Impact of Non-Agricultural Labor Transfer on the Ecological Efficiency of Cultivated Land: Evidence from China

Abstract

The ecological efficiency of cultivated land utilization is closely related to food security and the sustainable development of agriculture. As an important actor in the utilization of cultivated land, the transfer of labor to non-agricultural sectors and its impact on ecological efficiency remain underexplored. Taking China as an example, this study employs push–pull theory, technology factor substitution theory, and land scale economy theory to explore the motivations and mechanisms of non-agricultural labor transfer. An empirical analysis was conducted using provincial panel data from 2011 to 2023. The research methods include the super-efficiency SBM model, fixed effect model, mediating effect model, and threshold effect model. The results are as follows: (1) Non-agricultural labor transfer promotes improvements in the ecological efficiency of cultivated land utilization. A 1% growth in non-agricultural labor transfer is associated with a 0.615% improvement in the ecological efficiency of cultivated land utilization. The impact is especially evident in the main grain-producing areas and northern regions. (2) As a modern agricultural production factor, agricultural machinery plays a mediating role in factor substitution at the farmland stage, accounting for 39% of the effect. (3) The scale of agricultural land operation exhibits a single threshold effect with a threshold value of 1.1577. Against the backdrop of widespread non-agricultural labor transfer, this study provides a reference for further strengthening the utilization of agricultural machinery and promoting large-scale land operations.

1. Introduction

Food security is a critical pillar of national development [1]. As the foundation for ensuring food security, cultivated land is often referred to as “the lifeblood of food production”. However, global cultivated land resources are increasingly degraded, and the per capita cultivated land area continues to decline. According to the flagship report from the Food and Agriculture Organization of the United Nations (FAO), “The State of Land and Water Resources for Food and Agriculture: The System is at Its Limit”, land degradation is accelerating, with anthropogenic soil degradation affecting approximately 34% of agricultural land. Traditional grain production has historically relied on expanding cultivated land areas and increasing inputs such as pesticides and fertilizers to enhance productivity; however, while these inputs have contributed to higher grain yields, they have also exacerbated the problem of agricultural non-point source pollution [2]. Moreover, with rising environmental and economic costs, high-input, high-yield agricultural practices are increasingly unsustainable [3,4]. Enhancing the efficiency of cultivated land has become an important prerequisite for ensuring long-term food security and sustainable development [5,6], while addressing the challenge of population growth and shrinking land availability [7] The ecological efficiency of cultivated land use refers to the optimization of production factors to enhance land productivity while minimizing environmental impact, which is an important indicator of sustainable land management [8,9].

As the cultivated land system possesses both natural and social elements, the transfer of rural labor, as the main workforce in agricultural production, inevitably changes the allocation structure of labor, land, capital, and other resources within the agricultural system [10]. The behavior of rural labor is related to the output of crops [11]. When the number of agricultural laborers is insufficient, this may lead to an increase in the input of capital elements for cultivated land [12].

There are two primary viewpoints on the impact of the non-agricultural transfer of the labor force on the ecological efficiency of cultivated land use. On the positive side, non-agricultural labor transfer promotes the “land scale effect”, enhancing green grain production efficiency [13]. It also facilitates the adoption of agricultural machinery and advanced technologies, leading to more professionalized agricultural production while reducing non-point source pollution and carbon emissions [14,15]. The improvement of technical efficiency is conducive to increasing grain output [16]. In the long term, in China, the non-agricultural transfer of the labor force contributes to the sustainable utilization of cultivated land, the ecological efficiency of utilization, and the transformation of cultivated land utilization functions [17,18,19]. On the negative side, the transfer of agricultural labor can lead to workforce shortages, particularly in mountainous regions where terrain constraints limit the use of large-scale agricultural machinery. In these areas, labor loss is difficult to offset, which increases the risk of farmland abandonment and reduces grain production capacity [20,21].

Existing research has clarified the significance of cultivated land utilization for food security and its close relationship with the labor force. However, the motivations for the transfer of the labor force to the non-agricultural sector, the role of agricultural machinery in the cultivation process, the influence of land operation scale, and the reasons for regional heterogeneity all require further investigation. With 9% of the world’s arable land and nearly one-fifth of the global population, China has played an important role in strengthening global food security by taking concrete action. The proportion of non-agricultural employment in China has risen from 31.3% in 1980 to 77.2% in 2023, indicating a significant trend of the non-agricultural transfer of the labor force. Therefore, using China as a case study to explore the impact of non-agricultural labor transfer on the ecological efficiency of cultivated land utilization is highly constructive.

This study integrates push–pull theory, technology factor substitution theory, and land scale economy theory to explore the motivations and mechanisms of non-agricultural labor transfer. It constructs an ecological efficiency evaluation system for cultivated land utilization that includes factor input, expected output, and unexpected output, using the super-efficiency SBM (Slacks-Based Measure) model. The impact of non-agricultural labor transfer on the ecological efficiency of cultivated land utilization was analyzed using a fixed effect model, and heterogeneity analysis was conducted across selected regions. In addition, the mediating effect model and threshold effect model were used to examine the mediating role of agricultural machinery in land operation and the threshold effect of land scale.

This study is structured as follows: introduction, theoretical analysis, methodology (including model setting, variable selection, and data explanation), empirical results, discussion, and conclusions with recommendations.

2. Theoretical Analysis

2.1. Causes of Non-Agricultural Labor Transfer

Understanding the motivations behind non-agricultural labor transfer is essential for optimizing its impact on the ecological efficiency of cultivated land use. Since the primary industry mainly refers to agriculture, employees in the primary industry are usually defined as agricultural workers, and the proportion of non-agricultural labor force in the total number of workers is selected as the proxy variable for labor force transfer [22]. This study explores the impact of labor flow from the agricultural sector to the non-agricultural industrial sector on the ecological efficiency of cultivated land utilization. Changes in the proportion of non-agricultural labor transfer can intuitively reflect population mobility. A higher proportion indicates a larger scale of non-agricultural labor transfer.

