Impacts of the Extension of Cassava Soil Conservation and Efficient Technology on the Reduction of Chemical Fertilizer Input in China

Abstract

The reduction of fertilizer inputs can be considered as an effective policy tool for achieving Sustainable Development Goal 12 and for reaching carbon neutrality. In this study, we examine the impact of the extension of cassava (Manihot sculenta) soil conservation and efficiency (SCE) technology on the cost of chemical fertilizers used by China’s cassava industry. The SCE technology was developed by a Technology Integration and Demonstration (TID) Project of China’s National Key R&D Program, which is an innovative and official channel of agriculture technology extension. Based on data collected from cassava farmers in Guiping, Guangxi, China, the differences-in-differences with propensity score matching (PSM-DID) approach was conducted in this study. We found that farmers participating in the cassava SCE technology project reported a reduction in fertilizer costs per mu (15 mu = 1 ha) of approximately 24.0%. Consequently, the results demonstrate that the government can increase the number of TID projects in order to reduce chemical fertilizer inputs and to encourage the sustainable development of agriculture.

1. Introduction

Achieving the environmentally sound management of chemicals is one of the targets of Sustainable Development Goal 12 (SDG12) [1]. Though the agricultural production process relies heavily on chemical fertilizers in order to increase crop yield, fertilizers may cause water pollution (e.g., nitrates, ammonium) and air pollution (e.g., ammonia, nitrogen oxides), as well as greenhouse gas emissions (GHG), soil acidification, and biodiversity loss [2,3,4]. China is the world’s largest consumer of chemical fertilizers, accounting for around one-third of the world’s consumption of nitrogen fertilizers [5]. Due to China’s official commitment to reach a carbon peak by 2030 and carbon neutrality by 2060, the reduction of fertilizer inputs can be considered as an effective policy tool for reducing GHG emissions in agriculture [6].

Farmers using fertilizer is influenced by a variety of factors. In addition to macroeconomic factors, such as regional planting structure adjustments, fertilizer subsidies, fertilizer prices, and the prices of agricultural products [7,8,9,10], other factors can be categorized into three categories. Firstly, a farmer’s characteristics, such as his or her age, educational level, and previous training, have a significant impact on the amount of fertilizer he or she will apply [11,12,13,14]. Secondly, agricultural operations can be carried out on large scales with machinery, thereby reducing the amount of chemical fertilizer applied per unit area [15,16,17]. Thirdly, new technologies, such as biotechnology, nanotechnology, and new agricultural machinery, may directly or indirectly affect farmers’ behavior, including fertilizer consumption [18,19,20,21].

On the one hand, by developing and promoting new agricultural management techniques, fertilizer inputs can be directly reduced [22]. The plant-growth-promoting rhizobacteria (PGPR), for example, facilitate element management in the context of reduced chemical fertilizer usage, which directly affects crop growth through inoculation [18]. Using the Internet, on the other hand, can change farmers’ perceptions of green production through publicity, resulting in less fertilizer use [23,24]. Furthermore, the integrated soil–crop system management (ISSM) enhances fertilizer use efficiency through increased crop density while ensuring a constant amount of fertilizer is applied [14].

It is the objective of the agricultural extension services to transfer knowledge from researchers to farmers, to assist farmers in making better decisions, and to educate them on how to make better decisions. In addition to clarifying farmers’ own goals and possibilities, the extension services encourage desirable agricultural development [25]. Thus, the extension services provide a dual function in bridging the gap between scientists and farmers: they facilitate both the adoption of new technologies and their adaptation to the local environment [26]. Previous studies have shown that the impacts of agricultural extension services on the adoption of agricultural technologies and farm productivity are significant [27,28,29,30]. Some studies provided strong evidence of the impact of agricultural extension services on fertilizer use in China [31,32].

However, since the Chinese government began to cut its support for public research and extension in the late 1980s [33], it was found that between 1996 and 2002 more than 80% of farmers had no access to extension technicians in their villages [34]. Thus, the Chinese government has taken a number of new initiatives in order to promote a more demand-driven system of public agricultural extension. It was one of these efforts that added the Technology Integration and Demonstration (TID) Projects to China’s National Key R&D Program funded by the Ministry of Science and Technology of China. However, to the best of our knowledge, the economic impact of the TID Projects on farmers is still unknown.

