Outsourcing of Agricultural Machinery Operation Services and the Sustainability of Farmland Transfer Market: Promoting or Inhibiting?

Abstract

The agricultural machinery operation services (AMOS) market and the farmland transfer market are regarded by policymakers as complementary goals for promoting agricultural development in China. Nonetheless, the farmland transfer market in China is confronted with the threat of sustainable development. The relationship between AMOS and farmland transfer is not always complementary. To analyze the relationship between the AMOS market and the sustainability of the farmland transfer market, methods such as conditional mixed process, the Heckman two-step method, and the Sobel test were employed to explore the influence of AMOS on farmland transfer and its underlying mechanisms. The empirical results show the following: (i) AMOS inhibits farmland transfer-out but promotes farmland transfer-in, which will intensify the market competition of farmland transfer. This result remains valid after correcting for the potential endogenous bias and selective bias and is consistent across different variables and samples. This conclusion suggests that AMOS has emerged as a factor restricting the sustainable development of the farmland transfer market in China. (ii) The heterogeneity analysis results indicate that AMOS has a greater inhibitory effect on the farmland transfer-out of small-scale farmers, part-time farmers, and elderly farmers, and a greater incentive effect on the farmland transfer-in of large-scale farmers, professional farmers, and non-elderly farmers. (iii) Labor allocation and agricultural capital allocation are the potential mechanisms for AMOS to affect farmland transfer. AMOS indirectly inhibits farmland transfer-out and promotes farmland transfer-in by farmland operation ability of labor force and benefits of farmland operation. These results imply that there is a complementary relationship and substitution relationship between the AMOS market and the farmland transfer market. The substitution relationship may sometimes frustrate policies aimed at stimulating the farmland transfer market. The Chinese government is required to be wary of the potential menace that AMOS brings to the sustainability of China’s farmland transfer market.

1. Introduction

According to the data of the Third Agricultural Census in China (TACC), there are 207 million farmers in China, with an average cultivated land area of 0.65 ha. The World Bank defines small farmers as having a cultivated land area of less than 2 ha. According to World Bank standards, China is a typical country, dominated by small farmers. Countries such as the United States and Canada, where large farmers predominate, rapidly achieved agricultural mechanization in the aftermath of World War II. Nevertheless, the advancement of agricultural mechanization in East and Southeast Asian countries or regions dominated by small farmers is sluggish [1]. Scholars are gradually shifting their research perspective on agricultural mechanization to the relationship between agricultural mechanization and land scale. Agricultural machinery has the characteristics of strong asset specificity and low frequency of use [2,3]. Both the sunk cost and opportunity cost of investment in agricultural machinery are high, which makes small farmers lack the willingness or ability to invest in agricultural machinery [4,5]. Therefore, some studies believe that small farmers and agricultural mechanization are mutually exclusive [6,7].

However, based on small farmers, China has successfully realized agricultural mechanization and created a unique mode of agricultural machinery by relying on the rapidly developing AMOS market [8,9]. The judgment that small farmers and agricultural mechanization are mutually exclusive has been broken by China’s agricultural mechanization model [10]. Although agricultural machinery is inseparable in physical properties, the operation time of agricultural machinery is separable. The use of small farmers and large agricultural machinery does not conflict [11]. Small farmers are engaged in the division of labor economy by outsourcing AMOS and successfully use large agricultural machinery without bearing the sunk cost of purchasing agricultural machinery [12,13]. In order to promote the development of the AMOS market, the Chinese government has made tremendous efforts through policies and financial means, such as subsidies for purchasing agricultural machinery and for agricultural machinery operations [14,15]. According to the Yearbook of China’s Agricultural Machinery Industry, the total revenue of AMOS in China increased from 209.66 billion yuan to 481.62 billion yuan from 2004 to 2021. Correspondingly, according to the Statistical Bulletin on the Development of Agricultural Mechanization in China, the comprehensive mechanization rate of crops in China reached 72.03%, and the agricultural mechanization rate of grain crops in plain areas was close to 100% in 2021.

From the perspective of the input of agricultural production factors, the existing literature has proved that capital and land are two complementary production factors [8,16]. Agricultural machinery constitutes a significant form of agricultural capital, which implies that there exists a complementary relationship between agricultural machinery and farmland [17,18]. The expansion of farmland scale is considered to be an important driving force for agricultural mechanization [19]. Agricultural mechanization also helps promote the expansion of farmland scale [20,21]. However, in China, where smallholder farming prevails, farmers adopt the agricultural machinery technology by purchasing AMOS instead of agricultural machinery. AMOS has emerged as a crucial driving force in China’s grain production and the growth of agricultural productivity [22]. Some studies discussed the impact of AMOS on the matching relationship between labor and agricultural machinery. For example, AMOS improves land productivity, technical efficiency, and total factor productivity by replacing labor input and standardized operation [23,24,25,26]. AMOS can not only directly increase agricultural income by improving agricultural productivity [27] but also release agricultural labor force, promote farmers’ off-farm employment, and increase farmers’ off-farm income [28,29].

Whether there exists a complementary relationship between AMOS and the land factor remains a controversial question. Later studies that exploit the early stage of the implementation of China’s land transfer show mixed results. Q ian et al. (2022) and Xu et al. (2022) found that AMOS promotes land transfer by easing the capital investment constraints confronted by farmers, which corroborates the complementary relationship between the AMOS market and the land transfer market [18,30]. By contrast, Coelli et al. (1996) and Feng et al. (2019) found that the transaction costs of service outsourcing would exert a negative influence on farmers’ decision to enlarge the scale of farmland [31,32]. The existing studies mainly focus on the substitution relationship between agricultural machinery and labor. In fact, AMOS can also trigger changes in the matching relationship between land and agricultural machinery [14,18]. The question we are interested in is whether AMOS can affect farmland transfer and then affect the dynamic matching of land and agricultural machinery. The alterations in the relationship between AMOS and farmland might have a direct shock on the sustainability of the farmland transfer market.

