Effects of Off-Farm Employment on the Eco-Efficiency of Cultivated Land Use: Evidence from the North China Plain

Abstract

The effective allocation of labor and cultivated land resources to ensure food security is a global concern. Understanding the relationship between rural labor off-farm employment and the eco-efficiency of cultivated land use (ECLU) is critical, yet current research in this area remains insufficient. This study explores the dynamics between off-farm employment and ECLU using the North China Plain as a case study, analyzing panel data from 2001 to 2020 through spatial econometric models. The findings reveal significant temporal expansion and spatial differentiation in off-farm employment, with growth rates gradually slowing and spatial disparities diminishing. The average ECLU initially declined from 2001 to 2003, followed by fluctuating increases, with a notable acceleration in growth after 2017. A “U-shaped” relationship between off-farm employment and ECLU was identified, with a turning point at an off-farm employment ratio of 40.73%, occurring around 2003–2004 based on regional averages. Before this threshold, off-farm employment negatively impacted ECLU, while beyond this point, the impact became positive. The study also observed significant spatial spillover effects of off-farm employment on ECLU in the North China Plain. These findings underscore the complex interplay between rural labor migration and agricultural productivity. To maximize the benefits of off-farm employment, policies should encourage the reinvestment of income into sustainable agricultural practices. Furthermore, the significant spatial spillover effects call for enhanced regional coordination and tailored policy interventions to optimize labor allocation and improve ECLU.

1. Introduction

The interplay between agriculture and off-farm employment has been a prominent topic in global sociology and economics [1,2,3]. With the acceleration of industrialization and urbanization, rural labor forces across the world have increasingly migrated toward non-agricultural industries [4,5,6,7,8]. Since the initiation of China’s reform and opening up, particularly with the implementation of the Household Responsibility System in the late 1970s, farmers have gained the freedom to choose their employment paths, leading to a significant increase in off-farm employment. The number of migrant workers in China increased from 32 million in 1978 to 298 million in 2023. This shift in rural labor, characterized by off-farm employment, has profoundly influenced the socioeconomic landscape of the country, particularly in agricultural production [9,10,11,12]. Policymakers are now confronted with the dual challenge of supplying affordable labor to emerging industries while safeguarding national food security [13,14,15]. In this context, the effective allocation of labor and cultivated land resources has become a crucial focus of both academic inquiry and political strategy.

Cultivated land is a cornerstone of agricultural production, and its sustainable and efficient use is crucial for long-term agricultural development. Eco-efficiency offers a more comprehensive measure than traditional efficiency indicators (e.g., land productivity and crop yield per unit area) by incorporating both economic output and negative externalities, such as environmental pollution, thus providing a more accurate reflection of sustainable resource utilization [16,17]. Recently, the concept of eco-efficiency has been increasingly applied to cultivated land use to assess its overall efficiency [18,19]. The eco-efficiency of cultivated land use (ECLU) focuses on the rational management of cultivated land to maximize crop yields while minimizing adverse environmental impacts [20,21]. This approach seeks to balance agricultural economic development with environmental protection by improving resource utilization efficiency and reducing pollution. Therefore, improving the ECLU is particularly important in addressing global concerns about food security and environmental sustainability.

Off-farm employment disrupts the allocation of labor and capital within the agricultural production system, inevitably affecting the ECLU. The existing literature identifies two primary effects of off-farm employment on agricultural production: the negative labor loss effect and the positive income effect [22,23,24]. The labor loss effect refers to the reduction in agricultural labor supply as workers transition to off-farm employment. This is particularly evident in China, where many migrant workers are young men, leading to a decline in the rural labor force and hindering agricultural productivity. Labor loss can hinder productivity in several ways, such as through land abandonment, where unused fields degrade over time [13]; reduced cultivation intensity, which can lead to lower crop yields [25]; and decreased maintenance of agricultural infrastructure, resulting in inefficiencies and increased land degradation. Conversely, the income effect suggests that off-farm employment increases rural household income, enabling farmers to invest more in agricultural inputs such as fertilizers, pesticides, and machinery, thus enhancing agricultural specialization [26,27]. Although these investments are often believed to offset the negative impacts of labor loss by enhancing productivity and efficiency [28,29], improper use of these inputs may lead to non-point source pollution and increased emissions, ultimately undermining sustainable farmland use [30,31]. Thus, the trade-off between the labor loss effect and the income effect may result in a nonlinear impact of off-farm employment on the ECLU.

Several studies highlight the importance of spatial spillover effects in understanding the impact of off-farm employment on the ECLU [32,33]. Local rural labor migration affects migration patterns in neighboring areas through informal channels, such as social networks, indirectly influencing the scale of farmland operations and the input structure of land use in these surrounding regions [13,34]. These networks facilitate the spread of agricultural practices and technologies, extending the impact on agricultural production across regions. As China’s agricultural market becomes more integrated, the spatial mobility of rural labor and the cross-regional operations of agricultural machinery and other production factors have increased [35]. This enhanced mobility strengthens spatial linkages between agricultural activities, necessitating a regional perspective when assessing the effects of off-farm employment. Additionally, non-point source pollution from farmland exhibits spatial diffusion, impacting the sustainable use of farmland in neighboring areas [36]. The demonstration effects of off-farm employment, where higher incomes and practices adopted by off-farm workers influence others, may lead to improper use of agricultural inputs. This improper application of fertilizers and pesticides, driven by increased investments, can cause pollution that spreads beyond the immediate area of use, negatively affecting the ECLU in both directly connected and surrounding regions [37,38].

