H3). This study employs a confirmatory research design, utilizing spatial econometric methods to establish an empirical model and conducting verification analysis with panel data from 30 Chinese provinces (autonomous regions and municipalities) from 2011 to 2022. A detailed literature review and analysis will be presented in the subsequent section.

This study contributes three main innovations: (1) Methodologically, by constructing a spatial Durbin model, it examines the direction and magnitude of the impact of RLM on agricultural ecological efficiency from a spatial effects perspective. (2) In terms of measurement, based on sustainable development theory, it establishes an agricultural input-output framework, carefully selects input and output indicators, incorporates agricultural carbon emissions as an undesired output into the model, and employs the SBM-GML index to calculate agricultural GTFP, thereby quantifying agricultural ecological efficiency oriented toward low-carbon development and environmental sustainability. (3) Building on the confirmation of overall spatial spillover effect, this study goes beyond the traditional east-central-west regional partition paradigm (e.g., Yu et al. (2025) [30]), and, by integrating differences in agricultural conditions between northern and southern China, reveals the heterogeneity of the spatial spillover effect along two dimensions: “geographical region (north vs. south)” and “development stage (urbanization level)”. These effects exhibit systematic differences in both intensity and direction, clarifying their effective boundaries and context-dependence, and providing crucial evidence for designing differentiated regional collaborative intervention strategies.

2. Literature Review and Hypothesis Formulation

This chapter builds on the research hypotheses proposed in the introduction and, by reviewing relevant literature and theoretical foundations, systematically elaborates on the direct effects, spatial spillover effect, and heterogeneous mechanisms of RLM on agricultural ecological efficiency, thereby establishing the theoretical basis for subsequent empirical testing.

2.1. Direct Effects of RLM on Agricultural Ecological Efficiency

While the existing literature has not directly examined the impact of RLM on agricultural ecological efficiency, related findings from different research strands suggest competing inferences regarding its directional effect, creating a potential theoretical tension. This provides a critical foundation for this study to clarify the underlying mechanisms and construct core hypotheses.

Some studies suggest that labor outflow may inhibit agricultural ecological efficiency, primarily due to the negative shock to labor supply. Particularly in labor-intensive agricultural systems, labor shortages may constrain the effective implementation of labor-dependent green practices such as intensive farming and ecological pest control [13]. This pathway highlights the potential pressure of quantitative labor loss on ecological efficiency.

Nevertheless, a more representative body of literature emphasizes the positive role of RLM. We explain this through two main theoretical pathways, which lay the groundwork for constructing Hypothesis H1.

First, from the perspective of human capital theory, research confirms that RLM promotes the accumulation and diffusion of green production capacity. Li and Qian (2017) note that migrant workers acquire skills, technical knowledge, and modern management experience in cities, which they bring back to rural areas upon returning to agricultural production [31]. This “human capital spillover” effect directly facilitates the adoption of green technologies. Yu et al. (2023) further verify that RLM accelerates the spread of agricultural informatization and advanced management practices [32]. Xu et al. (2023) and Baležentis et al. (2021) find that labor optimization driven by RLM effectively enhances agricultural productivity [33,34]. Second, from a resource allocation theory perspective, studies emphasize that RLM can enhance ecological efficiency by optimizing factor combinations. Lei et al. (2023) argue that the optimal allocation of production factors is a prerequisite for improving agricultural green efficiency [7]. Zhang et al. (2023) and Li et al. (2023) observe that as RLM occurs, agricultural labor redundancy decreases, promoting the substitution of labor with machinery and technology [35,36]. Yuan and Chen (2019) further find that labor outflow facilitates land consolidation and large-scale operations, thereby reducing resource waste [37]. These studies collectively underscore the positive role of RLM, and multiple works demonstrate that measures such as technological substitution, farmland consolidation, and scaled operations contribute to higher agricultural ecological efficiency [1,8,38].

This body of literature provides theoretical support for establishing RLM as the core independent variable and agricultural ecological efficiency as the dependent variable in the research model. In summary, although there is discussion about potential inhibitory effects, the two mechanisms—human capital spillover and resource allocation optimization—constitute the key pathways through which RLM enhances agricultural ecological efficiency. Based on the above theoretical review and empirical support, this study proposes the following hypothesis:

H1. 

Rural labor migration can enhance agricultural ecological efficiency.

2.2. Spatial Spillover Effect of RLM on Agricultural Ecological Efficiency

Under the context of RLM, the movement of labor not only directly affects the development of green agriculture in the destination areas but may also influence the agricultural ecological efficiency in the origin areas through externality effects [39]. Existing studies have revealed the radiating effects of RLM on production and management patterns in neighboring regions from multiple dimensions. For instance, Zhang and Guo (2025) found that the spatial agglomeration of human capital brought about by labor mobility significantly enhances regional innovation capacity, and such innovation spillover provides developmental momentum for surrounding areas [16]. Zou et al. (2022), through empirical research on cultivated land use efficiency, further confirmed that RLM generates significant spatial spillover effects via factor recombination and technology diffusion [12]. These studies, from different perspectives, validate the spatial spillover effects induced by RLM, providing important theoretical support for this study’s exploration of the impact of RLM on agricultural ecological efficiency.

Based on the existing research foundation, we can further dissect the specific pathways through which RLM influences the agricultural ecological efficiency of neighboring regions. First, from the perspective of knowledge dissemination, the “technology-capital-knowledge” tripartite transmission mechanism formed alongside labor mobility establishes a green innovation network between origin and destination areas. This networked connection not only accelerates the cross-regional diffusion of green technologies such as water-saving irrigation and organic farming but also promotes the exchange and mutual learning of management experiences in ecological compensation and circular agriculture, reflecting the classic discourse of knowledge spillover theory on the spatial diffusion of innovation elements [40,41]. Second, from the market-driven dimension, labor mobility reshapes the supply and demand dynamics of agricultural products, triggering profound changes in agricultural production patterns in neighboring regions [42,43,44]. When the mobile population transmits green consumption concepts and demand for high-quality agricultural products to surrounding markets, it not only stimulates the extension and refinement of the green agricultural industry chain but also, through price signals and market competition mechanisms, encourages producers in adjacent areas to proactively adopt environmentally friendly production technologies. This process of demand-induced supply transformation perfectly illustrates the core tenets of externality theory [45,46]. Therefore, while improving local ecological efficiency, RLM also generates significant spatial spillover effects in neighboring regions through knowledge diffusion, market demand transmission, and the restructuring of incentive mechanisms.