AMOS is not only a production factor that replaces agricultural labor input but is also considered an important means of agricultural technology adoption and agricultural capital investment [33,34]. Labor and capital are generally considered to be important factors restricting farmland transfer [35]. The positive role of AMOS in easing the constraints on agricultural labor and capital investment has been demonstrated [36]. From this perspective, AMOS helps to facilitate farmland transfer and promote large-scale operation [18,37]. However, AMOS may have an adverse impact on farmland transfer-out by easing factor constraints. The two effects of farmland transfer-in and farmland transfer-out should be included in the analysis of the relationship between AMOS and farmland transfer. Unfortunately, there are few comparative studies on these two effects. To bridge this gap, this paper focuses on whether AMOS can affect farmland transfer. We will analyze the impact of AMOS on farmland transfer-in and transfer-out and investigate the impact mechanism from two aspects of capital and labor. Although the Chinese government has promulgated numerous policies intended to stimulate the development of the farmland transfer market, the statistical data from the Ministry of Agriculture and Rural Affairs of China reveals that the proportion of farmland transfer in China has not witnessed a remarkable escalation. Discussing the influence of AMOS on farmland transfer is conducive to interpreting why it is hard to make the development of the farmland transfer market in China sustainable.

The contribution of this paper can be generalized into two main aspects. First, the previous literature mainly investigated the impact of the substitution of labor and machinery on agricultural production. For example, studies have assessed the influence of purchasing AMOS on costs of agricultural production [38], input factor structure [23], labor allocation [36,37,39], and efficiency [26,40,41]. However, there are few studies focusing on the matching relationship between AMOS and farmland. This paper links AMOS with farmland transfer and provides systematic evidence on how AMOS affects farmland transfer-in and transfer-out. This research is conducive to explaining why China is confronted with the predicament of the sustainable development of the farmland transfer market.

Second, the existing literature regarding AMOS and production decisions usually regards AMOS as a precursor [3,18]. In actuality, the purchase decisions of AMOS by farmers could potentially be an outcome of land transfer, thereby indicating that endogenous biases ought to be rectified. Furthermore, the decisions of farmers in China typically have contagious effects [10]. Nevertheless, the selective bias of AMOS is seldom taken into account. In terms of empirical strategies, this paper attempts to overcome potential endogenous bias and optional bias and tries to test the impact path. The results of this paper show that AMOS inhibits farmland transfer-out but promotes farmland transfer-in, which will intensify the market competition and uneven development of farmland transfer-in. This conclusion suggests that AMOS constitutes a latent threat to the sustainability of the farmland transfer market in China. The conclusion of this paper also provides a new perspective to explain the slowdown in the development of China’s farmland transfer market after 2013. The findings of this study are highly interesting in a typical “big country–small farmer” context.

2. Theoretical Analysis and Research Hypothesis

2.1. The Sustainability of Farmland Transfer Market: AMOS and Farmland Transfer

Using the comparative static analysis model for reference [42,43], AMOS and farmland are regarded as the input factors of agricultural production. We build a farmers’ production decision model, which is used to analyze the dynamic matching process of AMOS and farmland. Suppose that the scale of farmers’ farmland is $L$, including self-owned, contracted farmland $\overline{L}$ and rented farmland $L_m$, in which the rent price of farmland is $r$. The outsourcing quantity of AMOS is $S$, and the AMOS price is $C_a$. The optimal AMOS $S^{*}$ and farmland scale $L^{*}$ are determined by farmers to pursue profit maximization. We assume that the agricultural production function is $Y=g(L, S)$. To guarantee that the production function attains a maximum value, we assume that the production function generally satisfies $g_L>0, g_S>0, g_{LL}<0, g_{SS}<0$. The production function meets the following conditions at the same time:

$$M=\left|\begin{array}{cc}{g}_{LL}& g_{LS}\\ g_{SL}& g_{SS}\end{array}\right|=g_{LL}\times g_{SS}-{g^2}_{LS}>0$$

(1)

The objective function for farmers to pursue the maximization of agricultural production profit is

$$\max \prod =p_{e}Y-r{L}_m-c_{a}S=p_{e}g(L, S)-r{L}_m-c_{a}S$$

(2)

The first order condition for the maximum value of the objective function (2) is

$$p_e{g}_L=r, p_e{g}_S=c_a$$

(3)

Equation (3) shows that, in an equilibrium state, 1 unit of marginal output of farmland is equal to its unit rent, and the revenue per unit of AMOS equals the unit cost of AMOS. The first order partial derivatives $g_L, g_S$ are functions of $L$ and $S$. We find the total differential equations of $L$, $S$, $r$, $c_a$, and $p_e$ for Equation (3). The result is expressed as follows:

$$\left\{\begin{array}{l}{p}_e{g}_{SL}dL+p_e{g}_{SS}dS+g_{S}d{p}_e-d{c}_a=0\\ p_e{g}_{LS}dS+p_e{g}_{LL}dL+g_{L}d{p}_e-dr=0\end{array}\right.$$

(4)

The result of solving the total differential Equation (4) is as follows:

$$dL={\displaystyle \frac{g_{L}d{c}_a-g_{S}dr+(p_e{g}_{LS}{g}_S-p_e{g}_{SS}{g}_L)dS}{p_e(g_{LS}{g}_L-g_{LL}{g}_S)}}$$

(5)

In order to investigate the relationship between AMOS outsourcing and farmland scale, we might as well assume $d{c}_a=0, dr=0, p_e=0$. The relationship between $L$ and $S$ is solved by Equation (5) as follows:

$${\displaystyle \frac{dL}{dS}}={\displaystyle \frac{g_{LS}{g}_S-g_{SS}{g}_L}{g_{LS}{g}_L-g_{LL}{g}_S}}$$

(6)

The sign of $dL/dS$ is determined by $g_{LS}$. If $g_{LS}>0$, then $dL/dS>0$, otherwise $dL/dS<0$. $g_{LS}$ reflects whether the two elements of agricultural machinery and land are complementary or alternative. Without considering technological progress, when the factor input of agricultural production reaches the initial equilibrium, the relative scarcity of farmland will be increased by the input of agricultural machinery, which will increase the marginal output rate of farmland [6,11]. The input of farmland needs to be increased to make the input of agricultural production factors reach a new and higher level of equilibrium [18]. Therefore, the two elements of agricultural machinery and farmland are complementary. $dL/dS>0$ is proved to be true. As a rational management subject pursuing the maximization of their own profits, farmers will fully consider the external constraints of AMOS when deciding whether to expand the cultivated land scale. The results of model derivation show that AMOS helps to expand the scale of farmland, which means that AMOS can increase the farmland transfer-in and inhibit the farmland transfer-out. This leads to the following hypothesis:

Hypothesis 1. 

AMOS promotes farmland transfer-in but inhibits farmland transfer-out.

