The Impact of Farm Structure on Agricultural Growth in China

Abstract

Farm structure has changed significantly since the land transfer policy was introduced in China. The quantity of and land area utilized by scale farmers and new agricultural entities are increasing, while the numbers of smallholders are decreasing. To analyze the impact of farm structure on agricultural growth in China, this study used provincial-level panel data from 2010 to 2021 and a fixed-effects model for empirical testing. The results show that (1) the structural change among farmers has a positive effect on agricultural growth, accounting for 16% of the contribution; (2) compared with scale farmers, new agricultural entities (NAEs) play a stronger role in improving agricultural growth, which is approximately five times higher than that of the former; and (3) unlike scale farmers, who only promote agricultural growth in central and major grain-producing areas, new agricultural entities have a positive effect on the entire area. Therefore, this study emphasizes the role of human capital in agricultural growth, especially in terms of promoting new agricultural entities, supporting land transfer, and optimizing farm structure.

1. Introduction

Since land transfer has been allowed in China, identifying new sources of agricultural growth, with the aim of linking land use to agricultural development, has become crucial. China’s household responsibility system (HRS) has achieved remarkable agricultural performance [1] but results in a pattern of small-scale decentralized management which is inefficient and does not meet the requirements of modernized agricultural development. In practice, agricultural development in various countries relies on large-scale management. In recent years, a series of policies announced by the Chinese government strongly supported land transfer to readjust human–land distribution. The Rural Land Contracting Law legalized land transfer and the reform of “separating three property rights” aimed to encourage this practice. Land flow from smallholders to professional farmers is spreading in rural areas, and the occurrence rate of land transfer increased from 4.4% in 2005 to 36.7% in 20221. Agricultural output continued to rise simultaneously: the gross output value of agriculture in China increased from CNY 19,613.37 billion in 2005 to CNY 84,438.58 billion in 20222. Land transfer is significant in promoting the agricultural economy [2], but how does it contribute to agricultural growth in China?

The promotion of agricultural growth is often attributed to land-scale management. However, previous studies have disagreed on the relationship between land scale and production efficiency. Studies have suggested positive, negative, U-shaped, and uncertain relationships [3,4,5]. An increased land-scale measurement only reflects an increase in the quantity of land a farmer holds [6]. The dichotomy used to analyze scale management undermines the complexity of land transfer. Prior analyses were limited to a single group of smallholders or large-scale farmers and cannot clarify the impact of any change from smallholding to large-scale farming. Moreover, the hypothesis that different farmers with the same land scale have consistent production efficiency is unconvincing. Farmers are important subjects in specific agricultural production [7]: they are distinguished by diverse types and heterogeneous human capital, make input factor decisions, and determine agricultural production performance. Rural economic development depends more on the effective utilization than the sufficiency of land resources [8]. This indicates that the scale effect has been exaggerated, and the effect of heterogeneous farmers has been ignored. In other words, the effectiveness of the farmers who conduct scale operations determines whether there are positive outcomes of land-scale management. It is important to focus on the role of farmers in land use and investigate the effect of heterogeneous farmers on agricultural growth.

Land transfer drives different types of farmers and changes their relative composition in the land use system. Structural change among farmers is the internal characteristic and result of land transfer, and the resulting economic efficiency merits attention. In China, the HRS enables a peasant household to become the basic unit of agricultural production. Land transfer aims to address the misallocation of land among peasant households caused by the HRS [9,10]. According to land area and operation methods, farmers can be divided into three types: smallholders3, scale farmers, and new agricultural entities (NAEs). Scale farmers and NAEs are professional farmers who carry out land-scale management, unlike smallholders. Scale farmers include family farms and large-scale professional farmers. NAEs are represented by agricultural enterprises or specialized cooperatives, and they always have a large operating area and intensive production.

Farm structure is most simply characterized by the size of farms and their characteristics [11,12]. Farm structure has changed rapidly with the acceleration of land transfer in China. Smallholders decreased from 238.89 million in 2012 to 221.58 million in 2022, scale farmers increased from 8.56 million in 2009 to 11.33 million in 2022, and NAEs increased from 0.23 million in 2009 to 2.03 million in 2021. Land transfer reallocates and optimizes land resources among different farmers [13,14]. The proportion of land area used by scale farmers and NAEs increased from 16.7% in 2010 to 35.7% in 20224. Nevertheless, few studies comprehensively consider the structural change among smallholders, scale farmers, and NAEs, and the economic effects are far from clear, with limited empirical evidence in the literature.

Therefore, the main purpose of this study is to explore the impact of the structural change among farmers on agricultural growth in the context of land transfer. This study supplements the existing literature in the following four ways: (1) we divide farmers into smallholders, scale farmers, and NAEs to identify and reflect the structural change among farmers; (2) we establish the relationship between land transfer and farm structure, enriching the understanding of the benefits of land transfer; (3) we focus on the attributes of farmers as production factors and emphasize the importance and heterogeneity of human capital; and (4) we empirically analyze the effect of farm structure on agricultural growth and clarify the human–land interactions and economic effect of farm structure.