In the process of industrialization and modernization, the transfer of labor from traditional agriculture to non-agricultural industries is a common phenomenon that all countries have faced or will face. The push–pull theory, proposed by American demographer Everett S. Lee in 1966, is a widely used framework for analyzing labor migration dynamics [23]. According to this theory, migration behavior is a rational decision influenced by multiple factors. “Push” factors are negative forces that drive individuals away from their place of origin; “pull” factors are positive incentives that attract them to a new location. Migration outcomes are also shaped by intermediate obstacles that either facilitate or hinder the transfer process.

From the perspective of pull factors, urbanization, industrialization, and higher economic returns in non-agricultural sectors drive labor migration. Lewis (1954) pointed out that the wage difference between agricultural and non-agricultural sectors was the internal cause of the migration of the agricultural labor force [24]. Such migration could not only increase the income of the labor force but also improve the efficiency of agricultural production by reducing the agricultural surplus labor force. Schulz pointed out that farmers in traditional agriculture would respond to market price changes and pursue the Pareto optimal allocation of production factors. When the non-agricultural marginal income is higher than the agricultural marginal income, farmers will prioritize the allocation of household labor to non-agricultural work that can earn more income [25]. Compared to traditional agriculture, non-agricultural industries offer comparative advantages, such as stable employment, better social security, and improved infrastructure. As rural populations achieve basic subsistence, there is a growing demand for higher incomes and improved living standards. Engaging in full-time or mixed non-agricultural employment significantly improves the well-being of rural residents by increasing household income and life satisfaction [26].

On the other hand, the push factors for non-agricultural labor transfer stem from structural agricultural constraints. China’s national conditions, characterized by a large population and limited land resources, lead to incompatibility between people and the land, particularly in small-scale land management. The scarcity of land intensifies this incompatibility, resulting in low per capita agricultural income [27]. The relatively low income in agriculture has led to income disparities. For instance, low productivity in the agricultural sector is the main cause of the income gap between residents in Turkey and those in other countries. In addition, agricultural production is highly susceptible to climate variability [28],causing income instability. Non-agricultural employment provides more reliable income sources, reducing livelihood vulnerability [29].

2.2. Non-Agricultural Labor Transfer and Ecological Efficiency of Cultivated Land Use

As the micro-main body of agricultural production, the agricultural labor force’s transfer behavior to the non-agricultural sector will inevitably change the allocation of factors in the agricultural production system. In the process of labor non-agricultural transfer, land resources will be concentrated in farmers with more productive capacity or new agricultural management subjects, and the labor marginal productivity of the traditional agricultural sector will be greatly improved, thus relieving the inefficient development situation [30]. Non-agricultural labor transfer brings an increase in wage income, which can improve the purchasing power of fertilizer, high-quality seeds, and agricultural machinery, among other resources and equipment, and make up for the negative effect of labor loss on agricultural production. The transfer of surplus labor is conducive to improving average labor productivity and average labor capital levels in agriculture [31].The income effect generated by non-agricultural employment encourages farmers to invest more capital in cultivated land [32]. If labor transfer is accompanied by the adoption of modern farming technologies and improvements in the scale efficiency and technical efficiency of cultivated land use, it can contribute to enhancing the ecological efficiency of cultivated land use [18].In addition, the effects of non-agricultural employment on fertilizer reduction may vary across regions [33]. Based on this discussion, the following hypothesis is proposed:

Hypothesis 1.

Non-agricultural labor transfer promotes improvements in the ecological efficiency of cultivated land use, and this effect exhibits regional heterogeneity.

2.3. The Mediating Effect of the Substitution Role of Agricultural Machinery Elements

Agricultural mechanization is an important means to improve agricultural labor productivity [34,35]. Existing studies have confirmed that non-agricultural labor transfer can, to a certain extent, lead to a decline in the quantity and overall quality of the agricultural labor force, thereby increasing the demand for agricultural machinery and inducing its substitution for labor to mitigate the impact of rural labor shortages on agricultural production [36,37].

Non-agricultural labor transfer increases non-agricultural income, strengthens purchasing power for production inputs such as agricultural machinery, and raises the adoption rate of agricultural technology. The contribution of agricultural machinery to agricultural production mainly operates through technology introduction and labor substitution effects. The machine-cultivated area is an important indicator of technological progress [38]. Improving the utilization rate of agricultural machinery in cultivated land can alleviate labor shortages on the one hand and improve agricultural output rates on the other. Mechanization helps to overcome human and financial constraints at various stages of food production, injecting new momentum into agricultural output. Farmers can use leased machinery or mechanized services to carry out agricultural operations, save labor, and reduce excessive fertilizer use [39,40]. These mechanisms are conducive to improving the ecological efficiency of cultivated land utilization. Based on this discussion, the following hypothesis is proposed:

Hypothesis 2.

The substitution role played by agricultural machinery serves as a mediating mechanism through which non-agricultural labor transfer promotes the ecological efficiency of cultivated land utilization.

2.4. Threshold Effect of Land Management Scale

Agricultural scale management is an important means to improve agricultural production efficiency [41]. According to the theory of economies of scale, a moderate scale of agricultural land management is conducive to the rational use of production resources by agricultural producers and operators in order to achieve the purpose of the scale effect. The expansion of land management scale is conducive to breaking the scale threshold for the adoption of agricultural green production technology [42]. An increase in per capita cultivated land area enhances large-scale intensive management, supporting the development of low-carbon and efficient agricultural models. Scale management and technological progress are fundamental drivers for green and low-carbon development [43,44].As rural labor continues to shift toward non-agricultural sectors, the phenomenon of “human–land separation” becomes prevalent, leading production entities to optimize land resource allocation through models such as “small fields and large fields”, which can improve agricultural scale management.