AMOS promotes the farmland transfer-in and restrains farmland transfer-out, which implies that AMOS can aggravate the competition in the farmland transfer market. The mismatch between the supply and demand of farmland transfer poses a threat to the sustainability of the farmland transfer market.

2.2. AMOS, Allocation of Labor, and Farmland Transfer

AMOS is the main way for Chinese farmers to use agricultural machinery technology, which is essentially a substitute for labor force [44]. Therefore, the allocation structure of farmers’ household agricultural production factors will be affected by AMOS. First, allocation of labor: The higher income of off-farm employment attracts agricultural labor to leave agriculture. Aging, feminization, and childrenization have gradually become the development characteristics of the agricultural labor force in China [20,45]. Labor transfer increases the price of agricultural labor. In accordance with the theory of induced technological change, farmers’ demand for agricultural machinery will increase [46]. AMOS outsourcing is considered to be a cost minimization decision for farmers to adopt agricultural machinery technology to replace labor [9,47]. With the transfer of labor force, labor shortage has gradually become a factor restricting land transfer [48]. AMOS has increased the farmland management ability of a single agricultural labor force at a lower cost. With AMOS, less labor is reallocated in agricultural production, which helps to alleviate the labor constraints of farmland transfer. Therefore, AMOS encourages farmers to rent farmland through labor allocation. Meanwhile, AMOS also indirectly increased the opportunity cost of farmland transfer-out, which formed a crowding-out effect on farmland transfer-out. This leads to the following hypothesis:

Hypothesis 2. 

AMOS affects farmers’ decisions about farmland transfer through the allocation of labor.

2.3. AMOS, Allocation of Agricultural Capital, and Farmland Transfer

Second, the allocation of agricultural capital: Purchasing agricultural machinery is the main form of agricultural capital investment. Agricultural machinery is characterized by high sunk costs and a long return cycle. Therefore, agricultural machinery is generally regarded as one of the most asset-specific agricultural capital investments [36]. Investment in agricultural machinery will not only increase the economic burden of farmers but also squeeze the cash flow of farmers’ families and indirectly increase the opportunity cost. By purchasing AMOS, farmers are involved in the division of labor economy and realize the substitution of agricultural machinery for labor without bearing the investment cost of agricultural machinery [10,11]. Therefore, AMOS helps alleviate the constraint of farmers’ capital investment by saving agricultural machinery capital investment. With the expansion of farmland management scale, agricultural capital investment also needs to increase [49]. AMOS alleviates the financial constraints of farmland transfer by reducing the cost of agricultural capital investment, which can increase the transfer of farmland. The reduction of capital input costs also indirectly increases the opportunity cost of farmland transfer-out, which will restrict farmland transfer-out. This leads to the following hypothesis:

Hypothesis 3. 

AMOS affects farmers’ decisions about farmland transfer through the allocation of agricultural capital.

The theoretical framework is constructed for the impact of AMOS on farmland transfer (Figure 1). First, this paper will examine whether AMOS will affect farmland transfer. It should be noted that the impact of AMOS on farmland transfer may exhibit asymmetric characteristics. Second, whether AMOS affects farmland transfer through the allocation of labor will be tested. Third, whether AMOS affects farmland transfer through the allocation of agricultural capital will be tested.

3. Materials and Methods

3.1. Data Sources

Two periods of mixed, cross-sectional data of China are used in this paper. The first is the data of 840 households surveyed in Hebei, Shandong, and Henan provinces in 2020. The second is the data of 780 households surveyed in Jilin Province in 2021. The grain output of Hebei, Shandong, Henan, and Jilin provinces in 2023 was 202.77 million tons, accounting for 29.17% of the total grain output of China. The four provinces are listed as important grain bases by the Chinese government. Meanwhile, the agricultural mechanization rate of grain crops in the four provinces has exceeded 90%. Therefore, selecting Hebei, Shandong, Henan, and Jilin as the research areas can better represent the development status of China’s AMOS market.

The stratified random sampling approach was utilized in the investigation of the sample. Firstly, 14 counties were randomly selected from the four provinces based on the grain sowing acreage and geographic location. Secondly, based on per capita GDP, the towns in each county were stratified into three levels: high, medium, and low. At each level, we randomly selected 1–2 towns. A total of 51 towns were included in the sampling box. Third, 2–3 administrative villages in each town were randomly selected. A total of 129 administrative villages were included in the sampling box. Fourth, 15–20 farm households were randomly selected from each village. Excluding samples with missing observation values, 1419 farmer samples were ultimately included in the analysis of this article. The content of our investigation includes household characteristics of farmers, agricultural management situation, adoption of agricultural machinery technology, etc.

3.2. Variables

The dependent variable is farmland transfer. Farmland transfer includes the farmland transfer-in and transfer-out. Whether to transfer in farmland and whether to transfer out farmland have been used in the existing literature to characterize farmland transfer variables [14,20]. However, using a selection variable may overlook the characteristics of the scale of farmland transfer. Therefore, the rate of farmland transfer-out of the total area of land contracted by farmers is used to characterize farmland transfer-out. The proportion of farmland transfer-in area to the total area of farmland operated by farmers is used to characterize farmland transfer-in.

AMOS outsourcing is the main independent variable. The wheat agricultural machinery operation process includes tillage, sowing, and harvesting. The corn agricultural machinery operation process includes machine tillage, machine sowing, machine harvesting, and the mechanized return of straw to the field. Corn is the main grain crop in Jilin Province, and it ripens once a year. Corn and wheat are the main grain crops in the Hebei, Shandong, and Henan provinces, while corn and wheat are intercropped and ripen twice a year. Considering the differences in grain varieties, the proportion of AMOS outsourcing in the agricultural machinery operation process is used to characterize the AMOS outsourcing variable.

Labor allocation and agricultural capital allocation are mediating variables. First, the farmland area corresponding to a single working day is used to characterize the allocation of labor. The larger the land area corresponding to a single working day, the stronger the substitution of machinery for labor, and the less allocation of labor for farmers in the agricultural field [5]. Second, the profit of agricultural production is considered the result of the allocation of agricultural capital [14]. The profit per unit of farmland is the result of agricultural capital allocation, which is used to characterize the allocation of agricultural capital. The higher the profit per unit of farmland, the greater the intensity of agricultural capital allocation.