The next section reviews the theoretical framework and hypotheses, explaining the differences between smallholders, scale farmers, and NAEs and analyzing the positive effect of farm structure on agricultural growth. Section 3 constructs a Cobb–Douglas production function based on the structural change among farmers and produces descriptive statistics on data for 30 provinces in China from 2010 to 2021. Section 4 presents a summary of empirical evidence regarding the effect of farmer categories on agricultural growth. Section 5 discusses the results, limitations, and prospects of this study. Section 6 proposes corresponding policy recommendations based on the research findings.

2. Theoretical Framework and Hypotheses

2.1. Farm Structure and Heterogeneous Human Capital

From the perspective of heterogeneous farmers, the development of human capital levels and the upgrading of production methods evolve with structural change in farming, which promotes agriculture development.

Endogenous Growth Theory and Human Capital Theory emphasize the effect of human capital on economic growth. Researchers have acknowledged that rural elites are the crucial endogenous actors in promoting rural development [15]. Johnson (1993) claimed that human capital is important for developing countries in agricultural transition [16]. Farmers make strategic decisions regarding the use of the farm resources and bear all risks associated with the farm [17]. The human capital of the farmer is reflected in their ability to manage or allocate production factors, such as land and labor. Regarding land factors, the human capital of the farmer is determined by whether they can make rational decisions to rent in or rent out land. To transform a small-scale agricultural operator into a large-scale agricultural operator, it is not just necessary to change the land scale [18]. Land scale expansion makes the skills and capabilities of farmers more important [19,20]. Regarding technology adoption, farmers have higher human capital and are more willing to adopt new technologies to earn higher profits [21]. Only large-scale operators with rich planting experience can significantly contribute to reducing production costs and developing the positive effect of scale operations [22]. The human capital of farmers in land use and modern agriculture cannot be ignored. Heterogeneous farmers can provide insights into farm structure and related economic effects.

In theory, with a complete factor market, farmers with higher human capital will continue to expand their scale until their marginal output is equal to their marginal cost, which means that the higher the human capital of farmers, the larger their land scale, and the two are positively correlated [23,24]. According to

Table 1. Differences among smallholders, scale farmers, and NAEs.

		Smallholders	Scale Farmers	NAEs
Average land area		<10 mu	≈400 mu	≈1200 mu
Labor status	The proportion aged 36–54	<47.3%	58.3%	61.2%
	The proportion with junior high school educational level or above	<56.7%	65.8%	74.6%

Note: The author calculated these statistics based on the data from the Third National Agricultural Census, the China Family Farm Report, and the China Micro and Small Enterprise Survey (CMES). Scale farmers include family farms and large-scale professional farmers. NAEs are represented by agricultural enterprises or specialized cooperatives. 1 mu = 0.067 hectares.

, the average land area used by smallholders is less than 10 mu, much smaller than the 400 mu averaged by scale farmers and the 1200 mu averaged by NAEs. The laborers hired by scale farmers and NAEs are younger and have a higher level of education. This means that the human capital level of scale farmers and NAEs is higher than that of smallholders. Land transfer in China show that farmers with higher agricultural productivity tend to transfer-in land [25,26]. The land rental market could cause land from low-skilled farmers to be transferred to high-skilled farmers [27,28]. With the continuous increase in land rental activities, land has been redistributed from less productive to more productive farmers [29]. The typical experience is that smallholders transfer land to scale farmers and NAEs. This indicates that scale farmers and NAEs are more productive than smallholders. Meanwhile, farmers have also expanded their land management capabilities by increasing their land area [30]. Scale farmers and NAEs could improve their human capital by learning by doing. Scale farmers and NAEs have stronger management capabilities than smallholders [31].

Compared with scale farmers, NAEs are likely to be more productive land users. Scale farmers are usually formed by smallholders who transfer a certain amount of land and are accustomed to traditional production methods. The labor of scale farmers is mostly supplied by their families. Table 1 shows that NAEs usually have younger laborers with higher levels of education. Scale operators are more vulnerable to production, market, institutional, personal, financial, and other kinds of risks [32]. Moreover, a lack of management ability and unreasonable operation modes of scale operators present severe challenges regarding sustainable survival [33]. The differences in the land productivity of farmers at different scales may be due to the heterogeneity of farmers and land, and land area expansion will not affect grain yield per unit area in general [34]. With the expansion of land area, the diversity and complexity of on-site processing of scale management may exceed the ability of scale farmers, which could lead to diseconomies of scale.

NAEs are agricultural operation organizations with significant advantages in advancing production technology and adopting modern management methods. The role of entrepreneurs is to discover, capture, and utilize profit opportunities in competition with other entrepreneurs in an uncertain environment [35,36], reform and innovate production methods [37], and effectively coordinate and judge scarce resources based on information and capability advantages [38]. This highlights that NAEs would increase agricultural productivity and generate more significant economies of scale than scale farmers.

