The Impact of Agricultural Machinery Services on Food Loss at the Producer Level: Evidence from China

Abstract

Reducing food loss can improve environmental sustainability, resource use, and food security. Agricultural machinery services have considerable advantages in enhancing the adaptability and competitiveness of farms, but little is known about its potential for addressing food loss. Here, this work attempts to reveal a strong yet under-discussed connection between agricultural machinery services and food loss. Using survey data covering 483 corn farmers in the Heilongjiang, Jilin, and Liaoning provinces of China from October to December 2024, this study examined the extent to which participation in agricultural machinery services reduced food loss. Our results confirmed the existence of this significant causal effect and estimated 0.864% and 0.862% reductions in weight and value losses in response to a 1% increase in the purchase of agricultural machinery services. The possible mechanisms driving this relationship, including factor allocation optimization and technology introduction, were further verified. A variety of robustness tests were conducted to validate the strength and reliability of the empirical results and address endogeneity issues. Also, to better contextualize the heterogeneous effects of agricultural machinery services on food loss, the differences across production links, land fragmentation, and service quality were explored. By highlighting the important roles of agricultural machinery services in reducing food loss, our analysis contributed to contemporary debates about the long-term linkage between the wide popularization of agricultural machinery services and achieving food security, particularly providing insights for developing countries.

1. Introduction

Food loss has become an important global issue and is recognized as a serious threat to food security, the economy, and the environment. Approximately one-third to one-fourth of the food produced was lost or wasted [1]. The Food and Agriculture Organization of the United Nations (FAO) evaluated that 931 million tons of edible food were wasted globally in 2019, while 868 million people experienced malnutrition and hunger [2]. Food loss happens throughout the food value chain, from field to table. Losses occur at the consumption stage, also defined as the term of “food waste”, which usually excludes losses before certain products become food [3]. Food waste is serious in developed countries, whereas food loss at the producer level caused by pests, diseases, technologies, etc., is serious in developing countries [4,5]. As estimated, losses are consistently the largest at the producer level, accounting for between 60 and 80 percent of the total value chain loss across the different estimation methodologies in six countries (Peru, Ecuador, Honduras, Guatemala, Ethiopia, and China), compared to the middleman, processor levels [6]. Therefore, reducing food loss at the producer lever is critical, particularly for developing countries.

China, which feeds 22% of the world’s population with only 7% of the world’s arable land, is experiencing an alarming increase in food loss [7]. Its small-scale agricultural management model reduces incentives for investments and limits the application of technologies. What is worse, with the development of urbanization, the agricultural labor force has become feminized and aged [8], and labor costs in the agricultural sector continue to rise [9]. There has been a decrease in both the quantity and quality of labor forces. Resource constraints and institutional limitations reduce the capacity of the primary production, resulting in food loss. Just in this situation, China has achieved a “mechanical revolution” in the agriculture sector in a short time [10]. In 2023, the comprehensive mechanization rate of crop cultivation and harvest has reached 74.3%, with tillage, sowing, and harvest links reaching 86.42%, 60.22%, and 64.66%, respectively. For small farmers, it is not cost-effective to purchase agricultural machinery by themselves [11]. They are more likely to purchase agricultural machinery services [12,13]. Under this demand, agricultural machinery services have developed rapidly in China. From 2017 to 2023, the number of agricultural socialized service organizations has increased from 0.227 million to 1.094 million, the number of farmers served has increased from 36.558 million to 94 million, and the service areas have increased from 230 million mu to 2.14 billion mu. As an important means of associating small farmers with modern agriculture, agricultural machinery services have changed the traditional methods of agricultural production and management. Therefore, clarifying whether and how the machinery services affect food loss at the producer level has important practical implications for improving the agricultural socialized service system and ensuring food security.

In recent years, there has been an increasing interest in the behavior and effects of farmers’ purchases of machinery services. Compared with purchasing expensive agricultural machinery, small farmers with insufficient capital and risk aversion are more inclined to purchase machinery services [14]. Agricultural machinery can replace the labor force. Families with high employment degrees, aged, and feminized labor force tend to buy agricultural machinery services [15]. Many studies have been conducted on its effects from various perspectives. In terms of the effects on agricultural production, many studies have confirmed that agricultural machinery services have the effect of increasing yield and productivity [16,17,18]. In terms of the effects on environment, the purchase of agricultural machinery services is conducive to reducing the amount of fertilizer input [19,20], carbon emissions [21], and improving the agricultural green production efficiency [22]. In addition, purchasing agricultural machinery services can also help farmers increase household income and narrow the income gap within rural areas [23,24,25].

In the literature on food loss, measurement and comparison have been the subject of considerable discussion. In terms of measurement indicators and scope, many studies have measured food quantity losses throughout the whole supply chain [26,27,28,29]. These studies generally adopt a macro approach, utilizing aggregated data from national or local authorities and large companies. The indicators used to measure and quantify food loss typically rely on mass or energy balances. While this method is cost-effective, it is limited by the lack of representative and high-quality data at specific stages of the value chain. Few studies have further examined the loss figures at specific parts of the food supply chain, such as harvesting, distribution, processing, and storage [30,31]. Given the attention to food loss at the producer level, Parfitt et al. included losses at the pre-harvest stage when calculating the global food loss [28]. Delgado et al. broke down the losses at the level of farmer, middleman, and processor by different estimation methodologies [6]. Furthermore, the reasons have been identified. Weather, pests, diseases, contractual practices, etc., are the important micro-causes of food loss [30,32,33]. The inefficient transport systems and lack of appropriate storage facilities are also considered as the causes of food loss [34]. Additionally, the digital technology has shown a significant positive impact on food loss in the circulation and sales stages of the food supply chain [35].