Drawing upon previous studies, we chose control variables from three dimensions [44,45,48,50]. First, gender, age, education level, health status, non-farm employment, agricultural training, member of the Communist Party of China (CPC), village cadre identity, and the farmland rent perceived by farmers were used to control the individual characteristics of decision-makers. Second, the labor force, agricultural labor, possession of agricultural machinery, aging labor, non-farm income, cooperative membership, and distance to county town were used to control the family characteristics of farmers. Thirdly, to control for the influence of unobservable variables within the region, 13 dummy variables set on a county basis were used to control for regional differences. The definitions and descriptions of the main variables are presented in

Table 1. Description of variables and descriptive statistics for variables.

Variable Name	Definition	Mean	SD
Farmland transfer-out	The proportion of farmland transfer-out area to the total contracted land area of farmers	0.126	0.297
Farmland transfer-in	The proportion of farmland transfer-in area to the total area of farmland operated by farmers	0.324	0.448
AMOS	The proportion of AMOS outsourcing in the agricultural machinery operation process	0.750	0.302
Gender	1 if the household head is male, and 0 otherwise	0.789	0.346
Age	Age of the household head (years)	55.273	13.215
Health	Self-reported health status of the household head: 1 = highly unhealthy, 2 = relatively unhealthy, 3 = moderate, 4 = comparatively healthy, 5 = highly healthy	4.045	0.819
Education	The educational attainment of the household head: 1 = illiterate, 2 = primary school, 3 = junior high school, 4 = high school or technical secondary school, 5 = junior college, 6 = university and above	2.522	0.728
Non-farm employment	1 if the household head has had the experience of working outside the hometown in the previous three years, and 0 otherwise	0.317	0.403
Agricultural training	1 if the household head has received agricultural training, and 0 otherwise	0.338	0.442
CPC membership	1 if the household head is a member of the CPC, and 0 otherwise	0.184	0.390
Village cadres identity	1 if the household head is a village cadre, and 0 otherwise	0.089	0.220
Farmland rent	Average farmland transfer rent perceived by farmers (10,000 yuan/hectare)	1.401	0.551
Labor force	The quantity of the population aged 16 and above and with labor capacity in farm household	3.039	1.307
Agricultural labor	The rate of the labor force mainly engaged in agricultural production	0.458	0.319
Aging labor	Proportion of aging labor in farm household	0.610	0.348
Possession of agricultural machinery	Original value of household agricultural machinery (10,000 yuan)	7.905	34.216
Non-farm income	The rate of non-agricultural income in the total income of the household	0.433	0.370
Cooperative membership	1 if the farmer joins the cooperative, and 0 otherwise	0.227	0.346
Distance to county town	The distance from the residence of farmers to the township government (km)	22.315	10.094
Labor allocation	The farmland area corresponding to a single working day (mu)	0.092	0.115
Agricultural capital allocation	The profit per unit of farmland (100 yuan/mu) a	3.214	1.932

Note: ^a The profit per unit of farmland = grain yield × grain price − seed cost − pesticide cost − fertilizer cost − fertilizer cost − farm machinery cost − labor cost.

.

3.3. Estimation Strategy

First, we need to test the influence of AMOS on farmland transfer. The model is expressed as follows:

$$Lrenta{l}_i={\alpha }_0+{\alpha }_{1}Amoservic{e}_i+{\displaystyle \sum _{i=1}^n{\kappa }_j{X}_{ij}}+{\epsilon }_i$$

(7)

where $Lrenta{l}_i$ represents farmland transfer, $Amoservic{e}_i$ represents AMOS outsourcing, $X_{ij}$ are the control variables, ${\alpha }_0$ is a constant term, and ${\alpha }_1,{\alpha }_2,{\alpha }_3,{\kappa }_j$ are the parameter to be estimated. ${\epsilon }_i$ is the random error.

Nevertheless, Model (7) may face a potential endogenous threat. There may be a reverse causal relationship between AMOS and farmland transfer. Previous studies have proved that the expansion of farmland scale is conducive to the adoption of AMOS [51,52]. The instrumental variable (IV) is used to correct the above endogenous problems. Referring to the idea of selecting tool variables from the regional level [5,10], the average number of AMOS outsourcing links of all farmers in the town where the farmers are located is utilized to serve as the IV. First, the average number of AMOS outsourcing links of all farmers in the town is related to AMOS outsourcing of farmers. Second, the average number of AMOS outsourcing links of all farmers in the town is a town-level variable. The town-level variable has no direct influence on the farmland transfer decisions of micro farmers. The endogenous variable is a truncated variable between 0 and 1. The 2SLS method based on continuous an endogenous variable is not suitable for truncated variables. A conditional mixed process (CMP) is used for IV regression. The CPM is applicable to classified variables, restricted variables, etc.

If farmers have farmland transfer behavior, the proportion of farmers’ farmland transfer can be observed. However, for farmers without farmland transfer behavior, we cannot observe the proportion of farmland transfer. Therefore, the variables of farmland transfer conform to the special “accidental tail breaking” feature selected by non-random samples, which will lead to selective bias in Model (7). The Heckman two-step model is used to correct the “accidental tail breaking” error. In the first step, the probability of farmers’ farmland transfer is estimated by probit model. The selection equation of farmland transfer is set as follows:

$$Ren{t}_i={\beta }_0+{\beta }_{1}Amoservic{e}_i+{\beta }_{3}Reco{g}_i+{\displaystyle \sum _{i=1}^n{\mu }_j{X}_{ij}}+{\epsilon }_i$$

(8)

where $Ren{t}_i$ represents farmers’ choice of farmland transfer, including whether farmland is transferred in and whether farmland is transferred out. $Reco{g}_i$ is the identifying variable. “Whether the village has a land transfer project organized by the government” is selected as the identification variable. If the village has farmland transfer projects, the probability of farmers’ spontaneous farmland transfer will be reduced. However, “Whether the village has a land transfer project organized by the government” will not directly affect the scale of farmers’ farmland transfer. The second step is to construct the resultant equation of farmland transfer as follows:

$$Rlrenta{l}_i={\chi }_0+{\chi }_{1}Amoservic{e}_i+{\displaystyle \sum _{i=1}^n{\theta }_j{X}_{ij}}+{\epsilon }_i$$

(9)

where $Rlrenta{l}_i$ represents the proportion of the scale of farmland transfer. Models (8) and (9) will be used together to correct “accidental tail breaking” error.

