Relationship of Cooperative Management and Green and Low-Carbon Transition of Agriculture and Its Impacts: A Case Study of the Western Tarim River Basin

Abstract

Clarifying the relationship between cooperative management and cultivated land use eco-efficiency (LUEE) is of great significance to promoting the green and low-carbon transition of agriculture. To explore the role of cooperative management in the green and low-carbon transition of agriculture of smallholder farmers in the western Tarim River Basin, in this study, based on the field survey data of 444 farmers in 2021, the carbon emissions of cultivated land were used to measure the LUEE with the slack-based model (SBM) with undesirable outputs. Then, propensity score matching (PSM) was used to test the relationship between cooperative management and LUEE. Additionally, the mediating effect of farmers’ green development willingness (FGDW) and the moderating effect of farmers’ part-time off-farm employment (POE) on the relationship was explored. The present study hypothesized that joining cooperatives has an improving effect on the LUEE, which can be achieved by increasing FGDW, and this effect can be enhanced by farmers’ POE. The results show that: (1) The LUEE was generally low (average LUEE value: 0.2678), and there was a significant difference between farmer households (the difference between the maximum and minimum values was as high as 2.8716). (2) Cooperative management had a significant improving effect on the LUEE. The LUEE of cooperative farmers (ACF) increased by 8.6% compared with that of non-cooperative farmers (NACF). (3) Joining a cooperative could improve the LUEE by improving FGDW. (4) POE could enhance the improving effect of cooperative management on the LUEE. Overall, all three hypotheses were supported: cooperative management could achieve scale effects that small farmers cannot achieve, which had a positive effect on improving the LUEE. This study provides a new ecological perspective for the analysis of the relationship between agricultural cooperatives and LUEE and decision-making reference for the rational utilization of cultivated land in northwest China.

1. Introduction

According to the Sixth Assessment Report (https://www.ipcc.ch/report/ar6/syr/, accessed on 16 April 2023) released by the Intergovernmental Panel on Climate Change (IPCC), human activities such as unsustainable energy consumption, land use, and production lead to continuous increases in the global greenhouse gas emissions. This ultimately causes global warming and adversely affects the atmosphere, oceans, cryosphere, and biosphere, which is hindering the sustainable development of humankind. To limit global warming, the report points out that it is necessary to reduce the carbon dioxide emissions caused by human activities and achieve net-zero carbon dioxide emissions.

According to the Special Report on Climate Change and Land (https://www.ipcc.ch/srccl/chapter/summaryfor-policymakers/, accessed on 16 April 2023) issued by the IPCC, cultivated land and woodland have contributed approximately 23% of the total global greenhouse gases from 2007 to 2016. Obviously, agricultural carbon emissions have become an important source of greenhouse gases. In response to the calls for carbon emission reduction and green agricultural development, scholars have introduced carbon emissions from cultivated land use as an environmental constraint to measure cultivated land use efficiency and proposed the LUEE based on the traditional economic efficiency of cultivated land use [1,2,3]. Previous research on LUEE has focused on two aspects: efficiency measurement and impact factor analysis. The widely used measurement methods are the modified SBM based on data envelopment analysis (SBM-DEA) [1,4,5] and the super-SBM [6]. Carbon emissions [6,7] and non-point-source pollution [8] have usually been used as undesired outputs for measurement. In most studies that have analyzed influencing factors, multiple variables have been simultaneously selected to conduct an overall analysis, but such studies often lack specificity due to the inclusion of too many influencing factors [8].

Some studies have analyzed the relationship between specific factors/core variables and LUEE; new ways, new features, and new phenomena emerging in rural and agricultural development have been the focus of previous studies in terms of the selection of core variables. For example, Li et al. [9] analyzed the data collected from 30 provinces in China, and found that the aging of the rural population had a negative impact on the LUEE, which was caused by reductions in land size and the number of rural laborers. However, this can be partially offset by the improvement in the mechanization level. Zhu et al. [10] reported that labor transfer had a positive impact on the LUEE, and could play a role through economy of scale. Hu et al. [11] pointed out that agriculture and tourism integration was beneficial to improving the LUEE.

With the rapid advancement of agricultural modernization, the relationship between new agricultural management forms and cultivated land use efficiency has become a hot topic. The relationship between cooperative management and cultivated land use efficiency has been the most widely discussed. Ma et al.’s study on apple growers in China [12], Dong et al.’s study on greenhouse vegetable farmers in China [13], Adewumi and Adebayo’s study on potato farmers in Nigeria [14], and Abate et al.’s study on smallholder farmers in Ethiopia [15] have all showed that joining agricultural cooperatives could improve cultivated land use efficiency. However, Huang et al. [16] found that the effect of joining cooperatives on cultivated land use efficiency was not significant for fruit farmers in Anhui, China. Similarly, Huang’s research, based on provincial panel data in China’s Yangtze River Economic Belt, also found that joining agricultural cooperatives had little or even negative impact on cultivated land use efficiency [17]. Therefore, there are two different conclusions about the impact of cooperative management on cultivated land use efficiency. Regarding whether the impact of cooperative management on cultivated land use efficiency is positive or negative after the inclusion of ecological dimension, there are few relevant studies at present. In addition, most of the existing studies have focused on changes in land use types in a region [18,19], and researchers have not yet paid attention to the LUEE in the basin. Due to the contradiction between the highly developed agriculture and the extremely fragile ecological environment in the Tarim River Basin, studying the agricultural management patterns and farmland utilization efficiency in the basin not only enriches the research results on the relationship between cooperative management and LUEE but also contributes to the green and low-carbon transition of agriculture in the region.