The existing research results have important reference significance for conducting this study, but there are some limitations as follows. First, although there have been increasing discussions about the measurement of food loss and its influencing factors, little empirical research has been performed to evaluate whether agricultural machinery services can reduce food loss. However, the low adoption of new technologies, such as agricultural machinery, is one of the primary reasons causing low farm productivity, particularly for small farmers [36]. The importance of popularizing agricultural machinery services in advancing agricultural modernization and alleviating poverty in developing countries has been widely documented [16,37]. Second, despite the body of literature measuring food loss, they primarily provide information on quantity losses, while ignoring another critical dimension of quality losses related to food security, which might underestimate the actual losses [6]. Third, the underlying mechanisms need to be further explored. Existing empirical evidence focuses on the direct and indirect determinants of food loss, including management operations, environmental and economic conditions, and human error [38]. Researchers in the existing literature have documented persistent gaps in effective interventions to reduce food loss and its underlying motivations [39].

The contributions of this study to the ongoing literature are threefold. First, this study primarily contributes to the literature by linking agricultural machinery services and food loss, expanding the understanding of the effects of machinery services. Second, based on the micro-level survey data in the northeast region of China, this research contributes to new empirical evidence capturing food loss in both quality and quantity aspects at the producer level. Third, this study contributes to the theoretical analysis by exploring the underlying mechanism driving the link between agriculture machinery services and food loss. Our empirical evidence on the underlying mechanisms from the pathways of factor allocation optimization and technology introduction provides a micro-level theoretical foundation for the effects of agricultural machinery services on food loss reduction.

In summary, this study estimates the loss reduction effect of machinery services, and its mechanisms and heterogeneity using survey data covering 483 corn farmers in the Heilongjiang, Jilin, and Liaoning provinces of China from October to December 2024. As a raw material for biomass energy, the global demand for corn has significantly increased, highlighting the importance of reducing corn losses. The northeast region of China is located in the world’s golden corn production belt, with abundant land endowment and high mechanization levels, providing a sample support for this study. Simultaneously, the framework of this study is also applicable to field crops such as rice and wheat.

The rest of the paper is organized as follows. Section 2 derives a theoretical framework that proposes testable hypotheses regarding the impacting mechanisms of agricultural machinery services on food loss. The data description, empirical model, and econometric analysis approaches are presented in Section 3. Section 4 summarizes the empirical results of causal effect estimation, robustness checks, and heterogeneity analysis. Section 5 shows the further discussions about the main results. Finally, Section 6 provides the conclusions and policy implications.

2. Theoretical Framework

The induced innovation theory suggests that farmers usually choose technologies to substitute for expensive and scarce production factors. Given the resource endowments in rural areas, there are demands for land-saving technologies (such as biochemical techniques like fertilizers and pesticides) and labor-saving technologies (such as machinery). According to the theory of rational small farmers, farmers are the microeconomic units of agricultural production. Scenario rationality determines that farmers’ production behaviors and goals will make choices conducive to maximizing household income according to the market environment, agricultural technical conditions, and the allocation of production factors.

With the development of industrialization and urbanization, the opportunity cost of engaging in the agricultural sector is increasing, resulting in rising wages for hired labor in China’s rural areas. Against this backdrop, farmers use agricultural machinery instead of labor. Due to the specificity of assets, purchasing machinery services is more effective for small farmers to reduce marginal costs than purchasing agricultural machinery [11,12,13]. Purchasing agricultural machinery services is the main choice for small farmers in rural China. According to the division of labor theory, the deepening division of labor constitutes the fundamental basis for performance enhancement. Farmers can share the division of the labor economy by purchasing agricultural machinery services, and using more advanced technologies provided by service organizations [40]. Purchasing agricultural machinery services can help small farmers overcome capital constraints and optimize factor allocation. This will increase agricultural productivity and reduce food loss. Meanwhile, the use of machinery leads to the standardization and specialization of agricultural production, which in turn, improves the quality of operations and reduces food loss. In particular, more specialized operators and advanced machinery are owned by service organizations, and the quality of operations is usually higher when machinery services are purchased [41]. In summary, purchasing agricultural machinery can improve farmers’ production and management levels, thus, effectively reducing the quantity and quality loss of food. On the basis of the analysis above, we proposed the following hypothesis:

Hypothesis 1.

Agricultural machinery services can reduce food loss.

The dual attributes belonging to the capital and technology of agricultural machinery services can optimize the allocation of agricultural resources and reduce food loss. Farmers can adjust their land scales by purchasing machinery services to achieve the optimal business scale. Considering China’s extensive agricultural population coupled with its constrained arable land resources, the contracted land scale is usually smaller than the optimal production scale. In order to maximize household income, farmers will transfer in land to adjust the scale of land management. However, the asset specificity of agricultural machinery is high, especially for large agricultural machinery. Small farmers frequently face the challenge of being unable to acquire all the necessary agricultural machinery owing to financial limitations. Small farmers can benefit from agricultural machinery services through low-cost leasing [42]. In addition, the shortage of agricultural labor also limits small farmers to from expanding their land scales. Agricultural machinery can substitute agricultural labor. When the market is perfect, it is more profitable to purchase machinery services than to hire labor. Therefore, agricultural machinery services could relieve the capital and labor constraints, thereby facilitating small farmers to transfer in land and optimize the factor allocation of households [43]. Land transfer can optimize not only the allocation of production resources to a farmer, but also the allocation of resources across farmers. Farmers with the comparative advantage of agricultural production will transfer in land, which reduces the efficiency loss caused by the misallocation of resources, improving land productivity [44]. Agricultural machinery services have changed the factor input and management modes of traditional agricultural production [45]. On the basis of the above theoretical analysis, we proposed the following hypothesis:

Hypothesis 2.

