1. Introduction
The core of agricultural economic growth lies in improving agricultural production efficiency (APE), and the key to solving the problem lies in realizing agricultural-scale operation. Agricultural scale management mainly includes land-scale operation (LSO) and service-scale operation (SSO); the former is mainly realized by land transfer, and the latter is mainly realized by agricultural socialization service. APE refers to the ratio relationship between the input of various production factors and the output of agricultural products in the process of agricultural production. Under the existing rural land system framework, China’s basic conditions with a small amount of cultivated land cannot be changed in a short period, and the LSO promoted by land transfer has become an important way to improve APE [
1]. In the late 1980s, the state issued relevant policies on land transfer, and governments at all levels began to issue a series of agricultural land policies one after another to encourage the orderly transfer of the contracted operation right of farmland and develop moderate agriculture-scale operation. However, due to the large number of small-scale farmers in China, it is impossible to solve the problem of the tense relationship between humans and land, which leads to the fact that the operation pattern of land decentralization has not changed basically [
2,
3]. During “The Twelfth Five-Year Plan” period, the national land transfer area increased by an average of 24.0% per year. After entering “The Thirteenth Five-Year Plan”, the land transfer area increased by an average of only 4.4% per year from 2016 to 2018. In recent years, the growth rate of land transfer areas has obviously slowed down [
4]. About 2/3 of the cultivated land in China is still operated by the original contracted farmers
1. In this way, farmers can actively participate in agricultural-scale operations and benefit from it. In the relevant policy documents, the state proposes to encourage various forms of moderate-scale operation and has repeatedly emphasized the development of agricultural socialization services. It can be seen that policymakers pay more and more attention to agricultural socialized services. By the end of 2020, there were 900,000 agricultural socialized service organizations in China, and the agricultural production custody service area exceeded 10.7 million ha. This means that the specific path to improve APE through agricultural-scale operation will not be unique.
However, the effect of agricultural-scale operation is not necessarily positive. A large number of empirical studies show that land transfer has a significant positive effect on improving APE [
5,
6,
7]. Land transfer rearranges land property rights among subjects with different behavioral abilities and decision-making preferences [
8]. Farmers participate in the division of labor activities according to their own endowment and comparative advantages to produce the division of labor efficiency and promote the overall optimal allocation of resources [
9]. For example, the farmers in Sweden want to evaluate the lessee’s ability to manage their own farmland according to the cleanliness of the existing farmland before transferring the land, and provide help to the lessee after transferring the land to make agricultural production more efficient [
10]. However, if the cost of land transfer increases or the scale of land transfer expands indefinitely due to the asymmetry of market information, there may also be invalid land transfer, resulting in an uneconomical scale [
11]. At the same time, because land transfer in China is mainly based on short-term contract arrangements at present [
12], the difference in property right duration and stability between the own plots and the transferred plots directly affects farmers’ production and investment behavior [
10,
13]. If farmers are not willing to invest in short-term operations, it will have a negative impact on agricultural production. With the weakening of the role of land-scale operation in agricultural production, SSO has found the “intersection” between farmers and scale economy under the current institutional framework [
2,
14]. Compared with LSO, SSO is less restricted by strong constraints such as human–land relationships and farmland systems [
15,
16]. It has the potential to quickly improve APE [
17]. In the process of agricultural production, if farmers only engage in relatively efficient production links and outsource relatively inefficient production links by purchasing services, the structural effect will improve the average efficiency of the whole agricultural production process [
18,
19]. However, some scholars have put forward different opinions. For example, according to the theory of scale economy, if all kinds of elements invested in agricultural production and operation cannot reach the optimal combination, farmers can neither achieve scale operation nor improve the average productivity of production links through the outsourcing of the production links [
20]. At the same time, there are still some problems in the current agricultural socialized service, such as the diversified demand is difficult to meet, the public welfare service system is not perfect, the connection mechanism between service subjects is not perfect, and the growth of business service subjects is slow, which will seriously affect the APE [
21,
22]. Therefore, the validity of LSO promoted by land transfer still needs further proof, and the relationship between LSO, SSO, and APE needs to be explored at the same time to make clear the concrete realization path of agricultural modernization in China.
Based on this, the paper uses meta-analysis to focus on agricultural-scale operations. By retrieving published journal articles, the observation period can be expanded, and a larger sample size can be obtained. By integrating and analyzing the existing empirical research literature on the relationship between LSO, SSO, and APE, 117,627 data from 68 articles are finally included in the research database. The innovation of this paper lies in that by deeply exploring the relationship between different agricultural scale management modes and APE, it can be determined whether LSO promoted by land circulation and SSO promoted by social services are conducive to the improvement of APE. This can solve the current differences in the path of agricultural scale management. On the basis of answering these questions, this study makes theoretical contributions through a moderator analysis, which is helpful to better understand the underlying economies of scale theory. It can explain the relationship between agricultural-scale operations and APE more scientifically and reasonably. The structure of this paper is as follows: The second part introduces the theoretical analysis and conceptual model. The third part introduces the data sources and research methods. The fourth part discusses the results. The fifth part summarizes the conclusions of this paper, puts forward proposals, and discusses the scope for future research.