Therefore, in this study, taking the carbon emissions generated by cultivated land use as the undesired output, the super SBM with undesirable outputs (hereinafter referred to as the super-SBM-undesirable model) was used to measure the LUEE in the western Tarim River Basin based on the survey data of 444 farmer households. Additionally, propensity score matching (PSM) was used to analyze the effect of joining cooperatives on the LUEE, and the mediating effect of farmers’ green development willingness (FGDW) and the moderating effect of farmers’ part-time off-farm employment (POE) were explored to determine the specific mechanism of action. This study aimed to provide decision-making support for the green and low-carbon transition of regional agriculture and rural revitalization.

The contributions of this study are as follows: Firstly, most previous research on agricultural cooperatives has been carried out from the aspect of economic benefits [12,13,14,15]. This study introduced a new dimension, the ecological dimension, for the study of cooperatives, by taking the LUEE as a dependent variable and carbon emissions as an ecological factor. Secondly, in this study, intermediate variables, such as consciousness and emerging agricultural production characteristics, were added to analyze their auxiliary role in the relationship between cooperative management and LUEE. This provided details for the analysis of the relationship between cooperative management and LUEE.

2. Theoretical Analysis and Research Hypothesis

Agricultural cooperatives are mutual-help organizations constructed by producers and operators of similar agricultural products or providers and users of similar agricultural services based on a household contracted responsibility system, characterized by voluntary associations and democratic management [20]. Compared with the small peasant approach, cooperative management has the advantages of scale, systematization, sufficient funds, and mature technology.

The LUEE is a comprehensive efficiency that has the connotations of both economic efficiency and ecological efficiency. According to Farrell’s theory [21], economic efficiency can be divided into technical efficiency and allocative efficiency. Technical efficiency refers to the ability to obtain the maximum output under a given combination of input elements; specifically, it refers to the production of as many outputs as possible under the given input conditions at the existing technology level or the use of as few inputs as possible to achieve the targeted output level at the given level of technology. Allocative efficiency refers to the ability of the production unit to obtain the maximum output by making the input elements in the optimal ratio at the given element price level. It indicates that under the condition of a certain level of production technology, the production unit can improve production efficiency with a reasonable combination of input elements. Joining agricultural cooperatives can help farmers obtain preferential treatment in agricultural material supply, credit financing, agricultural product sales, as well as training in management and farming technology, which can improve the technical efficiency of farmers. At the same time, joining agricultural cooperatives can enhance the marketing awareness of farmers, help farmers to obtain accurate supply and demand information on agricultural products in the market and to respond in a timely matter, and avoid farmers making experience judgment on market trends, thus improving allocative efficiency. Therefore, the technical efficiency and allocative efficiency of farmers can be improved at the same time through joining agricultural cooperatives, thus increasing the possibility of improving the LUEE [22]. The definition of eco-efficiency by the World Business Council for Sustainable Development [23] is “providing competitively priced products or services that meet human needs and ensure quality of life, while gradually reducing the impact on ecology and resource consumption intensity to be consistent with the approximate carrying capacity of the earth”. Therefore, eco-efficiency includes the two meanings, that is, achieving economic growth without increasing or even decreasing resource inputs and greatly reducing the amount of waste discharged to the environment under the condition that economic output remains unchanged or even increases. Cooperatives can provide farmers with the convenience of purchasing genuine organic fertilizers and biological pesticides that meet industry standards, which, to a certain extent, can reduce the carbon emissions during the farming process and promote the LUEE. Based on the above, the first hypothesis is proposed:

Hypothesis 1 (H1).

Joining agricultural cooperatives improves the eco-efficiency of cultivated land use.

Joining agricultural cooperatives not only optimizes material supply conditions but is also conducive to spreading new ideas, to a certain extent. Cooperatives often organize technical training. Through this platform, farmers can learn about green development and form green development awareness. Additionally, cooperatives connect farmers with the market by, for example, organizing farmers to supply and export agricultural products to domestic and foreign supermarkets. The market requires the quality of agricultural products to pass standard certifications, including pollution-free, green, and organic food certifications [24], which forces farmers to practice green production behaviors. Compared with peasant farmers, farmers in cooperatives with green development awareness have a lower time preference rate. This means that their future income has a lower discount rate compared with that on their current income. Therefore, they rarely discount future income streams when making decisions and devote more energy to environmental protection [25], including taking measures to ensure green production. This ultimately reduces the carbon emitted from cultivated land use and improves the LUEE. Therefore, joining cooperatives can enhance the LUEE by optimizing material conditions and enhancing farmers’ awareness of green development. Therefore, the second hypothesis is proposed:

Hypothesis 2 (H2).

Joining agricultural cooperatives increases the LUEE by improving farmers’ awareness of green development.

Many factors affect the relationship between cooperative management and LUEE. Among them, part-time off-farm employment (POE) has become a new factor that cannot be ignored in the study area. Therefore, it is necessary to discuss the effect of POE on the relationship between LUEE and cooperative management. POE refers to farmers’ engagement in both agricultural and nonagricultural production. In the rural areas in the study area, POE mainly has two forms. In the first, some family members are engaged full-time in nonagricultural production in the nearby factories while the remaining members are engaged full-time in agricultural production. In the second, family members who are engaged in agricultural production take part-time nonagricultural jobs in, for example, service industries and individual businesses, during the down time in the farming season. In both cases, urban areas with relatively developed economies and more employment opportunities are the main sites for farmers’ nonagricultural employment. Therefore, POE has become an important way to connect cities and rural areas and an important canal of information for farmers to understand the outside world. Based on this, POE may affect the relationship between cooperative management and LUEE. Specifically, the POE can improve the economic level and information level of farmers. The higher economic level increases the possibility of farmers investing in production factors and adopting new technologies. Additionally, the high information level helps farmers quickly access market information and make rational production decisions to avoid market risks. Therefore, compared with ACF with agriculture as the only economic income source, ACF with POE may create higher technical efficiency and allocative efficiency. Improvements in technical efficiency and allocative efficiency can improve the LUEE by improving the economic efficiency of farmers (Figure 1). Therefore, the third hypothesis is proposed:

Hypothesis 3 (H3).