The reviewed literature underscores the complex and multifaceted impacts of off-farm employment on the ECLU. Previous studies have explored how off-farm employment influences various aspects, such as land transfer [34], grain production [3], farmland abandonment [13], carbon emissions [39], and non-point source pollution [30]. However, few studies have comprehensively analyzed its impact on ECLUs. While some researchers have examined how rural labor migration affects land use or agricultural production efficiency, they often overlooked undesirable outputs, thereby missing critical eco-efficiency considerations [23,24,28]. Although Zou et al. investigated the effect of rural labor migration on the ECLU at the provincial level in China, there is a noticeable gap in down-scaled analyses tailored to specific regions, limiting the relevance of the findings to local contexts [33]. Thus, additional evidence is necessary to clarify the complex nonlinear and spatial effects of off-farm employment on ECLUs.

To address these research gaps, this study focuses on the North China Plain, exploring the nonlinear impact of off-farm employment on the ECLU. By analyzing panel data from 48 prefecture-level cities spanning 2001 to 2020, we assess the ECLU using the super-efficiency Epsilon-based measure (EBM) model, which accounts for undesirable outputs, such as carbon emissions and non-point source pollution. Spatial econometric models are then used to investigate the influence of off-farm employment on the ECLU. Specifically, the objectives of this study are to (1) investigate the spatiotemporal changes in off-farm employment and the ECLU, (2) explore the nonlinear effects of off-farm employment on the ECLU, and (3) determine whether these effects exhibit spatial correlation.

The article is structured as follows: Section 2 examines the mechanisms by which off-farm employment affects ECLU and introduces two key hypotheses. Section 3 presents the study area, methodology, and data sources. Section 4 provides the results. Section 5 discusses the findings and implications, and finally, Section 6 presents the conclusions.

2. Mechanism Analysis and Hypotheses

2.1. The U-Shaped Impact of Off-Farm Employment on the ECLU

Urbanization and industrialization have significantly transformed rural areas, notably through the large-scale migration of rural labor to non-agricultural sectors [40,41]. This shift disrupts traditional labor and capital inputs in cultivated land use, altering the input–output dynamics and influencing the ECLU [42,43,44]. This study explores the mechanisms by which off-farm employment impacts the ECLU from multiple perspectives, offering a comprehensive analysis of its effects.

Labor Dynamics and Agricultural Productivity: Schultz highlighted that farmers in developing countries are “poor but efficient”, optimizing their limited resources for agricultural activities [45]. The nature of traditional agriculture implies that hidden unemployment should not be present in rural areas; rather, farmers optimize scarce resources through efficient allocation in agricultural production [46]. Low productivity or poverty often stems from inadequate capital and technological access. When constrained by these factors, rural labor shifting to off-farm employment can lead to reduced agricultural output [47]. In China, where small-scale farming under the Household Contract Responsibility System prevails, the initial phase of rural labor migration tends to reduce labor-intensive activities essential for optimal land use, adversely affecting the ECLU, particularly in regions dependent on manual labor.

Economic Benefits and Technological Advancements: Over time, off-farm employment can boost household income [48], which may result in increased investments in agricultural technology and practices. These investments can include advanced machinery, upgraded irrigation systems, and more efficient farming techniques [26,49]. With increased capital input, agricultural operations transition from labor-intensive to specialized practices, reducing labor demand. Enhanced financial capacity enables the adoption of sustainable practices that improve the ECLU. Consequently, the economic benefits of off-farm employment can significantly increase the ECLU, especially once a critical threshold of labor migration is surpassed.

Scaling and Resource Allocation: Increased off-farm employment can drive economic growth and promote more effective resource allocation [50]. Income from off-farm employment reinvested in agriculture supports the expansion of operations and enhances land management practices, thereby improving the ECLU [32,34]. Furthermore, access to better resources, training, and innovations resulting from off-farm employment enhances agricultural productivity and sustainability. However, the expansion of operations, particularly when coupled with intensive farming practices, can lead to negative environmental impacts if resources are not properly allocated [31]. Practices such as the overuse of fertilizers and pesticides can degrade soil quality and harm ecosystems.

In summary, while the initial effects of labor migration on the ECLU may be detrimental due to reduced labor availability, the long-term impacts are often positive, as economic benefits and technological advancements contribute to improved land management and sustainability. Based on this analysis, we propose the following hypothesis:

H1. 

The impact of off-farm employment on the ECLU follows a U-shaped nonlinear relationship.

2.2. Spatial Spillover Effects of Off-Farm Employment on the ECLU

As rural labor shifts from agriculture to non-agricultural sectors, socioeconomic changes in one region can significantly impact neighboring areas through various channels [51,52]. Specifically, increased off-farm employment in a given area can boost local incomes [53], thereby stimulating the demand for agricultural products from adjacent regions. This change in demand may prompt neighboring areas to adjust their agricultural practices, which could affect their ECLU as they respond to new market pressures or increased competition [35]. Furthermore, regions with high levels of off-farm employment often accumulate knowledge and technological advancements, which can spill over to neighboring areas [54]. These neighboring regions can adopt more efficient farming practices and technologies through informal networks, trade relationships, and the migration of skilled workers, thereby enhancing their ECLU.

The reallocation of labor and resources due to off-farm employment can also lead to changes in resource availability and allocation in adjacent regions. For instance, if neighboring areas benefit from the migration of skilled agricultural workers or financial resources from regions with high off-farm employment, they may experience improvements in agricultural productivity and eco-efficiency [33]. However, economic shifts resulting from off-farm employment can also trigger environmental changes in neighboring regions. The increased demand for agricultural products may lead to intensified agricultural practices, including the increased use of fertilizers and pesticides [36]. This intensification can adversely affect environmental quality and ECLU in adjacent regions, potentially contributing to soil degradation, water pollution, and reduced biodiversity. Based on this analysis, we propose the following hypothesis:

H2. 

Off-farm employment exerts a spatial spillover effect on the ECLU.