Labor transfer inevitably influences land allocation structures and management scales [45]. Large-scale operation is conducive to the concentration of scattered planting land, attainment of centralized production, reduction of the application intensity of fertilizers and pesticides, reduction of agricultural pollutant emissions, and the improvement of agricultural productivity and green development level. However, in the process of realizing the large-scale management of land, it will show nonlinear characteristics. During the initial stage of large-scale cultivated land management, challenges arise in consolidating scattered land, and management entities require time to adapt to the new large-scale management mode, which can delay improvements in ecological efficiency. When the scale of land management is small, it is difficult to achieve the scale and specialization of agricultural production due to the limited land area, even if there is non-agricultural labor transfer, and there is little room for improving agricultural production efficiency [46]. However, when the scale of land management reaches a certain degree, business entities are inclined and equipped to invest in advanced agricultural technology and equipment, reducing resource waste and environmental pollution, thereby leading to a more substantial improvement in the ecological efficiency of cultivated land use. Based on this discussion, the following hypothesis is proposed:

Hypothesis 3.

The scale of land management exerts a threshold effect on the relationship between non-agricultural labor transfer and the enhancement of the ecological efficiency of cultivated land use.

Based on the above analysis, we developed a theoretical model illustrating the impact mechanism of non-agricultural labor transfer on the ecological efficiency of cultivated land use. Figure 1 presents the theoretical framework diagram.

3. Method

3.1. Model Setting

(1)

Super-Efficiency SBM model

The DEA (Data Envelopment Analysis) method is a commonly used method for measuring efficiency [47]. The super-efficiency SBM model is an improved model of DEA, which is used to evaluate the efficiency of decision-making units (DMUs) and was proposed by Tone [48]. The super-efficiency SBM model addresses the problem of slack variables and overcomes the limitation that the efficiency value in traditional models can only fall within the (0, 1) range [49]. It is now widely used in efficiency measurement [50,51,52].This article will adopt Dearun Tools v3.2.0.5 for the calculation of efficiency.

$${\rho }^{*}=min{\displaystyle \frac{\frac{1}{m}{\sum }_{i=1}^m\frac{\overline{x_i}}{x_{i0}}}{\frac{1}{S_1+S_2}\left({\sum }_{r=1}^{S_1}\frac{{\overline{y}}_r^g}{y_{r0}^g}+{\sum }_{r=1}^{S_2}\frac{{\overline{y}}_r^b}{y_{r0}^b}\right)}}$$

$$s.t.\left\{\begin{array}{c}\overline{x}\ge {\sum }_{j=1,\ne k}^n{\theta }_j{x}_i\\ {\overline{y}}^g\le {\sum }_{j=1,\ne k}^n{\theta }_j{y}_i^g\\ \begin{array}{c}{\overline{y}}^b\ge {\sum }_{j=1,\ne k}^n{\theta }_j{y}_i^b\\ \overline{x}\ge x_0,{\overline{y}}^g\le y_0^g,{\overline{y}}^b\ge y_0^b,{\overline{y}}^g\ge 0,\theta \ge 0\end{array}\end{array}\right.$$

(1)

where $n$ represents the number of DMUs. Each DMU consists of input, expected output ($S_1$), and unexpected output ($S_2$). $x$ denotes the elements in the input matrix, $y^g$ refers to the elements in the expected output matrix, and $y^b$ represents the elements in the unexpected output. The objective function value (${\rho }^{*}$) is the efficiency score of the DMU, and its value may exceed 1.

(2)

Benchmark Regression Model

The fixed effect model is a commonly used econometric approach for panel data analysis. It can control the individual and temporal effects of the model during the regression process, allowing for a more accurate estimation of the influence of explanatory variables on the dependent variable. This model is widely applied in panel data research [53,54]. In order to analyze the impact of non-agricultural labor transfer on the ecological efficiency of cultivated land use, a benchmark regression analysis was conducted using a fixed effect model:

$$Y_{it}=a_0+a_1{X}_{it}+a_2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(2)

where $Y_{it}$ represents the ecological efficiency of cultivated land use in province $i$ at year $t$, $X_{it}$ denotes the non-agricultural labor transfer, Controls includes a set of control variables, ${\mu }_i$ and ${\lambda }_t$ are fixed effects across time and individual dimensions, and ${\epsilon }_{it}$ is the random disturbance term.

(3)

Mediation Effect Model

The mediating effect model is designed to examine how the independent variable influences the dependent variable through the mediating variable. A three-step method was used to test the mediating effect [55], where $Z_{it}$ represents the mediating variable (the substitution of technical elements of agricultural machinery is measured by the rate of mechanical tillage). The models are specified as follows:

$$Z_{it}=b_0+b_1{X}_{it}+b_2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(3)

$$Y_{it}=c_0+c^{\prime }{X}_{it}+b_3{Z}_{it}+c_3{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(4)

(4)

Threshold Effect Model

The threshold effect model is employed to study nonlinear relationships between variables, where the influence of the independent variable on the dependent variable may differ across different threshold intervals [56]. The feature of the panel threshold model lies in that it can automatically identify the jump points based on the given threshold variables, avoiding subjective setting biases. In order to evaluate the threshold effect of land management scale on the ecological efficiency of cultivated land use in the context of non-agricultural labor transfer, this study adopts Hansen’s threshold effect model [57]. The regression function for a single threshold is specified as follows:

$$Y_{it}=d_0+{\beta }_1{X}_{it}I(L_{it}\le \gamma )+{\beta }_2{X}_{it}I(L_{it}>\gamma )+\sigma {\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(5)

where $L_{it}$ represents the threshold variable, r is the threshold value, and $I(.)$ denotes the indicator function. The influence coefficients for the two threshold intervals are ${\beta }_1$ and ${\beta }_2$.

3.2. Variable Selection and Explanation

(1)

Dependent Variable: Ecological Efficiency of Cultivated Land Use

The ecological efficiency of cultivated land utilization is generally defined as maximizing the expected output (agricultural output value and total grain output) and minimizing the unexpected output (carbon emissions) under a given level of production inputs. An ecological efficiency index system for cultivated land use is constructed based on three dimensions: input, expected output, and unexpected output [58,59,60]. The super-efficiency SBM model is used to calculate ecological efficiency, with details on variable selection and measurement provided in Figure 2.