POE may regulate the effect of cooperative management on LUEE.

3. Research Area, Data Sources, and Methods

3.1. Research Area

The Tarim River Basin is located in southern Xinjiang, China. Kizilsu, Kashgar, Hotan, Alar, Tumushuk, and Kunyu in the western Tarim River basin were included in this study (Figure 2).

The study area is located in the hinterland of the Eurasian continent, with an annual precipitation of less than 50 mm and evaporation of 2300~3000 mm [26]. It is extremely arid, with widespread deserts and a fragile ecological environment. However, the region has high annual sunshine hours of 2800~3100 h, a long frost-free period of 185~210 days, and an average annual temperature of over 10 °C [27], providing good conditions for agricultural development, especially cotton cultivation. With the water resources from the Tarim River and some of its tributaries, the region has developed into an important high-quality cotton production area in China. However, local farmers commonly use massive amounts of chemical fertilizers and pesticides to increase production, which seriously inhibits the transition to green and low-carbon agriculture.

With the promotion of the rural revitalization strategy in China, local governments have adopted a series of measures to increase the income of farmers, which have led to the emergence of many agricultural cooperatives, such as agricultural machinery cooperatives, agricultural marketing cooperatives, water user cooperatives, etc. Many farmers have joined them. Additionally, to increase family income, some family members \work or do business in the form of full-time or part-time nonagricultural work.

3.2. Data Sources

The data used were obtained from a field survey conducted from 11 January 2022 to 26 January 2022. This survey mainly took the form of a questionnaire and symposium. The participants in the symposium included government workers from villages, towns, and cities (counties); and the content of the symposium mainly included the overall development of local rural areas and the specific situation of cultivated land use of farmers. The questionnaire was completed in the form of household interviews involving farmers who joined agricultural cooperatives and those who did not. Through the probability proportionate to the sample size, more than 30 towns in the study area were selected to carry out questionnaire surveys. Then, based on stratified random sampling, a total of 465 questionnaires were distributed. Finally, 444 valid questionnaires were obtained, accounting for 95.5% of the total. Of the respondents, 125 farmers had joined agricultural cooperatives (ACF), and 319 farmers had not joined agricultural cooperatives (NACF).

3.3. Research Methods

3.3.1. Super-SBM-Undesirable Model

The super-SBM-undesirable model is a kind of data envelopment analysis. That is, based on the input and output data of the same type of decision-making unit (DMU), linear equations are used to find the optimal production frontier, and the production efficiency is determined by calculating the distance from the DMU to the production frontier. Traditional DEA models can only deal with the proportional reduction in inputs and outputs. When there are slack variables between the input and output, that is, there is input redundancy or insufficient output, the radial model easily overestimates its efficiency. In view of this, Tone [28] input slack variables to the objective equation and constructed an SBM model. This model solves the problem of the efficiency evaluation in the presence of variable relaxation and undesired outputs and is nonradial and nonangular, so can avoid the deviation caused by dimension and angle selection differences. However, because the efficiency values measured by the SBM model are all less than or equal to 1, it is not possible to compare multiple high-efficiency DMUs. To make up for its shortcomings, Tone [29] further proposed the super SBM-undesirable model (Equation (1)).

$$\mathrm{Min}\rho =\frac{1-\frac{1}{K}{{\displaystyle \sum }}_{k=1}^K{s}_k^{-}/x_{kd}}{1+\frac{1}{N+M}({{\displaystyle \sum }}_{M=1}^N{s}_n^{+}/y_{nd}+{{\displaystyle \sum }}_{m=1}^M{s}_m^{-}/u_{md})}$$

(1)

$$\begin{array}{c}\mathrm{s}.\mathrm{t}.{\displaystyle \sum }_{j=1}^j{\lambda }_j{x}_{kj}+s_k^{-}=x_{kd},k=1,2\dots ,K\\ {\displaystyle {\displaystyle \sum }_{j=1}^j}{\lambda }_j{y}_{nj}-s_n^{+}=y_{nd},n=1,2\dots ,N\\ {\displaystyle {\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{j}}}{\lambda }_j{u}_{mj}+s_m^{-}=u_{md},m=1,2\dots ,M\\ {\lambda }_j\ge 0,s_k^{-}\ge 0,s_n^{+}\ge 0,s_m^{-}\ge 0,j=1,2\dots ,n\end{array}$$

where ρ represents the eco-efficiency of cultivated land use (LUEE) of the DMU. In this study, each farmer was regarded as an independent DMU. K, N, and M represent the number of inputs, desired output, and undesired output factors, respectively; S_k⁻, S_n⁺, and S_m⁻ represent the slack of the input, desired output, and undesired output, respectively; x_kd, y_nd, and u_md represent the input, desired output, and undesired output values, respectively; $\lambda$ represents the weight; and x_kj, y_nj, and u_mj represent the kth input, desired output, and undesired output of the DMU J, respectively. When ρ ≥ 1, the DMU is efficient; when ρ < 1, the LUEE of the DMU still needs to be improved.

3.3.2. Propensity Score Matching (PSM) Test

This study examined the impact of cooperative management on the LUEE. In this study, cooperative farmers were regarded as the treatment group, and noncooperative farmers were regarded as the control group. By comparing the differences in the LUEE between the treatment group and the control group, the impact of cooperative management on LUEE could be obtained. Due to the relationship between cooperative management and LUEE being affected by other factors, it was necessary to find a treatment group and a control group with identical other factors for comparison to solve the problem of overt bias. However, due to the numerous other influencing factors and the large individual differences among farmers, it was difficult to find two farmer groups with all identical factors. Therefore, by referring to the approach of Rosenbaum and Rubin [30], all control factors were converted into a comprehensive propensity score. Based on the propensity scores of the farmers, the matching between the treatment group and the control group was carried out; that is, the PSM was used for analysis.