3. Materials and Methods

3.1. Study Area

This study focuses on the North China Plain, also known as the Huang-Huai-Hai Plain, a vast and fertile region in northern China (Figure 1). This extensive area encompasses parts of Hebei, Henan, Anhui, Jiangsu, and Shandong provinces, as well as the municipalities of Beijing and Tianjin. The North China Plain is the most densely populated plain in China, with approximately 254 million residents, 67.79% of whom are rural [55]. As one of China’s three major plains, it features extensive flat terrain that is highly conducive to agriculture, accounting for about one-fifth of the country’s cultivated land. Historically, the region’s agricultural productivity and population density have driven significant socioeconomic development. However, since the late 20th century, there has been a marked shift from agriculture to industry, leading to substantial labor migration from rural to non-agricultural sectors. This transition has impacted both the region’s agricultural practices and its land management strategies. The North China Plain’s representative nature in terms of population density, agricultural significance, and economic transformation makes it an ideal case study for understanding the complex interactions between labor migration and land use. Insights gained from this region can inform similar analyses in other developing regions and countries where rapid socioeconomic changes are influencing agricultural and environmental outcomes.

3.2. Super-Efficiency EBM Model

To measure the ECLU, we employed the super-efficiency EBM model, an advanced method within data envelopment analysis (DEA). Traditional DEA models are generally divided into radial and nonradial types. Radial models, such as the CCR (Charnes, Cooper, and Rhodes) and BCC (Banker, Charnes, and Cooper) models, require proportional adjustment of all inputs, which may not always align with practical scenarios [56]. In contrast, nonradial models, such as the slack-based measure (SBM), address this issue by considering input slack, although they may lose some proportional scaling details [57]. The EBM model, introduced by Tone and Tsutsui in 2010, integrates both radial and nonradial approaches through a hybrid distance function, thereby providing more nuanced efficiency assessments [58]. However, it only produces efficiency scores within the range of 0–1, which can limit the ability to distinguish among decision-making units (DMUs) with scores of 1. The super-efficiency EBM model overcomes this limitation by allowing efficiency scores to exceed 1, thereby enabling a more precise differentiation and ranking of DMUs that are fully efficient. The specific calculation of the super-efficiency EBM model is shown in Formula (1).

$$\begin{array}{l}{\rho }^{\ast }=\min {\displaystyle \frac{\theta -{\epsilon }_x{\displaystyle \sum _{i=1}^m\frac{w_i^{-}{s}_i^{-}}{x_{ik}}}}{\phi +{\epsilon }_y{\displaystyle \sum _{r=1}^q\frac{w_r^{+}{s}_r^{+}}{y_{rk}}+{\epsilon }_b{\displaystyle \sum _{z=1}^p\frac{w_z^{b-}{s}_z^{b-}}{b_{zk}}}}}}\\ s.t.\left\{\begin{array}{l}{\displaystyle \sum _{j=1,j\ne k}^n{\lambda }_j{x}_{ij}+s_i^{-}=\theta x_{ik},i=1,2,\cdots ,m}\\ {\displaystyle \sum _{j=1,j\ne k}^n{\lambda }_j{y}_{rj}-s_r^{+}=\phi y_{rk},r=1,2,\cdots ,q}\\ {\displaystyle \sum _{j=1,j\ne k}^n{\lambda }_j{b}_{zj}+s_z^{b-}=\phi b_{zk},z=1,2,\cdots ,p}\\ {\displaystyle \sum _{j=1}^n{\lambda }_j=1,}{\lambda }_j\ge 0,s_i^{-},s_r^{+},s_z^{b-}\ge 0,0<\theta \le 1,\phi \ge 1\end{array}\right.\end{array}$$

(1)

where ${\rho }^{\ast }$ is the optimal efficiency value; k denotes the $k\mathrm{th}$ DMU to be evaluated; $x_{ik}$, $y_{rk}$, and $b_{zk}$ are the $i\mathrm{th}$ input, $r\mathrm{th}$ desirable output, and $z\mathrm{th}$ undesirable output for the $k\mathrm{th}$ DMU, respectively, and m, q, p represent the number of corresponding inputs and outputs; ${\lambda }_j$ is the linear combination coefficient of DMU; $w$ is the spatial weight and $s$ is the slack variable; φ is the output expansion ratio; $\theta$ represents the radial planning parameter; and $\epsilon$ indicates the importance of the nonradial part in the EBM model.

3.3. Spatial Econometric Model

3.3.1. Global Spatial Autocorrelation Analysis

Before building a spatial econometric model, it is crucial to test for spatial autocorrelation. In this study, global Moran’s I was employed to assess the spatial autocorrelation of the ECLU and off-farm employment. Global Moran’s I evaluates whether the distribution of a variable is spatially clustered, dispersed, or random [59]. The index ranges from −1 to 1, with values near 1 suggesting strong positive spatial autocorrelation, indicating that similar values are geographically clustered, while values near −1 indicate strong negative spatial autocorrelation, meaning that dissimilar values are more spatially dispersed. Values near 0 suggest no spatial autocorrelation, which implies a random distribution. The specific calculation formula is as follows:

$$\mathrm{I}={\displaystyle \frac{n{\displaystyle {\sum }_{i=1}^n{\displaystyle {\sum }_{j=1}^n{w}_{ij}\left(x_i-\overline{x}\right)}}\left(x_j-\overline{x}\right)}{{\displaystyle {\sum }_{i=1}^n{\left(x_i-\overline{x}\right)}^2{\displaystyle {\sum }_{i=1}^n{\displaystyle {\sum }_{j=1}^n{w}_{ij}}}}}}$$

(2)

where n denotes the total number of observations; $x_i$ and $x_j$ are the observed values at locations i and j; $\overline{x}$ represents the mean of the observed values; and $w_{ij}$ represents the spatial weight between locations i and j.