Agricultural machinery services can reduce food loss by optimizing factors allocation.

In addition, agricultural machinery services can reduce food loss through the “technology spillover” effect. Agricultural machinery is one of the key and indispensable technologies for modern agriculture [46]. In particular, the technologies provided by professional service organizations are more advanced than those of farmer-owned machinery. Farmers can obtain advanced technologies by purchasing agricultural machinery services at each link of the agricultural production process [47]. In the land preparation link, “deep loosening” and “no-tillage” services can slow down the evaporation of soil moisture, control soil erosion, and enhance the water retention and moisture conservation capacity of the soil [48]. This improves the soil structure and promotes an increase in food production. Sowing with agricultural machinery is more standardized than sowing manually. The result is an increase in seed survival and a boost in crop growth. In the plant protection link, the amount and uniformity of fertilizer and pesticide usage can be controlled more effectively [49,50]. In the harvest link, agricultural machinery services can not only improve the speed to ensure the efficiency of the harvest, but also, the combined harvesting technology can reduce food loss compared to the manual methods [51]. In addition to technical equipment, the experience and skills of machinery operators are also superior to those of farmers [52]. It can be seen that the purchase of agricultural machinery services has not only solved the problem of labor shortage but also introduced advanced technology into agricultural production. In general, advanced technology can help reduce food loss more effectively. On the basis of the above analysis, we proposed the following hypothesis:

Hypothesis 3.

Agricultural machinery services can reduce food loss by introducing technologies. Farmers are likely to embrace advanced production technologies offered by machinery services, which will contribute to reducing food loss.

3. Data and Methodology

3.1. Data

According to the research objective, the selection of data acquisition areas should follow the following principles: On the one hand, to ensure planting large areas of food and having strong a foundation for agricultural development for a long period, the sampling area should have sufficient arable lands. On the other hand, to ensure good diversity among the samples, the sampling area should have a well-developed market for machinery services. The Heilongjiang, Jilin, and Liaoning provinces of China meet the above principles: First, they produce one-fifth of the country’s total food output every year. Corn has the largest planting area of all foods. Second, the level of agricultural mechanization is high, and many farmers purchase agricultural machinery services. Third, there are significant differences between the three provinces, Heilongjiang, Jilin, and Liaoning, in terms of land endowment, economic development, and other factors, which affect the degree of perfection in the machinery services market. The samples selected in these three provinces are well representative of the development level of machinery services.

The data came from a questionnaire survey conducted by the research group on farmers in the Heilongjiang, Jilin, and Liaoning provinces from October to December 2024, which included the pre-survey, formal survey, and supplementary survey. We adopted a survey sampling method that combines stratified sampling with random sampling. First, based on the differences in the comprehensive level of agricultural development and grain sown areas between provinces, Harbin and Qiqihar in the Heilongjiang Province, Changchun and Songyuan in the Jilin Province, and Tieling and Shenyang in the Liaoning Province were selected as sample areas. According to statistical data, these cities ranked at the forefront in terms of food sowing areas in 2023. In addition, the terrain of the cultivated land varies among them. Second, we selected 2–3 counties in each city by the random sampling method. Then, according to the per capita income level, all the towns (townships) in the sample county were categorized into three groups: high, medium, and low; two villages were randomly selected from each group. Finally, 8–10 farmers were randomly selected in each village for the survey. Corn farmers were used as the research samples to avoid the impact of food type differences on the consistency of food loss measurements.

The content of the questionnaire mainly included the following aspects: First, basic characteristics of farmer households, such as the age and education of the head of the household, the number of family laborers, the time allocation of labor, and the employment situation of family members. Second, agricultural production conditions, such as the planting area of corn, the number of plots, the situation of land circulation, and the yield of corn, were included. Third, the situation of purchasing agricultural machinery services, including whether the purchased services were applied in production processes such as land preparation, pesticide application, fertilization, and harvesting. Fourth, we investigated the loss of both food quality and quantity through farmers’ self-reported data. After obtaining the questionnaire, the data were cleaned as follows: First, the samples with missing agricultural machinery services and food loss were excluded. Second, we deleted the contradictory situation and the missing observed value of the study variables. Finally, we identified 483 effective corn farmers.

3.2. Model Setting

3.2.1. Ordinary Least Squares (OLS)

The primary objective of this research is to analyze the impact of agricultural machinery services on food loss. First, without considering the endogeneity issues caused by sample self-selection, we used the OLS model directly to test the impact of machinery services on food loss. The model is set up as follows:

$$ln{y}_i=a_0+a_{1i}{D}_i+a_{2i}{X}_i+{\epsilon }_i$$

(1)

$ln{y}_i$ represents the logarithm of food loss, and $D_i$ represents agricultural machinery services, set as a dummy variable. The value assigned to farmers who purchased the agricultural machinery services is 1; otherwise, it is 0. $X_i$ represents control variables, including four groups: production decision-maker characteristic variables, farmer household characteristic variables, agricultural operation characteristic variables, and external environmental variables. ${\mathrm{a}}_{1i}$ is the parameter to be estimated, and ${\epsilon }_i$ is the random disturbance term.

3.2.2. Propensity Score Matching (PSM)

The decision of farmers to purchase agricultural machinery services is based on their anticipated income. There are some factors that affect both the farmers’ decisions on machinery services and the level of food loss simultaneously, such as the area of cultivated land at the village level and village location. This means that the key to accurately identifying the impact of machinery services on food loss is crucial to addressing the endogeneity problem caused by sample “self-selection”. PSM constructs “counterfactual” to estimate the impact of machinery services on food loss. It can minimize the bias of the observed data by matching the resampling method to ensure that the observed data closely resemble those of a random trial, making the observed data as close as possible to the random trial data, and thus, effectively solving the biased estimation problem caused by sample “self-selection”. Therefore, we used the PSM model to estimate the average treatment effect (ATT) of agricultural machinery services on farmers’ food loss for a robust test.

