Impact of Conservation Tillage Technology Application on Farmers’ Technical Efficiency: Evidence from China

Abstract

Conservation tillage plays a crucial role in preventing soil erosion, improving soil quality, and encouraging sustainable agriculture. This study analyzes the effects of conservation tillage on farmers’ technical efficiency and its underlying mechanisms, using a sample of 853 households and 1706 land plots from China’s Jilin, Sichuan, Jiangsu, and Henan provinces. The results indicate that conservation tillage can enhance technical efficiency, resulting in an average increase of 0.022 units, The study further demonstrates that conservation tillage technology primarily enhances technical efficiency by increasing output and reducing production costs. This study proposes policy recommendations for promoting the use of conservation tillage technology to enhance farmers’ technical efficiency.

1. Introduction

China’s agriculture has undergone rapid growth since the implementation of the reform and opening-up policy over 40 years ago. However, this growth has resulted in increasingly severe contamination of the agricultural production environment, which poses a significant challenge to sustainable agriculture [1,2]. Data from China’s Second National Pollutant Source Census in 2017 showed that China’s agriculture had a demand for 10.6713 million tons of chemical oxygen, and emitted 1.4149 million tons of total nitrogen and 0.2120 million tons of total phosphorus, representing 49.77%, 46.52%, and 67.22% of their respective total pollutant emissions. Non-point source contamination from farmland is now a significant focus and challenge for environmental protection in China. To address the worsening state of the agricultural production environment, the Chinese government has proposed a transformation of the agricultural development mode. The proposal includes encouraging the use of conservation tillage techniques, such as soil testing and formula fertilization and straw returning, reducing agricultural waste, and enhancing the ecological environment of agricultural production. Nevertheless, can the promotion of conservation tillage technology help farmers to optimize their resource allocation and improve their technical efficiency?

Theoretically, the adoption of new technology should enhance production efficiency. However, in practice, production efficiency is closely linked to land conditions [3] and farmers’ endowments [4,5,6]. It is crucial to empirically verify whether the positive impacts of new technology in the laboratory still exist in actual agricultural production. Moreover, conservation tillage technology aims to conserve resources, decrease pollutant emissions, and potentially replace chemical element inputs essential for grain yield. Therefore, it is necessary to rigorously evaluate whether this substitution can increase yield and thereby improve production efficiency. In fact, some scholars have already studied the relationship between farmers’ technology adoption behavior and technology efficiency. Ismael used data from 1970 to 2014 to empirically analyze the interaction between agricultural technology factors and carbon emissions in Jordan [7]. Xiao et al. found that China’s agricultural green production efficiency has been increasing overall based on macro data from 30 provinces in China. Technological progress has a cumulative positive effect on improving agricultural green production efficiency, while a decrease in technological efficiency has a cumulative negative effect [8]. Xiong and Xu used panel data from Sichuan province to demonstrate that after implementing green agriculture, the environmentally friendly agricultural production efficiency in Sichuan province has maintained a trend of first decreasing and then increasing [9]. Guo et al. found that green production technology plays a mediating role between resource and environmental cognition and agricultural production efficiency [10]. Li et al. found that the adoption of protective cultivation technology in small-scale agriculture in China can significantly improve the technical efficiency of rice production, and this effect has obvious heterogeneity among small-scale farmers [11].

However, there is still a relative lack of in-depth analysis of the mechanism between the two, and there is still room for expansion. Firstly, most existing research has been conducted at either the macro-level or the household level, which may overlook the mechanisms and heterogeneity between different technologies. This study, on the other hand, focuses on micro-level data to investigate the mechanisms and heterogeneity between different technologies. Secondly, while some research has analyzed the relationship between green behavior and agricultural technology efficiency at the household level [11], more precise analysis at the plot level and for different crops is rare. Since plot characteristics and productivity relationships show diversity, it is necessary to discuss this more accurately based on the plot, which is a data advantage of this study. Thirdly, the decision of farmers to adopt protective tillage technology may be based on the quality of the land they cultivate and their own productivity, and such decisions may not be random. If traditional methods of comparison are used, serious endogeneity problems may arise, leading to estimation bias. Propensity score matching (PSM), as a method for approximating natural experiments, can effectively overcome sample selection bias and solve the endogeneity problem. This method is widely used by many scholars for policy effect evaluation [12].

To address the gaps in existing research, this paper utilizes plot-level data to systematically study the impact of conservation tillage technology adoption on farmers’ technical efficiency. Specifically, this study examines the mechanism by which conservation tillage may induce a change in technical efficiency, and evaluates how different conservation tillage technologies impact efficiency. The research findings provide answers to the research questions as to what the impact of using conservation tillage technology on farmers’ technical efficiency is, and how different conservation tillage technologies impact efficiency.

The remainder of this paper is structured as follows. Section 2 provides details on the model and data, while Section 3 presents the estimation results and conclusions. Finally, Section 4 outlines the policy implications of the research findings.