3.3.2. Model Specification

Spatial econometric models were used to analyze the impact of off-farm employment on the ECLU and its associated spatial spillover effects. These models include the spatial lag model (SLM), the spatial error model (SEM), and the spatial Durbin model (SDM). Unlike the traditional ordinary least squares regression, the SLM incorporates the spatial dependence of the dependent variable, while the SEM focuses on spatial autocorrelation within the error terms. The SDM further enhances the SLM by integrating the spatial lags of both the dependent and independent variables, providing a more comprehensive analysis [60]. Additionally, the adaptability of the SDM allows it to be reformulated into other spatial models under specific circumstances, making it the most adaptable of the spatial econometric models. The general form of the SDM is expressed as follows:

$$y_{it}=\rho {\displaystyle {\sum }_{j=1}^n{w}_{ij}{y}_{it}}+x_{it}\beta +{\displaystyle {\sum }_{j=1}^n{w}_{ij}{x}_{jt}}\theta +{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(3)

In Equation (3), when $\theta =0$, the SDM simplifies to the SLM; when $\theta +\rho \beta =0$, the SDM simplifies to the SEM. Based on these settings, the SDM expression constructed in this study is as follows:

$$\begin{array}{l}ECL{U}_{it}={\beta }_1\cdot OFE{R}_{it}+{\beta }_2\cdot {{OFER}_{it}}^2+{\beta }_3\cdot Co{n}_{it}+\rho {\displaystyle {\sum }_{j=1}^n{w}_{ij}\cdot ECL{U}_{it}}+{\alpha }_1{\displaystyle {\sum }_{j=1}^n{w}_{ij}\cdot OFE{R}_{jt}}\\ \hspace{1em}\hspace{1em}\hspace{1em}+{\alpha }_2{\displaystyle {\sum }_{j=1}^n{w}_{ij}·{{OFER}_{jt}}^2}+{\alpha }_3{\displaystyle {\sum }_{j=1}^n{w}_{ij}\cdot Co{n}_{jt}}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}\end{array}$$

(4)

where $ECL{U}_{it}$, $OFE{R}_{it}$ and $Co{n}_{it}$ represent the ECLU, the off-farm employment ratio, and control variables for region i in year t, respectively; $w_{ij}$ denotes the spatial weight between regions i and j; ${\beta }_i$ represents the linear coefficients; ${\alpha }_i$ represents the corresponding spatial spillover coefficients; $\rho$ represents the spatial spillover coefficient for the ECLU; n is the number of samples; ${\mu }_i$ and ${\lambda }_t$ represent the space fixed effect and time fixed effect, respectively; and ${\epsilon }_{it}$ is the error term.

Choosing the right spatial weight matrix is essential when building spatial econometric models. Typical matrices include those based on neighborhood contiguity and distance measures, such as geographical and economic distance matrices. After extensive testing, the geographical distance matrix, which demonstrated the best fit for the model, was selected to analyze the spatial impact of off-farm employment on ECLU.

3.4. Variable Definitions and Descriptions

3.4.1. Explained Variable

The explained variable in this study is the ECLU. Cultivated land use is conceptualized as an all-factor production process that incorporates multiple inputs and outputs [61]. The ECLU, in this context, is defined as the ratio of socioeconomic outputs to resource inputs, while also accounting for the environmental impacts associated with cultivated land use. The objective is to maximize socioeconomic outputs while minimizing resource consumption and environmental effects, thereby promoting sustainable agricultural development.

Based on existing studies, 11 indicators were selected to assess the ECLU, covering inputs, desirable outputs, and undesirable outputs (

Table 1. Index system for measuring the ECLU.

Index	Specific Index	Indicator Description
Input	Land	Crop sown area (103 hm2)
	Labor	Primary industry employees (104 people)
	Machinery	Agricultural machinery total power (104 kW)
	Irrigation	Effective irrigation area (103 hm2)
	Fertilizer	Agricultural fertilizer usage (t)
	Pesticide	Pesticide usage (t)
	Agricultural film	Agricultural film usage (t)
Desirable output	Social	Total grain yield (t)
	Economic	Agriculture output value (104 CNY)
Undesirable output	Pollution emission	Total residues of fertilizers, pesticides, and plastic films (t)
	Carbon emission	Total carbon emissions from cultivated land use (t)

) [19,20,21,62]. The input indicators encompass seven critical dimensions: land, labor, machinery, irrigation, fertilizers, pesticides, and agricultural films, each playing a vital role in the cultivation process. Outputs are categorized into desirable and undesirable categories; desirable outputs include social and economic benefits measured by total grain yield and agricultural output value. A detailed description of these indicators is provided in Table 1.

Undesirable outputs include non-point source pollution and carbon emissions generated during the cultivation process. Non-point source pollution is primarily associated with the residues of fertilizers, pesticides, and agricultural film, quantified by the loss amounts of these substances [20,62]. The specific calculation formula is as follows:

$$P_i={\alpha }_i{A}_i$$

(5)

where $P_i$, ${\alpha }_i$ and $A_i$ are the loss amount, loss coefficient, and consumption amount of fertilizers, pesticides, and agricultural film caused in the cultivation process, respectively. The loss coefficients are sourced from the First National Pollution Source Census: Handbook of Fertilizer Loss Coefficient, Pesticide Loss Coefficient, and Agricultural Film Residue Coefficient [20].

Another indicator of undesirable outputs is carbon emissions, represented as the total emissions produced during the cultivation process [54,63]. These emissions come from various agricultural activities, including tillage, machinery operation, application of fertilizers and pesticides, use of agricultural film, and irrigation. The carbon emissions are calculated using the following formula:

$$E={\displaystyle \sum E_i}={\displaystyle \sum T_i}\cdot {\delta }_i$$

(6)

where $E$ denotes the total carbon emissions associated with cultivated land use, while $T_i$ and ${\delta }_i$ represent the initial quantity and the emission coefficient of each carbon source, respectively. Details on carbon emission coefficients can be found in the studies by Yang et al. [19], Yin et al. [20], and Cao et al. [63].

3.4.2. Core Explanatory Variable

The primary explanatory variable in this study is the off-farm employment ratio (OFER), defined as the proportion of rural laborers engaged in off-farm employment (

Table 2. Descriptive statistics of variables.