Currently, the primary indicators used to quantify the negative externalities of cultivated land use mainly include non-point source pollution and carbon emissions from agricultural production. Compared to non-point source pollution, carbon emissions cover a broader scope and serve as more representative indicators [61]. In this study, total carbon emissions from sources such as fertilizers, pesticides, agricultural film, diesel fuel, plowing, and irrigation are considered the unexpected output. Carbon emission is the sum of the product of main carbon sources and the carbon emission coefficient in cultivated land utilization. The formula is as follows:

$$\mathrm{E}=\mathsf{\Sigma }{\mathrm{E}}_{\mathrm{i}}=\mathsf{\Sigma }{\mathrm{T}}_{\mathrm{i}}\text{×}{\mathsf{\delta }}_{\mathrm{i}}$$

(6)

where E represents the total carbon emissions from cultivated land use, Ei represents the carbon emissions of various carbon sources, and Ti and δi represent the carbon sources.

The corresponding carbon emission coefficients are 4.934 kg/kg, 0.896 kg/kg, 0.593 kg/kg, 5.18 kg/kg, 312.60 kg/km², and 25 kg/hm² [62,63].

(2)

Core Explanatory Variable: Non-Agricultural Labor Transfer

The ratio of the employed population in the secondary and tertiary industries to the total employed population is taken as the variable for non-agricultural labor transfer [22].

(3)

Mediation Variable: Machine Tillage Rate

In previous studies, the total power of agricultural machinery is often used as a substitute variable for the use of agricultural machinery, but the total power of agricultural machinery covers the sum of the power of a variety of agricultural machinery, and it is difficult to confirm the influence of machinery in each link.

The machine tillage rate is quantified as the ratio of machine tillage area to total crop sown area [64]. The machine tillage rate can directly reflect the application degree of machinery in cultivated land operation, directly reflect the influence of agricultural mechanization on cultivated land, and more accurately reflect the impact of mechanical operations on the ecological efficiency of cultivated land.

(4)

Threshold Variable: Land Management Scale

The scale of cultivated land management reflects the allocation of land resources. Small-scale farmers may be more inclined to use traditional and extensive agricultural methods of production. After the transfer of the labor force, land abandonment or semi-abandonment may occur. A larger cultivated land operation scale may mean more concentrated land use, which is conducive to the realization of large-scale and specialized agricultural production. Large-scale operation is often more effective in attracting capital and technology investment, which has various impacts on ecological efficiency. The land management scale is measured as the ratio of crop sown area to rural population.

(5)

Control Variables

The ecological efficiency of cultivated land use is affected by various factors, including the local economy, cultivated land conditions, natural environment, and household expenditures [65,66].Each regional economic development level is measured by per capita GDP. The disaster rate is measured as the effected area relative to the planted area of crops. Effective irrigation rate is calculated as the ratio of effective irrigated area to crop sown area. Fertilizer use intensity is measured by fertilizer conversion amount per unit of crop sown area. Science and technology expenditure intensity is determined by the proportion of expenditure on science and technology to local general financial budget expenditure. Agricultural labor productivity is measured as the gross agricultural output value per unit of agricultural employment. Cultural and educational expenditure intensity is assessed by the proportion of rural residents’ expenditure on education, culture, and entertainment relative to total household consumption expenditure. Agricultural structure is determined using the ratio of gross agricultural output value to the total output value of agriculture, forestry, animal husbandry, and fishery.

3.3. Data Sources and Descriptive Analysis

The data used in this study are sourced from the “China Statistical Yearbook”, “China Rural Statistical Yearbook”, and the EPS database. Due to the absence of provincial-level data on mechanical tillage area for 2011–2012, the mean value was applied to address the missing data. This approach is appropriate because mechanical tillage areas in agricultural production typically exhibit stability and regularity, without experiencing sharp fluctuations or anomalies. As a simple and intuitive statistical technique, the use of mean value helps maintain the continuity of time series data and prevents disruptions in the analysis from missing data. The descriptive statistical results are presented in

Table 1. Descriptive statistics of variables.

Variable	Variable Name	Observations	Mean	Standard Deviation
Dependent Variable	Ecological Efficiency of Cultivated Land Use	403	0.589	0.211
Explanatory Variable	Non-Agricultural Transfer of Labor Force	403	0.696	0.139
Mediation Variable	Machine Tillage Rate	403	0.713	0.186
Threshold Variable	Land Management Scale	403	3.063	2.154
Control Variable	Fertilizer Use Intensity	403	0.350	0.129
	Intensity of Science and Technology Expenditure	403	0.022	0.016
	Disaster Rate	403	0.129	0.109
	Effective Irrigation Rate	403	0.452	0.198
	Agricultural Labor Productivity	403	6.284	3.456
	GDP per Capita	403	6.226	3.177
	Agricultural Structure	403	0.529	0.087
	Intensity of Culture and Education Expenditure	403	0.095	0.026

. The standard deviation of all variables from 2011 to 2023 is less than the mean, indicating that the selected data exhibit good stability.

4. Results

4.1. Temporal Characteristics of Non-Agricultural Labor Transfer and Ecological Efficiency of Cultivated Land Use

With industrialization and urbanization accelerating in China, rural labor is increasingly shifting to non-agricultural industries [67,68].The concept of non-agricultural labor transfer refers to the movement of labor from the agricultural sector to non-agricultural sectors, representing a reallocation of labor resources. In this study, the proportion of employment in the secondary and tertiary industries relative to total employment is used as a proxy for non-agricultural labor transfer. The ecological efficiency of cultivated land use from 2011 to 2023 was calculated using the super-efficiency SBM model, and the temporal evolution trends of both indicators are presented in Figure 3.