2. Materials and Methods

2.1. Study Methods

To empirically analyze the effect of conservation tillage on technical efficiency, this study constructs the following econometric model, where technical efficiency is the dependent variable and conservation tillage technique application is the independent variable:

$$TE={\beta }_0+{\beta }_{1}adoption+{\beta }_{2}X+\epsilon$$

(1)

TE represents technical efficiency and environmental technical efficiency, adoption represents whether conservation tillage techniques are applied (1 = yes, 0 = no), and the conservation tillage techniques include soil testing and formula fertilization, straw returning, minimum tillage, and deep loosening. X represents other control variables, and $\epsilon$ represents the error term.

2.1.1. Stochastic Frontier Approach (SFA)

Technical efficiency is one of the main indicators for measuring the production and management level of the production decision unit. The measurement methods mainly include the stochastic frontier approach (SFA) and data envelope analysis (DEA) based on non-parametric analysis. This study adopts the SFA method based on parameter analysis for several reasons. First, this study needs to estimate the relationship parameters between inputs and outputs to reflect the marginal impact of household farm inputs on outputs and to verify the validity of the production function estimation. Second, there is only one target output of the household farm production function constructed in this study; thus, it is unnecessary to conduct a multi-input and multi-output analysis. Third, due to the great impact of natural conditions such as weather and disasters on grain production, and the significant interannual and regional variation, which can easily lead to abnormal data, it is more suitable to adopt the SFA method, which is more sensitive to outliers, than the DEA method. The stochastic frontier production function is formulated as follows:

$$y_i=f\left(x_i,\beta \right)\exp \left(v_i-u_i\right)$$

(2)

In the above equation, $y_i$ represents the total outputs of the household farm, $x_i$ represents variables such as agricultural inputs, farm machinery inputs, hired labor inputs, and land inputs, and subscript i represents the i-th sample household farm. The β represents the estimation coefficient. The error term consists of two parts: one part represents statistical errors, which are represented by $v_i$, and are assumed to follow an independent normal distribution, that is, $v_i=N:N\left(0,{\sigma }_v^2\right)$; the other part represents technical efficiency loss, which is represented by $u_i$, and is assumed to follow an independent half-normal distribution, that is, $u_i:N\left(0,{\sigma }_u^2\right)$. After obtaining the estimation parameters using the maximum likelihood estimation method, the technical efficiency loss term $u_i$ can be separated from the mixed error term $(v_i-u_i)$, and the technical efficiency of each sample household farm can be calculated. The calculation formula for technical efficiency is

$$T{E}_i=\exp \left[-E\left(\left.-u_i\right|{\epsilon }_i\right)\right]$$

(3)

Numerically, technical efficiency TE ranges from 0 to 1, and larger values indicate greater technical efficiency or smaller technical efficiency loss.

2.1.2. Tobit Regression Model

The technical efficiency values calculated by SFA are generally required to be within the range of [0, 1]. However, in reality, some samples may be less than 0 or greater than 1, all of which are merged into 0 or 1, and the probability distribution of efficiency values becomes a mixed distribution composed of discrete points and a continuous distribution. Therefore, this study uses the Tobit model with a restricted dependent variable to test the impact of conservation tillage technology application on farm household technical efficiency.

2.1.3. PSM Estimation Model

Considering that farmers’ decisions to adopt conservation tillage technology are based on their resource endowments, there is a self-selection problem in the estimation equation, which leads to biased estimation results. For instance, farmers with poor soil quality and low production efficiency may be more likely to adopt conservation tillage technology. On the other hand, only farmers with higher household economic levels, suitable climate and geographical conditions, and in key agricultural production areas may be willing to adopt conservation tillage technology, which can significantly affect its implementation. Failure to account for these factors can result in biased estimation outcomes and unreal estimates of conservation tillage technology’s impacts on environmental technical efficiency. Therefore, it is crucial to consider these variables when estimating the impacts of conservation tillage technology to avoid biased results.

To solve the self-selection problem, this study uses the propensity score matching method (PSM) to analyze the impact of the conservation tillage technology application on farm household technical efficiency. The basic idea of this method is to calculate the conditional probability of an individual being in the treatment group under given control variables, then select individuals with similar propensity scores in the treatment and control groups for matching, and calculate the individual treatment impacts. Finally, these average causal impacts are summed up as the total causal impact in a weighted average manner.

The general estimation steps of propensity score matching (PSM) are as follows:

First, select covariates.

Covariates, denoted by $x_{ij}$, are also known as matching variables. Generally, variables that affect the decision to adopt organic fertilizer and participate in insurance are selected. The specific selection and setting of variables are given in the next section.

Second, calculate propensity scores.