Therefore, unlike previous studies, this study aims to examine the impact of cassava (Manihot sculenta) soil conservation and efficiency (SCE) technology from a TID Project on the cost of chemical fertilizers used by China’s cassava industry. Cassava is a multifunctional cash crop that can be used in numerous ways, including as a raw material for starch, livestock feed, and related processed foods, and also as a bioenergy source through fermentation. Cassava production in China is limited by a shortage of domestic supplies and excessive fertilizer application worsens environmental degradation. Thus, the SCE technology was developed to reduce the input of chemical fertilizers.

The article contributes to the literature in two ways. First, an economic analysis of the effect of technology on cassava fertilizer use, based on data collected from cassava farmers in Guiping, Guangxi, China, has been conducted using the differences-in-differences with propensity score matching (PSM-DID) approach. Second, the study also explores the role of the TID Projects of China’s National Key R&D Program in providing demonstration and extension services to farmers.

2. Background

2.1. Cassava Production in China

With the increasing demand for feed, starch, modified starch, and biofuel ethanol in China [35,36,37], cassava-sown areas and production have increased rapidly since 2000 (Figure 1). As a consequence, the harvested area of cassava in China has increased from 239,000 ha in 2000 to 300,000 ha in 2020 and the amount of cassava produced has increased from 3,822,000 MT in 2000 to 4,885,000 MT in 2020. Furthermore, the ratio of cassava production to tubers (in the definition of the National Bureau of Statistics of China, tubers include potatoes and sweet potatoes but not cassava) increased to 16.4% in 2020 from 10.4% in 2000. In addition, as a result of China reducing its cassava import tariff from 10% to 5% in 2004, its cassava imports increased by 41.8% from 9,781,000 MT in 2003 to 13,872,000 MT in 2004. Since cassava is a tropical crop, its production is primarily limited to Guangxi, Guangdong, Hainan, Yunnan, Fujian, and Jiangxi in China. Among them, the largest amount of cassava is grown in Guangxi, which is responsible for about two-thirds of China’s total cassava production [38].

Since Chinese cassava production cannot meet demand, cassava in China has a serious supply and demand imbalance. According to the Food Balances of the FAO, China’s net cassava imports reached 21,294,000 MT in 2019, with a self-sufficiency rate of 19.0%. Thus, domestic farmers typically invest large quantities of fertilizer to increase their yields. As reported by Jiao et al., (2019) [37], the intensity of nitrogen fertilizer input for cassava production reached 207 kg/ha, compared with 115.95 kg/ha for wheat, 96.9 kg/ha for maize, and 125.85 kg/ha for potatoes. In order to sustain agriculture, it is important to control fertilizer use as much as possible [39].

2.2. Cassava Soil Conservation and Efficient Technology

In this study, the cassava soil conservation and efficiency (SCE) technology was founded by China’s National Key R&D Program and developed by the Tropical Crops Genetic Resources Institute, Chinese Academy of Tropical Agricultural Sciences, and the Rubber Research Institute, Chinese Academy of Tropical Agricultural Science. The program began in 2020 and Guiping, Guangxi, is the main location for its implementation.

From a technical standpoint, the cassava SCE technology focuses on soil conservation, sustainable production, and improving cassava farming operations. Specifically, the cassava SCE technology focuses primarily on cassava soil nutrient diagnosis and balanced fertilizer application. Based on the nutrient status of the cassava soil, it reduces operating farmers’ cassava fertilizer input by improving the fertilizer formula and reducing fertilizer use to improve fertilizer efficiency. With regard to the operational and environmental benefits of farmers, the cassava SCE technology is crucial in improving soil quality and in promoting sustainable cassava development.

3. Methods and Data

3.1. Methods

3.1.1. The OLS Regression

As shown in Equation (1), the main explanatory variable is whether the farmer participates in the cassava SCE technology project (program).

$$\ln Y_{it}=a+b\ast progra{m}_{it}+c{X}_{it}+{\epsilon }_{it}$$

(1)

where $Y_{it}$ denotes the fertilizer cost of farmer i in year t, $progra{m}_{it}$ denotes whether farmer i participates in the project in year t, $X_{it}$ denotes a series of control variables, and ${\epsilon }_{it}$ is a random error term.