In 2011, the degree of non-agricultural labor transfer was 0.65, rising to 0.77 in 2023. Despite a temporary decline in 2022, the upward trend continued in 2023, reflecting an overall increase in non-agricultural labor transfer. In terms of cultivated land use efficiency, efficiency values were significantly higher when agricultural carbon emissions were excluded than when they were included; however, both measures exhibited an increasing trend. The ecological efficiency of cultivated land use accounts for the social, economic, and environmental dimensions of land use, aligning more closely with the principles of sustainable agricultural development. The findings suggest that the sustainable utilization of cultivated land has improved over time.

4.2. Benchmark Regression Results

Before conducting the baseline regression, a collinearity test was performed using the variance inflation factor (VIF). The mean VIF was 2.31 (range: 1.24–4.86), well below the threshold of 10, indicating no collinearity issues.

Table 2. Benchmark regression results of non-agricultural employment on land use efficiency.

Variable Name	Ecological Efficiency of Cultivated Land Use
	(1)	(2)	(3)	(4)
Non-Agricultural Transfer of Labor Force	0.674 ***	0.615 ***	1.663 ***	0.498 ***
	(3.47)	(3.42)	(15.47)	(3.72)
Fertilizer Use Intensity		−0.667 ***		−0.799 ***
		(−3.99)		(−7.03)
Intensity of Science and Technology Investment		−1.767 **		−2.628 ***
		(−2.08)		(−3.16)
Disaster Rate		−0.064		−0.139 **
		(−1.30)		(−2.54)
Effective Irrigation Rate		0.273 **		0.038
		(2.14)		(0.45)
Agricultural Labor Productivity		0.027 ***		0.041 ***
		(5.38)		(12.96)
Culture and Education Expenditure		−0.528		−0.358
		(−1.60)		(−1.34)
GDP per Capita		−0.019 ***		0.000
		(−2.92)		(0.01)
Agricultural Structure		0.426 ***		0.504 ***
		(2.78)		(3.58)
Constant Term	0.120	0.090	−0.568 ***	0.090
	(0.90)	(0.62)	(−7.16)	(0.72)
N	403	403	403	403
R2	0.856	0.884
F	12.043	10.122

Note: t statistics in parentheses; ** p < 0.05, *** p < 0.01.

shows the baseline regression results of non-agricultural labor transfer on the ecological efficiency of cultivated land use. Columns (1) and (2) display the results of the bidirectional fixed effects model, where the estimated coefficients without control variables and with control variables are 0.674 and 0.615, respectively. The coefficients in the two cases are positive and significant at the 1% level, and, as more control variables are added, the R² value increases, improving the model’s fit. These results suggest that non-agricultural labor transfer has a significant positive impact on the ecological efficiency of cultivated land use. Columns (3) and (4) show the results of the random effects model, where the coefficients remain positive and significant at the 1% level. However, after conducting the Hausman test, the fixed effects model was determined to be more appropriate.

Overall, the baseline regression results confirm that non-agricultural labor transfer contributes to the improvement of the ecological efficiency of cultivated land use, thereby validating Hypothesis 1.

4.3. Robustness and Endogeneity Tests

(1)

Robustness Test

To ensure the robustness of the results, regression analysis was conducted using four approaches: a 1% truncation of all variables, replacement of core explanatory variables, exclusion of four municipalities, and adjustment of the sample period. To avoid the influence of extreme values on the results, the 1% extreme values at both ends of all variables are eliminated in this paper. The urbanization rate, which reflects the scale of non-agricultural labor transfer in agriculture, was used as a substitute variable for non-agricultural labor transfer. Excluding municipalities directly under the central government helps account for economic and policy differences between special cities and regular provinces. Due to significant fluctuations in the non-agricultural transfer of the labor force in 2022, the sample period was adjusted from 2011 to 2021 to eliminate the impact of this irregular change. The regression results for these four robustness tests are presented in

Table 3. Results of the robustness test.

Variable Name	Ecological Efficiency of Cultivated Land Use
	(1)	(2)	(3)	(4)
Non-Agricultural Transfer of Labor Force	0.585 ***
	(3.19)
Curban Rate		0.775 ***
		(2.67)
Non-Agricultural Transfer of Labor Force			0.612 ***	0.417 **
			(3.41)	(2.28)
Constant Term	0.099	−0.032	0.011	0.166
	(0.66)	(−0.15)	(0.08)	(1.13)
Control Variables/Time/Province	YES	YES	YES	YES
N	403	403	351	341
R2	0.881	0.884	0.896	0.877
F	9.071	8.251	8.557	7.666

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

(1)–(4).

The findings indicate that the impact of non-agricultural labor transfer on the ecological efficiency of cultivated land use remains positive and significant at the 1% level, confirming the robustness of the benchmark regression results.

(2)

Endogeneity Test

Considering the potential endogeneity issues arising from omitted variables and potential causal relationships within the model, the approach of adding control variables and instrumental variables was adopted for testing. First, two control variables, rural electricity consumption, and agricultural planting structure were added to Table 2 (2) of the model. The results are presented in

Table 4. Results of endogeneity tests.

Variable Name	Ecological Efficiency of Cultivated Land Use	1st	2nd
	(1)	(2)	(3)
Non-Agricultural Transfer of Labor Force	0.639 ***		0.215 **
	(3.36)		(2.22)
Non-Agricultural Transfer of Labor lag two periods		0.888 ***
		(44.38)
Constant Term	0.402	0.097 ***	0.442 ***
	(1.52)	(5.04)	(4.30)
Control Variables/Time/Province	YES	YES	YES
N	403	341	341
R2	0.887	0.970	0.563
F	8.778	2769.092
Underidentification Test Kleibergen–Paap rk LM Statistic		134.134
Weak Identification Test Cragg–Donald Wald F Statistic		2144.421
Kleibergen–Paap rk Wald F Statistic		1969.443
Hansen J Statistic Overidentification Test of all Instruments		0.000

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

(1). The second approach involved selecting the lagged second period of non-agricultural labor transfer as the instrumental variable to further verify the robustness of the results, as shown in Table 4 (2) and (3). The degree of non-agricultural labor transfer lagging by two periods does not directly impact the ecological efficiency of cultivated land use in the current year; however, it affects the willingness of non-agricultural labor transfer in the current year, thereby satisfying the requirements for instrumental variables. The two-stage least squares method (IV_2SLS) was applied for retesting, and the robustness and endogeneity test results are shown in Table 4. In the first-stage regression results, instrumental variables exhibit a significant positive correlation with endogenous variables at the 1% level, indicating a strong correlation. The LM value is 134.134, significant at the 1% level, confirming identifiability. The Cragg–Donald Wald F value (2144.42) far exceeds the critical threshold of 16.38, confirming that weak instrument concerns are negligible.