Based on the theoretical and empirical analysis of knowledge spillover and externality mechanisms presented above, this study can confirm the spatial spillover effect hypothesis:

H2. 

Rural labor migration has a positive spatial spillover effect on agricultural ecological efficiency.

2.3. Regional Heterogeneity in the Impact of RLM on Agricultural Ecological Efficiency

Significant differences exist across regions in terms of economic development levels, technological foundations, resource endowments, and policy environments, which determine that the impact of RLM on agricultural ecological efficiency exhibits regional heterogeneity [47]. Variations in agricultural production systems, urbanization levels, and green factor allocation approaches across regions mean that the ecological effects of labor migration are not homogeneous but manifest differences in direction and intensity through distinct structural mechanisms.

First, the north–south differences in agricultural production systems are an important source of heterogeneity in the ecological effects of RLM [48]. In southern regions, agricultural operations are relatively fragmented, plots are small, labor intensity is high, and production processes rely heavily on manual management and experiential practices [49]. Labor outflow is likely to weaken the implementation of precise cultivation and ecological management measures, increasing the substitutability of inputs such as fertilizers and pesticides [50], thereby inhibiting improvements in agricultural ecological efficiency. In contrast, northern regions feature larger-scale land holdings, higher levels of mechanization, and lower marginal dependence on labor [51]. Here, labor migration is more likely to enhance resource use efficiency and ecological performance by promoting land consolidation, capital investment, and the application of green machinery. Therefore, differences in operation scale, factor structure, and technological conditions between the north and south lead to significant differentiation in the direction and intensity of the ecological effects of labor migration.

Second, differences in urbanization levels constitute another key source of heterogeneity [52]. In regions with higher urbanization, urban economies have a strong capacity to absorb rural labor, resulting in labor migration that is large in scale and rapid in pace [53]. This large-scale, fast-moving migration, combined with well-developed infrastructure and higher levels of education and technology dissemination, provides favorable conditions for the flow and diffusion of green production factors [54]. Under such conditions, labor migration not only brings back income and capital but also facilitates the reintegration of green knowledge and management experience, making it easier for agricultural production to achieve energy conservation, emission reduction, and efficiency improvements. In contrast, in regions with lower urbanization, cities have limited capacity to absorb rural labor, so labor migration tends to be smaller in scale and slower in process [53]. Moreover, these regions generally have weaker capacities to adopt and transform green agricultural technologies [55], making it difficult for labor migration to effectively translate into improvements in ecological efficiency. Instead, the outflow of high-quality labor may lead to extensive management practices and stagnation in ecological efficiency.

Based on the above theoretical analysis of regional and development-stage heterogeneity, this study formulates the heterogeneity hypothesis as follows:

H3. 

The impact of rural labor migration on agricultural ecological efficiency exhibits regional heterogeneity; factors such as agricultural production characteristics across northern and southern regions and areas with different urbanization levels lead to significant differences in its effects.

3. Research Design

3.1. Modeling

3.1.1. Fixed Effects Model and SYS-GMM

The Fixed Effects Model is a commonly used method in panel data analysis, employed to control for heterogeneity among individuals. In this study, the Fixed Effects Model is primarily utilized to examine changes in agricultural productivity within different provinces (autonomous regions, municipalities), eliminating disturbances caused by inherent characteristics of each province (such as geographical location, cultural differences, etc.) and time-related factors.

Unlike the Fixed Effects Model, which is primarily used to control for time-invariant heterogeneity, SYS-GMM is particularly suitable for capturing dynamic relationships between variables. It can mitigate endogeneity arising from reverse causality or omitted variables [56,57], thereby providing more robust estimates of dynamic effects.

The fixed effects model and SYS-GMM expression are as follows:

$${\mathrm{G}\mathrm{T}\mathrm{F}\mathrm{P}}_{\mathrm{i}\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{R}\mathrm{L}\mathrm{M}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\alpha }}_3\mathrm{X}+{\mathrm{u}}_{\mathrm{i}}+{\mathrm{v}}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(1)

$${\mathrm{G}\mathrm{T}\mathrm{F}\mathrm{P}}_{\mathrm{i}\mathrm{t}}={\mathsf{\alpha }}_0{\mathrm{G}\mathrm{T}\mathrm{F}\mathrm{P}}_{\mathrm{i},\mathrm{t}-1}+{\mathsf{\alpha }}_1{\mathrm{R}\mathrm{L}\mathrm{M}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\alpha }}_3\mathrm{X}+{\mathrm{u}}_{\mathrm{i}}+{\mathrm{v}}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(2)

In the equation, GTFP represents agricultural GTFP. RLM represents rural labor migration. X is control variables.

3.1.2. SDM

(1) Pre-Test for Spatial Correlation

Before conducting formal spatial tests, it is necessary to first construct Moran’s I to examine the autocorrelation of the variables [58]. The formulas for the Global Moran’s I and Local Moran’s I are provided in Equations (3) and (4):

$$\text{Moran’s} \mathrm{I}={\displaystyle \frac{\mathrm{n}\sum _{\mathrm{i}=1}^{\mathrm{n}}\sum _{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}({\mathrm{X}}_{\mathrm{i}}-\overline{\mathrm{X}})({\mathrm{X}}_{\mathrm{j}}-\overline{\mathrm{X}})}{\sum _{\mathrm{i}=1}^{\mathrm{n}}{({\mathrm{X}}_{\mathrm{i}}-\overline{\mathrm{X}})}^2\sum _{\mathrm{i}=1}^{\mathrm{n}}\sum _{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}}}$$

(3)

$${\text{Moran’s} \mathrm{I}}_{\mathrm{i}}=\left({\mathrm{X}}_{\mathrm{i}}-\overline{\mathrm{X}}\right){\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}\left({\mathrm{X}}_{\mathrm{j}}-\overline{\mathrm{X}}\right)$$

(4)

where i, j, t represent city i, city j, and year t, respectively. ${\mathrm{w}}_{\mathrm{i}\mathrm{j}}$ is an element of the spatial weight matrix. In this study, two spatial weight matrices [59] are constructed. The formulas for the spatial adjacency matrix W1 and the spatial distance matrix W2 are as follows:

$${\mathrm{W}}_{1\mathrm{i}\mathrm{j}}=\left\{\begin{array}{c}1,\mathrm{if} \mathrm{i} \mathrm{is} \mathrm{a}\mathrm{d}\mathrm{j}\mathrm{a}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{t} \mathrm{to} \mathrm{j} \\ 0,\mathrm{if} \mathrm{i} \mathrm{is} \mathrm{not} \mathrm{a}\mathrm{d}\mathrm{j}\mathrm{a}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{t} \mathrm{to} \mathrm{j} \mathrm{or} \mathrm{i}=\mathrm{j}\end{array}\right.$$

(5)

$${\mathrm{W}}_{2\mathrm{i}\mathrm{j}}={\displaystyle \frac{1}{{\mathrm{d}}_{\mathrm{i}\mathrm{j}}}}$$

(6)

(2) Spatial Model Selection and SDM Specification

Common spatial econometric models include the SAR, SEM, SDM, SDPM, and SARAR. This study ultimately adopts the SDM as the core empirical specification for the following reasons.