PSM estimation is completed in three stages. In the first stage, the propensity score pi is calculated. The decision of farmers to adopt services can be represented by a discrete choice model. Then using a logit model estimates the conditional probability of each farmer’s service adoption, which is called the propensity score. In the second stage, sample pairing is performed. The control variables of the logit model and the covariates during the sample-pairing process include four groups of variables: characteristics of production decision-makers, characteristics of farmer households, characteristics of production and operation, and external environment. Theoretically, there are various matching methods that can achieve matching, and the results obtained by the different methods are asymptotically consistent. In practice, due to the different common support domains generated by different matching algorithms, the degree of sample loss varies, and the treatment group samples are matched with different control group samples, resulting in differences in the results of different methods. This project adopts three matching methods, including nearest neighbor, and compares the results. In the third stage, the matching estimator is used to measure the ATT. Referencing the study of Makate et al., the ATT is defined as follows [53]:

$$ATT=E\left(Y_{1i}-Y_{0i}|D_i=1\right)=E\left(Y_{1i}|D_i=1\right)-E\left(Y_{0i}|D_i=1\right)$$

(2)

$Y_{1i}$ stands for farmer $i$’ food loss when $i$ purchases the agricultural machinery services. $Y_{0i}$ stands for $i$’ food loss when $i$ does not purchase the agricultural machinery services. If $Y_{1i}$ and $Y_{0i}$ can be estimated simultaneously, the difference between them is the net impact of the agricultural machinery services on farmers’ food loss. However, if farmer $i$ adopts agricultural machinery services, we can only observe $E\left(Y_{1i}|D_i=1\right)$, and its counterfactual losses $E\left(Y_{0i}|D_i=1\right)$ are unobservable. Alternative metrics for $E\left(Y_{0i}|D_i=1\right)$ can be constructed using PSM.

3.2.3. Mediation Model

On the basis of the theoretical hypothesis above, this study introduced factor allocation optimization and technology introduction as mediating variables to examine how machinery services affect food loss through resource allocation and technology introduction. In this study, the three-step mediation effect model is used to test the mediation effect. The mediation effect model is set up as follows:

$$M_i={\beta }_0+{\beta }_{1i}{D}_i+{\beta }_{2i}{X}_i+{\epsilon }_i$$

(3)

$$ln{y}_i={\gamma }_0+{\gamma }_{1i}{D}_i+{\gamma }_{2i}{M}_i+{\gamma }_{3i}{X}_i+{\epsilon }_i$$

(4)

${\mathrm{M}}_i$ represents the mediating variables, which are the optimization of factor allocation and the introduction of technology. The remaining variables were defined as the same as in the model (1). The significance of the sobel-test was used to assess the mediation effect of the mediation variables and examine whether machinery services can reduce food loss by driving the optimization of factor allocation and the introduction of technology.

3.3. Variables and Summary Statistics

It should be emphasized that the dependent variables employed in this study are food losses, measured by both weight and value. Theoretically, the data measured by the field experiments can better reflect the accurate situation. However, field measurement is too expensive and too difficult and can only accurately calculate the level of the experimental site. Farmers have rich farming experience and can provide relatively accurate feedback from the practical level. The data obtained through questionnaire survey also have high credibility. Although farmers’ estimates may also have measurement bias, they are random in a large sample case and can reflect data more accurately [31]. The dependent variable is food loss, which is measured using farmers’ self-reported from the aspects of the weight and value at the producer level [6]. The weight of food loss (WFL) is relatively easy to quantify with the feedback from farmers, but it cannot reflect quality deterioration information. Poor quality is also an important part of the losses. Quality losses are difficult to directly detect due to various conditions. When farmers sell their foods, the acquirer will price according to the quality. Therefore, the difference between the ideal value and actual value sold by farmers can also reflect quality losses. Here, we measure the value losses (VFL) to take into account the quantity and quality losses. Moreover, food loss exists in every link of the supply chain. Considering that machinery services only involve different links during the production stage, we do not pay attention to the food loss post-harvest.

This study mainly examines the impact of machinery services on food loss, thus taking “machinery services” as the core explanatory variable. The value assigned to farmers who adopt is 1; otherwise, it is 0. In particular, we further explore the heterogeneity of different production links. In view of the endogenous problem that the core independent variables and dependent variables may be mutually causal, we selected “whether the farmers except the town purchase the average value of machinery services” as the tool variable [54].

We have sorted out the theoretical logic that machinery services promote a reduction in farmers’ food loss by optimizing the allocation of factors and introducing technology. Machinery services change the input ratio of production factors such as labor, capital, and land. Based on the constraint of land factors, this can be reflected in the efficiency of cultivated land use. In addition, agricultural machinery is the material foundation of corn cultivation and an important carrier of advanced technology. The more agricultural machinery is used, the more likely it is that advanced technologies will be introduced. Therefore, the former is measured by taking the logarithm of the average grain yield per mu. The latter is measured by taking the natural logarithm of the average labor input per mu across the processes of land preparation, sowing, field management, and harvesting. (

Table 1. Variable definitions and descriptive statistics.