2. Theoretical Analysis and Research Hypothesis
Regarding the relationship between agricultural scale management and APE, the theory of LSO advocates changing the decentralized operation pattern of small-scale farmers in China through the transfer of farmland. The theory of SSO advocates that through the division of labor in the production process and the subdivision of farmers’ operation rights, on the one hand, farmers are involved in the socialized division of labor [
23]; on the other hand, large-scale outsourcing service supply is formed. However, the two views are not contradictory. Promoting LSO through land transfer is considered the only way to transform traditional agriculture into modern agriculture to improve the input level and intensive degree per unit area. Land transfer improves the phenomenon of decentralization and fragmentation caused by household contracts and realizes large-scale and intensive farming to a certain extent [
24]. It is helpful for agricultural machinery to carry out large-scale operations [
25] and improve production efficiency. Although the LSO has reached the “deep water area” of reform at present, it has been widely accepted by farmers [
26]. At the same time, the agricultural socialized service provides important support for promoting the operation of agriculture scale and improving the comprehensive agricultural production capacity in China [
27]. It also has a great thrust on the transformation of traditional agricultural production methods and operation models. Although the agricultural socialized service system is not yet mature, it is undeniable that it helps small-scale farmers to integrate into the modern agricultural production and operation system. It helps to overcome the problem that traditional small-scale farmers have low endowments and cannot match the modern market economy. Based on this, this study puts forward a hypothesis:
H1a. Agricultural-scale operations can play a positive role in improving APE.
H1b. LSO and SSO can play a positive role in improving APE.
However, the research in different situations leads to different conclusions; that is, the research results are influenced by moderator variables. Arthur Jr et al. [
28] defined the moderating variable as any variable contained in the analysis that can explain or help explain more differences. Through a detailed review of the existing literature, this paper summarizes the moderator variables that may affect the relationship between LSO, SSO, and APE, which can be roughly divided into five aspects: the agricultural production and operation environment, characteristics of the agricultural location, degree of farmers’ participation, types of APE, and research situation.
Agricultural production and operation environment: In 2017, after the 19th Session of the National Congress of the Communist Party of China, China began to develop an agricultural socialized service system for small-scale farmers. This shows that China should not only adhere to the direction of LSO but also recognize the present situation, which is that small-scale farming is the basic form of agriculture [
29]. It is necessary to strengthen the socialized service for small-scale farmers and handle the relationship between moderate-scale operations and the development of small-scale farmers correctly. It can be seen that different agricultural production and operation environments are formed under different policy guidance. Because of the differences in the data acquisition time of different articles, this paper takes 2017 as the time node to represent different agricultural production and operation environments. Based on this, this study puts forward a hypothesis:
H2a. The agricultural production and operation environment can moderate the relationship between LSO and APE; that is, before 2017, this relationship is stronger.
H2b. The agricultural production and operation environment can moderate the relationship between SSO and APE; that is, after 2017, this relationship is stronger.
The characteristics of the agricultural location: Agricultural production activities are strictly restricted by time and space conditions [
30]. A scholar pointed out that if land is allowed to transfer freely, the average productivity of agriculture in most underdeveloped countries will be doubled, and if the indirect effects (such as human capital accumulation) are considered, the impact of land transfer on agricultural productivity will be more considerable. It can be seen that LSO has different effects in the areas with different levels of economic and agricultural development [
31]. Under the current agricultural production environment in China, making full use of the advantages of SSO can significantly improve the efficiency of production, especially in areas with large cultivated land per household and less agricultural labor [
32]. Due to the huge geographical differences among the provinces in China, the land and SSO cannot be generalized [
33]. For example, China has thirteen major grain-producing areas, and the provinces in the major grain-producing areas have superior natural conditions, which is very conducive to the development of agricultural production activities. Therefore, the characteristics of the agricultural location can affect the research results. Based on this, this study puts forward a hypothesis:
H3a. The characteristics of the agricultural location can moderate the relationship between LSO and APE; that is, this relationship is stronger in major grain-producing areas.
H3b. The characteristics of the agricultural location can moderate the relationship between SSO and APE; that is, this relationship is stronger in major grain-producing areas.