First, in terms of theoretical inclusiveness and specification robustness, the SDM is a general form of both SAR and SEM [60,61]. When certain parameter restrictions hold, the SDM can reduce to either model, thus minimizing the risk of misspecification under uncertain spatial dependencies. Compared with SDPM and SARAR, the SDM simultaneously includes spatial lags of both dependent and independent variables, directly capturing spatial diffusion of exogenous variables [62,63], which fits the purpose of analyzing the interregional spillover effects of RLM. In contrast, models such as SARAR or SDPM, while more comprehensive in form, require stronger distributional assumptions and more complex parameter structures. Their estimation precision—especially regarding error autocorrelation parameters—tends to deteriorate with moderate sample sizes [64]. Therefore, the SDM offers a balanced trade-off between theoretical generality, estimation robustness, and interpretability.

Second, regarding interpretability and alignment with research objectives, the study aims to quantify both direct and spatial spillover effects of RLM on agricultural GTFP. The SDM naturally accommodates these effects by including spatial lags of explanatory variables, allowing for direct hypothesis testing of interregional transmission mechanisms [61]. In comparison, models such as SAR or SARAR require indirect computation of partial derivatives or impact multipliers to infer spillover effects, which complicates interpretation [65].

Finally, empirical diagnostic tests also support the SDM specification (

Table 1. SDM Verification Results.

		Inspection Results
		W1	W2
Wald Test	Wald Test for SAR	95.97 ***	79.45 ***
	Wald Test for SEM	88.96 ***	90.41 ***
LR Test	Simplified as SAR	85.64 ***	71.69 ***
	Simplified as SEM	85.20 ***	66.06 ***
Hausman Test		41.93 ***	26.78 ***
LM Test	Moran’s I	10.112 ***	11.507 ***
	Lagrange multiplier	95.122 ***	110.710 ***
	Robust Lagrange multiplier	6.103 **	0.054 *
	Lagrange multiplier	141.350 ***	150.392 ***
	Robust Lagrange multiplier	52.331 ***	39.736 ***

* p < 0.1, ** p < 0.05, *** p < 0.01.

). The results of LM, Wald, and Likelihood Ratio tests all reject the null hypotheses that “the SDM can be simplified to SAR” or “to SEM” at the 1% level. The Hausman test further supports the use of fixed effects. Collectively, both theoretical and empirical evidence confirm that the SDM provides the most suitable framework for this study, consistent with prior spatial econometric findings in regional productivity and efficiency research [66].

Based on the above rationale, the Spatial Durbin Model constructed in this study is specified as follows:

$$\begin{gathered} {\mathrm{G}\mathrm{T}\mathrm{F}\mathrm{P}}_{\mathrm{i}\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\sum }_{\mathrm{j}}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}{\mathrm{G}\mathrm{T}\mathrm{F}\mathrm{P}}_{\mathrm{j}\mathrm{t}}+{\mathsf{\alpha }}_2{\mathrm{R}\mathrm{L}\mathrm{M}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\alpha }}_3{\sum }_{\mathrm{j}}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}{\mathrm{R}\mathrm{L}\mathrm{M}}_{\mathrm{j}\mathrm{t}}+{\mathsf{\alpha }}_4\mathrm{X}+{\mathsf{\alpha }}_5{\sum }_{\mathrm{j}}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}\mathrm{X}+ \\ {\mathrm{u}}_{\mathrm{i}}+{\mathrm{v}}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}} \end{gathered}$$

(7)

In Equation (7), ${\mathrm{u}}_{\mathrm{i}}$ and ${\mathrm{v}}_{\mathrm{i}}$ denote fixed effect. ${\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$ means a randomly perturbed error term.

3.2. Research Area and Time Frame

This study selects 30 provinces (autonomous regions and municipalities) in mainland China as the research area, including Beijing, Tianjin, Hebei, Shanxi, Inner Mongolia, Liaoning, Jilin, Heilongjiang, Shanghai, Jiangsu, Zhejiang, Anhui, Fujian, Jiangxi, Shandong, Henan, Hubei, Hunan, Guangdong, Guangxi, Hainan, Chongqing, Sichuan, Guizhou, Yunnan, Shaanxi, Gansu, Qinghai, Ningxia, and Xinjiang. These regions cover China’s major economic zones, encompassing both the economically developed coastal areas with higher levels of agricultural modernization and the inland and western regions characterized by diverse resource endowments and significant differences in agricultural foundations. Such a spatial distribution ensures strong representativeness and comparability of the sample in terms of economic structure, natural conditions, and stages of agricultural development, thereby facilitating a comprehensive depiction of regional disparities in agricultural ecological efficiency and RLM across China.

The research period is set from 2011 to 2022, covering 12 years. This period represents a crucial phase in China’s green agricultural transformation and urban–rural integration. Since the implementation of the 12th Five-Year Plan in 2011, the concept of green development has been systematically incorporated into the national strategy, accompanied by continuous promotion of agricultural modernization, ecological civilization, and rural revitalization policies, which have provided institutional and policy guarantees for improving agricultural ecological efficiency. Meanwhile, rural labor has continued to shift toward non-agricultural industries and urban areas, leading to significant changes in agricultural production methods, resource allocation, and ecological performance.

Overall, the selected research area and time frame reflect both spatial diversity and temporal evolution, providing a solid spatiotemporal foundation for systematically analyzing the relationship between RLM and agricultural ecological efficiency in China.