Variables		Definition	Mean	Std.	Min	Max
Food loss	(WFL)	Ln weight of food loss per mu + 1	3.930	1.855	0	7.447
	(VFL)	Ln value of food loss per mu + 1	3.878	1.854	0	7.447
Agricultural machinery service (AMS)		Agricultural machinery service: purchased = 1, not purchased = 0	0.789	0.409	0	1
Labor input per mu (LAP)		Ln labor input per mu	0.513	0.662	0	4.615
Grain yield per mu (GYP)		Ln grain yield per mu	3.660	0.146	2.538	5.064
Physical health condition (PHC)		Chronic diseases such as heart disease, hypertension: yes = 0, no = 1	0.251	0.434	0	1
Farming experience (FE)		Householder’s years of farming experience (years)	29.983	13.734	1	160
The attitude of risk (THR)		Willingness to try to grow new varieties: yes = 1, no = 0 first in the village	3.23	1.307	0	5
Communication expense (CE)		Fee of monthly communication: less than 50 = 1, 50–100 = 2, 100–200 = 3, more than 200 = 4	1.917	1.194	1	4
The percentage of non-farm (PNF)		Proportion of non-agricultural employment (%)	0.145	0.213	0	1
Annual gift money expenses (GME)		less than 1000 = 1, 1000–2000 = 2, 2000–5000 = 3, 5000–10,000 = 4, more than 10,000 = 5	3.298	1.131	1	5
The number of labor (TNL)		Number of labors engaged in agricultural production	2.046	0.766	0	6
The area of arable land (AAL)		Total farmland area of the farmer (mu)	94.671	198.507	1.5	3000
The quality of arable soil (QAS)		very poor = 1, poor = 2, general = 3, good = 4, very good = 5	2.491	0.683	1	5
Levelness of cultivated land (LCL)		very poor = 1, poor = 2, general = 3, good = 4, very good = 5	3.571	0.893	1	5
Village terrain (VTN)		flat land = 1, not plat land = 0	1.112	0.315	1	2
Village economy (NE)		Non-agricultural industries of the village: yes = 1, no = 0	0.319	0.467	0	1
Village location (VLN)		Distance between the village and the nearest county seat (km)	22.781	15.241	1	70
Natural disaster (ND)		Whether there was a natural disaster in the village in the past 3 years: yes = 1, no = 0	0.478	0.5	0	1

).

In addition to the agricultural machinery services decision, the individual characteristics, family characteristics, production and management characteristics, and village characteristics will also affect the food loss of farmers. Based on the theory of farmers’ behavior and the main practices of existing studies, we quoted the relevant literature and selected controlled variables as follows [55]. Personal characteristic variables of decision makers include planting life, risk preference, and physical health status variables. Family characteristic variables include the number of labor force, proportion of non-agricultural employment, and social network variables. Business characteristic variables include contracted land area, land soil quality, and land leveling degree. Village characteristic variables include village terrain characteristics, village non-agricultural industry situation, and village distance from county (city). Before empirical analysis, the VIF test is used to check for multicollinearity among independent variables. The resulting Mean VIF of 1.30 indicates that there is no evidence of multicollinearity among the independent variables.

4. Empirical Results

4.1. Baseline Regression Analysis

The OLS estimation results for the impacts of AMS on food loss are presented in

Table 2. Effect of the AMS on food loss.

Variables	WFL		VFL
	(1)	(2)	(3)	(4)
AMS	−0.872 ***	−0.864 ***	−0.839 ***	−0.862 ***
	(−4.29)	(−4.58)	(−4.13)	(−4.52)
PHC		0.351 *		0.382 **
		(1.81)		(1.97)
FE		0.004		0.005
		(0.77)		(0.98)
THR		0.042		0.054
		(0.60)		(0.78)
CE		0.142 **		0.176 ***
		(2.38)		(2.92)
PNF		−0.423		−0.417
		(−1.08)		(−1.07)
GME		−0.184 **		−0.171 **
		(−2.46)		(−2.30)
TNL		0.233 **		0.235 **
		(2.28)		(2.35)
AAL		−0.001 **		−0.001 **
		(−2.04)		(−2.08)
QAS		0.236 *		0.243 *
		(1.87)		(1.94)
LCL		0.019		0.027
		(0.20)		(0.28)
VTN		0.233		0.098
		(0.81)		(0.34)
NE		−0.605 ***		−0.575 ***
		(−2.92)		(−2.79)
VLN		−0.010 *		−0.010 *
		(−1.69)		(−1.78)
ND		1.310 ***		1.283 ***
		(8.01)		(7.88)
Constant	4.618 ***	3.164 ***	4.539 ***	3.044 ***
	(25.59)	(4.14)	(25.14)	(4.03)
Observations	483	483	483	483
R-squared	0.037	0.211	0.034	0.213
Adjusted R-squared	0.0348	0.186	0.0322	0.188
F-value	18.40	9.151	17.04	9.405

Note: The estimated parameters’ t-values are in parentheses. * p < 0.1, ** p < 0.05, and *** p < 0.01.

. Columns (1) and (3) report that when the core explanatory variable is machinery services, the impact coefficient is significantly positive at the 1% level, indicating that the AMS has positive effects on reducing food loss. Columns (2) and (4) are based on columns (1) and (3) to add some control variables, such as household characteristics, family characteristics, agricultural production and operation characteristics, and village characteristics. The results show that no matter which control variables are added, the impact of AMS on food loss is significantly negative at the level of 1%. The coefficients of the impact on the weight and value of food loss (WFL and VFL) are −0.864 and−0.862, which show that AMS can reduce food loss. Hypothesis 1 is verified.

4.2. Robustness Tests

4.2.1. Robustness Test: Replacing the Model

The OLS results show the overall impact of agricultural machinery services on food loss. Considering the endogenous issue caused by self-selection, we further use the PSM model to examine the treatment effect of machinery services on food loss under the counterfactual framework. The nearest neighbor matching, nearest neighbor caliper matching, and kernel matching are used to calculate the average processing effect. The significance test was obtained with 500 repeated sampling times with the bootstrap method (Bootstrap). The results are presented in

Table 3. Results of PSM.