The degree of farmers’ participation: Because of the heterogeneity of farmers’ participation in LSO or SSO, the different forms of variables in the model can also affect the research results [
34]. Represented by the form of independent variables in the sample literature model, the variable in the form of 0–1 only distinguishes whether to participate in land transfer or social services. Such variables may not be enough to describe the actual situation in detail, which leads to the overestimation of the empirical results, while continuous variables can measure the participation level more accurately. Therefore, the degree of farmers’ participation can affect the research results. Because different forms of variables are used in the measurement of farmers’ participation in the various literature, this study puts forward a hypothesis:
H4a. The degree of farmers’ participation can moderate the relationship between LSO and APE; that is, when 0–1 is taken as the independent variable, LSO has a positively significant impact on APE.
H4b. The degree of farmers’ participation can moderate the relationship between SSO and APE; that is, when 0–1 is taken as the independent variable, SSO has a positively significant impact on APE.
The types of APE: It is impossible to separate the two dimensions of “yield” and “income” for discussion when studying the land or SSO [
35]. The former is the main concern of the government, while the latter is the demand for micro-operation subjects. As market subjects and rational economic men, farmers’ behavior is driven by the motivation of profit maximization [
36]. It is difficult to keep consistent with the government’s public welfare goal of ensuring agricultural product supply when making land transfer choices. In contrast, in the construction of an agricultural socialized service system, more emphasis is placed on systematic and supporting services, and it is easier to focus on public welfare services, especially government-led public welfare services [
37]. It can be seen that LSO and SSO can have different effects on the “production efficiency” and “output value” of farmers and the “yield” of the government. Because the indexes with different connotations are used in the measurement of APE in the various literature, this study puts forward a hypothesis accordingly:
H5a. The types of APE can moderate the relationship between LSO and APE; that is, when the type is “efficiency or output value”, LSO has a positively significant impact on APE.
H5b. The types of APE can moderate the relationship between SSO and APE; that is, when the type is “yield”, SSO has a positively significant impact on APE.
Research situation: The data for studying LSO or SSO are mainly divided into micro data and macro data. Different data types can affect the empirical results, leading to completely different conclusions on the same issue [
38]. Micro data are more targeted, but the survey results are influenced by the subjective cognition of the respondents, which affects the objectivity and repeatability of the empirical analysis results. At the same time, there are cases in which the respondents do not cooperate, resulting in a biased estimation of data projections. Comparatively speaking, macro data are more objective and complete, and the measurement results of variables are repeatable and more convincing [
39]. Therefore, the research situation can affect the research results. Based on this, this study puts forward a hypothesis:
H6a. The research situation can moderate the relationship between LSO and APE; that is, this relationship is stronger under the macro situation.
H6b. The research situation can moderate the relationship between SSO and APE; that is, this relationship is stronger under the macro situation.
To sum up, due to the complexity of agricultural-scale operations, the existing research on the influence of LSO and SSO on APE by theoretical analysis or econometric methods has not reached a consistent conclusion. Therefore, this paper adopts the method of meta-analysis to analyze the findings of multiple independent empirical studies comprehensively and quantitatively in the existing literature, clarify the moderator variables, and answer the dispute about whether different agricultural-scale operation modes can improve APE. The conceptual model of this study is shown in
Figure 1.
4. Results and Discussion
4.1. Overall Effect
The comprehensive effect value can reflect the correlation between the variables, and the meta-analysis results using the random effect model are shown in
Table 2. The effect value of the research on LSO and APE is 0.174. The effect value of the research on SSO and APE is 0.162. The effect value of the research on agricultural-scale operation and APE is 0.168. This shows that there is a positive correlation between LSO, SSO, and APE; that is, agricultural-scale operation contributes to the improvement of APE, and the impact of LSO is stronger. H1a and H1b are established. LSO concentrates land resources and replaces land reasonably, which can avoid small-scale diseconomy caused by scattered plots [
42]. It has played an important role in improving APE for a long time. However, the new agricultural socialized service system is still in the process of continuous improvement, and the specialized division of labor is not very detailed. The division of labor makes it difficult to carry out labor supervision, leading to higher transaction costs. At the same time, due to the diversification of the farmers’ economic characteristics, the development of agricultural socialized services is more difficult, and its role in agricultural production is not fully highlighted. However, this does not mean that the effect of agricultural SSO will always be lower than that of LSO in the future. If the problems of the existing socialized service system can be gradually solved in the future, SSO will have a stronger effect on improving APE. Therefore, improving APE is inseparable from the cross-integration of the two.
4.2. Meta-Regression Analysis
As seen from
Table 3, from the perspective of the agricultural production and operation environment, the coefficient of this variable is positive in the regression results of the sample documents of LSO, and it is significant at a 10% level (
β0 = −0.195,
p = 0.054). H2a is established. However, it is insignificant in the regression results of the sample documents of SSO. H2b is not verified. From a practical point of view, it shows that the positive relationship between LSO and APE is stronger in the early research. This may be related to the increase in the transfer cost of land contractual operation rights [
43] and the slowdown in the growth rate of land transfer areas in recent years. However, after 2017, the agricultural socialized service continued to develop [
44], and the SSO was favored, which made this kind of sample literature data more distributed, resulting in the variable not playing a moderating role.