Indicator	Variables	Symbol	Definition	Obs.	Mean	S.D.
Explained variable	Eco-efficiency of cultivated land use	ECLU	ECLU calculated by super-efficiency EBM model	960	0.71	0.16
Explanatory variables	Off-farm employment ratio (%)	OFER	Rural off-farm employees/rural general employees	960	48.58	11.59
	Off-farm employment ratio squared (%)	OFER2	(Rural off-farm employees/rural general employees)2	960	24.94	11.49
Control variables	Crop sown area per laborer (hm2/person)	CSA	Crop sown area/rural general employees	960	0.62	0.24
	Multiple cropping index (%)	MCI	Crop sown area/cultivated land area	960	165.23	32.63
	Effective irrigation rate (%)	EIR	Effective irrigated area/cultivated land area	960	47.24	18.60
	Agricultural machinery level (kW/hm2)	AML	Agricultural machinery total power/cultivated land area	960	14.43	5.42
	Per capita disposable income of rural households (104 CNY)	RDI	Per capita disposable income of rural households	960	0.88	0.55
	Financial support for agriculture (%)	FSA	Government agriculture expenditure/government general public expenditure	960	10.14	8.22

). This variable ranges between 0 and 1, with higher values signifying a greater extent of rural labor migration to off-farm activities. To examine the potential nonlinear effects of off-farm employment on ECLU, the squared term of the OFER is also included as an explanatory variable.

3.4.3. Control Variables

Based on existing research, several factors that can influence the ECLU at the prefecture level are included as control variables, including endowments of resources, production conditions, economic status, and government support [64,65,66]. The specific variables considered are the crop sown area per laborer (CSA), multiple cropping index (MCI), effective irrigation rate (EIR), agricultural mechanization level (AML), per capita disposable income of rural households (RDI), and financial support for agriculture (FSA). Detailed definitions of these variables are provided in Table 2. The average RDI is 8800 CNY with a standard deviation of 5500 CNY, reflecting notable regional disparities in economic development. The descriptive statistics for the other variables are also provided in Table 2.

3.5. Data Source

To ensure data availability and consistency, a balanced panel data set of 46 prefecture-level cities and 2 municipalities in the North China Plain from 2001 to 2020 was used for empirical analysis, excluding cities with substantial data gaps. Data were sourced primarily from the China Rural Statistical Yearbook (2002–2021), the China City Statistical Yearbook (2002–2021), the Provincial and Municipal Statistical Yearbooks (2002–2021), and the Statistical Bulletins of National Economic and Social Development (2002–2021) for the relevant regions. Missing data for specific cities and years were supplemented through requests to local governments or by applying average estimation and linear interpolation methods. The descriptive statistics of the main variables suggest that the statistical properties of the sample data are appropriate for econometric modeling (Table 2).

4. Empirical Results

4.1. Probability Distribution of Off-Farm Employment

From 2021 to 2020, the OFER in the North China Plain exhibited a clear trend of temporal expansion and spatial differentiation, although the growth rate gradually slowed. Figure 2 shows the OFER kernel density curve at five different time points, highlighting three notable characteristics. First, the center of the curve consistently shifted to the right, indicating an overall increase in the OFER over time. Second, the relative distance between the kernel density curves gradually decreased, suggesting a slowdown in the growth rate of off-farm employment. Third, from 2001 to 2015, the kernel density curves transitioned from smooth to steep, suggesting a reduction in spatial differences in the OFER. However, between 2015 and 2020, these spatial differences began to increase.

4.2. Spatial–Temporal Pattern of the ECLU

Figure 3 shows the temporal evolution of the ECLU in the North China Plain from 2001 to 2020, highlighting an initial decline followed by a fluctuating upward trend. From 2001 to 2003, the average ECLU fell from 0.623 to 0.512, reflecting a compound annual growth rate (CAGR) of −9.35%. Around 2000, especially after China joined the WTO, there was a rapid shift of rural labor toward off-farm employment, resulting in reduced agricultural labor input [52]. Meanwhile, capital investments in agricultural machinery, irrigation, and fertilizers had not yet risen sufficiently, leading to a low ECLU during this period. From 2003 to 2020, the ECLU exhibited a generally upward but fluctuating trend, with the average increasing from 0.512 to 0.909. Growth was particularly robust from 2003 to 2010, with a CAGR of 4.76%, fueled by increasing household incomes from rural labor migration, which facilitated increased capital investment in land and technological advances in agriculture [49]. However, between 2010 and 2017, the CAGR decreased to 1.88% due to the adverse ecological impacts of the overuse of inputs, such as pesticides, fertilizers, and agricultural films, which restricted improvements in the ECLU [36]. After 2017, the ECLU rebounded with a CAGR of 4.00%, fueled by the detailed implementation of China’s ecological civilization strategy, which intensified government efforts to mitigate environmental pollution in agricultural production, leading to steady improvements in the ECLU. Overall, the structure of the ECLU shifted toward higher efficiency levels over time. In 2001, only about 10% of cities had an efficiency value greater than 0.8, while by 2020, this proportion had increased to 81% (Figure 3).

Figure 4 illustrates the spatiotemporal pattern of the ECLU. Over time, high-value ECLU regions have consistently expanded, while low-value regions have gradually decreased. These high-value regions transitioned from a scattered distribution to a clustered pattern and eventually to a contiguous distribution. In 2001, cities with ECLU values greater than 0.8 were distributed sporadically in northern Henan and southern Hebei. By 2015, these cities had formed clusters, primarily at the borders of Henan and Anhui as well as at Jiangsu and Shandong. By 2020, they had coalesced into a contiguous distribution across the entire study area. During the past 20 years, despite the ongoing increase in rural labor migration (Figure 2), the ECLU in the North China Plain has generally improved, in large part due to improved agricultural production conditions and advances in agricultural technology driven by rapid economic growth.