After conducting robustness and endogeneity tests, the coefficients remained positive and statistically significant. These results indicate that the regression results are robust and reliable.

4.4. Heterogeneity Analysis

Due to differences in the endowment and functional roles of cultivated land resources in different regions, variations may exist in the ecological efficiency of arable land use resulting from non-agricultural labor transfer. In order to examine these differences, this analysis categorizes the regions into major grain-producing areas, and non-major grain-producing areas, as well as northern and southern regions. A grouped regression approach is applied to assess the heterogeneity in the ecological efficiency of arable land use associated with non-agricultural labor transfer. The results are presented in

Table 5. Analysis of regional heterogeneity.

Variable Name	Ecological Efficiency of Cultivated Land Use
	(1) Major Grain-Producing Areas	(2) Non-Major Grain-Producing areas	(3) North	(4) South
Non-Agricultural Transfer of Labor Force	0.874 **	0.494 **	0.946 ***	0.133
	(2.48)	(2.39)	(3.40)	(0.58)
Constant Term	−0.018	0.179	−0.001	0.334
	(−0.05)	(0.99)	(−0.00)	(1.26)
Control Variables/Time/Province	YES	YES	YES	YES
N	169	234	195	208
R2	0.915	0.905	0.887	0.900
F	7.494	11.126	9.772	4.563

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

.

From Table 5 (1) to (2), the results indicate that non-agricultural labor transfer plays a significant role in enhancing ecological efficiency in both major and non-major grain-producing areas, with a stronger effect observed in major grain-producing areas. A one-percentage-point increase in non-agricultural labor transfer correlates with a 0.874 percentage-point increase in the ecological efficiency of cultivated land use.

One possible explanation is that major grain-producing areas are essential for national food security, possessing more abundant land resources and receiving greater government support in infrastructure development and agricultural subsidies. Following labor transfer to non-agricultural sectors, large-scale growers and new agricultural enterprises contribute to efficiency gains through large-scale operations. In non-grain-producing areas, fewer laborers are engaged in grain production, and a greater diversity of crops is cultivated, resulting in a less pronounced effect in terms of non-agricultural labor transfer compared to grain-producing areas.

Table 5 (3) and (4) indicate that non-agricultural labor transfer has a stronger positive impact on ecological efficiency in the northern region. The northern region’s dry land and plains facilitate centralized, contiguous farming, making large-scale mechanization more efficient. These factors help alleviate labor shortages and improve ecological efficiency. In contrast, the southern region is characterized by paddy fields, complex terrain, and highly fragmented arable land. Labor transfer to non-agricultural sectors makes land consolidation challenging, imposing significant constraints on large-scale management and intensive land use.

The potential for technological substitution through large-scale agricultural machinery is not fully realized, and, in some areas, farmland abandonment may occur, reducing the benefits of labor transfer.

These findings highlight regional disparities in the impact of non-agricultural labor transfer on the ecological efficiency of arable land use. The impact of non-agricultural transfer of labor force on the ecological efficiency of cultivated land utilization is influenced by factors such as land resource endowment, local topography, and agricultural infrastructure.

4.5. Mechanism Verification

To further examine the mechanism through which non-agricultural labor transfer affects the ecological efficiency of cultivated land use, the machine tillage rate variable is introduced and analyzed using a mediation effect model. First, a three-step mediation effect model is applied, and the results are presented in

Table 6. Test of mediating effect.

Variable Name	Ecological Efficiency of Cultivated Land Use	Machine Tillage Rate		Ecological Efficiency of Cultivated Land Use
	(1)	(2)		(3)
Non-Agricultural Transfer of Labor Force	0.615 ***	1.012 ***		0.375 **
	(3.42)	(5.77)		(2.28)
Machine Tillage Rate				0.237 ***
				(3.86)
Constant Term	0.090	0.070		0.074
	(0.62)	(0.45)		(0.55)
Control Variables/Time/Province	YES	YES		YES
N	403	403		403
R2	0.884	0.870		0.889
F	10.122	9.925		12.145
Sobel Test
Coefficient	Est	Std_err	z	P > |z|
a_coefficient	1.012	0.175	5.773	0.000
b_coefficient	0.237	0.062	3.857	0.000
Indirect_effect_a*b	0.240	0.075	3.207	0.001
Direct_effect_c’	0.375	0.165	2.275	0.023
Total_effect_c	0.615	0.180	3.415	0.001

Proportion of total effect that is mediated: 0.390. ** p < 0.05, *** p < 0.01.

.

The coefficient for non-agricultural labor transfer in Table 6 (1) is 0.615, which is significantly positive at the 1% level, indicating that non-agricultural labor transfer contributes to the improvement of ecological efficiency in cultivated land use. The regression coefficient for non-agricultural labor transfer on machine tillage rate in Table 6 (2) is 1.012, which is positive and significant at the 1% level, suggesting that non-agricultural labor transfer increases the utilization rate of agricultural machinery in farmland operations. The regression coefficients for machine tillage rate and non-agricultural labor transfer on the ecological efficiency of cultivated land use in Table 6 (3) are 0.237 and 0.375, respectively, which is statistically significant at the 1% and 5% levels, respectively.