The propensity score of the $j-\mathrm{th}$ plot of the $i-\mathrm{th}$ farmer is the conditional probability that the $j-\mathrm{th}$ plot of the $i-\mathrm{th}$ farmer enters the treatment group given $x_{ij}$, denoted as $p\left(x_{ij}\right)\equiv P\left(D_{ij}=1|x=x_{ij}\right)$. $D_{ij}$ is the treatment variable that reflects whether the plot uses conservation tillage technology. $D_{ij}=1$ means that the plot applies conservation tillage technology and is in the treatment group, and conversely, $D_{ij}=0$ means that the plot does not apply conservation tillage technology and is in the control group. Propensity scores are generally estimated using a logistic model, that is,

$$P{\left(D_{ij}=1\left|x=x_{ij}\right.\right)}_i=F\left(x_{ij},\beta \right)$$

(4)

where $F$ is the cumulative distribution function of the logistic distribution, and $\beta$ is the corresponding parameter vector.

Third, perform propensity score matching.

There are different estimation methods for propensity score matching, but there is no clear consensus on which method to use in practical estimation. This paper uses the most commonly used radius matching method.

Fourth, calculate the average treatment impact.

The average treatment impact on the treated (ATT) of the plots that adopt conservation tillage technology is expressed as

$$ATT=E\left(y_{1ij}-y_{0ij}\left|D_{ij}=1\right.\right)$$

(5)

After matching, $\widehat{ATT}$, the estimate of the average treatment impact on production efficiency of the treated plots (applying conservation tillage technology) is expressed as

$$\widehat{ATT}=\frac{1}{N_{ij}}\sum i:D_{I=1}\left(y_{1ij}-{\widehat{y}}_{0ij}\right)$$

(6)

where $y_{1ij}$ is the production efficiency of the $j-\mathrm{th}$ plot of the $i-\mathrm{th}$ farmer that adopts conservation tillage technology, $y_{0ij}$ is the production efficiency of the $j-\mathrm{th}$ plot of the $i-\mathrm{th}$ farmer that does not adopt conservation tillage technology, ${\widehat{y}}_{0ij}$ is the production efficiency of the $j-\mathrm{th}$ plot of the $i-\mathrm{th}$ farmer that does not adopt conservation tillage technology after matching, $N_{ij}=\sum D_{ij}$ is the number of individuals in the treatment group, and $\sum i:D_{i=1}$ is the sum of only the number of individuals in the treatment group. By examining $\widehat{ATT}$, the probability differences between the treatment group and the control group due to differences in conservation tillage adoption can be observed.

2.2. Variables Selection

In terms of the production function in stochastic frontier analysis, the existing research commonly uses the Cobb–Douglas (C-D) production function and the transcendental logarithmic production function for analysis. Each method has its own advantages and disadvantages. The C-D production function is concise, easy to decompose, and has clear economic meaning, but it assumes that there is no substitution between factors. On the other hand, the transcendental logarithmic production function is more complex in form, but allows for variable factor substitution. To ensure the objectivity of the research results, this paper will first use the C-D production function to measure technical efficiency, and then use the transcendental logarithmic production function in subsequent empirical analysis to calculate technical efficiency and conduct robustness tests. Following the research methods of Battese and Coelli [13,14], four input indicators at the plot level, agricultural input, labor input, agricultural machinery input, and land input, were selected, and the yield at the plot level was used as the output indicator for measurement. The agricultural input includes seed, fertilizer, and pesticide input. The agricultural machinery input includes both self-owned and hired machinery, and the self-owned machinery input cost is converted based on the hired machinery price. The labor input also includes self-owned and hired labor input, and the self-owned labor input is converted based on the hired labor price. The land input is measured using the area of the plot.

In the PSM empirical model, based on the framework of total factor productivity analysis and the theory of household economic behavior, the decision of households to adopt conservation tillage technology is based on their household characteristics and the characteristics of the technology. Conservation tillage technology has high costs, high risk, and slow payoff under intertemporal externalities. On this basis, households will consider factors such as land, labor, capital, management capacity, and other endowments before selecting a suitable route for the application of conservation tillage technology [15,16,17]. Therefore, control variables should consider not only plot-level characteristics, but also household-level characteristics.

Table 1. Descriptive statistics of the main variables.

Variables	Variable Definition	Mean	Std. Error
Technical efficiency	TE of the plot	0.737	0.154
Plot irrigability	Whether the land is irrigable (1 = Yes; 0 = No)	0.757	0.429
Plot quality	The quality of the plot (1 = Good; 2 = Moderate; 3 = Poor)	1.625	0.642
Plot slope	The slope of the plot (1 = Flat land; 2 = Slope land; 3 = Concave land; 4 = Others)	1.225	0.517
Soil type	The slope of the plot (1 = Sandy soil; 2 = Clay; 3 = Loam; 4 = Others)	2.259	0.796
Training	Number of technical training sessions in the village	1.165	2.997
Plot transfer	Whether the plot is transferred land (1 = Yes; 0 = No)	0.425	0.494
Cultivated land scale	The logarithm of the cultivated area	3.398	1.426
Crop type	1 = Rice; 2 = Maize	0.533	0.499
Family income	Total household income in the previous year (yuan)	83,047.8	306,861.8
Gender	0 = Female; 1 = Male	0.965	0.183
Education level	Years of education of household head (years)	6.791	3.071
Years of planting	Years of planting of household head (years)	31.870	13.702

shows definition and descriptive statistics of variables used in this study.