Variables	WFL			VFL
	Neighbor Match	Neighbor Caliper	Nuclear Match	Neighbor Match	Neighbor Caliper	Nuclear Match
AMS	−0.904 ***	−1.069 ***	−0.940 ***	−0.902 ***	−1.079 ***	−0.943 ***
	(−3.37)	(−3.66)	(−3.57)	(−3.33)	(−3.69)	(−3.55)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes
Observations	483	483	483	483	483	483

Note: PSM reports the average treatment response (ATT) of the treatment group. *** p < 0.01. The standard errors are presented in brackets. The nearest neighbor matching approach follows the procedure performed by Abadie et al. (2004) [56]. A one-to-four matching approach is applied, generally aiming to minimize the mean squared error. Regarding the nearest neighbor matching with calipers, the caliper radius is specified as 0.01. According to the existing literature, kernel matching method adopts the Gaussian function, meanwhile setting a bandwidth value of 0.06.

. Although various matching methods are used, the results are consistent. The PSM results and the OLS benchmark regression consistently show the following: agricultural machinery services can reduce food loss. The impact coefficient of food loss in the OLS is slightly lower than the PSM, which indicates that the potential endogenous problems of the model underestimate the impact of agricultural machinery services on food loss.

4.2.2. Robustness Test: Using Instrumental Variable Method

The food loss may also reversely affect the purchase of agricultural machinery services. Agricultural machinery services and food loss may interact with each other which may lead to errors in model estimation parameters because of the endogenous issues. To eliminate the endogenous problem of the model, we draw the average purchase rate of machinery services by other farmers in the town (townships) as the instrumental variable. This variable as the instrumental variable meets two necessary conditions: On the one hand, it is exogenous. Whether other farmers in the town purchase machinery services in theory will not directly affect the food loss of this farmer. On the other hand, it has a correlation with the core variables. There is a widespread “herd effect” in farmers’ behavior in the same town. Whether other farmers in the town purchase machinery services will affect this farmer through the demonstration effect. Therefore, the instrumental variable meets the criteria. The findings presented in

Table 4. Results of the instrumental variable method.

Variables	WFL	VFL
	(1)	(2)
AMS	1.285 **	1.099 *
	(1.98)	(1.70)
Control variables	Yes	Yes
Observations	483	483
R-squared	0.189	0.189
r2_a	0.163	0.163
F	7.545	7.836

Note: The t-statistic values estimated are in parentheses. * p < 0.1 and ** p < 0.05.

indicate that machinery services remain a significant positive factor in food loss reduction. The values in Column (1) show that the influence of AMS on the weight of food loss is significant, with a coefficient of 1.285 at the 5% significance level. The finding according to Column (2) demonstrates the role of AMS in reducing the value of food loss with the coefficient of 1.099 at a significance level of 10%. These findings are consistent with the baseline regression results, which indicate that agricultural machinery services have a substantial role in reducing food loss after solving the endogenous issue. Therefore, the conclusions of this study remain robust.

4.3. Mechanism Analysis

According to the above theoretical analysis, farmers outsource some links of agricultural production to machinery service organizations with advanced technology that alleviates the constraints of labor, capital, and land. It affects food loss through the optimization of factor allocation and the introduction of technology. With the aim to clarify the internal mechanism, the mediation effect of machinery services on food loss are verified from the perspectives of average yield per mu and average labor input time per mu.

The Sobel test in

Table 5. The mediation effect of factor allocation.

Variables	GYP	WFL	VFL
	(1)	(2)	(3)
GYP		0.489 **	0.487 *
		(2.00)	(1.96)
AMS	0.064 *	−0.895 ***	−0.893 ***
	(1.79)	(−4.80)	(−4.74)
Control variables	Yes	Yes	Yes
Observations	483	483	483
R-squared	0.084	0.217	0.218
Adjusted R-squared	0.0542	0.190	0.192
F-value	1.977	8.939	9.242

Note: The t-statistic values estimated are in parentheses. * p < 0.1, ** p < 0.05, and *** p < 0.01.

rejects the original hypothesis, indicating that the mediation effect is notable. Column (1) reports that machinery services have a remarkably positive effect on factor allocation at the level of 10%, which suggest that machinery services promote the optimization of factor allocation. Columns (2) and (3) add the variable of average yield per mu on the basis of baseline regression. They exhibit that the coefficient of machinery services changes from −0.862 to −0.489 and −0.864 to −0.487, respectively. The results indicate that the optimization of factor allocation plays a partial mediation effect on the food loss reduction effect of machinery services. This conclusion supports Hypothesis 2.

The results of the Sobel test in

Table 6. The mediation effect of technology introduction.

Variables	LAP	WFL	VFL
	(4)	(5)	(6)
LAP		0.257 **	0.263 **
		(2.32)	(2.33)
AMS	−0.649 ***	−0.697 ***	−0.691 ***
	(−7.76)	(−3.47)	(−3.38)
Control variables	Yes	Yes	Yes
Observations	483	483	483
R-squared	0.249	0.217	0.220
Adjusted R-squared	0.225	0.191	0.193
F-value	8.482	9.100	9.371

Note: The t-statistic values estimated are in parentheses. ** p < 0.05, and *** p < 0.01.

reject the original hypothesis, which indicates that the mediation effect is significant. Column (4) reports that machinery services have a considerably positive effect on the introduction of technology at the level of 1%. These results indicate that machinery services promote the introduction of technology in agricultural production. On the basis of the baseline regression, Column (5) supplements the variable of average labor input per mu. The results show that the coefficients of machinery service vary from −0.862 to −0.697 and −0.864 to −0.691, respectively. The results indicate that the introduction of technology plays a partial mediation effect on the food loss reduction effect of agricultural machinery services. This conclusion supports Hypothesis 3.