A farm’s land area and hired labor status are influenced by the human capital of farmers. Differences in average land area and labor status among smallholders, scale farmers, and NAEs indicate that smallholders have the lowest level of human capital, while NAEs have the highest level of human capital.

2.2. Human Capital of Farmers’ Influence on Agricultural Growth

Based on the model of analyzing the size distribution of firms [39] and allocating capital based on land quality [40], we constructed a model of allocating land based on farmers’ human capital. Agricultural goods are produced by farmers with heterogeneous human capital in managing a farm. We assumed that the land area of the farmers was endogenous to their human capital. Farmers with a relatively high level of human capital will have higher returns. Considering that the quantity of land remains relatively stable, the model is constrained by fixed land amounts. Thus, maximizing agricultural output depends on allocating limited land to farmers with different levels of human capital.

Specifically, the human capital level of each farmer is recorded as $h$. $h\ge z$. $z$ is the marginal value of human capital determined by the model. $L\left(h\right)$ represents the number of farmers with a human capital level of $h$. In total, the number of farmers is $A={{\displaystyle \int }}_z^{\infty }L\left(h\right)dh$. The total human capital level of farmers is $H={{\displaystyle \int }}_z^{\infty }hL\left(h\right)dh$. $S\left(h\right)$ is the amount of land allocated to farmers with a human capital level of h. $s\left(h\right)={\displaystyle \frac{S\left(h\right)}{L\left(h\right)}}$ is the ratio of allocated land area to the number of farmers with a human capital level of $h$. A given distribution of human capital level also means a known land area distribution. The total quantity of land is $S={{\displaystyle \int }}_z^{\infty }S\left(h\right)dh={{\displaystyle \int }}_z^{\infty }s\left(h\right)L\left(h\right)dh$. The production function of farmers with human capital level of $h$ is $Y\left(h\right)=F\left[hL\left(h\right),S\left(h\right)\right]=hL\left(h\right)f\left({\displaystyle \frac{s\left(h\right)}{h}}\right)$, where $hL\left(h\right)f\left({\displaystyle \frac{s\left(h\right)}{h}}\right)$ is the output of an agricultural entity with human capital level of $h$. $F\left(·\right)$ is the quadratic differentiable, increasing and strictly concave. The total output of agriculture is $Y={{\displaystyle \int }}_z^{\infty }Y\left(h\right)dh={{\displaystyle \int }}_z^{\infty }hL\left(h\right)f\left({\displaystyle \frac{s\left(h\right)}{h}}\right)dh$.

Based on previous theoretical models, this study focuses on how farm structure and agricultural growth change parallel with land transfer development in China. The policy implementation process of land transfer in China can be simply divided into the following three stages: prohibition (1978–1983), gradual authorization (1984–2007), and rapid development (2008–2024).

In 1978, the HRS was announced to distribute land evenly to farmers based on their household population, and land transfer was not allowed. It could be considered that all farmers were smallholders; their land area was not determined by human capital level. In this situation, the land area of farmers (smallholders) can be considered to be at the same level, which is lower than the area standard for scale management, as shown by $S_1$ in Figure 1. Therefore, even if there are differences in the human capital level of farmers (smallholders), the output between farmers (smallholders) is limited to a certain range, as shown by $Y_1$ in Figure 1.

From 1984 to 2007, land transfer was gradually legalized, and land transfer transactions mainly occurred between smallholders. Smallholders were allowed to expand their land size to achieve the standard of scale operation and become scale farmers. For all farmers, it was equivalent to no longer being constrained by a fixed land area. The relationship between the human capital of farmers and their land size changed from $S_1$ to $S_2$. At the same time, farmers had a threshold of $z^{*}$. As some smallholders became scale farmers, their human capital was released and improved, signifying an overall increase in the human capital level of farmers. The production function of farmers changed from $Y_1$ to $Y_2$, increasing total agricultural output.

Since 2008, land transfer development has accelerated with the formation and expansion of NAEs. The entry of NAEs with higher human capital can be seen as having an exogenous impact, further improving the overall human capital level of farmers. In this stage, $z_1$ represents the threshold for farmers. The relationship between the human capital and land size of farmers has changed from $S_2$ to $S_3$, and the production function of farmers has changed from $Y_2$ to $Y_3$. Subsequently, NAEs’ rapid formation and development have further promoted agricultural growth.

Therefore, the structural change among farmers stems from the transfer of land from smallholders with lower human capital levels to scale farmers and NAEs with higher human capital levels, which is conducive to improved agricultural production. In summary, this study proposes the following hypotheses to be tested:

H1. 

Farm structure positively affects agricultural growth in the land transfer context.

H2. 

Compared to scale farmers, NAEs in farm structure have a stronger promoting effect on agricultural growth.