Land characteristics. Land characteristics include seven variables, namely whether the land can be irrigated, whether it is transferred land, the slope of the land, the quality of the land, the soil type, the area of the land, and the type of crops planted on the land. Specifically, first, whether the land is irrigable is a dummy variable. This is mainly because irrigable land has better quality and has a significant impact on technical efficiency. Second, whether the land is transferred land is a dummy variable. This variable reflects the stability of land tenure. Owned land is more stable than transferred land, and farmers may be more inclined to invest more, which may prompt higher technical efficiency. Third, the slope of the land is a dummy variable. Generally, flat land has higher quality than slope land and other types of land, and is more suitable for cultivation and mechanized production, which affects input behavior and technical efficiency, and also affects the application of conservation tillage technology. Fourth, the quality of the land. The quality of the land is the subjective evaluation of the interviewees of the land, compared with land in the same village. Fifth, soil type includes sandy soil, clay, loam, and four other types. Clay and loam are set as dummy variables, with sandy soil is the control group. Different soil types differ in water and fertilizer retention, and loam has better water and fertilizer retention and is more suitable for crop growth compared to the other two types. Farmers may make input decisions based on different soil types, which also affect the application of environmentally friendly technology and technical efficiency. Sixth, the area of the land also affects farmers’ technical efficiency. Generally, the larger the land area, the more suitable it is for large-scale production, and the more farmers tend to invest in large plots to improve technical efficiency [18,19]. Seventh, the type of crop planted on the land. Different crops have different physiological properties, and farmers adopt different conservation tillage technologies for different crops. The crops selected in this paper are mainly maize and rice.

The variables of farmer characteristics consist of total household income, level of education of the head of the household, years of farming, and years of planting. Since the application of conservation tillage technology demands certain costs, the level of household income has a substantial impact on technical efficiency [20]. The remaining three variables are applied to measure the individual farming ability. Gender, education level, and years of planting are significant variables for measuring household production capacity, particularly education level, which directly measures the individual capacity of farmers; all of them affect agricultural production, thereby influencing the level of technical efficiency [21,22].

In addition to the characteristics of the households, this study also includes the number of technical training sessions within the village in the model. Agricultural technology training provides an effective way to increase farmers’ investment in human capital and introduce new land protection technologies into agricultural production. Through technology training, external conditions are created for farmers to learn land protection techniques [23]. The more training sessions there are in the village, the more opportunities farmers have to understand and learn about the economic and ecological benefits of land protection techniques [24], which can have a subtle influence on their adoption of land cultivation techniques [25]. In addition, technology training can enable farmers to use new technologies more scientifically and rationally, which has a direct impact on the level of technical efficiency. Ignoring this variable would result in significant endogeneity problems.

Furthermore, to control the impacts of different natural geography conditions, climate conditions, and economic development levels in different regions, this study also uses provincial dummy variables to control for provincial fixed impacts, to obtain more consistent estimation results.

2.3. Data Source

The data used in this study come from a micro survey of farmers conducted in 2015. The field survey used a multi-stage random sampling method to ensure the objectivity and representativeness of the survey data. First, based on the comprehensive consideration of geographical factors, economic development level, and climatic and geographical conditions, the sample provinces were selected. Second, two counties were randomly selected as sample counties in each province, and four townships were selected in each county. Two villages were selected in each township, and approximately twelve households were selected in each village. After sorting, a total of 853 sample households were obtained. At the same time, to analyze the input–output and technical efficiency level at the plot level, the field survey also randomly selected two plots in each household’s planting land, obtaining an effective plot number of 1706. When selecting plots, a certain principle was followed to randomly select the two pieces of land that the household remembered most clearly, and detailed input–output information of the plot was recorded.

3. Results

3.1. Estimation Results of Technical Efficiency Based on Stochastic Frontier Production Function

The results based on the Cobb–Douglas production function estimated by stochastic frontier analysis are shown in

Table 2. Estimation of the stochastic front production function.

Variables	Coefficient	Standard Error
Agricultural input (Logarithm)	0.068 ***	0.013
Labor input (Logarithm)	−0.062 ***	0.006
Agricultural machinery input (Logarithm)	0.0181 ***	0.006
Land input (Logarithm)	0.0297 ***	0.054
Constant	6.9842 ***	0.073
lnsig2v	−4.3469 ***	0.110
lnsig2u	−1.6421 ***	0.049
Wald Chi2	283.52
Observation	1706

Note: *** p < 0.01.

. It can be seen from the table that both the technical inefficiency term and the random disturbance term in the residual term are significant at the 1% level, indicating that the fitting effect of the method for estimating technical efficiency is good. Among the four input factors, agricultural inputs, labor, machinery, and land, all are significant at the 1% level. The agricultural input has a positive effect on unit area yield, indicating that with the increase in inputs such as seeds, fertilizers, and pesticides, the unit area yield also increases. It should be noted that the coefficient of the labor variable is significantly negative at the 1% level, indicating that with the increase in labor input, the yield per unit area shows a decreasing trend. This indicates that the marginal return of the labor factor decreases under the condition that other input variables remain unchanged. This phenomenon also conforms to the current situation of China’s agricultural production, which features surplus labor and high labor costs.