3.3. Variables and Data

3.3.1. Explained Variable: Agricultural Ecological Efficiency

Drawing on previous research [10,67], this paper uses agricultural GTFP to measure agricultural ecological efficiency, which reflects the efficiency of various resources used in agricultural production. In this process, the SBM–GML index method is employed to calculate GTFP [68,69], with input and output indicators listed in

Table 2. Agricultural Green Total Factor Productivity Index System.

	Criteria Layer	Indicator Layer	Unit
Input	Land investment	Seeded area	1000 hm2
	Labor input	Number of employed individuals in the primary industry	Ten thousand people
	Mechanical investment	Mechanical investment	10,000 kW
	Resource investment	Fertilizer usage	10,000 t
		Agricultural film usage	Tons
		Agricultural irrigation area	1000 hm2
Output	Expected output	Primary production value	RMB 100mn
	Unexpected output	Agricultural carbon emissions	Ten thousand tons

. The SBM model and GML index model are constructed as follows:

$$min\theta ={\displaystyle \frac{1-\frac{1}{m}{\sum }_{i=1}^m\frac{s_i^{-}}{x_i}}{1+\frac{1}{q+p}({\sum }_{r=1}^q\frac{s_r^g}{y_r^g}+{\sum }_{k=1}^q\frac{s_k^b}{y_k^b})}}$$

(8)

$$st.\left\{\begin{array}{c}X\lambda +s^{-}=x_0\\ Y^g\lambda -s^g=y_0^g\\ Y^b\lambda +s^b=y_0^b\\ \lambda ,s^{-},s^g,s^b\ge 0\end{array}\right.$$

$${GML}^{t,t+1}\left(x^{t+1},y^{g,t+1},y^{b,t+1};x^t,y^{g,t},y^{b,t}\right)={\displaystyle \frac{1+D_G^T(x^t,y^{g,t},y^{b,t})}{1+D_G^T\left(x^{t+1},y^{g,t+1},y^{b,t+1}\right)}}\ast ⌊{\displaystyle \frac{\frac{1+D_G^T(x^t,y^{g,t},y^{b,t})}{1+D^t(x^t,y^{g,t},y^{b,t})}}{\frac{1+D_G^T\left(x^{t+1},y^{g,t+1},y^{b,t+1}\right)}{1+D^t\left(x^{t+1},y^{g,t+1},y^{b,t+1}\right)}}}⌋$$

(9)

In Equations (8) and (9), x represents the input. s^g denotes the desirable output. s^b represents the undesirable output. $s^{-}$ refer to the input slack. $s_k^g$ and $s_k^b$ represent the expected output shortfall and unexpected output surplus, respectively. $\lambda$ is a weight vector. ${\mathrm{x}}_0,{\mathrm{y}}_0^{\mathrm{g}},{\mathrm{y}}_0^{\mathrm{b}}$ represent the actual input variables, expected output, and unexpected output variable values corresponding to the decision-making unit, respectively. $X,Y^g,Y^b$ represent the input variables, expected output, and unexpected output variable values required for the decision unit estimation, respectively. m, q and p represent the number of input, expected output, and unexpected output indicators, respectively. $min\theta$ is the efficiency value. $D_G^T(x^t,y^{g,t},y^{b,t})$ represents the reference and direction vector, referring to the direction of the output vector specified by the distance SBM function between x and the production structure.

Figure 1 presents the average agricultural GTFP of 30 provinces (autonomous regions, municipalities) over a 12-year period. The highest agricultural ecological efficiency are observed in Beijing, Guangdong, Fujian, Hainan, and Shaanxi, whereas Gansu, Shanxi, and Anhui show relatively low efficiency. Geographically, the agricultural GTFP is higher in the southeastern coastal regions and relatively lower in the northern regions, exhibiting an overall increasing trend from north to south and showing signs of regional clustering. These differences can be attributed to variations in economic structure, industrial development, and access to advanced technologies across regions. Southeastern coastal provinces benefit from greater economic openness, better infrastructure, and higher investment in high-tech industries, while northern regions may experience slower development due to lower economic diversification and regional imbalances.

3.3.2. Explanatory Variable: Rural Labor Migration

Traditionally, rural labor has mainly focused on the primary industry. Urban labor force is generally engaged in the secondary and tertiary industries. With the dynamic changes in the labor market, labor is gradually flowing from traditional agricultural sectors to more modern industrial and service sectors. Building on the methodology of Liu et al. (2021) [70], the ratio of the total number of rural labor transfers to the total rural labor force is used as a proxy indicator for RLM. The total number of rural labor transfers is calculated as the difference between the total rural employment and the number of rural employment engaged in the primary industry.

Figure 2 illustrates the average RLM in the 30 sampled provinces (autonomous regions, municipalities) over the 12-year period. The results show that the speed of rural labor migration in the southeastern coastal regions is significantly higher than in the inland areas, particularly in Shanghai, Beijing, Jiangsu, Zhejiang, and Fujian. Overall, the trend exhibits a gradual increase from north to south and from the central and western inland regions toward the eastern coastal areas, with a pronounced clustering pattern. The higher migration ratios in these southeastern regions indicate that non-agricultural sectors have absorbed a large portion of the rural labor force. In contrast, some provinces in the western and northern regions, such as Yunnan, Inner Mongolia, and Jilin, show slower RLM with lower migration ratios. These differences are likely closely related to regional variations in economic development, industrial structure, and urbanization progress. The southeastern coastal areas have experienced rapid economic growth, well-developed infrastructure, and strong absorption capacity of non-agricultural sectors, providing more opportunities for RLM. In comparison, inland regions lag behind economically, with lower levels of urbanization and fewer non-agricultural employment opportunities, resulting in slower labor migration. Additionally, local government policy support and labor training initiatives also play a certain role in influencing RLM across regions.

3.3.3. Control Variable

In order to ensure the reliability of the study, this article selects regional average living standards, education levels, agricultural industry agglomeration levels, government intervention levels, and levels of openness to the outside world as control variables.

The detailed definition and descriptive statistics of variables are shown in

Table 3. Variable definitions and descriptive statistics.

Variable	Name	Abbreviation	Definition	N	Mean	SD
Dependent variable	Agricultural GTFP	GTFP	Refer to 3.3.1	360	0.969	0.200
Explanatory variables	Rural labor migration	RLM	Refer to 3.3.2	360	0.622	0.109
Control variable	Average standard of living	ENG	Engel coefficient	360	33.452	5.769
	Educational level	EDU	Per capita years of education	360	7.974	0.848
	Agricultural industry agglomeration level	IA	(Provincial primary output value/National primary output value)/(Provincial GDP/National GDP)	360	0.026	0.039
	Degree of government intervention	GOV	Fiscal expenditure/regional GDP	360	0.249	0.103
	Level of openness to the outside world	OPEN	Total import and export volume/regional GDP	360	0.263	0.285

.