4.4. Heterogeneity Analysis

4.4.1. Heterogeneity of Production Links

There are differences in the demand for labor, knowledge, and technology in various production links. According to the differences in labor intensity and mechanization degree required in various links, production links are divided into labor-intensive links (including land preparation, sowing, and harvesting operations) and technology-intensive links (filed management), respectively. The machinery services of harvesting links represent labor-intensive services, while the machinery services of field management links represent technology-intensive services to examine the influence of purchasing machinery services of different links on the food loss.

The results shown in

Table 7. Effects of AMS on food loss with different production links.

Variables	WFL	WFL	VFL	VFL
	(1)	(2)	(3)	(4)
Labor intensity	−0.273 *		−0.286 *
	(−1.69)		(−1.78)
Technology-intensive		−0.533 ***		−0.537 ***
		(−3.26)		(−3.27)
Control variables	Yes	Yes	Yes	Yes
Observations	483	483	483	483
R-squared	0.188	0.198	0.190	0.200
Adjusted R-squared	0.161	0.172	0.164	0.174
F-value	7.384	8.167	7.798	8.521

Note: The t-statistic values estimated are in parentheses. * p < 0.1 and *** p < 0.01.

document that both labor-intensive and technology-intensive machinery services are positive in reducing food loss. Labor-intensive machinery services significantly affect food loss at the statistical level of 1%, with coefficients of −0.533 and−0.537. Technology-intensive machinery services are significant at the statistical level of 10%, with coefficients of −0.273 and−0.286. However, compared with labor-intensive machinery services, the effect of technology-intensive machinery services on reducing food loss is smaller.

4.4.2. Heterogeneity of Land Fragmentation

We further investigate the impact of machinery services on food loss based on the differences in land fragmentation. We divided the study subjects into two groups based on the mean land fragmentation. The results are displayed in

Table 8. Effects of AMS on food loss with different farmland fragmentation.

Variables	WFL	WFL	VFL	VFL
	(5)	(6)	(7)	(8)
AMS	−1.242 ***	−0.764 ***	−1.256 ***	−0.751 ***
	(−3.98)	(−2.74)	(−4.01)	(−2.70)
Control variables	Yes	Yes	Yes	Yes
Observations	163	320	163	320
R-squared	0.396	0.183	0.389	0.184
Adjusted R-squared	0.334	0.143	0.327	0.143
F-value	6.424	4.544	6.246	4.560

Note: The t-statistic values estimated are in parentheses. *** p < 0.01.

. The influence of machinery services on food loss is significantly negative with coefficients of −0.764 and −0.751 in the high group. The impact is also significantly negative with coefficients of −1.242 and −1.256 in the low group. It can be seen that when the degree of land fragmentation is low, the effect of machinery services on reducing food loss is more significant.

4.4.3. Heterogeneity of Service Quality

Machinery service belongs to the typical principal–agent operation model [50]. The behavior of the service supplier and the advanced technology level of operation machinery affect the service quality. The impact of machinery services on food loss is further verified based on the different quality of service. We evaluate the service quality from multiple dimensions by investigating the satisfaction of farmers with machinery services based on the research of Cai and Liu [57]. We select operation efficiency, operation cost, operation attitude, operation machinery refinement, operation standard standardization, and operation intelligence to construct the comprehensive evaluation index system of service quality and use the factor analysis method to calculate the comprehensive index score. We divide the samples into two groups based on the mean of service quality. The results are shown in

Table 9. Effects of AMS on food loss with different service quality.

Variables	WFL	WFL	VFL	VFL
	(9)	(10)	(11)	(12)
AMS	−0.857 ***	−1.092 ***	−0.828 ***	−1.125 ***
	(−3.33)	(−3.40)	(−3.17)	(−3.52)
Control variables	Yes	Yes	Yes	Yes
Observations	203	280	203	280
R-squared	0.305	0.237	0.308	0.239
Adjusted R-squared	0.249	0.194	0.252	0.196
F-value	5.468	5.466	5.541	5.531

Note: The t-statistic values estimated are in parentheses. *** p < 0.01.

. In the high group, machinery services have a higher impact on food loss, with coefficients of −1.092 and −1.125. In the low group, machinery services have a lower impact on food loss, with coefficients of −0.857 and−0.828.

5. Discussion

Agricultural machinery as one of the key technologies that plays a crucial role in promoting agricultural production and ensuring food security. Due to different natural resources and supportive policies, its development level and route vary in different countries. Purchasing socialized machinery services has become an important way to realize modern agriculture in China. This study provides a more comprehensive analysis of the impact of machinery services on food loss at the producer level, and the findings have important theoretical and practical value.

According to the induced technology innovation, the farmers will choose to use agricultural machinery instead of using labor force. Although the effects of the purchase of machinery services have been widely examined [20,22,23], the connection between agricultural machinery services and food loss is still underdiscussed. To the best of our knowledge, this study is the first to clarify the causal effect of machinery services on food loss. The findings of this study are conducive to a further understanding of the impact of socialization services. Significantly, some scholars argue that the information asymmetry between farmers and agricultural machinery service providers may lead to lower service quality by providers and “excessive supervision” by farmers, resulting in efficiency losses [50]. This risk is more likely to occur in cross-regional operation service models. In recent years, local service providers have grown rapidly with the agricultural machinery subsidy policies. Moreover, rural China is a typical relational society which will reduce opportunism and make the positive externalities of agricultural services. The findings indicate that the asymmetrical information between and incomplete contract does not hinder the food loss reduction effect of machinery services. This article confirms the economic effect of agricultural machinery, but it should be noted that it is considered to possibly increase the energy burden and lead to environmental pollution [58]. To promote the sustainable development of economy, energy, and environment, it is necessary to strengthen the R&D and application of agricultural machinery using clean energy, such as solar and electric power.