3.2. Baseline Regression

As shown in

Table 3. Tobit model estimation results of the impact of conservation tillage technology application on production efficiency.

Variables	Coefficient	Standard Error
Conservation tillage technology application	0.020 **	(0.009)
Plot irrigablity	0.039 ***	(0.011)
Plot quality
Moderate	−0.021 ***	(0.007)
Poor	−0.067 ***	(0.013)
Plot slope
Slope land	−0.062 ***	(0.011)
Concave land	−0.019	(0.019)
Others	0.016	(0.050)
Soil type
Loam	−0.010	(0.010)
Clay	0.014	(0.010)
Others	−0.056 ***	(0.022)
Training	−0.003 ***	(0.001)
Plot transfer	−0.019 ***	(0.007)
Cultivated land scale	−0.018 ***	(0.003)
Crop type	−0.051 ***	(0.012)
Family income	0.000	(0.000)
Gender	−0.007	(0.019)
Education level	0.003 **	(0.001)
Years of planting	0.000	(0.000)
Area dummies
Jiangsu	−0.126 ***	(0.015)
Henan	−0.061 ***	(0.013)
Sichuan	−0.123 ***	(0.013)
Sigma	0.020 ***	(0.001)
constant	0.848 ***	(0.031)
Observation	1706
LR chi2	328.98

Note: *** p < 0.01, ** p < 0.05.

, the results of the Tobit model show that the impact of conservation tillage techniques’ application on technical efficiency is positive and significant at the 5% significance level, indicating that the average technical efficiency of farmers who use conservation tillage techniques is 0.0196 units higher than those who do not use any conservation tillage techniques when other variables are constant. The land irrigability variable is significant at the 1% level, indicating that the average technical efficiency of irrigated plots is higher than that of non-irrigated plots. Similarly, with high-quality plots as the control group, the average technical efficiency of plots with medium and poor quality is significantly lower than that of high-quality plots; the results of the plot slope variable are also significant, with the average technical efficiency of sloped plots being lower than that of flat plots at the 1% significance level. In addition, variables such as transferred plots, household operating scale, and education level also have significant impacts on the average technical efficiency level. These results are consistent with theoretical expectations and actual agricultural production, indicating that the overall results of this study model are reasonable.

Surprisingly, the variable of the number of technical training sessions in the village is significantly negative at the 1% significance level, indicating that an increase in the number of technical training sessions in the village reduces the average technical efficiency level. This may be due to the self-selection problem of the average number of technical training sessions in the village, as villages with overall lower technical efficiency may be more motivated to conduct more technical training sessions, so villages with more training sessions may have lower average technical efficiency. Omitting this variable may lead to estimation bias in the decision to adopt conservation tillage techniques, as it also has a significant impact on this decision. Therefore, the number of technical training sessions in the village still needs to be included in the model. However, the solution to this self-selection problem is relatively complex, and the variable is not a core variable of interest in this study. As long as the self-selection problem of the core variable of whether to use conservation tillage techniques can be properly addressed, this problem will not have a significant impact on the model estimation results and research conclusions. Therefore, the following will focus on addressing the self-selection problem of the core variable of whether to use conservation tillage techniques.

3.3. PSM Estimation Results

To ensure the validity of the matching, overlap and balance tests should be conducted for the propensity score matching (PSM). Figure 1 displays a common range of propensity scores, depicting that the majority of observations lie within this range with only a few outliers. This indicates a small sample loss and good overlap between the treatment and control groups, and validates the possibility of PSM estimation.

To ensure the quality of the matching, it is generally required that the standardized bias between the two groups does not exceed 10%.

Table 4. Results of the balance test for the matching variables.