3. Method, Data, and Variables

3.1. Model

Based on the Cobb–Douglas production function, the empirical model introduces variables to measure the structural change among farmers. The estimation equation is as follows:

$$\begin{array}{l}\ln \left(Y_{it}\right)={\alpha }_0+{\alpha }_{1}NA{E}_{it-1}+{\alpha }_{2}S{F}_{it-1}+{\alpha }_3\ln \left(Lan{d}_{it}\right)+{\alpha }_4\ln \left(Labo{r}_{it}\right)+{\alpha }_5\ln \left(Capita{l}_{it}\right)+{\alpha }_6\ln \left(Fer{t}_{it}\right)\\ +{\alpha }_{7}St{r}_{it}+{\alpha }_{8}W{a}_{it}+{\alpha }_{9}Fin{a}_{it}+{\mu }_i+{\epsilon }_{it}\end{array}$$

where Y indicates the agricultural output used to measure agricultural growth. The key explanatory variables used to measure the structural change among farmers in the equation are the proportion of land area used by NAEs to the total land area (NAE) and the proportion of land area used by scale farmers to the total land area (SF). A one-stage lag is applied to reduce the possibility of reverse causality. Four conventional inputs are included in the function: land, labor, capital, and fertilizer (Fert) [1]. The output and the four conventional inputs are in a natural logarithm form. Three other variables are included in the function: agricultural industrial structure (Stru), the impact of weather on agriculture (Wa), and the degree of financial support for agriculture (Fina) [2]. Finally, i denotes the province, t denotes the year, α denotes the parameter to be estimated, μ denotes the provincial fixed effect, and ε denotes the error term.

3.2. Variables and Data

This study uses provincial-level panel data from 30 provinces in China from 2010 to 2021. The China Rural Policy and Reform Statistical Annual Report statistics related to scale farmers and NAEs were updated in 2021, and there is a lack of indicators in the China Rural Business Management Statistical Annual Report before 2009. We selected the following variables:

(1) Explained variables. The gross output value of agriculture (Y) is used to measure agricultural growth5. This value (CNY 100 million) was retrieved from the China Agricultural Statistical Yearbook, calculated at comparable prices for 2010.

(2) Key explanatory variables. Farm structure is reflected through the proportion of smallholders, scale farmers, and NAEs within the land use system. Small farms account for 84% of all farms worldwide but use only approximately 12% of all agricultural land [41]. The land area is comparable and reflects the proportion and relative composition of various farmer categories. Therefore, the proportion of land area used by NAEs to the total land area (NAE) and the proportion of land area used by scale farmers to the total land area (SF) are used to indicate the structural change among farmers.

The land area used by NAEs equals the sum of the land area flowing into specialized cooperatives, enterprises, and other NAEs. Because the land used by NAEs is obtained through land transfer, scale farmers are defined as those with land area of more than 50 mu. We separated the scale farmers into three land size sections: 50–100 mu, 100–200 mu, and more than 200 mu. The land area used by scale farmers is equivalent to the mean value of each land size section multiplied the number of scale farmers in each section. The data sources on farmers are the China Rural Business Management Statistical Annual Report (2009–2018) and the China Rural Policy and Reform Statistical Annual Report (2019–2020). The total land area data are from the China Land and Resources Statistical Yearbook and the provincial statistical yearbook. Missing data were supplemented with the linear interpolation method.

(3) Control variables. Land input is measured using the total sown area of farm crops (1000 hectares). Labor input is measured using the total number of employed persons in primary industry (10,000 persons) times the ratio of gross output value of agriculture to gross output value of agriculture, forestry, husbandry, and fishery. The method of measuring capital input involves multiplying the total power of agricultural machinery (10,000 kw) by the ratio of the gross output value of agriculture to the gross output value of agriculture, forestry, husbandry, and fishery. Fertilizer input is measured using the volume of effective components of chemical fertilizer (10,000 tons). The agricultural industrial structure (Stru) is defined as the ratio of the gross output value of agriculture to the gross output value of agriculture, forestry, animal husbandry, and fishery. The impact of weather on agriculture (Wa) is defined as the ratio of the area covered by natural disasters to the total land area. The degree of financial support for agriculture (Fina) is the ratio of agriculture, forestry, and water conservancy expenditure to general budgetary expenditure from local governments. The relevant data came from the annual data of the China Agricultural Statistical Yearbook and the provincial statistical yearbook. A summary of all these variables is provided in

Table 2. Descriptive statistical results.

Variables	Mean	S.D.	Min	Max	Observations
Y	6.83	1.02	4.21	8.52	360
NAE	10.47	11.40	0.29	81.51	360
SF	15.20	10.87	1.87	52.74	328
Land	8.18	1.15	4.48	9.62	360
Labor	5.59	1.13	2.46	7.21	360
Capital	7.01	1.16	3.77	8.86	360
Fert	4.81	1.14	1.59	6.57	360
Stru	51.96	8.20	33.40	72.05	360
Wa	15.36	12.09	0.00	69.59	360
Fina	11.39	3.29	4.11	20.38	360

Note: The SF variable removes samples with missing data and outliers.