According to Figure 2, the theoretical intercept b between participating and non-participating projects represents the net benefits of participating projects for farmers growing cassava. However, Equation (1) is endogenous, since unobservable factors such as farmers’ ability in $X_{it}$ cannot be controlled, resulting in biased and inconsistent estimates of coefficient b that do not reflect the effect of farmers’ participation.

3.1.2. Difference-In-Difference

Under the difference-in-difference method (DID), the incremental difference between the treatment group and the control group is calculated under the policy intervention in order to identify the effect of the policy [40,41,42]. The cassava SCE technology optimization can be an independent “quasi-natural experiment” in which participating farmers serve as the treatment group and non-participating farmers as the control group. There is an incremental change in the treatment group $T=A_2-A_1$, an incremental change in the control group $C=B_2-B_1$, and a net effect $Diff=T-C$.

The DID model is shown in Equation (2), where ${\alpha }_2$ is the main focus. In Figure 3, $Diff=\left(A_2-A_1\right)-\left(B_2-B_1\right)$ indicates the difference in cassava fertilizer costs between participating and non-participating farmers. This can also be called the average treatment on the treated (ATT).

$$\ln Y_{it}={\alpha }_1+{\alpha }_{2}di{d}_{it}+{\alpha }_3{X}_{it}+{\mu }_i{D}_i+{\gamma }_t{T}_t+{\epsilon }_{it}$$

(2)

where $D_i$ and $T_i$ are dummy variables. $D_i$ indicates whether farmer i participates in the project (if farmer i participates, then $D_i=1$, otherwise $D_i=0$). $T_i$ indicates whether year t is before or after project implementation ($T_t=1$ if year t is after project implementation, otherwise $T_t=0$). $Y_{it}$ denotes the fertilizer cost of farmer i in year t, $X_{it}$ denotes a series of control variables, and ${\epsilon }_{it}$ is a random error term. Since the farmers are not totally randomized whether to participate in the cassava soil conservation and efficient technology optimization project, Equation (2) suffers from selection bias and the resulting estimated ${\alpha }_2$ is also biased.

3.1.3. Propensity Score Matching

Using a discrete probability selection model, the propensity score matching (PSM) method estimates the propensity scores for individuals in the control and treatment groups. Then, the PSM method selects individuals from the control and treatment groups with similar propensity scores for sample matching to obtain unbiased estimates of the ATT [43,44,45]. In order to analyze the factors influencing farmers’ participation in the project, a logit or probit model (in this paper, we choose the logit model) is constructed to analyze these factors. Based on the estimated results of the model, a propensity score is used to predict farmers’ participation in the project. Given the control variable $X_i,$ Equation (3) represents the propensity score of farmer i to participate in the project.

$$P\left(X_i\right)=Prob\left(D_i=1|X_i\right)=e^{b{X}_i}/\left(1+e^{b{X}_i}\right)$$

(3)

where $X_i$ indicates the variables involved in the matching, $D_i$ indicates whether the farmer participates in the project, and b is the parameter to estimate.

In the following step, a matching algorithm is selected according to the propensity score. Common matching algorithms include the nearest neighbor method, the kernel matching method, the radius matching method, etc., [46]. As a final step, ATT is calculated using the matched samples and Equation (4).

$$ATT=E\left. [\ln Y_i^T-\ln Y_i^C|D_i=1,P\left(X_i\right)\right]=E\left. [\ln Y_i^T|D_i=1,P\left(X_i\right)\right]-E\left. [\ln Y_i^C|D_i=1,P\left(X_i\right)\right]$$

(4)

where $Y_i^T$ and $Y_i^C$ represent the fertilizer costs of the treatment and control groups, respectively.

Due to the difficulty of controlling for unobservable factors, such as farmers’ ability in $X_i$, Equation (3) is also endogenous in the matching process. Consequently, the estimation results of coefficient b [47] are also biased and inconsistent and the ATT calculated in Equation (4) may also be biased.