3.4. Data Sources

The data are primarily sourced from the China Statistical Yearbook, China Rural Statistical Yearbook, China City Statistical Yearbook, China Regional Economic Statistical Yearbook, provincial statistical yearbooks, and provincial rural economic statistical data. For a few missing values, imputation methods were applied to ensure data completeness and analytical accuracy.

4. Empirical Results

4.1. Analysis of Direct Impact

Based on the estimation results of the fixed effects model and the system GMM model (see

Table 4. Benchmark Regression Results.

	Fixed Effects Model		SYS-GMM
RLM	1.888 ***	0.588 **	0.592 *	1.949 **
	7.59	2.39	1.64	1.95
ENG		−0.007 ***		0.004
		−5.20		1.16
EDU		0.106 ***		0.064 *
		11.41		1.83
IA		−6.626 ***		−3.984 **
		−3.41		−2.59
GOV		−0.013		0.708 ***
		−0.06		3.16
OPEN		−0.227 **		0.102
		−2.57		0.74
_cons	−0.785 ***	−0.352 *	0.021	−1.583 ***
	−5.07	−1.83	0.09	−3.44
N	360	360	330	330
R2	0.149	0.446
AR(1)			2.20	−0.72
AR(2)			2.27	−0.64
Hansen Test			29.98	24.88

z-statistics are reported below the coefficients, * p < 0.1, ** p < 0.05, *** p < 0.01.

), RLM has a significant positive effect on agricultural ecological efficiency, regardless of whether control variables are included, thus supporting Hypothesis 1. This result indicates that, after controlling for regional differences and time effects, RLM can still significantly enhance agricultural ecological efficiency. As the migration of rural labor progresses, agricultural production gradually adjusts toward scale and intensification, and farmers tend to adopt mechanization and green technologies to compensate for labor shortages, thereby improving resource allocation efficiency and environmental performance. Overall, RLM promotes the transformation and upgrading of agricultural production practices and drives the adoption and diffusion of green technologies, demonstrating a significant positive impact on agricultural ecological efficiency.

4.2. Analysis of Spatial Effects

4.2.1. Moran’s I

Before conducting SDM analysis, it is necessary to verify the spatial correlation between variables, which is a necessary step to ensure the accuracy and scientificity of the model. Therefore, this article selects two different spatial weight matrices: adjacency matrix (W1) and economic distance matrix (W2). Based on these two matrices, the spatial correlation of agricultural production efficiency was tested using the global Moran index (see

Table 5. Global Moran Index.

Year	Moran’s I (W1)	Moran’s I (W2)
2011	0.238 **	0.091 ***
2012	0.223 **	0.071 ***
2013	0.207 **	0.108 ***
2014	0.155 **	0.060 ***
2015	0.204 **	0.079 ***
2016	0.262 ***	0.136 ***
2017	0.295 ***	0.136 ***
2018	0.349 ***	0.016 *
2019	0.301 ***	0.075 ***
2020	0.331 ***	−0.009
2021	0.219 **	0.038 *
2022	0.220 **	0.065 *

* p < 0.1, ** p < 0.05, *** p < 0.01.

for specific results). The test results indicate that almost all variables passed the test at a high level of significance, and there is significant spatial correlation between these variables. This indicates that the agricultural production efficiency in different regions has a clear spatial dependence and spatial spillover effect.

To further analyze the spatial characteristics of agricultural ecological efficiency, this study conducts a local Moran’s I analysis on agricultural production efficiency data for 2012, 2016, and 2022, with the results shown in Figure 3. The scatter plots reveal a clear spatial clustering pattern of agricultural production efficiency across regions, indicating strong similarity in agricultural production efficiency among neighboring areas. Significant “high–high” or “low–low” clustering patterns are observed, confirming the pronounced spatial clustering effect of agricultural ecological efficiency.

4.2.2. SDM Regression Results

Table 6. SDM Regression Results.

		W1		W2
Wx	RLM	3.043 ***	3.000 ***	2.124 ***	4.376 ***
		10.90	8.89	5.38	5.62
	ENG		0.002		0.009 ***
			1.06		2.86
	EDU		0.006		0.012
			0.50		0.71
	IA		−8.130 **		−8.336 *
			−2.44		−1.94
	GOV		−1.059 ***		−2.424 ***
			−4.00		−4.57
	OPEN		−0.335 ***		−0.680 ***
			−3.35		−3.09
Spatial	rho	0.511 ***	0.334 ***	0.694 ***	0.311 ***
		11.86	6.21	13.14	2.98
Variance	sigma2_e	0.005 ***	0.005 ***	0.006 ***	0.006 ***
		13.20	13.30	12.69	12.71
Direct effect	RLM	0.209	0.105 *	0.135 *	0.076 **
		1.49	1.61	1.79	2.21
Indirect effect	RLM	5.517 ***	4.152 ***	6.061 ***	6.179 ***
		14.52	9.70	7.67	7.93
Total effect	RLM	5.726 ***	4.257 ***	6.196 ***	6.103 ***
		13.67	8.65	7.48	7.68
N		360	360	360	360
R2		0.587	0.720	0.496	0.694

z-statistics are reported below the coefficients, * p < 0.1, ** p < 0.05, *** p < 0.01.

presents the regression results of the SDM. The results indicate that under both W1 and W2 spatial weight matrices, the spatial lag effect of RLM is positive and significant at the 1% level, suggesting that labor migration in neighboring regions also exerts a significant positive spatial spillover effect on local agricultural ecological efficiency. The spatial autocorrelation coefficient is significantly positive, indicating strong positive correlation of agricultural ecological efficiency among neighboring regions. These findings demonstrate that the impact of RLM on agricultural ecological efficiency is not limited to local areas but exhibits clear spatial diffusion effects, thus supporting Hypothesis 2.

Further decomposition of the results shows that the direct, indirect, and total effects of RLM on agricultural ecological efficiency are all positive, with most being statistically significant, confirming that RLM not only enhances agricultural ecological efficiency locally but also promotes improvements in neighboring regions through spatial spillovers.