.

4. Empirical Analysis

4.1. Econometric Results

The Hausman test results in

Table 3. Hausman test.

Regression	Chisq Value	p-Value
Only include NAE	109.19	0.0000
Only include SF	128.08	0.0000
Include NAE and SF	104.34	0.0000

indicate that regression should be performed using a fixed-effects model. The VIF value of variables is 8.82, less than 10, which indicates that the problem of multicollinearity is not severe [42]. Empirical results are shown in

Table 4. Effect of farm structure on agricultural growth.

	FE (1)	FE (2)	FE (3)	FE-GLS (4)	FRONTIER (5)
NAE	0.0088 ***		0.0081 ***	0.0082 ***	0.0087 ***
	(7.4917)		(6.2241)	(8.9261)	(7.0959)
SF		0.0023 **	0.0014	0.0008	0.0015 *
		(2.4733)	(1.5355)	(1.5369)	(1.7043)
Land	0.6429 ***	0.6788 ***	0.7338 ***	0.7167 ***	0.7440 ***
	(12.0274)	(9.7006)	(11.0479)	(11.1875)	(12.7431)
Labor	−0.4423 ***	−0.4927 ***	−0.4013 ***	−0.4359 ***	−0.3886 ***
	(−14.2804)	(−14.6633)	(−11.5136)	(−19.9585)	(−12.3885)
Capital	0.1109 ***	0.1115 ***	0.1124 ***	0.1335 ***	0.1189 ***
	(3.9857)	(3.2907)	(3.5262)	(7.0643)	(3.7830)
Fert	0.1317 ***	0.0551	0.0814	0.0560 *	0.1091 **
	(2.6343)	(0.9840)	(1.5405)	(1.7165)	(2.0988)
Stru	0.0246 ***	0.0262 ***	0.0238 ***	0.0188 ***	0.0224 ***
	(9.1919)	(8.8391)	(8.4377)	(10.2827)	(8.1778)
Wa	−0.0018 ***	−0.0020 ***	−0.0017 ***	−0.0012 ***	−0.0018 ***
	(−4.5342)	(−4.4524)	(−3.9565)	(−3.4883)	(−4.2232)
Fina	0.0087 ***	0.0115 ***	0.0081 **	0.0040 *	0.0069 **
	(2.7833)	(3.3371)	(2.4509)	(1.8883)	(2.2069)
N	360	328	328	328	328
R2	0.8440	0.8201	0.8414	N.A.	N.A.
F-test value	153.60	126.76	140.11	N.A.	N.A.
Log-likelihood	N.A.	N.A.	N.A.	444.9594	331.1052

Note: The values in parentheses are t-statistics. ***, **, and * are expressed at 1%, 5%, and 10% significance levels, respectively. The values have been controlled for provincial fixed effects.

. FE (1), (2), and (3) represent the estimates using the fixed-effects model. With exceptions for traditional factors, FE (1) only includes the proportion of land area used by NAEs, and FE (2) only includes the proportion of land area used by scale farmers. FE (3) covers the proportion of land area used by NAEs and scale farmers. Considering the common heteroscedasticity and serial correlation problems of panel data, we adopt the generalized least squares method developed by Nicholas M. Kiefer [43] and the stochastic frontier regression developed by Dennis J. Aigner et al. [44] in order to improve the consistency of estimates. The estimates are reported in FE-GLS (4) and FRONTIER (5) in Table 4. The estimated coefficients differ slightly from those estimated using FE.

FE (1) shows that when the proportion of land area used by NAEs in a province increases by 1%, agricultural output will significantly increase by 0.88%, indicating that the development of NAEs promotes agricultural growth. FE (2) shows that the development of scale farmers promotes agricultural growth. When the proportion of land area used by scale farmers in a province increases by 1%, agricultural output will increase by 0.23%. The coefficients of the two explanatory variables in columns (3), (4), and (5) are all positive, but the coefficients of the proportion of land area used by NAEs are larger than those of scale farmers. This indicates that structural change among farmers promotes agricultural growth, and NAEs in farm structure have a stronger effect on promoting agricultural growth. Hypothesis 1 and Hypothesis 2 are tested. When the proportion of land area used by NAEs in a province increases by 1%, agricultural output will be significantly increased by 0.81~0.88%; when the proportion of land area used by scale farmers in a province increases by 1%, it will significantly increase the total agricultural output value by approximately 0.15~0.23%.

Control variables also contribute to agricultural growth. The number of agricultural employees (Labor) has a significant negative impact on agricultural growth, indicating that the productivity of agricultural labor is distinct from equipment efficiency, and agricultural mechanization should continue to substitute agricultural labor. Fertilizer input (Fert) has a significant positive effect on agricultural growth. The impact of weather on agriculture (Wa) is not conducive to agricultural growth, suggesting that more attention should be paid to the impact of environmental change on agricultural production. The regression results of control variables support the current agricultural policy orientation that aims to improve agricultural mechanization equipment and promote rural labor outmigration.