To test why AMOS outsourcing can affect farmland transfer, we assessed the mediating effect of the labor allocation as follows:

$$Labo{r}_i={\delta }_0+{\delta }_{1}Amoservic{e}_i+{\displaystyle \sum _{i=1}^n{\varpi }_j{X}_{ij}}+{\epsilon }_i$$

(10)

$$Lrenta{l}_i={\phi }_0+{\phi }_{1}Amoservic{e}_i+{\phi }_{2}Labo{r}_i+{\displaystyle \sum _{i=1}^n{\rho }_j{X}_{ij}}+{\epsilon }_i$$

(11)

where $Labo{r}_i$ represents the labor allocation of farmers. Models (9) and (10) combined with Model (7) constitute the mediating effect test model of labor allocation. It should be noted that the same model is used to test the mediating effect of agricultural capital allocation. Because of the disadvantage of the stepwise regression method in testing accuracy, Sobel and bootstrap are used to test the mechanism of action.

4. Results

4.1. The Impact of AMOS on Farmland Transfer

The baseline regression results are reported in

Table 2. Impacts of AMOS outsourcing on farmland transfer.

Variable Name	Farmland Transfer-Out		Farmland Transfer-In
	OLS	Tobit	OLS	Tobit
AMOS	−0.463 ***	−0.267 ***	0.593 ***	0.357 ***
	(0.092)	(0.031)	(0.110)	(0.086)
Gender	0.026	0.045	0.067	0.051
	(0.149)	(0.113)	(0.216)	(0.144)
Age	0.067 ***	0.042 ***	−0.045 ***	−0.032 ***
	(0.021)	(0.011)	(0.011)	(0.012)
Health	−0.036 ***	−0.031 ***	0.025 **	0.020 **
	(0.008)	(0.005)	(0.012)	(0.010)
Education	−0.037	−0.032	0.014 ***	0.021 ***
	(0.054)	(0.058)	(0.003)	(0.004)
Non-farm employment	−0.112	−0.092	0.195	0.161
	(0.128)	(0.097)	(0.314)	(0.226)
Agricultural training	−0.137 ***	−0.116 ***	0.129 ***	0.145 ***
	(0.034)	(0.031)	(0.040)	(0.043)
CPC membership	−0.236	−0.203	0.062 ***	0.049 ***
	(0.191)	(0.178)	(0.023)	(0.012)
Village cadres identity	0.018	0.011	0.042	0.041
	(0.074)	(0.021)	(0.052)	(0.061)
Farmland rent	0.007 ***	0.005 ***	−0.011 ***	−0.017 ***
	(0.001)	(0.001)	(0.002)	(0.004)
Labor force	0.095	0.071	−0.048	−0.024
	(0.114)	(0.087)	(0.120)	(0.087)
Agricultural labor	−0.385 ***	−0.291 ***	0.418 ***	0.431 ***
	(0.103)	(0.076)	(0.144)	(0.106)
Aging labor	0.112	0.097	−0.106 ***	−0.079 ***
	(0.231)	(0.205)	(0.031)	(0.015)
Possession of agricultural machinery	−0.029 ***	−0.020 ***	0.098 ***	0.075 ***
	(0.005)	(0.003)	(0.022)	(0.020)
Non-farm income	0.108	0.071	−0.321 **	−0.298 **
	(0.151)	(0.184)	(0.152)	(0.137)
Cooperative membership	−0.158 *	−0.116 *	0.045	0.023
	(0.089)	(0.062)	(0.083)	(0.034)
Distance to county town	0.008	0.004	0.045	0.031
	(0.051)	(0.029)	(0.082)	(0.067)
County dummies	yes	yes	yes	yes
Constant	−0.210 ***	−0.184 ***	0.112 ***	0.093 ***
	(0.063)	(0.022)	(0.016)	(0.027)
Pseudo R2/R2	0.455	0.417	0.628	0.610
Observations	1419	1419	1419	1419

Note: Robust standard errors are presented in parentheses; *, **, and *** indicate significance levels at 10%, 5%, and 1%, respectively.

. The estimation results in Columns (1) and (2) indicate that the coefficient of AMOS is significance negative. The results of OLS regression that considers farmland transfer-out as a continuous variable are consistent with the Tobit model that considers farmland transfer-out as a truncated variable. The estimated results all indicate that AMOS has a negative impact on farmland transfer-out. The results in Columns (3) and (4) indicate that AMOS significantly positively affects farmland transfer-in. AMOS encourages farmers to transfer in farmland but inhibits farmers from transferring out farmland, which to some extent will exacerbate the supply–demand contradiction in the farmland transfer market. Hypothesis 1 is supported by baseline regression results. The sustainability of the farmland transfer market is being challenged by AMOS. The result explains to some extent why the development speed of China’s farmland transfer market has become slow in recent years with the development of the AMOS market.

Among the other control variables, we found that the older the decision-maker, the more inclined they are to increase the farmland transfer-out and reduce the farmland transfer-in. A good health condition helps to motivate farmland transfer-in and prevent farmland transfer-out. The level of education has a positive impact on farmland transfer-in. Farmers who participate in agricultural production training tend to increase farmland transfer-in and reduce farmland transfer-out. Farmers with CPC or village cadre status are more likely to increase farmland transfer-in to develop into a large-scale farm. The perceived rent level has a negative impact on farmland transfer-in and a positive impact on farmland transfer-out, which is in line with our intuition. The ratio of agricultural labor restrains the inflow of farmland and boosts the outflow of farmland. Agricultural machinery has strong characteristics of asset specificity, which hinders farmland transfer-out and encourages farmland transfer-in. The higher the proportion of household non-agricultural income, the less farmland transfer-in. Joining the cooperative inhibits farmland transfer-out but does not have an impact on farmland transfer-in.

4.2. Robustness Check

4.2.1. Correcting Endogenous Biases

The estimated results of the CMP based on IV are reported in

Table 3. Results of CMP estimation.

Variable Name	First-Stage	Farmland Transfer-Out	Farmland Transfer-In
		Second-Stage	Second-Stage
AMOS$\_I{V}^{*}$		−0.415 ***	0.348 ***
		(0.079)	(0.064)
IV	0.119 ***
	(0.025)
atanhrho_12		−0.713 ***	0.287 ***
		(0.116)	(0.050)
Control variables	yes	yes	yes
County dummies	yes	yes	yes
Observations	1419	1419	1419

Note: Robust standard errors are presented in parentheses; *** indicates significance levels at 1%.