The results of the mechanism analysis indicate that agricultural machinery services can reduce food loss by optimizing the allocation of production factors and technology introduction. First, agricultural machinery services can break the capital and technical constraints of small farmers, and help small farmers transfer land to reach an appropriate operation scale [59]. Second, the rural laborers who are left behind are generally feminized and aged, which limits the adoption and diffusion of advanced agricultural technologies, resulting in reduced agricultural productivity [8]. Agricultural machines, particularly those with higher horsepower, are more conducive to promoting agricultural productivity [60]. During the survey, we found that agricultural machineries provided by service organizations were more advanced than those owned by small farmers. This is also highlighted in the existing research [61]. In summary, these findings contribute to a more comprehensive understanding of how agricultural machinery services impact food loss. Policy makers should further improve land circulation, the financial market, and provide institutional guarantee for the optimal allocation of production factors in rural China. In addition, efforts should be made to improve technological research on agricultural machinery equipment.

Heterogeneity in terms of production link, land fragmentation, and service quality are further analyzed in this study. First, the adoption of services in labor-intensive production links is more conducive to reducing food loss. Agricultural machinery services have labor substitution effects [61]. The high development level of the labor-intensive service market is driven by farmers’ greater demand. The economies of scale have lowered the threshold for entry into the market in return [62]. The effect of machinery services on reducing food loss varies depending on the fragmentation of the land and the quality of services. It is more significant when the degree of fragmentation is low. The fragmentation of land increases operating costs and restricts the application of large-scale agricultural machineries [7,63], thus reducing the effect. In addition, the quality of services also has heterogeneous impacts on food loss reduction. The level of the standardization of agricultural machinery services is low in China, due to the lack of formal contracts. It is difficult to monitor and evaluate the quality of services [64]. Therefore, it is necessary to further strengthen farmers’ contract awareness, clarify service requirements, and improve service quality.

To address the food loss issue, it is necessary to break down the loss figures and investigate the influencing factors at specific stages of the food value chain as the causes are different. This study focuses on food loss at the micro level and identifies effective interventions to reduce it. Further research is needed at the meso- and macro-level. This study uses weight and value indicators to estimate food loss, highlighting that food loss includes both quantity and quality. While food loss is measured by farmers’ self-report, researchers may explore objective measurement methods like filed experiments and suitable calculations to obtain more accurate data in future research.

6. Conclusions and Policy Implications

In order to ensure national food security, it is necessary to focus not only on increasing production, but also on reducing losses. Under the background of the limited potential of food production, more attention should be paid to the control of food loss. Mechanization is the key way to agricultural modernization. In recent years, the rapid development of agricultural productive service organizations has provided innovative solutions to reduce farmers’ food loss. Therefore, it is of great practical significance and value to deeply study the potential impact of machinery services on food loss.

Three main conclusions were drawn from this study: First, food loss need to be measured in terms of both quantity and quality. Food loss is an important issue that requires attention. Second, we found that machinery services can significantly reduce food loss, affirming the importance of machinery services in promoting agricultural production. This study gradually controlled the basic characteristics of household heads, family characteristics, agricultural production and management characteristics, and the characteristics of villages, and confirmed the important impact of machinery services on food loss reduction. Third, we verified the mechanism of machinery services on food loss. The optimization of factor allocation and the introduction of technology played a partial mediation effect in the food loss effect of machinery services. Fourth, different types of production links, degrees of land fragmentation, and service quality have an impact on the role of machinery services in reducing food loss. Specifically, compared with technology-intensive links, machinery services in labor-intensive links have a more significant effect on reducing losses. Moreover, the higher the degree of land fragmentation, the more limited the positive impact of machinery services on food loss reduction. For farmers who have a higher evaluation of service quality, machinery services can more effectively reduce food loss.

This study discussed the impact of machinery services on reducing food loss. The policy implications of the research findings on promoting food loss reduction through machinery services are as follows: First, we should improve the level of agricultural machinery services; expand the breadth and depth of machinery services in agricultural production; develop a full-linkage service supply to achieve diversified social services in areas with conditions; inject various technologies into agricultural production through outsourcing services; promote various business entities to improve agricultural production efficiency; and enable small farmers to better integrate into modern agriculture. Second, it is necessary to increase the intensity of financial investment; accelerate the renewal and upgrading of agricultural machinery equipment by service suppliers through subsidies and other preferential policies; increase funding support for the research and development of agricultural machinery equipment; encourage the innovation of technology and the application of technological achievements; and improve the level of research, development, and manufacturing of agricultural machinery equipment. Third, it is important to focus on the quality of agricultural machinery service supply and regulate the service market. We should explore specific quantitative standards for different service contents according to local conditions and improve service supervision and management measures, especially in terms of increasing the standardization of technical–intensive links such as fertilization and pest control, which are difficult to supervise and are prone to moral hazards. This will alleviate farmers’ concerns about service effectiveness and guide the development of machinery services towards standardization and high quality. Fourth, for countries and regions where small-scale operations are predominant, the government should encourage farmers to use land exchange and other forms of circulation to achieve an appropriate scale and continuous operation and reduce the land constraints faced by socialized agricultural machinery services. We should increase public investment in infrastructure and create favorable conditions for agricultural mechanization and reducing food loss by improving irrigation and water conservancy, land leveling, and optimizing machinery farming road conditions.