4.3. Spatial Effect of Off-Farm Employment on the ECLU

4.3.1. Spatial Autocorrelation Test

Global Moran’s I, based on a geographical distance matrix, was used to detect the spatial autocorrelation of the ECLU and the OFER at the prefecture level, with the annual value changes presented in Figure 5. Excluding 2001, Moran’s I values for ECLU were above 0.1 before 2008, gradually decreasing to 0.016 by 2012. Moreover, the Moran’s I values before 2012 were all significant at the 5% level. After 2013, Moran’s I fluctuated around 0 and lost statistical significance, indicating that the ECLU was spatially positively correlated before 2012, but then became randomly distributed. For the OFER, Moran’s I demonstrated a downward trend from 2001 to 2008, with values consistently significant at the 1% level. After 2008, the index fluctuated around zero and was significant at the 10% level in only a few years. Overall, neither the ECLU or the OFER showed a consistent spatial correlation over time. Consequently, spatial econometric models are used for a more detailed analysis.

4.3.2. Model Identification

Before conducting a spatial econometric analysis, it is crucial to select the appropriate model through a series of diagnostic tests. The results of the Lagrange multiplier (LM) tests reveal that all statistics except the robust LM-spatial lag coefficient are significant at the 1% level (

Table 3. Test results of model selection.

Test	Statistic	Value
LM test	Moran’s I	36.443 ***
	LM-spatial error	1156.230 ***
	Robust LM-spatial error	656.069 ***
	LM-spatial lag	502.664 ***
	Robust LM-spatial lag	2.503
LR test	LR-spatial error	77.41 ***
	LR-spatial lag	65.12 ***
Wald test	Wald-spatial error	28.87 ***
	Wald-spatial lag	18.26 **
Hausman test	Hausman	49.09 ***

Note: *** and ** denote significance at the levels of 1% and 5%, respectively.

). This suggests the presence of both spatial error and spatial lag effects, justifying the use of spatial econometric methods to analyze the impact of off-farm employment on the ECLU. Subsequently, the likelihood ratio (LR) test and the Wald test were applied to determine the most appropriate spatial model [67]. Both tests were significant at the 5% level (Table 3), indicating that the SDM does not simplify to the SLM or SEM models, confirming its robustness and suitability for further analysis. Finally, the Hausman test was conducted to decide between fixed- and random-effects models. The results show that the Hausman statistic is significant at the 1% level, leading to the rejection of the random-effects model (Table 3). Therefore, the fixed-effects SDM model was selected to analyze the spatial spillover effects in this study.

4.3.3. Model Estimation

Table 4. Spatial regression results.

Variable	SLM	SEM	SDM
			Main	Wx
OFER	−0.726 ***	−0.715 ***	−0.457 *	−3.708 **
	(0.222)	(0.231)	(0.241)	(1.674)
OFER2	0.723 ***	0.706 ***	0.558 **	4.939 ***
	(0.224)	(0.231)	(0.237)	(1.808)
CSA	0.145 ***	0.143 ***	0.148 ***	−0.024
	(0.018)	(0.017)	(0.018)	(0.199)
MCI	0.021	0.022	0.010	−0.557 ***
	(0.015)	(0.015)	(0.015)	(0.155)
EIR	0.075 ***	0.074 ***	0.077 ***	−0.172
	(0.016)	(0.016)	(0.017)	(0.232)
AML	−0.006 ***	−0.006 ***	−0.004 ***	−0.002
	(0.001)	(0.001)	(0.001)	(0.007)
RDI	−0.004	−0.001	−0.012	−0.351
	(0.030)	(0.031)	(0.032)	(0.299)
FSA	−0.167 ***	−0.167 ***	−0.176 ***	0.375
	(0.042)	(0.042)	(0.044)	(0.520)
ρ	0.577 ***	0.548 ***	0.480 ***
	(0.080)	(0.086)	(0.096)
sigma2_e	0.005 ***	0.005 ***	0.005 ***
	(0.000)	(0.000)	(0.000)
Log-L	1139.2302	1135.5291	1161.0240
Time fixed	Yes	Yes	Yes
Spatial fixed	Yes	Yes	Yes
N	960	960	960
R2	0.006	0.036	0.319

Note: ***, ** and * denote significance at the levels of 1%, 5% and 10%, respectively. Standard errors are in parentheses.

provides the estimation results for the SLM and SEM models using fixed effects alongside the SDM results. The SDM exhibits the highest R² and logarithmic likelihood values, supporting its suitability as the chosen model. Across all three spatial econometric models, the coefficients for the OFER are statistically significant, underscoring the significant influence of off-farm employment on the ECLU. The negative coefficient of the OFER and the positive coefficient of its squared term indicate a U-shaped relationship between off-farm employment and the ECLU. Specifically, during the early stages of rural labor migration to non-agricultural sectors, the ECLU decreases with increasing OFER, but as migration intensifies and surpasses a certain threshold, the ECLU improves with further increases in the OFER. This finding confirms Hypothesis 1. Additionally, the results show that control variables, such as the crop sown area per laborer, the effective irrigation rate, the agricultural mechanization level, and the financial support for agriculture, are significantly associated with the ECLU.

4.3.4. Spatial Spillover Effect

The spatial spillover effects of off-farm employment on the ECLU were examined using the SDM estimation results. The spatial spillover coefficient for the ECLU is 0.48 and is significant at the 1% level (Table 4), indicating that a 1% increase in the ECLU in a prefecture city leads to a 0.48% increase in the ECLU in adjacent cities. This result implies that improving the ECLU in one region has beneficial spillover effects in the surrounding areas. Furthermore, the spatial spillover coefficients for the OFER and its squared term are significant at the levels of 5% and 1%, respectively (Table 4), highlighting that off-farm employment in one region affects the ECLU of neighboring regions, which supports research Hypothesis 2. However, due to feedback effects, the coefficients from the SDM for both the main and spillover effects do not directly capture the marginal impact of explanatory variables on the ECLU [68]. To gain a clearer understanding of the effects of off-farm employment on the ECLU, further analysis using effect decomposition is necessary.