4.2.3. Robust Test

This paper mainly conducts robustness testing through two methods. (1) Replace with economic distance matrix (W3) to test the original model to ensure the impact of different spatial weight matrices on the model results. The W3 expression is as follows: $W_{3ij}={\displaystyle \frac{1}{|{GDP}_i-{GDP}_j|}}$. (2) Replace with SAR and SEM models for regression analysis again (based on W1), which, respectively, consider spatial lag and spatial error effects, further verifying the accuracy and robustness of the models.

Table 7. Results of robustness test.

		Other Matrices	Other Models
		W3	SAR	SEM
Main	RLM	0.181	0.319 ***	0.311 ***
		0.98	2.88	2.56
Wx	RLM	3.172 ***
		6.99
Spatial	rho	0.341 ***	0.399 ***
		5.42	7.00
	lambda			0.341 ***
				4.88
Variance	sigma2_e	0.006 ***	0.178 ***	0.019 ***
		12.68	13.25	13.22
Direct effect		0.029	0.339 ***
		0.16	2.82
Indirect effect		4.513 ***	0.208 **
		8.08	2.14
Total effect		4.542 ***	0.547 ***
		7.57	2.62
Controls		YES	YES	YES
N		360	360	360
R2		0.675	0.443	0.345

z-statistics are reported below the coefficients, ** p < 0.05, *** p < 0.01.

reports the robustness test results. The findings are as follows:

(1) Consistency of the core variable: The coefficient of RLM is positive and significant at the 1% level in both the SAR and SEM models, consistent with the main coefficient in the SDM, indicating that the positive effect of labor migration on agricultural ecological efficiency holds across different model specifications.

(2) Analysis of spatial effect differences: In the SDM (W3) model, RLM exhibits a significant spatial spillover effect (indirect effect = 4.513 ***), suggesting that labor flows between regions with similar economic levels influence each other through knowledge diffusion, technology spillovers, and other channels. In the SAR model, although the spatial autocorrelation coefficient is significantly positive, its indirect effect (0.208 **) is much smaller than that in the SDM, mainly because the SAR model only accounts for the spatial dependence of the dependent variable, whereas the SDM captures the spatial interactions between independent and dependent variables. In the SEM model, the spatial error coefficient is significantly positive, indicating the potential presence of unobserved spatially dependent factors, yet the model still supports a significant positive impact of RLM on agricultural ecological efficiency.

In summary, although different models characterize spatial effects differently, the conclusion that RLM promotes agricultural ecological efficiency remains robust, and the results are highly reliable.

4.3. Heterogeneity Analysis

To explore the regional and urbanization heterogeneity of labor migration effects, this study conducted a subgroup regression analysis.

(1) North–South regional heterogeneity: Significant differences exist between southern and northern China in terms of agricultural production types, climatic conditions, and cultivation systems. In this section, the study area is divided into southern and northern regions using the Qinling–Huaihe line as the boundary. For provinces spanning both regions (Jiangsu, Anhui, Shaanxi, Shandong, Henan), the division is based on the proportion of their area within each region. The results (see

Table 8. Heterogeneity Analysis Results.

		(1)	(2)	(3)	(4)
		Southern Region	Northern Region	High Level of Urbanization	Low Level of Urbanization
Main	RLM	−0.949 ***	0.427 *	0.845 ***	−1.192 ***
		−4.22	1.68	2.83	−5.18
Wx	RLM	1.275 ***	3.314 ***	1.809 ***	3.486 ***
		2.65	7.65	5.79	6.28
Spatial	rho	0.482 ***	0.240 ***	0.242 ***	0.302 ***
		6.85	2.92	3.92	3.66
Variance	sigma2_e	0.004 ***	0.004 ***	0.005 ***	0.005 ***
		9.26	9.43	9.69	9.02
Direct effect	RLM	−0.808 ***	0.680 ***	1.094 ***	−0.832 ***
		−3.29	2.80	3.67	−3.51
Indirect effect	RLM	1.439 *	4.236 ***	2.398 ***	3.437 ***
		1.68	8.13	6.88	6.09
Total effect	RLM	0.631	4.916 ***	3.492 ***	2.605 ***
		0.65	8.05	6.81	3.89
Controls		YES	YES	YES	YES
N		180	180	192	168
R2		0.744	0.731	0.721	0.737

z-statistics are reported below the coefficients, * p < 0.1, *** p < 0.01.

(1)–(2)) indicate that in southern regions, the direct effect of RLM is negative while the indirect effect is positive and significant at the 10% level, suggesting that local labor outflow may suppress agricultural ecological efficiency due to the high concentration of farming activities, but technological diffusion and capital inflows from neighboring areas can partially offset this negative impact. In northern regions, the direct, indirect, and total effects are all significantly positive, indicating that in the more mechanized north, RLM helps promote land consolidation and large-scale operations, thereby enhancing agricultural ecological efficiency.

(2) Urbanization heterogeneity: Based on the median urbanization rate, the sample is divided into high and low urbanization groups. The results (see Table 8(3)–(4)) show that in highly urbanized areas, the direct, indirect, and total effects of RLM are all significantly positive, indicating that when cities have a strong capacity to absorb labor, RLM can simultaneously improve ecological efficiency locally and in surrounding areas. In low-urbanization regions, the direct effect is negative while the indirect effect is positive, suggesting that local areas may face short-term production adjustment costs, but neighboring areas can still benefit through the spatial spillover effect.

Overall, Hypothesis 3 is supported.

5. Discussion

5.1. Discussion of Direct and Spatial Spillover Effects

The findings of this study indicate that RLM has significantly enhanced agricultural eco-efficiency overall and has generated positive spatial spillover effects on neighboring regions, thereby confirming Hypotheses 1 and 2. This result deepens our understanding of the systemic role of RLM in promoting agricultural green transformation, suggesting that RLM not only reshapes the allocation structure of agricultural production factors but also facilitates the diffusion and optimization of green production practices across broader spatial scales. Existing research, such as Kapoor and Das (2021) and Hao et al. (2023), has primarily focused on the impacts of labor mobility on agricultural labor efficiency, rural household income structures, and factor substitution [71,72], while systematic investigations into its effects on agricultural eco-efficiency remain relatively scarce. In contrast, this study reveals another economic function of RLM from the perspective of eco-efficiency—namely, promoting the green transformation of agricultural systems—thereby extending the research frontier of the economic effects of labor mobility. Specifically, the supply constraints induced by labor outflow are likely to compel farmers to substitute labor with capital and technological factors. Li et al. (2024) provide empirical support for the significant role of RLM in agricultural mechanization [25]. Furthermore, a study based on a sample of 1122 rice farmers in China found that households with migration experience had significantly higher agricultural machinery expenditures than those without such experience [73], offering evidence for the “induced technological change” process proposed in this study. Together, these findings outline a theoretical framework for how labor migration influences agricultural eco-efficiency, providing robust support for understanding its role in green transformation.