Variables	Matching Status	Mean		Bias Reduction Range (%)	t-Test
		Treatment Group	Control Group		t Value	p Value
Plot irrigability	Unmatched	0.67	0.73	14.80	−2.34	0.02
	Matched	0.67	0.62		2.23	0.03
Plot quality
Medium	Unmatched	0.43	0.44	23.20	−0.28	0.78
	Matched	0.43	0.44		−0.25	0.80
Poor	Unmatched	0.12	0.09	10.00	1.55	0.12
	Matched	0.11	0.14		−1.50	0.13
Plot slope
Slope land	Unmatched	0.20	0.16	41.40	1.60	0.11
	Matched	0.20	0.22		−1.04	0.30
Concave land	Unmatched	0.03	0.04	66.80	−0.35	0.73
	Matched	0.03	0.03		0.14	0.89
Soil type
Loam	Unmatched	0.34	0.34	−288.20	−0.22	0.83
	Matched	0.33	0.36		−0.99	0.32
Clay	Unmatched	0.44	0.41	69.10	1.12	0.26
	Matched	0.44	0.45		−0.40	0.69
Other	Unmatched	0.02	0.04	78.30	−1.18	0.24
	Matched	0.02	0.03		−0.31	0.76
Number of technical training sessions in the village	Unmatched	1.32	1.36	−1281.40	−0.26	0.80
	Matched	1.33	1.98		−3.71	0.00
Plot transfer	Unmatched	0.46	0.42	89.20	1.59	0.11
	Matched	0.45	0.45		0.20	0.84
Household operation scale	Unmatched	3.52	3.31	49.80	2.54	0.01
	Matched	3.49	3.59		−1.40	0.16
Crop type	Unmatched	0.40	0.37	−123.20	1.02	0.31
	Matched	0.40	0.46		−2.61	0.01
Total household income	Unmatched	96,898	69,990	95.50	1.34	0.18
	Matched	80,503	81,716		−0.11	0.91
Gender	Unmatched	0.97	0.99	−189.00	−1.65	0.10
	Matched	0.97	0.93		3.87	0.00
Education level	Unmatched	6.44	6.53	−342.70	−0.50	0.62
	Matched	6.41	6.81		−2.43	0.02
Years of planting	Unmatched	32.08	31.96	−1504.90	0.16	0.87
	Matched	32.20	30.18		2.85	0.00
Area
Jiangsu	Unmatched	0.28	0.29	−302.60	−0.51	0.61
	Matched	0.28	0.23		2.46	0.01
Sichuan	Unmatched	0.42	0.29	83.40	4.83	0.00
	Matched	0.42	0.40		0.91	0.36

provides the results of the balance test for the matching variables, which shows that the standardized bias for all matching variables is less than 10% after matching. This suggests that the matching process has significantly reduced the standardized bias for most matching variables, indicating good matching results compared to the results obtained before matching. In addition, the t-test results for all variables are not significant at the 10% level, indicating that there are no significant differences between the treatment group and the control group, and the balance test was passed.

This study estimates the average treatment effect (ATT) of using conservation tillage technology on the treated land by utilizing the radius matching method in propensity score matching (PSM) with a matching radius set at 0.01. The ATT value quantifies the difference in average technical efficiency between using conservation tillage technology and not using it. The standard errors reported by the Psmatch2 program in Stata 14 do not take into account the fact that the propensity score is estimated, i.e., the score is assumed to be the true value, and then the standard errors are derived [26]. Following previous studies [27], this paper obtains standard errors through bootstrap, while trying to overcome the small sample bias caused by the loss of a small number of unmatched samples in the matching process.

Table 5. PSM estimation results of the impact of conservation tillage technology decisions on technical efficiency.

Groups	Results
Control group	0.7154
Treatment group	0.7376
ATT	0.0222 ** (0.0110)

Note: Bootstrap standard errors are shown in parentheses, ** p < 0.05.

displays the outcomes of the PSM estimation, which reveal that the average technical efficiency of the control group, who did not utilize conservation tillage technology, was 0.7154. Meanwhile, the average technical efficiency of the treatment group, who employed conservation tillage technology, was 0.7376. Both results are statistically significant at the 5% level, with an ATT of 0.0222. This indicates that when controlling for other variables, the average technical efficiency of farmers who utilized conservation tillage technology was 0.0222 higher than those who did not use it. The Tobit estimation result of 0.0196 does not significantly differ from this finding, further reinforcing the robustness of the estimation outcomes.

3.4. Heterogeneity Analysis Conservation Tillage Techniques

To further analyze the impacts of different conservation tillage techniques on farmers’ technical efficiency, this study will analyze the impacts of the straw returning technique, soil testing and formula fertilization technique, minimum tillage technique, and deep loosening technique on farmers’ technical efficiency, respectively. Considering that the use of different specific conservation tillage techniques also involves self-selection problems, this study continues to use the PSM method to deal with self-selection problems to obtain consistent estimates.

Table 6. PSM estimation results of the impacts of different conservation tillage techniques application on technical efficiency.

Groups	Straw Returning	Deep Loosening	Minimum Tillage	Soil Testing and Formula Fertilization
Control group	0.748	0.762	0.729	0.746
Treatment group	0.740	0.762	0.759	0.769
ATT	−0.008 (0.02)	0.004 (0.017)	0.030 * (0.017)	0.023 ** (0.011)

Note: Bootstrap standard errors are shown in parentheses, ** p < 0.05, * p < 0.1.

presents the PSM estimation results of the effects of various conservation tillage techniques on technical efficiency. Since the balance and overlap tests for this part of the matching are almost the same as the previous part, they are not repeated here to avoid repetition. The findings reveal that the average treatment impacts of straw returning and deep loosening techniques are not statistically significant at the 10% level, indicating that their use does not significantly affect farmers’ technical efficiency, which is in line with the results of the Tobit model. However, the use of minimum tillage and soil testing and formula fertilization techniques has a statistically significant impact at the 10% and 5% levels, respectively, with average treatment impacts of 0.030 and 0.023. These results are similar to the Tobit estimation results (0.034 and 0.022, respectively), indicating that minimum tillage technique and soil testing and formula fertilization technique can improve farmers’ technical efficiency level, and further indicating that the estimation results of this study are robust.