3.1.4. Differences-In-Differences with Propensity Score Matching

The differences-in-differences with propensity score matching (PSM-DID) approach combines the advantages of DID and PSM by using the matched samples to construct a double difference model to assess policy effects [48,49,50]. Therefore, in this paper, the PSM-DID approach is used to study the effects of farmers’ participation in the project of cassava SCE technology.

$$\ln Y_{it}^{psm}={\beta }_0+{\beta }_{1}di{d}_{it}+{\beta }_2{X}_{it}+{\mu }_i{D}_i+{\gamma }_t{T}_t+{\epsilon }_{it}$$

(5)

where $Y_{it}^{psm}$ indicates the fertilizer cost of farmer i in year t after matching (). ${\beta }_1$ is the core parameter focused on in this paper, indicating the ATT of cassava fertilizer cost of farmers in the treatment and control groups after matching. $D_i$ indicates whether farmer i participates in the project (if farmer i participates, then $D_i=1$, otherwise $D_i=0$). $T_i$ indicates whether year t is before or after project implementation ($T_t=1$ if year t is after project implementation, otherwise $T_t=0$). $Y_{it}$ denotes the fertilizer cost of farmer i in year t, $X_{it}$ denotes a series of control variables, and ${\epsilon }_{it}$ is a random error term.

3.2. Data

3.2.1. Data Collection

We obtained our data from a 2022 cassava industry research conducted in Guiping City, Guangxi Zhuang Autonomous Region (Figure 4). A total of 113 households in five townships of Guiping City were randomly selected for the research activity, and questionnaire surveys were conducted between 2019 and 2021 to collect information about household members, household heads, land status, tropical crop cultivation, cassava planting areas, and input costs. The data are therefore balanced panel data with 339 observations. We divided the sample into two categories: treatment groups and control groups. Treatment groups participated in the project, while control groups did not.

3.2.2. Key Variables

The logarithm of fertilizer cost (lnfcost) for cassava was chosen as the explained variable in this study since the core objectives of the cassava SCE technology project are to reduce farmers’ fertilizer costs. Since fertilizer prices were almost identical in the research area, fertilizer cost could be used as a proxy variable for fertilizer input quantity. According to the National Bureau of Statistics of China, the explained variable lnfcost excludes inflation based on a fixed-base CPI index.

We use the ATT of cassava fertilizer cost (did) as the key explanatory variable. Furthermore, farmers’ household management characteristics and land cultivation status were introduced as control variables in this study to minimize estimation error due to omitted variables.

Table 1. Variable description.

Variables	Descriptions
Explained variable
lnfcost	Logarithm of fertilizer cost per mu (15 mu = 1 ha)
Key explanatory variable
did	$did=D\times T$
Control variables
acre	Sown area (mu)
rlcacres	Contracted farmland area (mu)
age	Age of household head (year)
edu	Education level of household head(year)
D	Participating in the project (yes = 1, no = 0)
T	Time dummy variable (Year 2021 = 1, otherwise = 0)

presents the names and definitions of the variables used in this study.

3.2.3. Descriptive Statistics

Based on a quasi-natural experiment, farmers’ participation in the cassava SCE technology project was tested as a treatment group for the sample farmers from 2019 to 2021. Statistical analysis showed that treatment group farmers spent less per mu on fertilizer inputs than control group farmers (

Table 2. Descriptive statistics of the main variables.

Variables	Full Sample		Treatment Group		Control Group		Diff. in Means
	Mean	SD	Mean	SD	Mean	SD	Treatment Control
lnfcost	3.532	0.403	3.507	0.374	3.568	0.440	−0.061 ***
did	0.294	0.456	0.497	0.501	0.000	0.000	0.497
acre	5.764	12.730	6.076	13.420	5.313	11.710	0.763 **
rlcacres	10.210	14.740	10.650	15.590	9.562	13.450	1.008 **
age	54.040	8.150	54.070	8.307	54.000	7.949	0.070
edu	7.564	2.221	7.554	2.197	7.578	2.264	−0.024 *
D	0.591	0.492	1.000	0.000	0.000	0.000	1.000
T	0.333	0.472	0.497	0.501	0.096	0.296	0.401 *
Obs.	339		195		144

Note: *, **, and *** indicate significance at 10, 5, and 1 percent levels, respectively.