The PSM test is based on observable variables. If there are unobservable variables, hidden bias can be introduced. To solve this problem, following the method of Abate et al. [15,31], this study used the boundary method proposed by Rosenbaum et al. [30] to test the sensitivity of the PSM estimation results to hidden bias. The specific model construction was as follows:

Assume that the LUEE of farmer i is Yi, and the management type of farmer i is $D_i$. If $D_i$ = 1, the farmer has joined cooperatives and belongs to the treatment group; if $D_i$ = 0, the farmer has not joined cooperatives and belongs to the control group. $X_i^T$ represents other factors affecting the management type and LUEE of farmer i. Then, the following model could be constructed:

$$Y_i=X_i^T{\beta }_x+{\alpha }_d{D}_i+{\epsilon }_i=\left\{\begin{array}{c}{Y}_{0i}=X_i^T{\beta }_x+{\epsilon }_i\left(if D_i=0\right)\\ Y_{1i}=X_i^T{\beta }_x+{\alpha }_d+{\epsilon }_i\left(if D_i=1\right)\end{array} \right.$$

(2)

where ${\epsilon }_i$ is the error term, $Y_{1i}$ is the LUEE of cooperative farmer i, and $Y_{0i}$ is the LUEE of noncooperative farmer i. Because there is only one selection for a farmer, $Y_{0i}$ and $Y_{1i}$ cannot be observed at the same time; that is, $Y_{0i}$ is the counterfactual result of $Y_{1i}$ [30].

Based on the counterfactual framework, $Y_{1i}-Y_{0i}$ is the treatment effect brought about by the cooperative management of farmer i, $\mathrm{E}(Y_{1i}-Y_{0i})$ is the average treatment effect of all farmers, and $\mathrm{E}(Y_{1i}-Y_{0i}|D_i=1)$ is the average treatment effect of the farmers in the treatment group, that is, ATT, which can intuitively express the cooperative-management-induced difference in the LUEE. The ATT can be expressed as follows:

$$\begin{array}{l}\mathrm{ATT}=\mathrm{E}(Y_{1i}-Y_{0i}|D_i=1)=\mathrm{E}(Y_{1i}|D_i=1)-\mathrm{E}(Y_{0i}|D_i=1)\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}=\mathrm{E}(Y_{1i}|D_i=1)-\mathrm{E}(Y_{0i}|D_i=0)+\mathrm{E}(Y_{0i}|D_i=0)\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}-\mathrm{E}(Y_{0i}|D_i=1)\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}=\mathrm{E}(Y_{1i}|D_i=1)-\mathrm{E}(Y_{0i}|D_i=0)+selection bias\end{array}$$

(3)

where $\mathrm{E}(Y_{0i}|D_i=0)-\mathrm{E}(Y_{0i}|D_i=1)$ is the selection bias, that is, the cooperative management choice of farmers induced by the influence of other factors. When other factors are controlled so that other factors are independent of the potential outcome, i.e., $Y_{1i}, Y_{0i}⫫D_i, X_i^T$, selection bias can be considered eliminated, in which case ATT = $\mathrm{E}(Y_{1i}|D_i=1)-\mathrm{E}(Y_{0i}|D_i=0)$. In other words, in practice, it was necessary to find cooperative and noncooperative farmers with the same characteristics to achieve matching. To eliminate the influence of confounding factors as much as possible, according to the principle proposed by Stuart and Robin [32], this study selected the control variables as comprehensively as possible and selected the influencing factors that were highly correlated with the potential results.

Due to the large number of influencing factors, many samples could not be matched, so this study adopted the propensity score proposed by Rosenbaum and Rubin to solve the multidimensional matching problem. The propensity score was the conditional probability that farmers chose cooperative management after controlling other influencing factors. The propensity score P of farmer i to join a cooperative could be expressed as follows:

$$P=\mathrm{Pr}\left(D_i=1|X_i^T\right)$$

(4)

The logit model was used to estimate the conditional probability P of farmers joining a cooperative. After obtaining P, the same or similar P was used to match the farmers in the treatment group and the control group. In this case, the average treatment effect (ATT) of the LUEE brought about by cooperative management could be expressed by Equation (5). The value of ATT is the difference in the LUEE caused by different management types.

$$\mathrm{ATT}=\mathrm{E}\left(Y_{1i}|D_i=1,p\left(X_i^T\right)\right)-\mathrm{E}\left(Y_{0i}|D_i=0,p\left(X_i^T\right)\right)$$

(5)

It should be noted that to make the matching better, the treatment group and the control group were required to have a relatively large common support domain, as shown in Equation (6).

$$\mathrm{ATT}=\frac{1}{N_T}{{\displaystyle \sum }}_{i\in I_1\cap CS}[Y_{i0}-{{\displaystyle \sum }}_{j\in I_0}w\left(i,j\right)Y_{j0}]$$

(6)

where CS is the sample range of the common support domain, NT is the number of samples in the treatment group of the common support domain, and W(i,j) is the matching weight. ${\displaystyle \sum }_{j\in I_0}w\left(i,j\right){\mathrm{Y}}_{j0}$ A is the weighted average of the propensity score of the control group within the common support domain. The average value obtained from the weighted average within the common value range is the average treatment effect [33].

3.3.3. Variable Selection

To measure the LUEE of farmers, it was necessary to select appropriate input and output indicators. By referring to previous research [34,35,36], 8 input indicators were selected from 3 aspects including cultivated land, labor, and agricultural materials; and 2 output indicators were selected from both desired and undesired aspects (

Table 1. Input- and output-related explanatory indicators.