Most importantly, this study reveals the spatial spillover effects of labor transfer on agricultural eco-efficiency (H2), which are underpinned by the dynamic coupling of production factors across regions and knowledge diffusion. RLM not only alters the spatial pattern of factor allocation but also facilitates the spread of green production concepts and eco-friendly management practices through the cross-regional flow of human capital, information, and technology, thereby generating spatial spillover effects in agricultural eco-efficiency. Research by He and Li (2024) indicates that labor mobility promotes interregional knowledge dissemination and optimized factor allocation through social networks and learning effects [74], providing theoretical and empirical support for the spatial spillover of eco-efficiency uncovered in this study. Furthermore, it is evident that labor mobility does not merely enhance production efficiency at the local level but also forms a “demonstration-imitation-diffusion” chain for green technologies at the regional level, enabling agricultural eco-efficiency to exhibit synergistic improvement across spatial dimensions. Simultaneously, the return of capital and experience brought about by labor mobility strengthens the impetus for green transformation in sending areas. Remittances from migrant workers enhance the capital accumulation capacity of rural households, providing them with the economic foundation to adopt mechanized, energy-saving, and environmentally friendly agricultural technologies. Meanwhile, the knowledge, perspectives, and management experience brought back by migration promote the renewal of production methods and ecological awareness, driving the greening and intensification of agricultural transformation. The interaction of these flows of funds, knowledge, and ideas makes labor mobility a crucial ecological link connecting urban and rural areas and bridging regions. Compared to previous studies that primarily emphasized the role of labor mobility in promoting economic growth, income distribution, or labor productivity [75], this study further reveals its systemic function in enhancing eco-efficiency and diffusing green technologies. Overall, through technological diffusion, return flows of capital, and cognitive renewal, labor mobility not only improves agricultural eco-efficiency in inflow areas but also generates significant green spatial synergy effects.

5.2. Discussion on Heterogeneity

This study finds that the impact of RLM on agricultural eco-efficiency exhibits significant regional heterogeneity, supporting Hypothesis 3.

First, this heterogeneity is reflected in north–south differences. This represents an innovative approach, as previous research, such as Yu et al. (2025) [30], has primarily considered the eastern, central, and western characteristics of rural labor. This paper adopts a north–south perspective that better aligns with agricultural characteristics: in labor-intensive agricultural regions of southern China, labor outflow may temporarily suppress eco-efficiency in the short term. This is mainly because production cycles and farming activities in these areas are highly concentrated, mechanization substitution capacity is limited, and staple crops like rice heavily rely on manual labor. Under these circumstances, labor outflow may constrain farmers’ production arrangements and increase the difficulty of implementing intensive farming and green management practices, thereby creating short-term pressure on eco-efficiency. Empirical research by Liu et al. (2025) shows that in labor-intensive rice-growing areas, population outflow may lead to decreased production efficiency [76], which aligns with the short-term pressure on eco-efficiency observed in southern regions in this study, supporting the mechanism that labor loss may bring short-term negative impacts. In contrast, in northern dryland farming areas, where mechanization levels are higher and land consolidation and scale operations are more mature, labor mobility does not significantly constrain production. Research by Yangchen et al. (2025) found that high mechanization levels can buffer the impact of labor outflow on production efficiency [77], and indirectly enhance eco-efficiency, which resonates with this study’s findings of improved eco-efficiency in northern regions.

Second, the influence of urbanization level differences is evident. In regions with high urbanization levels, cities’ strong absorptive capacity facilitates large-scale and rapid labor mobility. This enables the easier diffusion of advanced agricultural technologies, green capital, and management knowledge to rural areas, achieving efficient flows of knowledge and production factors between urban and rural sectors. This not only enhances production efficiency in inflow areas but also provides financial and experiential support for agricultural structural adjustment, low-carbon transformation, and sustainable development in outflow areas, forming a virtuous cycle. In contrast, in regions with lower urbanization levels, constrained by limited urban absorptive capacity, labor migration occurs at a slower pace and smaller scale. Agriculture in these areas may experience a transitional phase of “labor loss-production adjustment-eco-efficiency fluctuation” in the short term. However, with the development of policies, infrastructure, and social networks, these regions retain the potential to gradually realize the positive ecological effects of RLM. These findings are consistent with existing studies affirming the positive role of urbanization in agricultural production efficiency or eco-efficiency (e.g., Chen et al., 2023 [78]; Jie et al. (2020) [79]). However, this study further examines the heterogeneous effects across regions with different urbanization levels through median-based grouping tests, revealing significant differences in eco-efficiency improvement between high- and low-urbanization areas, thereby expanding the understanding of urbanization’s impact.

Overall, this study underscores the spatially driven effects of RLM on agricultural green development, thereby broadening the previous research paradigm that primarily explained the consequences of RLM from productivity or income perspectives [19,29,80,81]. RLM fundamentally reshapes the structure of rural production factors, technological systems, and spatial interconnections, thereby generating a systematic driving force for the enhancement of eco-efficiency. This finding suggests that the economic and ecological implications of labor mobility should be understood from the perspective of regional coordination and factor linkage, rather than being analyzed solely through the narrow lens of labor supply changes.

6. Conclusions, Policy Recommendations, and Research Limitations

6.1. Conclusion and Policy Recommendations

This study examines the impact of RLM on agricultural ecological efficiency from the perspective of agricultural GTFP, providing empirical evidence through the spatial Durbin model. The results indicate that RLM not only enhances the ecological efficiency of agriculture in the local region but also exerts a positive spatial spillover effect on neighboring areas. The study further reveals that agricultural ecological efficiency exhibits significant spatial clustering, following “high–high” or “low–low” patterns, reflecting a strong spatial dependence across regions. In the heterogeneity analysis, differences between northern and southern regions and variations in urbanization levels significantly influence the effects of RLM on agricultural ecological efficiency. In southern regions, the positive impact of labor migration on ecological efficiency is relatively modest, whereas in northern regions, the effect is notably stronger. Moreover, in areas with higher urbanization levels, the positive effect of labor migration is more pronounced, as the greater capacity of cities to absorb rural labor further promotes improvements in agricultural ecological efficiency.