). At a 1% significance level, the t-test results revealed significant differences between the treatment and control groups in fertilizer costs. Accordingly, the mean value of fertilizer input per mu was CNY 33.35 for the treatment group and CNY 35.44 for the control group. This indicates that participating in the program might reduce the cost of fertilizer for farmers.

Additionally, the results of the basic characteristics of the sample farmers showed that those who participated in the project had greater sown and contracted land areas than those who did not. The average size of the treatment group was 6.076 mu, while the size of the control group was 5.313 mu. It appears that farmers with larger operations may be more willing to participate in the project, to adopt environmentally friendly technologies, and to engage in cassava production and management. In addition, the average age of household heads in the treatment group was very similar to that of the control group, both around 54 years old.

4. Results

4.1. PSM-DID Estimation

A PSM-DID method was used to estimate the parameters in this paper, which is a theoretically most accurate representation of farmers’ participation in the cassava SCE technology project. The results are outlined in

Table 3. Estimation results.

Variables	(1)	(2)	(3)
	PSM-DID	DID	OLS
did	−0.240 *	−0.178
	(0.116)	(0.141)
acre	−0.008	−0.008 **	−0.008 **
	(0.005)	(0.003)	(0.004)
rlcacres	0.008 *	0.009 ***	0.008 ***
	(0.005)	(0.003)	(0.003)
age	0.005 **	0.005 **	0.005 **
	(0.003)	(0.002)	(0.002)
edu	0.037 ***	0.035 ***	0.035 **
	(0.020)	(0.013)	(0.014)
D	−0.075	−0.081	−0.062
	(0.052)	(0.053)	(0.045)
T	0.329 ***	0.268 **
	(0.119)	(0.131)
Constant term	2.973 ***	2.946 ***	2.975 ***
	(0.290)	(0.194)	(0.200)
Obs.	339	339	339

Note: *, **, and *** indicate significance at 10, 5, and 1 percent levels, respectively; robust standard errors are in parentheses; the 1 nearest neighbor matching algorithms was selected for PSM estimation.

(Column 1) as regression results. The estimated results show that DID is significant at the 10% significance level and the estimated coefficient is −0.240, indicating a −0.240 average treatment effect of farmers participating in the cassava SCE technology project. As a result, the farmers’ participation in the project resulted in a 24.0% reduction in fertilizer cost per mu.

Additionally, contract farmland area, age of household head, education level of household head, and the time dummy variable are significant control variables. As the contracted farmland area of farmers increases by 1 mu, the cost of fertilizer increases by 0.8%. Furthermore, with each additional year of a household head’s age and education level, fertilizer costs increase by 0.5% and 3.7%, respectively.

In addition, we calculated the elasticities of the contract farmland area, age of the household head, and the education level of the household head. The results show that the elasticities of contract farmland area, age of household head, and education level of household head are 0.087, 0.288, and 0.268, respectively. It suggests that for every 1% increase in contract farmland area, age of household head, and education level of household head, fertilizer costs increase by 8.7%, 28.8%, and 26.8%, respectively.

4.2. Robustness Check

To examine the robustness of the PSM-DID results, we compared the results of different estimation techniques, i.e., the DID approach. Table 3 (Column 2) shows that using the DID approach, the variable did is not significant at a significance level of 10%. This result and the estimation results of other variables are not significantly different from those obtained from the PSM-DID estimation. It suggests that the PSM-DID results are robust.

Additionally, the results indicate that DID estimates underestimate the average treatment effect of farmers participating in the project. The primary reason for this is the fact that the DID approach may be capable of controlling some unobservable variables but, since the decision to participate in the project is made by the farmers on their own, it is not exogenous. Thus, the results of the DID approach may be biased due to the endogeneity caused by farmers’ self-selection.