Variable Type	Variable	Variable Interpretation	Cooperative Farmers (n = 125)		Noncooperative Farmers (n = 319)		Test of Difference
			Mean	Standard Deviation	Mean	Standard Deviation	Difference in Means	Difference in Means
Input	Land input	Crop area (ha)	1.741	2.348	2.109	3.049	5.523	5.523
	Labor input	Labor force	1.64	0.574	1.699	0.642	0.059	0.059
	Agricultural material input	Tillage input (USD)	550.861	1165.909	620.723	1619.426	479.825	479.825
		Fertilizer input (USD)	547.471	897.518	909.375	1619.832	2458.601 **	2458.601 **
		Pesticide input (USD)	399.862	795.648	410.462	1191.802	72.807	72.807
		Seed input (USD)	188.288	327.771	215.277	509.468	185.366	185.366
		Agricultural film input (USD)	547.471	897.518	168.497	595.357	257.585	257.585
		Irrigation input (USD)	557.890	824.684	842.836	2055.322	1957.049	1957.049
Desired output	Material output	Total crop yield (kg)	14,795.898	23,484.673	19,790.66	64,633.78	4994.757	4994.757
Non-desired output	Ecological constraints	Carbon emissions from cultivated land use (kg)	1120.187	2029.953	1338.362	3033.183	218.175	218.175

Note: All the data in the table were from the authors’ on-site research on farmers in the study area. ** represents p < 0.05.

). According to Feng et al. [34], this study used the land area actually cultivated by farmers to represent land input and the number of labors engaged in planting to represent labor input. Due to the diversity of input elements in the process of cultivated land use, according to Lu et al. [35], six subitems including pesticides, fertilizers, seeds, agricultural films, irrigation, and mechanical services were selected in this study.

Affected by the regional price differences in agricultural products, the total crop output is uncertain; so, according to Coluccia et al. [36], the total crop output was selected as the desired output. In addition, according to Lu et al. [35], the total carbon emissions in the process of cultivated land use was used as undesired output, and the carbon emission estimation coefficients were calculated according to West et al. [37] and Post et al. [38].

In the selection of the control variables, by referring to Dong et al. [13], control variables were selected from four aspects: household head characteristics, household characteristics, cultivated land characteristics, and management conditions. In the selection of the mechanism test variables, because consciousness is subjective and hidden, it is difficult to measure accurately, so this study selected the level of farmers’ willing to invest in green development to measure farmers’ green development awareness; that is, the willingness of farmers to invest in green development was used as the mediating variable (

Table 2. PSM test for eco-efficiency of cultivated land use by farmers and variables for mechanism test.

Variable Type	Variable	Cooperative Farmers (n = 125)		Noncooperative Farmers (n = 319)		Test of Difference
		Mean	Standard Deviation	Mean	Standard Deviation	Difference in Means	Difference in Medians
Dependent variable	Ecological efficiency of cultivated land use	0.314	0.41	0.25	0.33	−0.064 *	4.91 **
Core explanatory variable	Agricultural Cooperative	1	0	0	0	NA	NA
Controlled Variables	Health condition	1.712	1.054	1.821	1.123	0.109	0.607
	Educational level	2.792	0.944	2.633	1.381	−0.234 **	7.635 ***
	Income	5.043	5.081	5.012	5.152	0	0.011
	Labor situation	1.64	0.574	1.7	0.642	0.059	2.841 *
	Fragmentation degree	2.888	1.833	2.382	1.57	−0.506 ***	6.161 **
	Irrigation conditions	0.56	0.498	0.536	0.499	−0.024	NA
	Information acquisition	0.864	0.344	0.893	0.309	0.029	NA
	Antirisk measures	0.048	0.215	0.11	0.313	0.062 **	4.081 ***
Mediating variable	Households’ Willingness to pay for green development	4.6	3.923	3.674	3.881	−0.926 **	2.388
Moderating variable	Part-time nonagricultural work	0.656	0.477	0.727	0.446	0.071	NA

Note: All the data in the table were from the authors’ on-site research on farmers in the study area. *** p < 0.01; ** p < 0.05; * p < 0.1.

and

Table 3. Definition of variables.

Type	Variable	Interpretation
Dependent variable	Ecological efficiency of cultivated land use	Calculated by Super-SBM-undesirable model
Core explanatory variable	Agricultural Cooperative	Whether to join cooperatives (yes = 1, no = 0)
Controlled Variables	Health condition	Health condition of the head of household (very healthy = 1, relatively healthy = 2, healthy = 3, unhealthy = 4, very unhealthy = 5)
	Educational level	Educational level of the head of household (primary school = 1, junior high school = 2, high school/technical secondary school = 3, junior college/higher vocational school = 4, undergraduate = 5, master degree and above = 6)
	Income	Logarithm of household income
	Labor situation	Number of people engaged in agricultural production in the household
	Fragmentation degree	Number of cultivated land blocks
	Irrigation conditions	Drip irrigation used (yes = 1, no = 0)
	Information acquisition	Proficiency with smartphone use (Yes = 1, No = 0)
	Anti-risk measures	Agricultural insurance purchase (yes = 1, no = 0)
Mediating variable	Households’ Willingness to pay for green development	The willingness to pay per hectare for green production (PFG, in USD) (PFG = 0, 0; PFG < 10.92, 1; 13.104 < PFG < 54.6, 2; 56.784 < PFG < 109.2, 3; 111.384 < PFG < 163.8, 4; 165.984 < PFG < 218.4, 5; 220.584 < PFG < 327.6, 6; 329.784 < PFG < 436.8, 7; 438.984 < PFG < 546, 8; 548.184 < PFG < 655.2, 9; 655.2 < PFG, 10)
Moderating variable	Part-time nonagricultural work	Whether family members are part-time nonagricultural workers (yes = 1, no = 0)

).