Based on the research conclusions, the following policy recommendations are proposed:

(1)

RLM is not merely a population movement issue but also provides an important opportunity for advancing green agricultural development. While facilitating labor outflow, efforts should be made to equip the migrating workforce with modern agricultural management skills, mechanized operation capabilities, and eco-friendly production techniques, enabling them to contribute to the enhancement of agricultural green total factor productivity. Specific measures may include: first, providing targeted training for migrant workers, with a focus on organic agriculture, green farming, ecological pest management, and precision fertilization; second, expanding the adoption of smart agricultural equipment and mechanized production facilities to mitigate the negative effects of labor shortages on production efficiency and ecological performance; third, establishing technology demonstration sites and mobile training stations, where hands-on practice and experience sharing allow laborers to acquire practical skills during migration. These initiatives not only improve local agricultural ecological efficiency but also facilitate the diffusion of green technologies and management experience through labor returning to their hometowns or across upstream and downstream links in the agricultural value chain.

(2)

RLM not only affects local agricultural ecological efficiency but also generates a spatial spillover effect through the transfer of technology, capital, management experience, and industrial chain linkages. To fully leverage this potential, cross-regional cooperation and technology sharing should be encouraged: first, establish regional agricultural cooperation alliances to promote high-efficiency agricultural practices in areas with lower ecological efficiency through experience exchange and technical training; second, support collaboration along regional agricultural product value chains so that the technological and managerial advantages brought by labor migration can diffuse throughout upstream and downstream segments; third, encourage local governments to incorporate considerations of neighboring regions’ coordinated development into policy-making, by enhancing infrastructure, sharing information platforms, and linking public services to strengthen the synergy of green agricultural development across regions. Through these measures, RLM can not only improve the agricultural ecological efficiency of individual regions but also drive the optimization of green total factor productivity across entire regions, contributing to the sustainable development of regional agriculture.

(3)

In the southern regions, where paddy rice cultivation dominates, labor demand is concentrated during peak farming seasons, and land is fragmented with a high level of refined management. As a result, labor outflows can cause short-term disruptions to green agricultural production. To address this, it is essential to promote the adoption of eco-friendly planting technologies, precision management practices, and agricultural machinery, while encouraging cooperatives, family farms, and agricultural extension services to provide labor substitutes or technical support during critical production stages. In contrast, northern regions feature higher levels of mechanization, and labor outflows may even facilitate land consolidation and large-scale operations, thereby enhancing agricultural ecological efficiency. In these areas, further efforts should focus on promoting smart agriculture, information-based management, and precision fertilization technologies, as well as strengthening the transfer of agricultural technological achievements to farmers and enterprises to drive the transformation from traditional to green modern agriculture. In highly urbanized regions, governments should reinforce urban–rural collaboration mechanisms to optimize the positive impact of labor mobility on agricultural production efficiency; in less urbanized areas, infrastructure development, technology promotion, and public service investment should be increased to enhance local attractiveness, enabling labor flows to more evenly support green agricultural development.

6.2. Practical and Theoretical Contributions

From a theoretical perspective, this study primarily contributes by expanding the research perspective on the impact of RLM on agricultural ecological efficiency. While previous studies have largely focused on the effects of labor mobility on agricultural productivity or income, this study systematically examines the influence of RLM on agricultural ecological efficiency using a spatial Durbin model. The findings indicate that RLM not only enhances local agricultural ecological efficiency but also generates positive spatial spillover effects in neighboring regions. In addition, agricultural ecological efficiency exhibits clear spatial clustering patterns, specifically “high–high” and “low–low” clusters. These results enrich the theoretical framework of labor economics and sustainable agriculture and provide new empirical evidence for understanding the role of labor reallocation in promoting green agricultural development. Furthermore, by conducting heterogeneity analyses across northern and southern regions and different levels of urbanization, this study reveals the regional differences in RLM’s effects on agricultural ecological efficiency, offering valuable references for future research on geographical and development-level heterogeneity.

From a practical perspective, the results provide actionable insights for policymakers and regional agricultural planners. The study demonstrates that facilitating rational RLM can improve agricultural ecological efficiency and generate positive inter-regional spillover effects, providing empirical support for policies on cross-regional labor allocation and green agricultural development. Moreover, the heterogeneity effects related to urbanization levels suggest that policy formulation should consider regional development differences and coordinate labor flows between urban and rural areas to maximize ecological benefits.

6.3. Research Limitations and Future Research Suggestions

In analyzing the impact of RLM on agricultural productivity, this study uses the ratio of the total number of rural labor transfers to the total rural labor force as a proxy for RLM. Specifically, this indicator reflects the scale of RLM, measuring the extent to which labor moves from the agricultural sector (particularly the primary industry) to non-agricultural sectors (such as the secondary and tertiary industries). However, this indicator also has certain limitations: it only considers the quantity of RLM while ignoring the quality differences among laborers. The indicator used in this paper focuses only on changes in total employment and employment in the primary sector, and does not fully account for the characteristics of labor mobility across industries, temporary labor, and seasonal labor.

Future research can be further advanced in the following two directions: First, refining the measurement system of RLM. On the basis of existing macro-level indicators, future studies could develop a multidimensional index that incorporates factors such as educational attainment, skill composition, occupational type, migration frequency, and income variation. This would better capture the structural transformation of labor mobility from a “quantitative shift” to a “qualitative reallocation”. Introducing such quality dimensions would enable a more precise assessment of how different types of labor migration—such as technical, service-oriented, or seasonal migration—affect agricultural ecological efficiency, thereby providing a more nuanced analytical framework for understanding labor reallocation in the context of sustainable agricultural transformation. Second, integrating micro-level data and field investigations to uncover underlying mechanisms. Future research may employ household survey data, labor migration tracking data, or enterprise-level production information, and utilize multilevel regression, structural equation modeling, or mediation analysis to explore how labor migration affects agricultural ecological efficiency through channels such as technology adoption, capital return, knowledge diffusion, and organizational adjustment. Such micro-level empirical analysis would offer deeper insights into the transmission mechanisms and dynamic effects of labor mobility in promoting green agricultural development and enhancing ecological efficiency.