From the characteristics of the agricultural location, in the regression results of the sample literature of LSO, it has a significant positive moderator effect (
β0 = 0.114,
p = 0.053). H3a is established. It indicates that the effect of LSO is easily influenced by the agricultural location. The main grain-producing provinces have excellent climate, soil, and precipitation conditions [
45]. Moreover, the scale of cultivated land per household is large, and the degree of land concentration is high, making the correlation between LSO and APE in the major grain-producing areas stronger than in the non-major grain-producing areas. The moderate effect of this variable is insignificant in the regression results of the SSO sample literature. H3b is not verified. The reason is that the form of service scale can avoid the constraints of human–land relationships or farmland systems to a certain extent. With the emergence of intelligent agricultural machinery, more agricultural machinery with precise positioning, terrain analysis, path planning, and other functions are put into agricultural productive services [
46,
47,
48,
49,
50], which break the limitation of regional natural conditions and make the SSO develop more widely.
From the perspective of farmers’ participation, the literature meta-regression of the LSO samples results show that farmers’ participation is significantly positive at the 10% level (
β0 = 0.130,
p = 0.058). H4a is established. On the contrary, it is insignificant in the regression results of the SSO sample literature. H4b is not verified. It shows that the degree of farmers’ participation affects the relationship between LSO and farmers’ production efficiency. If the research is based solely on whether they participate in land transfer, it will produce results higher than the actual impact level. Comparatively speaking, SSO is more flexible, and social services can be divided into many different service links [
51,
52]. Therefore, if the variable in the form of 0–1 is used to measure whether farmers buy services in different links, respectively, the degree of their participation is also measured in detail. This can explain why the moderator variable is insignificant.
The types of APE are an important factor affecting the regression results of the model. The meta-regression results of the two types of samples show that the types of APE are significantly positive at the 1% level and significantly negative at the 10% level (β0 = 0.264, p = 0.008; β0 = −0.186, p = 0.057). H5a and H5b are established. It shows that the correlation between LSO and APE is stronger when “efficiency or output value” is the dependent variable, and the correlation between SSO and APE is stronger when “yield” is the dependent variable. Because the motivation of profit maximization drives farmers’ behavior, the goal of land transfer is not to ensure food supply but to improve production efficiency or income. In contrast, the government dominates agricultural socialization services, and the services provided are of a public welfare nature. This makes farmers’ agricultural production goals consistent with government goals, and “yield” is more easily concerned.
The research situation’s meta-regression results of the two kinds of samples show that the moderator variable is significant at 1% and 10% levels, respectively (β0 = −0.488, p = 0.006; β0 = −0.240, p = 0.054). H6a and H6b are established. They show that the correlation between LSO, SSO, and APE is stronger under the macro situation.
4.3. Subgroup Regression
According to the results of meta-regression, the significant moderator variables are grouped to find more accurate heterogeneity information, and the results are shown in
Table 4 and
Table 5.
1. LSO: From the agricultural production and operation environment, the effect value of document data collection time before 2017 is higher than that after 2017, and the heterogeneity is significant. From the characteristics of the agricultural location, the effect value of the major grain-producing areas is higher than that of the non-major grain-producing areas, and the heterogeneity is significant. From the degree of farmers’ participation, the effect of the variable in the form of 0–1 is higher than that of the continuous variable, and the heterogeneity is significant. From the different types of APE, “efficiency or output value” has a significant impact, while “yield” has no significant impact, and heterogeneity is significant. From the research situation, the effect value of using the macro situation is higher than that of the micro situation, and the heterogeneity is significant.
2. SSO: From the types of APE, “efficiency or output value” has a significant impact, while “yield” has no significant impact, and the heterogeneity is significant. From the research situation, the effect value of using the macro situation is higher than that of the micro situation, and the heterogeneity is significant.
4.4. Sensitivity Analysis
In order to test the robustness of the estimation results of the previous model, the sensitivity analysis is carried out by using the leave-one-out method [
53]. The effects of individual studies are eliminated one by one, and the remaining 34 and 32 studies are synthesized, respectively, and get the corresponding effect value. By integrating the remaining effect value, the difference between the variation interval of the effect values and the results of the overall effect analysis is observed. The results show that the literature range of the LSO sample is 0.114–0.180, and the literature range of the SSO sample is 0.134–0.167, which is not much different from the corresponding overall effect analysis results, so it can be considered that the analysis results obtained above are robust.