These results indicate that a higher machine tillage rate enhances the ecological efficiency of cultivated land use. Machine tillage rate serves as a partial mediator in the relationship between non-agricultural labor transfer and ecological efficiency. The Sobel test confirms that the machine tillage rate mediates 39% of the total effect of non-agricultural labor transfer on ecological efficiency.

The mediating effect was gradually tested by the three-step method and both Sobel tests were passed. These results demonstrate the robustness of the mediating effect, thus validating Hypothesis 2 of this paper.

4.6. Threshold Effect Analysis

(1)

Threshold Effect Test Results

This study employs land management scale as the threshold variable. Before estimating the threshold value, the presence of a threshold effect is tested using the bootstrap method. The stata17 statistical software combined with the bootstrap method was used to repeat the sampling 300 times to obtain the p-values corresponding to the test statistics and to determine the number of thresholds. First, the existence of a single threshold effect was analyzed. If confirmed, double-threshold and multi-threshold tests were conducted (see

Table 7. Test results of the threshold effect.

Variable Name	Threshold Number	Threshold Value	F	P	BS	Critical Value
						10%	5%	1%
Land Management Scale	Single Threshold	1.1577	41.63	0.01	300	25.1062	29.4966	39.1286
	Double Threshold	2.0667	7.79	0.7633	300	28.1740	42.1125	61.3274

).

The results indicate that the double threshold test was not significant, while a single threshold was significant at the 5% level. This suggests that a single threshold effect exists in the relationship between non-agricultural labor transfer and the ecological efficiency of cultivated land use. The estimated threshold value (1.1577) falls within the 95% confidence interval (1.1294, 1.1752), confirming its reliability.

(2)

Regression Results of the Threshold Effect Model

Using the threshold effect model, a single-threshold regression analysis was conducted, and the results are presented in

Table 8. Regression results of threshold effect.

Variable Name	Ecological Efficiency of Cultivated Land Utilization
Non-Agricultural Transfer of Labor Force * I (Land Management Scale ≤ r)	0.428 *
	(1.72)
Non-Agricultural Transfer of Labor Force * I (Land Management Scale > r)	0.615 **
	(2.63)
Constant Term	−0.047
	(−0.31)
Control Variables/Time/Province	YES
N	403
R2	0.792

* p < 0.1, ** p < 0.05.

. When the land management scale is ≤1.1577, the regression coefficient is 0.428, which is significant at the 10% level. Once the threshold is exceeded, the promotion effect strengthens, with the coefficient increasing to 0.615, significant at the 5% level. This suggests that the influence of cultivated land scale management gradually expands.

Several factors may explain this pattern. The non-agricultural transfer of labor creates conditions for large-scale land management; however, in the early stages of large-scale land management, small-scale farmers may struggle to offset labor shortages, increasing the risk of land abandonment and inefficient use. These findings suggest a need for policies supporting smallholder adaptation, such as land transfer or the use of agricultural machinery services to prevent land abandonment. As large-scale land management matures, land resources released through non-agricultural labor transfer can be absorbed by large farms and emerging agricultural enterprises. With greater access to capital, technology, and management expertise, large-scale operators can continuously improve cultivated land utilization, optimize resource allocation, and enhance the ecological efficiency of cultivated land use. Thereby validating Hypothesis 3.

Figure 4 presents the likelihood ratio function of the threshold model. The likelihood ratio function graph can help to clarify the estimation of the threshold value and the construction process of the confidence interval. The threshold value for large-scale management falls within the 95% confidence interval, represented by the area below the critical value of 7.35 (dashed line). This confirms the reliability and validity of the estimated threshold value.

5. Discussion

5.1. Overview and Comparison

Food security is a global concern [69,70,71]. China has the national condition of a large population and limited land. This study integrates the push–pull theory to examine the driving forces behind non-agricultural labor transfer, providing a deeper understanding of this phenomenon. The ecological efficiency of cultivated land utilization is significantly higher when agricultural carbon emissions are excluded from the assessment. The ecological efficiency of cultivated land utilization, considering non-expected outputs, covers the issue of carbon emissions generated during the process of cultivated land utilization and has a higher degree of applicability to the actual situation of cultivated land utilization, consistent with previous research conclusions [72].

In the long term, in China, the non-agricultural transfer of the labor force contributes to the sustainable utilization of cultivated land. This conclusion is consistent with earlier research, indicating a “U-shaped” relationship between non-agricultural employment and the ecological efficiency of cultivated land use. Non-farm employment has a positive effect on the ecological efficiency of cultivated land use when non-farm employment exceeds 40.73% [73]. To ensure the robustness of the results, various tests were carried out, including the tail-shortening of data, substitution of core explanatory variables, exclusion of four municipalities, adjustment of the sample period, and the use of instrumental variables. These robustness and endogeneity tests further confirm the reliability of the study’s conclusions. The heterogeneity analysis across the selected regions reveals that non-agricultural labor transfer exerts a more pronounced promoting effect on the ecological efficiency of cultivated land use in major grain-producing areas and northern areas dominated by plains. This finding offers valuable insights for developing region-specific policies. Cai et al. (2024) [74] similarly concluded that the importance of non-agricultural labor transfer was significant, noting that a 1% increase in agricultural labor transfer raised the resilience level of grain production in major grain-producing areas by an average of 1.7120%, with the effect being more significant in major grain-producing areas in northern China.

Agricultural machinery serves as a crucial substitute for labor shortages, playing a vital role in enhancing agricultural productivity [75]. Expanding large-scale operations and advancing agricultural mechanization are key strategies for reducing the intensity of fertilizer and pesticide use in agricultural land transfers [76]. While previous studies have largely focused on total agricultural machinery power, this research specifically investigates which agricultural processes benefit from mechanization. The measurement of the machine tillage rate provides a more precise examination of the role of machinery in cultivated land use and its close relationship with ecological efficiency. This study also highlights the critical role of large-scale agricultural land management. Large-scale management not only contribute to carbon reduction [77], but also exhibit threshold effects. Land scale is a key constraint in the effective use of agricultural machinery, as larger operational areas enable farmers to increase both agricultural output and income [78]. Large-scale management not only significantly improves the efficiency of agricultural production; it is also the most effective way to achieve agricultural modernization.