3.5. Mechanism Analysis

To further study the mechanism of the impact of conservation tillage technology on technical efficiency, this paper constructs an econometric model to empirically analyze the impact of the application of conservation tillage technology on yield and cost. Theoretically, the application of conservation tillage technology can affect technical efficiency through two mechanisms: increasing yield and reducing production costs. Specifically, conservation tillage can reduce two to three operations, lower operating costs by about 20%; increase dryland crop yields by more than 5%; and increase farmers’ income by 20–30% [28,29]. For example, deep loosening technology refers to the tillage technique of deep soil cultivation through large-scale deep tillage machinery. The feature of this technique is that it breaks the hard plow bottom layer, thickens the loosened soil layer, and improves the soil tillage structure without overturning the soil and disrupting the original soil layer structure, thus effectively enhancing the soil’s water retention, drought resistance and flood control capacity, thereby improving the basic production capacity of food, increasing the agricultural resilience to natural disasters, and achieving the goal of reducing costs and increasing income [30,31].

3.5.1. Impact of Conservation Tillage Technology Application on Yield and Cost

Since both unit yield and cost are continuous variables, ordinary least squares (OLS) regression is used to estimate the above regressions. However, similar to the technical efficiency equation, the unit yield and cost equations also have self-selection issues, and PSM method is used to regress to obtain consistent estimates.

Table 7. PSM estimation results of the impact of conservation tillage technology application on yield and cost.

Group	Yield	Cost
Control group	6.803	6.845
Treatment group	6.858	6.7575
ATT	0.054 ** (0.023)	−0.088 ** (0.042)

Note: Bootstrap standard errors are shown in parentheses, ** p < 0.05.

presents the PSM estimation results of the impact of conservation tillage technology on yield and cost. The table reveals that the decision to implement conservation tillage technology is statistically significant at the 5% level for both PSM regression results, signifying a notable increase in unit yield level and reduction in cost. Moreover, the direction and degree of impact are not substantially distinct from the Tobit model results, thereby reiterating the strength and robustness of the research outcomes.

3.5.2. Impact of Different Conservation Tillage Techniques’ Application on Yield and Cost

This study further analyzes and validates the impacts of different conservation tillage techniques on yield and cost.

Table 8. The impacts of different conservation tillage techniques application on yield and cost.

Variables	Yield				Cost
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Straw returning	0.022				−0.013
	(0.017)				(0.024)
Deep loosening		0.038				−0.022
		(0.027)				(0.047)
Soil testing and formula fertilization			0.074 ***				−0.104 ***
			(0.018)				(0.033)
Minimum tillage				0.047 **				−0.094 **
				(0.021)				(0.040)
Other variables	controlled	controlled	controlled	controlled	controlled	controlled	controlled	controlled

Note: Standard errors are shown in parentheses; *** p < 0.01, ** p < 0.05.

shows the impacts of different techniques’ adoption on yield and cost. Regarding yield, the results show that straw returning and deep loosening have no significant impact on yield, while minimum tillage and soil testing and formula fertilization are significantly positive at 5% and 1% significance levels, respectively, indicating that these two techniques can significantly improve yield and promote the increase of technical efficiency. Regarding cost, the results show that straw returning and deep loosening have no significant impact on unit production cost, while minimum tillage and soil testing and formula fertilization are significantly negative at 5% and 1% significance levels, respectively, indicating that these two technologies can significantly reduce production costs and also significantly improve yield, further enhancing farmers’ technical efficiency.

Table 9. PSM estimation results for the impacts of different conservation tillage techniques application on yield and cost.

Dependent Variables	Groups	Straw Returning	Deep Loosening	Minimum Tillage	Soil Testing and Formula Fertilization
Yield	Control group	0.748	6.992	6.839	6.922
	Treatment group	6.861	7.002	6.924	6.973
	ATT	−0.023 (0.057)	0.034 (0.017)	0.085 ** (0.038)	0.051 ** (0.023)
Cost	Control group	6.884	6.351	6.587	6.584
	Treatment group	6.806	6.334	6.592	6.485
	ATT	0.056 (0.114)	−0.017 (0.063)	−0.287 ** (0.144)	−0.099 ** (0.045)

Note: Standard errors are shown in parentheses; ** p < 0.05.

displays the PSM estimation outcomes for the effects of diverse conservation tillage methods on both yield and cost. The table reveals that straw returning and deep loosening practices do not significantly influence either yield or cost. In contrast, minimum tillage and soil testing and formula fertilization techniques have a significant impact at the 5% level. These findings are consistent with previous studies, supporting the notion that minimum tillage and soil testing and formula fertilization techniques can considerably enhance yield and lower costs, thus enhancing farmers’ technical efficiency. Additionally, the PSM estimation results reinforce the Tobit model results in terms of the direction and magnitude of the effects, indicating the robustness of the research findings.