Therefore, the spatial spillover effects on the ECLU were analyzed by breaking them down into direct, indirect, and total effects using partial differentiation. This method enables a comprehensive evaluation of how off-farm employment affects the ECLU by examining its influence both within the local region and in surrounding areas [69]. Specifically, the direct effect coefficient assesses how changes in a variable impact the local ECLU, the indirect effect coefficient gauges the influence on adjacent regions, and the total effect coefficient represents the aggregate of these impacts. The coefficient size, direction, and significance of direct, indirect, and total effects are shown in Figure 6, with specific values detailed in

Table A1. Decomposition results of spatial effects.

Variable	Direct	Indirect	Total
OFER	−0.563 **	−7.784 ***	−8.347 ***
	(0.234)	(3.000)	(2.986)
OFER2	0.691 ***	10.326 ***	11.017 ***
	(0.231)	(3.438)	(3.454)
CSA	0.152 ***	0.146	0.297
	(0.021)	(0.421)	(0.432)
MCI	−0.004	−1.109 ***	−1.113 ***
	(0.018)	(0.391)	(0.399)
EIR	0.070 ***	−0.331	−0.261
	(0.024)	(0.519)	(0.536)
AML	−0.004 ***	−0.007	−0.010
	(0.001)	(0.015)	(0.016)
RDI	−0.022	−0.749	−0.771
	(0.029)	(0.621)	(0.623)
FSA	−0.177 ***	0.496	0.319
	(0.042)	(0.991)	(1.007)

Note: *** and ** denote significance at the levels of 1% and 5%, respectively. Standard errors are in parentheses.

.

The direct effect coefficients for the OFER and its squared term are −0.563 and 0.691, respectively, with statistical significance at the levels of 5% and 1% (Figure 6a). These results indicate a U-shaped relationship between off-farm employment and the local ECLU, with a turning point at an OFER of approximately 40.73%, occurring roughly between 2003 and 2004, based on regional averages. Prior to this turning point, off-farm employment had a negative impact on the ECLU; however, beyond this threshold, its impact turned positive. This finding aligns with the pattern observed in Figure 3, where the average ECLU in the North China Plain began to rise after 2003, following a period of decline. A comparison between the direct effect coefficients and the main effect coefficients for the OFER and its squared term reveals only minor differences (see Table 4 and Table A1). This minimal variance is likely attributed to feedback effects between neighboring regions. Specifically, changes in local off-farm employment can affect the ECLU in adjacent areas, which then feeds back to influence the local ECLU. However, the economic significance of these feedback effects is generally minimal because of their small magnitude.

For indirect effects, the coefficients for the OFER and its squared term are −7.784 and 10.326, respectively, both significant at the 1% level (Figure 6a). This confirms a U-shaped relationship between off-farm employment and the ECLU of adjacent regions. Initially, an increase in off-farm employment negatively impacts the ECLU of neighboring areas. However, this effect eventually turns positive. This shift is influenced by the movement of labor, capital, and other resources through social networks, which subsequently enhances the ECLU in surrounding regions.

The decomposition analysis of the control variables reveals that direct effects are predominantly significant, while indirect effects are generally not (Figure 6b). For example, the direct effect coefficient of the crop sown area per laborer is notably positive and significant at the 1% level. On the contrary, the indirect and total effects are positive but lack statistical significance (Figure 6b). This suggests that expanding the crop sown area per laborer improves large-scale farming operations, thereby improving the local ECLU. Furthermore, the effective irrigation rate has a positive and significant impact on the local ECLU, with a 1% increase in the rate resulting in a 0.07% improvement in the ECLU (Table A1). On the contrary, the level of agricultural mechanization has a significantly negative impact on the local ECLU. This may be due to redundant investment in agricultural machinery under the current smallholder farming model, where excessive mechanization and the resulting pollutant emissions, such as exhaust gases, adversely affect the ecological environment of the cultivated land. Financial support for agriculture also shows a significantly negative impact on the local ECLU, likely due to substantial subsidies for fertilizers, pesticides, and agricultural machinery, which lead to their increased use and a consequent reduction in the ECLU. These findings suggest that the allocation structure of financial support for agriculture needs further optimization.

4.4. Robustness Test

To verify the robustness of our findings, we re-estimated the model using both the inverse distance matrix and the economic geography nested matrix. The results are presented in

Table 5. Results of robustness tests.

Variable	(1)	(2)
OFER	−0.594 **	−0.631 ***
	(0.237)	(0.245)
OFER2	0.539 **	0.735 ***
	(0.234)	(0.238)
CSA	0.145 ***	0.143 ***
	(0.017)	(0.018)
MCI	0.017	0.026 *
	(0.015)	(0.015)
EIR	0.067 ***	0.076 ***
	(0.017)	(0.017)
AML	−0.005 ***	−0.006 ***
	(0.001)	(0.001)
RDI	0.050	−0.146 ***
	(0.034)	(0.042)
FSA	−0.153 ***	−0.167 ***
	(0.043)	(0.044)
Time fixed	Yes	Yes
Spatial fixed	Yes	Yes
R2	0.259	0.317

Note: ***, ** and * denote significance at the levels of 1%, 5% and 10%, respectively. Standard errors are in parentheses.

, columns (1) and (2), respectively. The direction and statistical significance of the coefficients for the OFER and its squared term remain consistent with those observed in previous models, suggesting that the impact of off-farm employment on the ECLU is stable across different spatial weight matrices. Furthermore, the coefficients for the other control variables show minimal variation. These robustness checks support the reliability of our initial empirical results.