4. Results and Discussion

4.1. Analysis of the LUEE

Data envelopment analysis was performed to calculate the LUEE of 444 farmers in the study area using Matlab (R2018b MathWorks, Inc., Massachusetts, USA). The average LUEE value of the whole sample was small (0.2678). This indicated that the overall level of the LUEE was low in the study area. Secondly, the LUEE of most was is low. The LUEE value of 376 households (accounting for 84.69%) was less than 0.5, and the samples with an LUEE value less than 0.1 accounted for 51.13% of the total. The lowest LUEE value was only 0.0002, far behind the frontier of the LUEE (

Table 4. Distribution of eco-efficiency of cultivated land use by farmers in the study area.

LUEE Value Range	Number			Proportion
	All Farmers	Cooperative Farmers	Noncooperative Farmers	All Farmers	Cooperative Farmers	Noncooperative Farmers
[0, 0.1)	227	52	175	51.13%	41.60%	54.90%
[0.1, 0.3)	90	29	61	20.27%	23.20%	19.10%
[0.3, 0.5)	59	23	36	13.29%	18.40%	11.30%
[0.5, 1)	9	2	7	2.03%	1.60%	2.20%
[1, 3]	58	19	40	13.06%	15.20%	12.50%

). Only a small number of farmers (13.3%) had an LUEE value greater than or equal to one; that is, LUEE reached the regional best level under fixed cultivated land use carbon emissions conditions. Additionally, there were large individual differences in the LUEE in the study area, and the highest LUEE value was 2.8716 higher than the lowest. In general, the LUEE in the study area was low, and the gap between households as large, indicating the slow development of low-carbon ecological agriculture in the study area.

Farmers were classified by management type (

Table 5. Differences in eco-efficiency of cultivated land use between farmers participating in agricultural cooperatives (ACFs) and farmers who had not joined agricultural cooperatives (NACFs).

Group	Count	Minimum	Maximum	Mean	Standard Deviation	Proportion of Samples with an LUEE Value Greater Than or Equal to 1
Cooperative farmers	125	0.0171	2.8718	0.3138	0.4096	15.2%
Noncooperative farmers	319	0.0002	1.5604	0.2498	0.3299	12.5%
Total sample size	444	0.0002	2.8718	0.2678	0.3548	13.3%

). The average LUEE of ACFs was 0.3138 (Table 4 and Figure 3), ere 0w.064 higher than that of NACFs, and the maximum and minimum values of the former are greater than those of the latter. Additionally, the proportion of the ACFs achieving the optimal efficiency was greater than that of the NACFs. Therefore, there was a clear difference in the LUEE between the ACFs and NACFs, that is, compared with the NACFs, the ACFs had a higher LUEE.

4.2. Analysis of the Impact of Cooperative Management on Eco-Efficiency of Cultivated Land Use by Farmers

4.2.1. PSM Test Results

The results of the PSM test (

Table 6. PSM test results under different matching methods.

Matching Method	Sample	Treatment Group	Control Group	ATT	Standard Error	T-Statistic
k-nearest neighbor matching (n = 4)	Before matching	0.314	0.250	0.064	0.037	1.72 *
	After matching	0.314	0.212	0.101	0.045	2.25 **
Radius matching (radius = 0.05)	After matching	0.314	0.240	0.073	0.042	1.73 *
kernel matching radius = 0.06)	After matching	0.314	0.238	0.076	0.042	1.78 *
Mean value of the three matchings		0.314	0.236	0.086

Notes: ** p < 0.05; * p < 0.1.

) show that the results of k-nearest neighbor matching, caliper matching, and kernel matching were all significantly positive. This indicates that cooperative management had a significant role in promoting the eco-efficiency of cultivated land use by farmers, and the results were robust. Specifically, the mean value of the four matchings was 0.086; that is, after excluding other factors including household head characteristics, household characteristics, cultivated land characteristics, and management situation, the LUEE of the ACFs was 8.6% higher than that of the NACFs. In addition, the ATT values of the treatment group and the control group after the three matchings were greater than those before matching. This indicates that the results before matching underestimated the promoting effect of joining cooperatives on the LUEE. In other words, the results after matching were closer to the actual situation. Therefore, the hypothesis 1 was supported; that is, joining cooperatives can improve the LUEE.

4.2.2. Matching Quality Inspection

Most samples in the treatment group and the control group had a common value range. This indicated that not too many samples were lost during the matching process and that the matching situation was good (Figure 4).

The equilibrium test results of the PSM matching (

Table 7. Equilibrium test of PSM matching.

Matching Method	Pseudo R2	LR chi2	p > chi2	Mean Deviation (%)	Median Deviation (%)	B (%)	R
Before matching	0.044	23.27	0.003	14.7	9.9	51.7	0.96
k-nearest neighbor matching (n = 4)	0.002	0.8	0.999	3.5	2.7	11.1	0.55
Radius matching (Radius = 0.05)	0.002	0.78	0.999	3.7	3.2	11.2	0.94
Kernel matching (radius = 0.06)	0.002	0.74	0.999	3.8	3.7	10.8	0.78

Notes: B is the bias; R is the ratio of ACF to NACF variances.

) show that after matching, the pseudo R² and LR chi2 significantly decreased, the B value was less than 25%, and the R values were in the range of 0.5–1. According to Caliendo and Kopeinig [1], the worse the explanatory capacity of the pos-match variables on the probability of project participation, the smaller the difference in the variables between the treatment group and the control group after matching (i.e., the lower the pseudo R² and LR chi2 of the logit model based on the postmatching sample estimate, the worse the significance of the test results of the explanatory variables), and the higher the matching quality. According to Rubin [39], when the B value is lower than 25% and R is between 0.5 and 2, the equilibrium hypothesis is fully met. This shows that the PSM model selected in this study had good matching results; that is, the overall estimation of the model passed the equilibrium test. Because the B value of kernel matching was the smallest, the results of kernel matching were selected for further analysis.