4.3. Results of the Non-DID Approach

We also performed parameter estimation based on the OLS and PSM methods in order to evaluate the differences between the DID approach and the non-DID approach. Table 3 (Column 3) presents the results of the OLS estimation and

Table 4. Estimation results of the PSM.

Matching Algorithm	ATT	S.E.
1 nearest neighbor matching	0.024	0.072
5 nearest neighbors matching	0.057	0.058
Kernel matching	0.042	0.051
Radius matching	−0.048	0.037

presents the results of the PSM estimation.

According to the regression results in Table 3 (Column 3), farmers who participate in the program do not see a significant reduction in fertilizer costs as a result of their participation. There is a possibility that this may be due to the fact that OLS estimation cannot control for unobservable variables that influence the cost of fertilizer for farmers. The result may be biased as a result of the endogeneity problem that is caused by OLS estimation.

Based on the results of the regression with the PSM method, Table 4 indicates that the estimated ATT is not statistically significant. It should be noted that, although PSM controls for selection bias, there is still an endogeneity risk owing to the inability to control for “invisible bias” that may be introduced by unobservable variables.

5. Discussion

Though some research found that the flow of information via these official channels of the agricultural extension services may be restricted [51,52], most of the world’s agricultural extension services are publicly funded [53]. With respect to the reform of China’s agricultural technology extension system, a number of innovative and official channels of agriculture technology extension, such as the “science-and-technology backyard” (STB), have recently attracted considerable attention. According to Zhang et al., (2016) [54] and Li et al., (2022) [55], the STB provides farmer field schools, local field demonstrations, and case-to-case counseling to promote sustainable agricultural production.

In addition, China’s National Key R&D Program also provides demonstration and extension services to farmers. The Technology Integration and Demonstration (TID) Projects in the agricultural field occupy an important position in the National Key R&D Program, which is funded by the Ministry of Science and Technology of China. Compared with the STB, the TID project focuses on a specific technology or a variety of technologies, is more targeted, covers a broader demonstration area, and involves a greater number of farmers.

With the support of the TID project, the R&D and extension team from the Tropical Crops Genetic Resources Institute, Chinese Academy of Tropical Agricultural Sciences and the Rubber Research Institute, Chinese Academy of Tropical Agricultural Sciences constructed a 10,500-mu demonstration base. During the period from 2020 to 2022, the R&D and extension team conducted three rounds of technical training (Figure 5). Based on the results of this study, it is concluded that the TIP project has achieved significant results and is an effective means of reducing farmers’ fertilizer input. Consequently, the TIP projects can play an important role in the diffusion of agricultural technology and in promoting sustainable agricultural development.

The findings of the study have implications for policymakers. First, the government can increase the number of TID projects in order to reduce chemical fertilizer inputs and to encourage sustainable development of agriculture. Second, it is necessary to establish a special technical extension service team within a TID project in order to increase the efficiency of the technology extension. Third, for the effective implementation of TID projects, economists should be encouraged to participate and to use scientific methods.

A few limitations should be noted in this study. First, the demonstration had a relatively limited scope, which resulted in a small sample size for this study. Second, before the launch of the project, a baseline survey should be conducted; a follow-up survey should be conducted at the end of the project. Due to the COVID-19 epidemic, however, this study conducted only one round of research, with farmers filling in baseline data based on recollections, which may be biased. It is important to follow farmers’ participation in cassava SCE technologies over time and to maximize sample sizes in future studies. Lastly, future studies are suggested to continue to explore methods for evaluating the effects of technology adoption and to formulate more accurate policy recommendations.

6. Conclusions

In summary, based on panel data collected from 113 cassava farmers in Guiping, Guangxi from 2019 to 2021, this study examined the average treatment effects of farmers’ participation in cassava soil conservation and efficient (SCE) technologies, using the differences-in-differences with propensity score matching (PSM-DID) approach. We found that farmers participating in the cassava SCE technology project reported a reduction in fertilizer costs per mu of approximately 24.0%. Meanwhile, our result suggests that inappropriate research methods may result in biased results regarding evaluating the effects of the Technology Integration and Demonstration (TID) Projects. Therefore, the cassava SCE technology will not only save costs and increase efficiency but will also promote the sustainability of the cassava industry.