5. Discussion

5.1. Major Findings

This study, focusing on the North China Plain, provides new insights into the nonlinear effects of off-farm employment on the ECLU and its spatial spillover effects. Our findings reveal a U-shaped relationship between off-farm employment and the ECLU. Specifically, when the OFER is below 40.73%, off-farm employment negatively impacts the ECLU. However, once the OFER exceeds this threshold, the impact becomes positive. This suggests that in prefecture-level cities with low OFERs, rural labor engaging in non-agricultural activities may initially reduce the ECLU. However, as the OFER increases, this negative effect weakens and eventually turns positive, reflecting a transition from weak to strong substitution effects of capital and technology for rural labor [29,48].

These findings are consistent with those of Zhao et al., who also identified a nonlinear U-shaped impact of off-farm employment on agricultural land-use efficiency using the Driscoll and Kraay standard error fixed-effects model [24]. However, our study extends Zhao’s work by incorporating undesirable outputs into the measurement of cultivated land-use efficiency, offering a more comprehensive analysis of how off-farm employment affects the ECLU. This approach improves our understanding of the impact of rural labor migration on agricultural systems, particularly within the context of ecological security.

Moreover, our study demonstrates that off-farm employment not only impacts the local ECLU but also influences the ECLU in neighboring regions, highlighting the broader regional implications of rural labor migration [35]. Similarly, Wang et al. assessed the impact of rural labor transfer on the ECLU on the provincial scale in China and confirmed significant spatial spillover effects [33]. However, their analysis at the provincial level is limited in its applicability to localized contexts, especially given the large size of China’s provinces and the substantial differences in their socioeconomic development stages. In contrast, our study, conducted at the prefecture-city level, provides a more nuanced and policy-relevant understanding of the relationship between off-farm employment and the ECLU, offering crucial insights for effective policy-making and practical applications in similar developing regions.

5.2. Policy Implications

Based on the findings of our study, several policy recommendations can be drawn up to enhance the ECLU while managing the impacts of off-farm employment.

(1) Promote Balanced Rural Labor Allocation: The U-shaped relationship between off-farm employment and the ECLU indicates that while off-farm employment initially reduces the ECLU, this effect reverses beyond a specific threshold. Policymakers should promote balanced rural labor allocation, ensuring that as off-farm employment increases, sufficient labor remains in agriculture to maintain and improve the ECLU, particularly during the early stages of labor migration. Although many regions in China have passed this phase, these insights are particularly relevant for other developing countries that are beginning to experience rural labor migration [7,8].

(2) Support Sustainable Agricultural Practices and Training: To maximize the positive impact of off-farm employment on the ECLU, policies should encourage reinvesting income from off-farm employment into sustainable agricultural practices. This can be achieved through incentives or subsidies for advanced technologies, such as precision farming tools, improved irrigation systems, and environmentally friendly machinery, that reduce manual labor while improving the ECLU [48,70]. Furthermore, strengthening agricultural training and extension services is crucial. Programs should focus on educating farmers about the integration of advanced technologies and sustainable practices, highlighting the benefits of eco-efficient farming and responsible land management.

(3) Strengthen Regional Coordination and Resource Sharing: The significant spatial spillover effects of off-farm employment on the ECLU highlight the need for enhanced regional coordination and tailored policy interventions. Establishing regional coordination mechanisms can facilitate the sharing of best practices, technological advances, and resources between regions, amplifying the positive spillover effects. Promoting interregional collaboration and joint agricultural initiatives can further enhance these benefits [71,72]. Given the spatial variations in the ECLU, policies should be adapted to regional characteristics: Areas heavily dependent on manual labor may require targeted support to retain agricultural workers, while regions that have surpassed the turning point of off-farm employment should focus on fostering technology adoption and improving the ECLU.

5.3. Limitations and Outlooks

While this study offers valuable insights into the effects of off-farm employment on the ECLU, several limitations highlight areas for further research. First, the analysis is based on panel data from 2001 to 2020 for 48 cities in the North China Plain. This temporal scope may not capture recent trends or the long-term effects of off-farm employment on the ECLU. Future research could extend the timeframe to include more recent data and assess the impact of recent policy changes on the ECLU. Second, while the study identifies a U-shaped relationship between off-farm employment and the ECLU, it does not fully explore the mechanisms driving these threshold effects. Investigating factors such as technological adoption, income reinvestment in agriculture, or changes in land-use practices could offer deeper insights into these nonlinear impacts [29,53]. Future research should examine these microlevel dynamics to better inform targeted policy interventions. Finally, empirical studies are needed that compare the effects of off-farm employment on the ECLU in various agricultural and socioeconomic contexts. Such research would improve the generalizability of the findings and provide more robust policy guidance.

6. Conclusions

Understanding the complex effects of off-farm employment on the ECLU is crucial for optimizing labor and land resource allocation, ensuring sustainable agricultural development, and maintaining food security. By analyzing panel data from 2001 to 2020, we assessed the ECLUs of 48 cities in the North China Plain using the super-efficiency EBM model. Spatial econometric models were then employed to examine the impacts of off-farm employment on the ECLU, including spatial spillover effects.

The results show that between 2001 and 2020, off-farm employment in the North China Plain exhibited significant temporal expansion and spatial variation, with growth rates gradually decelerating and spatial disparities narrowing. The average ECLU initially declined between 2001 and 2003, followed by fluctuating increases, with a notable acceleration in growth after 2017. During the study period, high-ECLU regions evolved from dispersed patterns to clustered and, eventually, continuous distributions. A “U-shaped” relationship was observed between off-farm employment and the ECLU, with a turning point at an OFER of 40.73%, occurring around 2003–2004 based on regional averages. Before this point, off-farm employment negatively impacted the ECLU, but its influence became positive after the threshold was crossed. Additionally, the study found significant spatial spillover effects of off-farm employment on the ECLU in the North China Plain.

In conclusion, this work advances our understanding of how off-farm employment shapes the ECLU, reinforcing the theoretical knowledge of the human–land relationship in the context of labor migration. The insights gained from this study provide valuable guidance for policymakers in China and other developing countries as they seek to balance rural labor migration with sustainable agricultural practices.