. The first-stage estimation results show a significant positive correlation between IV and farmers’ decisions about AMOS outsourcing, indicating that IV meets the correlation condition. The results in Columns (3) and (4) show that the endogenous test parameters (atanhrho_12) of the farmland transfer-out and farmland transfer-in equations both pass the significance test. Therefore, we can judge that the potential endogenous problem of the baseline regression does exist, and the CMP estimation results are better than the baseline regression. The second stage estimation results show that the negative impact of AMOS on farmland transfer-out and the positive impact on farmland transfer-in have passed the significance test at the 1% level. After correcting the potential endogenous bias, the inhibitory effect of AMOS on farmland transfer-out and the incentive effect on farmland transfer-in have been proved to still exist, which supports Hypothesis 1 again.

4.2.2. Correcting Selectivity Bias

The estimated results of the Heckman two-step method are reported in

Table 4. The estimated results of Heckman two-step.

Variable Name	Farmland Transfer-Out		Farmland Transfer-In
	Selection Equation	Outcome Equation	Selection Equation	Outcome Equation
AMOS	−0.681 ***	−0.219 ***	0.952 ***	0.216 ***
	(0.072)	(0.025)	(0.214)	(0.055)
Identifying variable	0.715 ***		−0.087 ***
	(0.113)		(0.025)
Control variables	yes	yes	yes	yes
County dummies	yes	yes	yes	yes
Constant	3.755 ***	0.283 ***	2.009 ***	0.168 ***
	(0.872)	(0.056)	(0.061)	(0.048)
IMR	−0.197 **		0.359 ***
	(0.071)		(0.053)
Observations	1419	1419	1419	1419

Note: Robust standard errors are presented in parentheses; ** and *** indicate significance levels at 5% and 1%, respectively.

. The inverse mills ratio (IMR) of farmland transfer-out and farmland transfer-in pass the significance test, which proves that the selectivity bias of the baseline regression does exist. The estimation results of the selection equation in Columns (1) and (3) show that AMOS outsourcing significantly increases the probability of farmland transfer-in but reduces the probability of farmland transfer-out. The estimation results of the outcome equation in Columns (2) and (4) show that AMOS outsourcing significantly increases the proportion of farmland transfer-in but decreased the proportion of farmland transfer-out. After correcting the potential selective bias caused by “accidental tail breaking”, the inhibitory effect of AMOS on farmland transfer-out and the incentive effect on farmland transfer-in have been proved to still exist, which supports Hypothesis 1 again.

4.2.3. Replace Core Variables

First, replace dependent variables: The willingness of farmers to participate in farmland transfer is used as an alternative variable. “Whether you have the intention to transfer out of farmland” and “whether you have the intention to transfer into farmland” are used as dependent variables. Columns (1) and (2) of

Table 5. Estimated results of replaced core variables.

Variable Name	Willingness to Transfer Farmland Out	Willingness to Transfer Farmland In	Farmland Transfer-Out	Farmland Transfer-In
AMOS	−0.852 ***	0.619 ***
	(0.077)	(0.041)
Links of AMOS outsourcing			−0.210 ***	0.175 ***
			(0.034)	(0.039)
Control variables	yes	yes	yes	yes
County dummies	yes	yes	yes	yes
Constant	−2.760 ***	2.118 ***	−0.234 ***	0.287 ***
	(0.713)	(0.552)	(0.026)	(0.039)
Pseudo R2/R2	0.722	0.846	0.467	0.492
Observations	1419	1419	748	748

Note: Robust standard errors are presented in parentheses; *** indicates significance levels at 1%.

report the results of Probit model estimation. The results showed that AMOS significantly reduced farmers’ willingness to transfer farmland out but significantly increased farmers’ willingness to transfer farmland in. Second, replace independent variables: The number of AMOS outsourcing links is regarded as the core independent variable to re-estimate the results. Moreover, considering the differences in crop cultivation, we re-estimated the relationship between AMOS and farmland transfer after replacing independent variables using samples from Henan, Hebei, and Shandong provinces. Columns (3) and (4) of Table 5 report the results of Tobit model estimation. The results show that the number of links for farmers to purchase AMOS is significantly negatively correlated with farmland transfer-out and significantly positively correlated with farmland transfer-in. The estimation results of replacing the core variables show that the conclusion that AMOS promotes farmland transfer-in and inhibits farmland transfer-out is robust and reliable.

4.3. Heterogeneity Analysis: Who Can Benefit More from AMOS?

First, heterogeneity of land scale: Large-scale farmers have stronger bargaining power in AMOS outsourcing. The marginal cost for large-scale farmers to obtain AMOS is lower than that for small-scale farmers. Therefore, the impact of AMOS on farmers’ farmland transfer may show the heterogeneity of land scale. The criteria of the third China Agricultural Census are used as the boundary of heterogeneity discussion. Taking 50 mu as the dividing line, the sample is divided into small-scale farmers and large-scale farmers. The group estimation results of the Tobit model are reported in Rows (1) to (4) of

Table 6. The results of heterogeneity analysis.

Heterogeneous Projects	Variable Name	Group	Coefficient	Standard Errors	Observations
Land scale	Farmland transfer-out	Small-scale farmers	−0.281 ***	(0.065)	871
		Large-scale farmers	—	—	—
	Farmland transfer-in	Small-scale farmers	0.116 ***	(0.029)	871
		Large-scale farmers	0.602 ***	(0.118)	548
Type	Farmland transfer-out	Professional farmers	−0.174	(0.205)	366
		Part-time farmers	−0.485 ***	(0.083)	1053
	Farmland transfer-in	Professional farmers	0.510 ***	(0.094)	366
		Part-time farmers	0.187 ***	(0.067)	1053
Age	Farmland transfer-out	Elderly farmers	−0.193 ***	(0.056)	815
		Non-elderly farmers	−0.348 ***	(0.072)	604
	Farmland transfer-in	Elderly farmers	0.422 ***	(0.097)	815
		Non-elderly farmers	0.280 *	(0.157)	604

Note: Robust standard errors are presented in parentheses; * and *** indicate significance levels at 10% and 1%, respectively.