3.6. Robustness Test

In order to assess the robustness of the aforementioned estimation results, this study performed tests by altering the independent variables and redefining them.

3.6.1. Quantity of Conservation Tillage Adopted

The regression results above employed dummy variables to measure the adoption of conservation farming technology. However, in practical production, farmers may use various conservation tillage techniques concurrently. It is challenging to capture the collective impact of different conservation tillage techniques using dummy variables. As such, this study conducted an additional analysis of the quantity of conservation tillage techniques applied as an independent variable, and used it to test the robustness of the results.

Table 10. Estimation results of the quantity of conservation tillage technologies.

Variables	Results
Quantity of conservation tillage techniques applied	0.018 ***
	(0.005)
Constant	0.858 ***
	(0.033)
Other variables	controlled
Observations	1706
LR chi2	343.11

Note: Standard errors are shown in parentheses; *** p < 0.01.

shows the estimation results of the quantity of conservation tillage techniques applied. From the table, it can be seen that the quantity of conservation tillage techniques applied is significantly positive at the 1% significance level, indicating that the more conservation tillage techniques used, the higher the level of technical efficiency. These research results are similar to the previous research results, indicating that the research results of this study are robust.

3.6.2. Technical Efficiency Estimation Using Translog Production Function

The simplicity of the Cobb–Douglas production function led to its initial use in measuring technical efficiency. However, the transcendental logarithmic production function has the advantage of capturing the non-linear relationship between input and output and the elasticity of factors. To verify the robustness of the technical efficiency calculation method used in this paper, the dependent variable was replaced with technical efficiency, calculated by the transcendental logarithmic production function.

Table 11. PSM estimation results using technical efficiency estimated by translog production function.

Group	Conservation Tillage Technology Application	Straw Returning	Deep Loosening	Minimum Tillage	Soil Testing and Formula Fertilization
Control group	0.716	0.749	0.766	0.727	0.752
Treatment group	0.736	0.739	0.768	0.764	0.772
ATT	0.012 * (0.011)	−0.010 (0.027)	0.002 (0.016)	0.037 ** (0.017)	0.020 * (0.011)

Note: Standard errors are shown in parentheses; ** p < 0.05, * p < 0.1.

shows the PSM estimation results using the transcendental logarithmic production function. Due to the self-selection problem in the model, no robustness test was conducted for the Tobit model, and the PSM method was directly used for estimation.

The estimation coefficients, significance, and other results do not differ significantly from those obtained using the Cobb–Douglas function to calculate technical efficiency, irrespective of the dummy variable for conservation tillage technology application or specific technique variables. This suggests that the research findings in this paper are highly robust.

4. Conclusions

This study employed stochastic frontier analysis to measure farmers’ technical efficiency based on micro-level survey data of their land plots, and explored the impact of the application of conservation tillage techniques on technical efficiency. The findings suggest that the application of conservation tillage techniques can significantly improve farmers’ technical efficiency by optimizing the allocation of agricultural production factors, effectively reducing agricultural production costs and improving crop yields. Moreover, different conservation tillage techniques have significantly different impacts on technical efficiency. Straw returning and deep loosening have no significant impact, while minimum tillage and soil testing and formula fertilization can significantly improve technical efficiency. This is mainly because the operation of straw returning and deep loosening requires additional investment in machinery and labor costs, and their yield-increasing impacts take a long time to develop. In contrast, minimum tillage and soil testing and formula fertilization do not require high costs for labor and machinery, and minimum tillage can effectively reduce land costs while also significantly promoting yield.

5. Policy Implications

Based on the above research conclusions, this study proposes the following policy suggestions to effectively promote the application of conservation tillage techniques:

First, enhancing the promotion of conservation tillage techniques is deemed necessary. Conservation tillage techniques are not only conducive to reducing agricultural non-point source pollutant emissions, but are also beneficial to improving technical efficiency, achieving a win-win situation for rural environmental protection and agricultural production. However, currently, the adoption rate of some conservation tillage techniques is still relatively low, especially the soil testing and formula fertilization technique and minimum tillage technique, which have significant impacts on technical efficiency, with the highest adoption rate being only slightly over 30%. Therefore, it is necessary to continue to promote popularization of and education on conservation tillage technology, improve farmers’ scientific awareness and ability to use conservation tillage technology, and promote the integration of agricultural production and productive service industries with the aid of public opinion guidance and herd behavior among farmers. To encourage more farmers to adopt more conservation tillage techniques, targeted promotion and marketing on mature internet marketing and e-commerce platforms should be used.

Second, the promotion of conservation tillage technology should be in line with local conditions. The technical efficiency of different conservation tillage techniques varies, and promotion of the suitability of straw returning and deep loosening technology should be fully considered to maximize the yield-increasing and cost-saving impact of conservation tillage technology. Additionally, conservation tillage techniques vary in their effectiveness across different regions due to variations in geography and climate. Some techniques have a more significant impact on technical efficiency in some areas, but may not be valid in other areas.