Additionally, the further equilibrium test of the control variables showed that the bias ratio of all covariates was less than 20% (

Table 8. Balance test of control variable matching.

Variable	Before/After Matching	Treatment Group	Control Group	Bias Ratio (%)	Bias Reduction Ratio (%)	t	p > |t|
Health condition	Before matching	1.712	1.8213	−10.0	66.0	−0.94	0.348
	After matching	1.712	1.7492	−3.4		−0.28	−0.783
Educational level	Before matching	2.792	2.558	25.8	−76.6	2.49	−0.013
	After matching	2.792	2.7373	6.0		0.45	0.651
Agricultural labor	Before matching	1.64	1.6991	−9.7	59.0	−0.9	0.370
	After matching	1.64	1.6642	−4.0		−0.32	0.748
Household income	Before matching	1.1	1.0	5.7	9.7	0.52	0.603
	After matching	1.1	1.2	−5.1		−0.34	0.737
Fragmentation of cultivated land	Before matching	2.888	2.3824	29.6	95.5	2.91	0.004
	After matching	2.888	2.9107	−1.3		−0.1	0.924
Irrigation conditions	Before matching	0.56	0.5361	4.8	47.68	0.45	−2.711
	After matching	0.56	0.5295	6.1		0.48	−0.958
Antirisk measures	Before matching	0.048	0.1097	23.0	88.3	−2.03	0.043
	After matching	0.048	0.0408	2.7		0.28	0.784
Information level	Before matching	0.864	0.8934	−9.0	77.8	−0.87	0.383
	After matching	0.864	0.8575	2.0		0.15	0.882

Notes: t is the value of the t-test, and p >| t | is the p-value corresponding to the t-value.

). Paul et al. [40] reported that if the absolute value of the standard deviation of each variable after matching is less than 20%, the matching is good. The results of the t-test also showed that there was no significant difference between the two groups. This indicated that the systematic variance in the distribution of explanatory variables between the treatment group and the control group was significantly eliminated after matching, the sample selection bias as minimized, and the propensity score estimation and sample matching were successful.

4.2.3. Robustness Test

The validation results of ordinary least squares regression (OLS) and Tobit regression (Table 8) show that cooperative management had a significant positive impact on the LUEE, both in controlled and uncontrolled situations.

The PSM test results (

Table 9. Ordinary least squares regression and Tobit regression results.

Variable	OLS		Tobit
Agricultural cooperative	0.0641 *	0.0676744 *	0.0641 *	0.0676744 *
Informatization level	Uncontrolled	Controlled	Uncontrolled	Controlled
Health condition
Educational level
Fragmentation of cultivated land
Irrigation conditions
Antirisk measures
Agricultural labor
Household income

Notes: * p < 0.1.

) show that the results of the PSM test based on the Probit model were consistent with those based on the logit model. Specifically, regardless of the matching method adopted, the results based on Probit regression show that cooperative management had a significant effect on the LUEE. The mean ATT of the three matching methods was 0.081. This indicated that the LUEE of ACFs was 8.1% higher than that of NACF. Additionally, the postmatching ATT was greater than that before matching. In summary, the matching results in PSM test were good (

Table 10. PSM test results based on the Probit model.

Matching Method	Before/After Matching	Treatment Group	Control Group	ATT	Standard Error	T-Statistic
k-nearest neighbor matching (n = 1)	Before matching	0.314	0.250	0.064	0.037	1.72 *
	After matching	0.314	0.217	0.097	0.053	1.85 *
k-nearest neighbor matching (n = 4)	After matching	0.314	0.220	0.094	0.045	2.09 *
Radius matching (radius = 0.05)	After matching	0.314	0.239	0.075	0.042	1.77 *
Kernel matching (radius = 0.06)	After matching	0.314	0.239	0.075	0.042	1.77 *
Mean value of four matching methods		0.314	0.233	0.081

Notes: * p < 0.1.

).

Table 11. Sensitivity analysis results.

Γ	k-Nearest Neighbor Matching (n = 1)		k-Nearest Neighbor Matching (n = 4)		Radius Matching (Radius = 0.05)		Kernel Matching (Radius = 0.06)
	sig+	sig−	sig+	sig−	sig+	sig−	sig+	sig−
1	0	0	0	0	0	0	0	0
6	0.00004	0	0.00004	0	0.00004	0	0.00004	0
11	0.00172	0	0.00172	0	0.00179	0	0.00172	0
16	0.00765	0	0.00765	0	0.00785	0	0.00765	0
25	0.02617	0	0.02617	0	0.02664	0	0.02617	0
30	0.03826	0	0.03826	0	0.03885	0	0.03826	0
35	0.05051	0	0.05051	0	0.0512	0	0.05051	0

presents the results of the sensitivity analysis of the PSM testing to bias. According to Rosenbaum and Rubin [30], Γ indicates the odds of bias allocation due to unobserved factors. The higher the Γ, the lower the sensitivity, and the less sensitive the results to potential hidden biases. From Table 11, it can be seen that when analyzing the impact of cooperative management on the LUEE, Γ was not significant at the p = 0.05 level until it approached 35. This indicates that the ATT results of the model were not sensitive to unobserved variables; that is, the estimation results of the PSM were robust. Therefore, the conclusion of this study is valid.

4.3. Mechanism Analysis of Cooperative Management Promoting Eco-Efficiency of Cultivated Land Use

4.3.1. Analysis of the Mediating Effect of Farmers’ Green Development Awareness

Wen’s three-step method [41] was used to test the mediation effect. According to the results of models 1 and 2 (

Table 12. Mediation effect test results.