. The results show that AMOS promotes the farmland transfer-in of both small-scale farmers and large-scale farmers. However, the incentive effect of AMOS on large-scale farmers’ farmland transfer-in is greater than that on small-scale farmers. Meanwhile, AMOS prevents small-scale farmers from transferring out farmland. Since there is almost no behavior of farmland transfer-out among large-scale farmers, the estimated results have not been reported. The demise of small-scale farmers is delayed by AMOS, which undoubtedly hinders the process of agricultural scale operation in China.

Second, heterogeneity of farmers’ types: Non-farm employment is considered to be an important way to improve the income of farmers [29]. The existing literature has proved that non-agricultural employment can affect the allocation of agricultural production factors [28,53]. The proportion of non-farm employment income in total income is used to distinguish the types of farmers. Farmers whose non-agricultural income accounts for less than 10% are defined as professional farmers, and farmers whose non-agricultural income accounts for more than 10% are defined as part-time farmers. The group estimation results of the Tobit model are reported in Rows (5) to (8) of Table 6. The estimation results show that AMOS inhibits the farmland transfer-out of part-time farmers, but the positive incentive effect on the farmland transfer-in of professional farmers is greater than that of part-time farmers. The demise of part-time farmers is delayed by AMOS, which undoubtedly hinders the process of agricultural scale operation in China.

Third, heterogeneity of farmers’ age: With the continuous transfer of the agricultural labor force to urban sectors, China’s agricultural labor force has shown an aging trend [54,55]. There are significant differences in agricultural production decisions between the elderly labor force and the young labor force. We divide the sample into elderly farmers and non-elderly farmers based on whether the respondents are over 60 years old. The group estimation results of the Tobit model are reported in Rows (9) to (12) of Table 6. The results show that the inhibitory effect of AMOS on farmland transfer-out of elderly farmers is greater than that of non-elderly farmers, while the incentive effect on farmland transfer-in of elderly farmers is less than that of non-elderly farmers. The demise of elderly farmers is delayed by AMOS, which undoubtedly hinders the process of agricultural scale operation in China.

4.4. The Mediating Effect Test of Labor Allocation and Agricultural Capital Allocation

The results of the mechanism test are reported in

Table 7. Estimation result of the mediating effect of labor allocation and agricultural capital allocation.

Effective Path	Sobel Test	95% Conf. Interval of Bootstrap
AMOS → Labor allocation → Farmland transfer-out	−0.093 ***	[−0.528,−0.237]
	(0.011)
AMOS → Labor allocation → Farmland transfer-in	0.170 ***	[0.154,0.376]
	(0.042)
AMOS → Agricultural capital allocation → Farmland transfer-out	−0.106 **	[−0.186,−0.041]
	(0.094)
AMOS → Agricultural capital allocation → Farmland transfer-in	0.227 **	[0.079,0.205]
	(0.075)

Note: Robust standard errors are presented in parentheses; ** and *** indicate significance levels at 5% and 1%, respectively.

. First, Rows (1) and (2) report the test results of the mediating effect for labor allocation. The results of the Sobel test show that the mediating effect of labor allocation between AMOS and farmland transfer-out is significantly negative at the 1% level, while the mediating effect of labor allocation between AMOS and farmland transfer-in is significantly positive at the 1% level. Meanwhile, the 95% confidence interval of the bootstrap test does not contain 0. Therefore, the mediating effect of labor allocation is proved to be true. AMOS not only directly affects farmland transfer but also indirectly affects farmland transfer through labor allocation, which supports Hypothesis 2. Specifically, AMOS indirectly inhibits farmland transfer-out and promotes farmland transfer-in by improving the farmland management ability of the labor force, thus intensifying the market competition of farmland transfer. Second, Rows (3) and (4) report the test results of the mediating effect for agricultural capital allocation. The method to determine the mediating effect of agricultural capital allocation is the same as that of labor allocation. According to the estimation results, we can judge that the mediating effect of agricultural capital allocation is tenable, which supports Hypothesis 3. AMOS indirectly inhibits farmland transfer-out and promotes farmland transfer-in by improving the profit per unit of farmland, thus intensifying the market competition of farmland transfer.

5. Discussion

5.1. The Relationship Between AMOS and the Sustainability of the Farmland Transfer Market

This paper makes an empirical investigation on the impact of AMOS on the sustainability of the farmland transfer market. From the perspective of factor allocation, a theoretical framework based on comparative static analysis is established. Based on the survey data of Chinese farmers, we discussed the impact of AMOS on farmland transfer and its potential mechanism. The Sobel test, the Heckman two-step model, and other methods are used to correct potential endogenous errors and sample self-selection errors. The results show that AMOS inhibits farmland transfer-out but encourages farmland transfer-in, which further intensifies the market competition of farmland transfer and poses a potential threat to the sustainability of the farmland transfer market in China. The findings of this study echo the research of Wei and Lu (2023) [5], who believe that the AMOS market and the land transfer market are not isolated from each other.

The conclusion of this paper is helpful to understand the debate on the mode and scale of agricultural operations in China. Some scholars believe that agricultural socialized service outsourcing is an important way for small farmers in China to apply new technologies, improve production efficiency, and reduce production costs [25,56,57]. They believe that the Chinese government should encourage service outsourcing and promote the division of labor, which will help promote the large-scale operation of services. However, other scholars believe that high transaction costs are induced by service outsourcing, which will have a negative impact on agricultural production [27,58]. They believe that the transfer of farmland should be encouraged by the Chinese government, which will help promote the large-scale operation of land. The debate on how to realize the large-scale operation of agriculture in China still exists today. The research findings of this paper are conducive to explaining the controversies in recent research papers.

The results in this paper show that AMOS encourages farmland transfer-in but inhibits farmland transfer-out, which indicates that there is a complementary relationship and substitution relationship between the large-scale operation of services and the large-scale operation of land. On the one hand, AMOS encourages farmers to transfer in farmland and promotes the large-scale operation of land, which is in line with the findings from Yu et al. (2021) and Qian et al. (2022) [14,18]. On the other hand, AMOS prevents farmers from transferring out farmland and hinders the large-scale operation of land. However, it is regretful that the potential threat posed by AMOS to the sustainable development of the farmland transfer market in China has not received adequate attention from existing studies and policymakers. The Chinese government should not only pay attention to the complementary relationship between the AMOS market and the farmland transfer market but also pay attention to the substitution relationship between them. It is indispensable for the Chinese government to adjust the development policies of AMOS to facilitate the sustainable development of the farmland transfer market.