Based on the estimated coefficients6 in FRONTIER (5) in Table 4, we further quantitatively calculated the contribution of structural change among farmers and input factors on agricultural growth from 2010 to 2021. The results are shown in

Table 5. Contribution of farm structure to agricultural growth.

Variables	Estimated Coefficient (1)	Change in Explanatory Variables (2)	Contribution to Growth (3) = (1) × (2)	Contribution to Growth (%) (4)
NAE	0.87	9.50	8.27	13.48
SF	0.15	10.49	1.57	2.56
Land	0.74	6.35	4.70	7.66
Labor	−0.39	−40.83	15.92	25.95
Capital	0.12	15.98	1.92	3.13
Fert	0.11	−6.66	−0.73	−1.19
Stru	2.24	1.45	3.25	5.29
Wa	−0.18	−14.81	2.67	4.34
Fina	0.69	−0.01	−0.01	0.00
Residual			23.8	38.78
Growth		61.36		100.00

Note: To calculate the contributions of these variables to output growth in terms of percentage, the estimated coefficients of these variables are multiplied by 100. For land, labor, capital, and Fert, “Change in explanatory variables” refers to the percentage growth of that variable. For NAE, SF, Str, Wa, and Fina, “Change in explanatory variables” refers to the difference in magnitude of that variable. The numbers in parentheses are the percentage shares of contribution to growth, with growth set at 100.

. During this period, agricultural output increased by 61.36% in China. Farm structure contributed significantly to agricultural growth, with scale farmers and NAEs contributing 16% of output growth. NAEs contribute 13.48%, more than five times higher than that of scale farmers.

4.2. Endogeneity Analyses and Robustness Checks

Based on using a one-period lagged variable, we further employed the instrumental variable method to address potential endogeneity issues. In 2013, the Chinese government announced policies to encourage the transfer of contracted management rights to large professional households, family farms, cooperatives, and agricultural enterprises. Since then, the central government has issued a series of policies to accelerate the development of scale farmers and NAEs, that is, to promote structural change among farmers. Agricultural growth in the provinces will not affect policies implemented by the central government. This study devised an exogenous policy impact variable (Policy), with a value of 1 after 2013 and 0 before 2013. A pre-determined state can affect the implementation effect of the policy; thus, in order to enhance the effectiveness of the policy dummy variable (Policy), we used the following interaction term: policy dummy variable times the proportion of land area used by NAEs in 2012 (Policy*NAE) as an instrumental variable for the proportion of land area used by NAEs. We used the interaction term of the policy dummy variable (Policy) and the proportion of land area used by scale farmers in 2012 (Policy*SF) as instrumental variables for the proportion of land area used by scale farmers.

Table 6. Results of endogeneity analyses.

	2SLS (1)	2SLS (2)
NAE	0.0194 ***
	(2.8124)
SF		0.0107 **
		(2.4880)
Land	0.7491 ***	0.6163 ***
	(8.2815)	(6.6073)
Labor	−0.2840 ***	−0.2774 ***
	(−2.6575)	(−2.9325)
Capital	0.1054 ***	0.1232 **
	(3.3644)	(2.4484)
Fert	0.1730 ***	−0.0451
	(2.7961)	(−0.5467)
Stru	0.0222 ***	0.0265 ***
	(6.6110)	(7.0830)
Wa	−0.0015 ***	−0.0025 ***
	(−2.9782)	(−4.0715)
Fina	0.0030	0.0088 **
	(0.5836)	(1.9988)
N	360	259
R2	0.8046	0.8041

Note: The values in parentheses are t-statistics. *** and ** are expressed at 1% and 5% significance levels, respectively. The values have been controlled for provincial fixed effects.

reports the results of using instrumental variables to address endogeneity. The Anderson LM statistic for the two models is 11.99 and 16.54, respectively, which strongly refutes the null hypothesis of under-identification. The Cragg–Donald Wald F-statistic for the two models is 12.14 and 17.18, respectively, indicating no issues with weak instruments. This suggests that we chose suitable instrumental variables that can be used to address the endogeneity problem in the model. The estimation results show that farm structure promotes agricultural growth.

This study performed a variety of checks to test the robustness of the above empirical results. The corresponding estimates are listed in

Table 7. Results of robustness checks.