Independent Variable	Model 1	Model 2	Model 3
Cooperative	0.068 * (0.037)	1.062 ** (0.416)	0.057 (0.037)
Expected green investment			0.010 ** (0.004)
Controlled variable	Controlled	Controlled	Controlled
Constant term	0.405 *** (0.094)	0.405 *** (1.048)	0.351 *** (0.096)
Prob > F	0.0001	0.0026	0

Notes: *** p < 0.01; ** p < 0.05; * p < 0.1.

), joining cooperatives had a significant positive impact on the LUEE and green development willingness of farmers. According to the results of model 3, the green development willingness of farmers had a significant positive effect on the LUEE. Additionally, the impact of cooperative membership on the LUEE changed from significant to insignificant after adding the mediating variable. This indicated that the green development willingness of farmers played a completely mediating role in the improvement in LUEE brought about by joining cooperatives.

The above results support hypothesis 2. Joining cooperatives can improve the LUEE by raising farmers’ awareness of green development.

4.3.2. Analysis of the Moderating Effect of Farmers’ Part-Time Off-Farm Employment (POE)

The moderating effect model was used to test the influence of farmers’ POE on the improvement in LUEE brought about by cooperative management. The results of model 4 (

Table 13. Moderating effect test results.

Independent Variable	Model 3	Model 4
Whether or not to join cooperatives	0.070 * (0.038)	0.074 ** (0.037)
Whether there is a part-time nonagricultural employment	0.025 (0.037)	0.021 (0.037)
Whether engaged in part-time nonagricultural employment*whether a cooperative member		0.190 ** (0.079)
Controlled variable	Controlled	Controlled
Constant term	0.384 *** (0.099)	0.363 *** (0.099)
Prob > F	0.0002	0

Notes: *** p < 0.01; ** p < 0.05; * p < 0.1.

) show that the interaction between POE and cooperative membership had a significant positive impact on the LUEE. Additionally, after adding the interaction item, joining cooperatives still had a significant impact on the LUEE, and the coefficient increased by 0.004. This indicated that POE cold enhance the improvement effect on LUEE brought about by joining cooperatives; that is, the joining of cooperatives by farmers with POE had a greater improvement effect on the LUEE than that of farmers without POE.

The regression analysis results (

Table 14. Impact of part-time nonagricultural employment on the income level and information level of farmer households.

Dependent Variable	Model 1	Model 2	Model 3
Income level	0.380 *** (0.079)
Proficiency in using smart phones		0.055 * (0.033)
Broadband			0.581 * (0.043)
Constant term	2.80 *** (0.066)	0.846 *** (0.028)	0.723 *** (0.036)
R-squared	0.0504	0.0062	0.0076
Prob > F	0	0.0978	0.0664

Notes: *** p < 0.01; * p < 0.1.

) show that POE had a significant positive impact on the income level of farmer households and the possibility of farmers to use smartphones and install broadband. This confirmed the above conclusion that POE not only increased household income but also brought rich information to family members who remained in the local area to farm. This ultimately improved the LUEE.

Thus, hypothesis 3 was supported. That is, POE had an enhancement effect on the improvement in LUEE brought about by cooperative management. POE had a positive impact on the improvement in the allocative efficiency of farmers, thereby enhancing the improvement effect of joining cooperatives on the LUEE. Based on the above analysis of the results, the relationship between cooperative management and LUEE and the mechanism were obtained (Figure 5). That is, cooperative management had an improving effect on the LUEE, which is consistent with the research results of Samuel et al. [42], Eric et al. [43], Dong et al. [13], Adewumi et al. [14], and Abate et al. [15].

Because the LUEE incorporates carbon emissions on the basis of traditional economic efficiency, compared with the above previous studies, this study further verified that cooperative management can still improve the cultivated land use efficiency under the premise of considering carbon emissions. This is an extension of previous research results and provides an ecological perspective to the study of the relationship between cooperative management and cultivated land use efficiency. Additionally, this study explored the mechanism through which cooperative management influences the LUEE, which provides more details for us to understand the relationship between the two. That is, (1) cooperative management can improve the LUEE by improving FGDW, which echoes the research result of Haldar et al. [44] in that farmers who join cooperatives are more inclined to choose organic production technology. (2) The improving effect of cooperative management on the LUEE can be enhanced by improving the information level and economic level of farmers (POE). This is consistent with Wang’s study result [45] that cooperatives have a positive impact on farmers’ intention to adopt information technology.

5. Conclusions

Cooperative management is an effective way forward for the green and low-carbon transition of agriculture. Cooperatives not only bring convenience for farmer members in obtaining materials, technologies, and market information but also drive the dissemination of new ideas such as green development and low-carbon production among farmers. That is, cooperative management can enhance the LUEE by enhancing FGDW. The stronger FGDW, the more likely farmers are to invest in green agricultural development, and farmers are more willing to buy expensive eco-friendly fertilizers and pesticides and adopt new environmentally friendly agricultural production technologies. This will inevitably lead to a reduction in carbon emissions, thereby improving the LUEE. Additionally, this indirectly verifies the environmental benefits of cooperative management. However, the environmental benefits of cooperative management needs further in-depth analysis in the future.

The POE can also promote the green and low-carbon transition of agriculture. The POE can improve the economic and informatization levels of farmers. The better economic level can provide sufficient financial support for the adoption of expensive eco-friendly fertilizers, pesticides, and agricultural production technologies. In addition, because some family members perform work or engage in business, farmer households with POE have wider access to green development information than those without POE. The above factors are all conducive to improving the LUEE.

This study only evaluated the improving effect of agricultural cooperatives on the LUEE, without distinguishing the type of cooperatives or considering the impact of new-type cooperatives emerging in the research area. In the future, the differences in the impact of different types of cooperatives on the LUEE will be explored to provide support for the green and low-carbon transition of agriculture.