5.2. Limitations and Future Directions

Despite efforts to analyze the impact of non-agricultural labor transfer on the ecological efficiency of cultivated land use through theoretical and empirical approaches, certain limitations remain. In this paper, the empirical test is mainly carried out through macro panel data, and the investigation of different types of non-agricultural labor transfer is insufficient. For example, the various impacts of the transfer of agricultural labor force to different non-agricultural industries, different regions, and different seasons on the ecological efficiency of cultivated land use need to be further analyzed. The current sample mainly uses China as an example and lacks comparisons with other countries.

To address these limitations, future research will conduct farmer surveys to classify different types of non-agricultural labor transfer and explore their impacts on the ecological efficiency of cultivated land use. In addition, expanding the scope of the literature review and empirical analysis to incorporate comparative studies from other countries will enhance the broader applicability of the findings.

6. Conclusions and Recommendations

6.1. Conclusions

This study investigated the motivations behind non-agricultural labor transfer and its impact on the ecological efficiency of cultivated land use. Using provincial panel data from 2011 to 2023, empirical analyses were conducted through the SBM model, fixed effects model, mediation effect model, and threshold effect model, leading to the following main conclusions:

(1) In 2011, the degree of non-agricultural labor transfer was 0.65, rising to 0.77 in 2023. Overall, the labor force still shows a trend of shifting from the agricultural sector to the non-agricultural sector. The efficiency measured without accounting for the unexpected output of agricultural carbon emissions was significantly higher than the ecological efficiency of cropland use when such outputs were considered. The observed upward trend in the ecological efficiency of cultivated land use indicates that China has achieved notable progress in the ecological protection of cultivated land. (2) Non-agricultural labor transfer contributes to improvements in the ecological efficiency of cultivated land use, although regional differences are evident. While keeping other factors affecting the ecological efficiency of cultivated land utilization unchanged, for every one-unit increase in the degree of non-agricultural transfer of the labor force, the ecological efficiency of cultivated land utilization will increase by an average of 0.615 units. This indicates that there is a positive relationship between non-agricultural transfer of the labor force and the ecological efficiency of cultivated land utilization. This underscores the necessity of guiding the transfer of surplus agricultural labor to non-agricultural industries in a balanced manner. (3) The use of agricultural machinery in cultivated land helps mitigate labor shortages, with the machine tillage rate serving as a mediating factor, accounting for 39% of the effect on ecological efficiency. This conclusion highlights the need to prioritize the popularization and use of agricultural mechanization against the backdrop of non-agricultural labor transfer. (4) Non-agricultural labor transfer supports large-scale land management, which in turn enhances the ecological efficiency of cultivated land use. The effect follows a single-threshold pattern, becoming more pronounced once the threshold value is surpassed. This conclusion reinforces the importance of large-scale land management.

6.2. Recommendations

In order to enhance the ecological efficiency of cultivated land use, the following policy measures are suggested:

(1)

Improve the relevant laws and regulations, and strengthen the ecological protection of cultivated land use at the policy level to help the sustainable development of agriculture. China has introduced a series of policies to coordinate the protection of food security and the ecological environment. In 2024, the Opinions of the CPC Central Committee and The State Council on Accelerating the Comprehensive Green Transformation of Economic and Social Development proposed to “promote green and low-carbon development of economic and social development, and promote the reduction of agricultural inputs such as fertilizers and pesticides to increase efficiency”, which plays an important role in promoting the improvement of the ecological efficiency of cultivated land use.

(2)

Rationally guide the orderly transfer of agricultural surplus labor force. The Decision of the Central Committee of the Communist Party of China on Further Comprehensively Deepening Reform and Promoting Chinese-Style Modernization, adopted in 2024, proposes comprehensively deepening reform to improve agricultural labor productivity, eliminate the dual structure between urban and rural areas and within cities, and to promote the transfer of agricultural labor to cities and the citizenization of the agricultural migrant population. In order to better improve the adaptability of the labor force from the agricultural sector to the non-agricultural sector, measures can be formulated from the following two proposals: (1) Provide employment training and information services for non-agricultural transfer personnel to help the labor force better adapt to the non-agricultural sector work; (2) Increase support for the development of enterprises, especially township enterprises that absorb more rural migrants, and improve the comprehensive ability of enterprises to absorb agricultural surplus labor.

(3)

Strengthen the support policy for agricultural machinery. China’s “14th Five-Year” National Agricultural Mechanization Development Plan emphasizes the following conclusions: Agricultural machinery operation services replacing human and animal power operation is an important mechanism for agricultural modernization that can enhance the technical support of agricultural machinery for agricultural green development requirements. Increasing agricultural science and technology input can effectively improve the ecological efficiency of cultivated land use [79,80]. Technological progress is often considered to be the main driver of efficiency [81,82], which can be achieved by incentivizing or subsidizing advanced technologies. “Subsidies for the purchase of agricultural machinery and tools” significantly promoted the mechanization level of the villages under management and maximized the input of the agricultural labor force [83]. We should further increase research and development and investment in small- and medium-sized machinery suitable for hills and increase subsidies for the grain compensation mechanism in hills and mountains to stimulate the vitality of grain production. In order to alleviate the phenomenon of terraced fields being left uncultivated, the government should attach importance to the labor substitution effect of agricultural service outsourcing [84].

(4)

Improve the rural land transfer market and strengthen the large-scale management of land. Land concentration and agricultural scale management should be reasonably realized in an environment dominated by market mechanisms and supplemented by government regulation. Land ownership confirmation promotes the transfer of agricultural land, which is conducive to realizing a large-scale economy and improving the utilization rate of cultivated land.