	FE (1)	FE (2)	FE (3)	FE (4)
NAE	0.0155 ***		0.0148 ***	0.0068 ***
	(9.3410)		(8.0217)	(4.3435)
SF		0.0032 ***	0.0013 *	0.0021 *
		(4.2851)	(1.8549)	(1.9248)
Land	0.5459 ***	0.6833 ***	0.6982 ***	0.8421 ***
	(10.8897)	(9.9651)	(11.2346)	(9.7852)
Labor	−0.4132 ***	−0.4642 ***	−0.3622 ***	−0.3368 ***
	(−13.7979)	(−14.1977)	(−11.2351)	(−7.7299)
Capital	0.1009 ***	0.1015 ***	0.1000 ***	0.1882 ***
	(3.7672)	(3.0588)	(3.3252)	(4.4903)
Fert	0.1979 ***	0.0522	0.1330 ***	−0.0676
	(4.0340)	(0.9522)	(2.6237)	(−1.0702)
Stru	0.0217 ***	0.0256 ***	0.0211 ***	0.0156 ***
	(8.2989)	(8.8047)	(7.8199)	(4.9656)
Wa	−0.0018 ***	−0.0020 ***	−0.0016 ***	−0.0018 ***
	(−4.6138)	(−4.5006)	(−3.9648)	(−3.2653)
Fina	0.0063 **	0.0116 ***	0.0049	0.0067 *
	(2.0715)	(3.4252)	(1.5586)	(1.7269)
N	360	328	328	328
R2	0.8559	0.8273	0.8587	0.7707

Note: The values in parentheses are t-statistics. ***, **, and * are expressed at 1%, 5%, and 10% significance levels, respectively. The values have been controlled for provincial fixed effects.

. (1) Replace the proportion of land area used by NAEs. Only professional cooperatives and enterprises were considered, and the proportion of land area used by NAEs was recalculated. (2) Replace the proportion of land area used by scale farmers. Farmers with more than 30 mu land area were regarded as scale farmers, and the proportion of land area used by scale farmers was recalculated. (3) Replace the variables according to (1) and (2), ensuring that they are incorporated into the model at the same time. (4) Adopt the proportion of land area used by scale farmers and the proportion of land area used by NAEs with lag two periods and adjust the data from 2011 to 2022. The significance of key explanatory variables and the sign of the estimation coefficients are robust.

4.3. Heterogeneity Analysis

Due to the obvious differences between regions in China, regression was carried out for eastern, central, western, and major grain-producing and non-major grain-producing provinces to analyze the heterogeneity effects of NAEs and scale farmers. The results are shown in

Table 8. Results of heterogeneity analysis.

	Eastern (1)	Central (2)	Western (3)	Major Grain-Producing Area (4)	Non-Major Grain-Producing Area (5)
NAE	0.0068 ***	0.0073 ***	0.0188 ***	0.0089 ***	0.0073 ***
	(3.3377)	(3.0211)	(7.2388)	(6.1836)	(3.8750)
SF	0.0028	0.0043 ***	0.0015	0.0029 ***	0.0019
	(1.3962)	(3.1344)	(0.8708)	(3.3568)	(1.1326)
Land	0.4675 ***	1.2048 ***	0.6847 ***	1.1376 ***	0.6398 ***
	(4.4703)	(4.0404)	(4.0080)	(6.5495)	(7.8077)
Labor	−0.4103 ***	−0.3052 ***	−0.3548 ***	−0.2259 ***	−0.4833 ***
	(−6.2103)	(−5.2704)	(−6.0266)	(−5.8823)	(−9.1782)
Capital	0.2238 ***	−0.0503	0.0584	0.0846 ***	0.1327 ***
	(3.0146)	(−1.1911)	(0.9137)	(2.7482)	(2.7146)
Fert	0.3575 ***	−0.1644	0.0282	−0.2916 ***	0.2246 ***
	(2.9695)	(−1.3946)	(0.4607)	(−4.1426)	(3.1318)
Stru	0.0252 ***	0.0199 ***	0.0229 ***	0.0208 ***	0.0250 ***
	(4.7589)	(4.0241)	(4.4957)	(6.3810)	(5.8268)
Wa	−0.0013 *	−0.0002	−0.0015 **	−0.0008	−0.0015 **
	(−1.8320)	(−0.1794)	(−2.3315)	(−1.5966)	(−2.4119)
Fina	−0.0017	0.0048	0.0023	−0.0007	0.0180 ***
	(−0.1650)	(0.8644)	(0.4985)	(−0.1900)	(3.5859)
N	115	85	128	142	186
R2	0.7548	0.9039	0.9173	0.9355	0.8231

Note: The values in parentheses are t-statistics. ***, **, and * are expressed at 1%, 5%, and 10% significance levels, respectively. The values have been controlled for provincial fixed effects.

. There is regional heterogeneity between the role of NAEs and scale farmers. Only NAEs contribute to agricultural growth in the eastern and western regions and non-major grain-producing areas. This might be because the comparative income of grain or agriculture production in these regions is low, and only farmers with a higher human capital level can obtain a higher agricultural income. NAEs and scale farmers have contributed to agricultural growth in the central region and major grain-producing areas. NAEs have a stronger performance in promoting agricultural growth.