5.2. Heterogeneity of AMOS on the Sustainability of the Farmland Transfer Market

The research results of this paper indicate that the impact of AMOS on farmland transfer presents asymmetric characteristics. The inhibitory effect of AMOS on the farmland transfer from elderly farmers, part-time farmers, and small-scale farmers is greater, which prolongs the survival time of elderly farmers, part-time farmers, and small-scale farmers. This conclusion suggests that AMOS may become a factor hindering the improvement of China’s agricultural competitiveness. In fact, with the continuous transfer of agricultural labor to urban sectors, China’s agricultural labor force is facing developmental challenges such as aging, part-time operation, small-scale operation, women’s operation, and children’s operation [59,60], which means that China’s agricultural labor force is characterized by weak development. The withdrawal of a weak agricultural labor force and the expansion of the farmland scale of individual farmers are considered to be an important method to enhance agricultural competitiveness [61]. However, the development of the AMOS market has provided technical and capital support for weak labor to operate agriculture, which leaves the weak agricultural labor force stranded in agricultural production in China. Therefore, AMOS is a double-edged sword [5]. While AMOS enhances the production efficiency of small-scale farmers, elderly farmers, and part-time farmers, it simultaneously disrupts the supply–demand equilibrium, thereby threatening the sustainable development of the farmland transfer market. The Chinese government should not only pay attention to the positive impact of AMOS on older farmers, part-time farmers, and small-scale farmers but also be more vigilant about the negative impact of AMOS market development on China’s long-term agricultural competitiveness.

5.3. The Mediating Effect of Labor Allocation and Agricultural Capital Allocation

Our results show that labor allocation and agricultural capital allocation are the potential mechanisms of AMOS affecting farmland transfer, which partly supports the conclusions of Yu et al. (2021) and Xu et al. (2024) [14,51]. First, the agricultural labor force is reallocated in the agricultural and non-agricultural fields by farmers who choose AMOS outsourcing, which indicates that AMOS will induce a high-quality agricultural labor force to leave agriculture and enter the non-agricultural fields. The transfer of high-quality agricultural labor will lead to the decline of human capital of agricultural labor, which will have a negative impact on agricultural production efficiency [62]. Second, the capital is reallocated in the agricultural and non-agricultural fields by farmers who choose AMOS outsourcing, which will also lead to the decline of long-term capital investment in agriculture. The growth of capital investment is regarded as an important engine that promoted China’s agricultural development in the past 40 years [63]. Therefore, the labor allocation and agricultural capital allocation have the potential risk of inducing the decline of agricultural productivity. The Chinese government needs to be alert to the adverse effects of these potential risks on the sustainability of the farmland transfer market.

5.4. Limitations

Our study, of course, has its limitations. We mainly discussed the agricultural machinery operation links of field crops such as tillage, planting, harvesting, and straw returning. We found that the pesticide spraying service using unmanned aerial vehicles has been applied, although this application is very limited. Due to data limitations, the field plant protection link was not included in this analysis. Moreover, in the process of sampling, we could not survey migrant farmers, which inevitably brings sample bias to the research. Fortunately, we used a variety of strategies to correct potential deviations, and we believe that the impact of such deviations is limited. However, providing more accurate and informative farmer data is crucial for future research.

6. Conclusions

The data of 1419 households in the North China Plain and the Northeast China Plain were used to test the impact of AMOS on farmland transfer and its mechanism. The following three conclusions can be drawn:

First, AMOS inhibits farmland transfer-out but promotes farmland transfer-in, which supports Hypothesis 1. This conclusion demonstrates that AMOS poses a threat to the sustainability of the farmland transfer market. Specifically, a 1% increase in the proportion of AMOS service outsourcing resulted in a 0.267% increase in the proportion of farmland transfer-in relative to the total area of farmland, while the proportion of farmland transfer-out relative to the area of contracted farmland decreased by 0.357%. However, the impact of AMOS on farmland transfer shows non-countermeasure characteristics. AMOS has a greater inhibitory effect on the farmland transfer-out of small-scale farmers, part-time farmers, and elderly farmers and a greater incentive effect on the farmland transfer-in of large-scale farmers, professional farmers, and non-elderly farmers. In the context of the rapid rise in the relative factor prices of agricultural labor, AMOS has extended the survival time of small-scale farmers, part-time farmers, and elderly farmers, which may hinder the sustainable development of China’s farmland transfer market. Second, AMOS has improved the farmland operation ability of unit labor through labor allocation, and thereby promotes the transfer-in of farmland and inhibits the transfer-out of farmland, which supports Hypothesis 2. Third, AMOS has improved the operating income of unit farmland through agricultural capital allocation, and thereby promotes the transfer-in of farmland and inhibits the transfer-out of farmland, which supports Hypothesis 3. Our conclusion indicates that labor allocation and agricultural capital allocation are the potential mechanisms for AMOS to affect the sustainability of the farmland transfer market.

Based on the above research conclusions, we draw some policy inspirations. First, the government is required to be vigilant about the potential threat that AMOS poses to the sustainability of the farmland transfer market and seeks the coordinated development of the two. The universal subsidy policy of the government for farmland transfer needs to be adjusted. We propose that more subsidies should be allocated to AMOS suppliers, since AMOS suppliers possess a comparative advantage in farmland transfer-in. Meanwhile, the government needs to continue to promote the reform of the farmland property rights system, reduce the transaction cost of farmland circulation, and improve the efficiency of farmland circulation. Second, in order to reduce the potential negative impact of AMOS on the sustainability of the farmland transfer market, the flexibility of the agricultural machinery purchase subsidy policy needs to be improved. With the variation of the number of items of agricultural machinery in a region, the agricultural machinery purchase subsidy policy in this region needs to be adjusted. The scope and proportion of agricultural machinery purchase subsidies should be in line with the saturation degree of the agricultural machinery market. If the number of items of agricultural machinery has been saturated, the government should encourage the exchange of old agricultural machinery for new ones. If the number of items of agricultural machinery is insufficient, the subsidy needs to be increased. Third, the government should encourage large-scale farmers, professional farmers, and non-elderly farmers to establish long-term cooperative relationships with AMOS suppliers, which will help to ensure the positive role of AMOS on farmland transfer-in. Meanwhile, the orderly transfer out of farmland by small farmers, part-time farmers, and elderly farmers needs to be encouraged and protected by the government, which will help promote a benign interaction between the AMOS market and the farmland transfer market.