5. Discussions

This study focuses on agricultural growth in China. Unlike Lin (1992), who revealed the early contributions of rural reforms [1], this study introduces farm structure as a new perspective to understand this issue in the present day. The results show that an increase in the land area used by scale farmers and NAEs could explain China’s 16% agricultural growth from 2010 to 2021. Although smallholders make up the majority of farmers in China, the tendency of farm structure is towards an increase in scale farmers and NAEs and a decrease in smallholders. Research on farm structure and agricultural growth has important reference value for developing and transforming other countries with smallholders as the largest group of farmers, such as Kenya, Ethiopia, Vietnam. As in China, some Western countries, e.g., Germany and the United States, also follow the doctrine of ‘one thing, one right’ and the principle of ‘real right and usufructuary right’, especially in a land property rights setting. Although the notion of a land real right is protected by collective ownership in China, which seems different from private land ownership in Western countries at first glance, the usufructuary right aims to promote the utility of land resources.

In addition, decreasing labor input and increasing capital input positively impact agricultural growth. This result is the same as the finding by Li (2022) [45]. Decreasing labor input can optimize factor allocation and promote the agricultural transition from labor- intensive to capital-intensive production, ultimately improving the land output. Fertilizer input has a positive effect on agricultural growth. Appropriate and scientific use of fertilizers can significantly boost production [2,46]. However, a large increase in fertilizer input is undesirable because of the negative externality effect on the ecological environment. Organic fertilizer will help diminish environmental degradation and encourage sustainable agricultural development [47]. The residual error of the model identifies 38% hidden contributions to agricultural growth, which can be provided by environmental institutions [1], technical progress [48], agricultural research and development [49], agricultural infrastructure [50], and other hidden factors. If detailed data are available in the future, further analysis can focus on isolating each factor’s contribution to agricultural growth.

In present-day China, we cannot ignore the fact that the proportion of land area used by scale farmers and NAEs in the total land area of agriculture has reached as high as 36%. Previous studies, such as Lin (1992) [1], Fan (1997) [48], and Huang (1996) [51] only cover the phenomenon prior to 2000, which was dominated by smallholders, and therefore cannot reveal the contribution of scale farmers and NAEs to agricultural growth. Farm structure also influences both ecology and society. Large-scale farms have a positive spillover effect on small-scale farms around the adoption of improved technologies and increased crop yield [52]. NAEs have a positive impact on the level of farmers’ interest in green production practices [53]. Limited to the available data, this study uses the land area used by scale farmers and NAEs to reflect farm structure and compares their heterogeneity effects. Considering this limitation, it is necessary to improve the construction of a relevant database of all types of farmers. Further studies are required to reveal how to support farm structure to achieve better outcomes in agricultural economics, social welfare, and the ecological environment.

6. Conclusions and Policy Suggestions

At the macro level, land transfer has driven the structural change among farmers in China. The quantity of and utilized land area worked by scale farmers and NAEs are increasing, improving the human capital of farmers and optimizing land use methods. This study has constructed a production function based on farm structure, has analyzed the impact of farm structure on agricultural growth, and has empirically tested the research hypotheses using provincial panel data.

The results indicate the following: (1) Structural change among farmers has promoted agricultural growth. Both scale farmers and NAEs play a positive role in agricultural growth. (2) Unlike scale farmers, NAEs contribute more to agricultural growth. Every 1% increase in the proportion of land area used by NAEs will significantly increase the total agricultural output by 0.81–0.88%. Every 1% increase in the proportion of land area used by scale farmers will increase the total agricultural output by 0.15–0.23%. (3) During the research period, the total contribution of structural change among farmers to agricultural growth was 16%. The contribution of scale farmers and NAEs to agricultural growth is 2.56% and 13.48%, respectively. (4) The roles of NAEs and scale farmers show regional heterogeneity. Scale farmers only promote agricultural growth in central and grain-producing areas; new agricultural entities positively affect the entire area.

Based on the research findings, this study proposes the following: (1) Emphasize the role of farmers and their human capital and improve the quantity and quality of NAEs and scale farmers. Farmers directly determine the sustainability and modernization level of agricultural development. It is necessary to cultivate new types of professional farmers, including by carrying out agricultural production skills training, guiding scale farmers in achieving the transformation of agricultural production methods, supporting talents who are returning to rural areas to establish NAEs, and improving public services and supporting incentive measures for NAEs. (2) The development of human capital and farm structure varies from region to region, but they follow the same overall tendency of decreasing numbers of smallholders while increasing numbers of scale farmers and NAEs. Policies should take into account region-specific agricultural practices. In some areas with lower levels of human capital and more dispersed land distribution among farmers, land consolidation and land planning policies are necessary to create better conditions for development. Cooperation and resource sharing can also be strengthened to reduce regional disparities and optimize farm structure. (3) Encourage the development of the land transfer market and promote the concentration of land towards NAEs and scale farmers. Land resources are the foundation for farmers to provide the function of human capital. Although governments encourage land transfer to scale farmers and NAEs and even direct administrative intervention in some areas, the reality remains that transfers between smallholders have always dominated. Therefore, related institutions should strengthen management and services by constructing and improving land transfer transaction platforms, and then encourage and prioritize land flow to NAEs and scale farmers.