Economic and Environmental Effects of Farmers’ Green Production Behaviors: Evidence from Major Rice-Producing Areas in Jiangxi Province, China

Abstract

This study examines the economic and environmental impacts of green production practices among farmers. It aims to contribute to sustainable agricultural development, mitigate agricultural non-point source (NPS) pollution, and align environmental protection with economic growth. This paper utilizes survey data from 1345 farm households in the main rice production areas of Jiangxi Province, China, using the example of reduced fertilizer application (RFA) among rice farmers. This study constructs a slack-based measure data envelopment analysis (DEA—SBM) model with undesirable outputs to measure environmental effects and applies an endogenous switching regression model (ESRM) to test the economic and environmental effects of farmers’ adoption of green production technologies. We found the following: (1) The RFA behavior of farmers has a significant positive impact on their net profit per hectare (NPH), helping farmers increase their income, with the increase ranging from 2.05% to 6.54%. (2) Farmers’ RFA behavior has a significant positive impact on agricultural green productivity (AGP), contributing to the improvement of the environment, ranging from 44.09% to 45.35%. (3) A heterogeneity analysis found inconsistencies in the income-enhancing and environmental-enhancing effects at different quantiles of NPH and AGP. Therefore, attention should be placed on improving the agricultural product quality supervision system under the market circulation mechanism, creating land scale conditions conducive to the promotion and application of fertilizer reduction technologies and promoting the implementation of externality internalization compensation systems.

1. Introduction

Promoting green agricultural development and improving agricultural ecological efficiency are essential pathways to achieving the organic integration of environmental protection and economic development. They are also crucial for advancing supply-side structural reforms in agriculture and implementing the strategies of green agriculture and quality agriculture [1,2,3,4]. In the past few years, China’s agricultural economy has developed rapidly, with fertilizers playing an indispensable driving role as a key agricultural input [5]. However, fertilizer input has surpassed the optimal level, which balances economic returns and ecological efficiency [6]. According to the law of diminishing marginal returns to factor inputs, as fertilizer application exceeds the optimal level, its marginal output declines [7,8]. Over-reliance on fertilizers hampers agricultural economic growth and leads to severe ecological crises, including soil fertility decline, water pollution, and increased greenhouse gas emissions. These issues threaten agricultural product safety and hinder sustainable agricultural development1. Sustainable agricultural development refers to practices that maintain productivity and profitability while minimizing environmental impacts, conserving natural resources, and ensuring long-term agricultural viability for future generations [9].

Under the dual pressures of economic slowdown and tightening resource and environmental constraints, the Chinese government highlights the importance of comprehensive prevention and control of agricultural non-point source (NPS) pollution. It calls for strengthening rural ecological civilization construction and solid promotion of fertilizer and pesticide reduction and efficiency actions. Agricultural production must shift from over-reliance on resource consumption to an environmentally friendly and sustainable development model, striving to achieve the integrated goals of improving agricultural product quality, increasing agricultural output, and enhancing the ecological environment. To strive to meet the goals set by the Ministry of Agriculture and Rural Affairs of the People’s Republic of China in the “Action Plan for Reducing Fertilizer Use by 2025”, the nutrient structures of nitrogen, phosphorus, potassium, and micronutrients will become more balanced, the national application of agricultural fertilizers will stabilize and decrease, and the fertilizer utilization rate of the three major grain crops will reach 43%.

It is thus clear that under the context of the new era, farmers’ green production behaviors, which are crucial for improving agricultural product quality, preventing agricultural NPS pollution, and ensuring food safety, serve as a critical driving force for achieving sustainable agricultural development. Reduced fertilizer application (RFA) is a typical green agricultural production behavior and has received widespread attention from the government and academia [10]. Many farmers actively cooperate with national agricultural reduction-related work and social development goals, vigorously implementing RFA. To ensure food security for nearly one-fifth of the world’s population, China uses more fertilizers than any other country2. At the same time, China leads the world in rice production and consumption, with its rice planting area and yield accounting for 20% and 33% of the global total, respectively3. The current literature on the environmental impact of China’s rice production finds that nitrogen fertilizer accounts for approximately 50% of the total carbon emissions, which stem from rice production, mainly due to improper fertilizer application [11]. Jiangxi Province, a traditional agricultural hub and a major rice-producing region in China, had a rice planting area accounting for 11.44% of the national total and a yield accounting for 10.33% in 20204. Therefore, we focused on studying the RFA among rice farmers in Jiangxi Province.

Currently, the academic community has been actively exploring the factors influencing farmers’ RFA behavior, primarily focusing on treating this behavior as an outcome variable and investigating the barriers to achieving RFA by farmers. Existing research on RFA mainly focuses on viewing this behavior as an outcome variable, exploring the obstacles to achieving RFA. The research scope primarily involves individual-level behavioral inertia changes, technical cognition, and risk preferences of farmers [12]; the allocation of production factors (land, capital, and labor) at the household level [13]; and socialized services, technical training, and subsidies for green production technologies at the agricultural production level [14]. Although RFA is influenced by many factors, farmers only adopt fertilizer reduction technologies if the net profit exceeds that of traditional fertilization methods [15]. Thus, whether the goal of maximizing farmers’ economic benefits can be met is key to the successful promotion of RFA. Meanwhile, under the severely increasing trend in agricultural NPS pollution, whether RFA can address ecological environment shortcomings is also a matter of concern.

So, can RFA successfully achieve both economic and environmental goals? In the context where the contradiction between economic rationality and ecological rationality cannot be effectively reconciled, the debate on this goal continues. From the perspective of natural sciences, most studies are based on field experiments, and RFA plays an important role in increasing agricultural yield and improving the environmental situation. For example, reducing fertilizer use by 10% in the wheat production drought–drought model and by 24% in the dry–wet rotation mode can produce a certain “reduction and efficiency enhancement” effect. Farmers’ economic returns vary significantly under different irrigation and fertilization models for fruit trees [16]. Some studies point out that the yield improvement effect of RFA is not significant in the initial year, but yields increase over several years of adoption [17]. Additionally, RFA helps improve soil fertility, increase soil biodiversity, enhance soil health, ensure crop yields, improve agricultural product quality, control agricultural NPS pollution, and reduce greenhouse gas emissions [18]. A comprehensive evaluation of 85 green production technologies across 24 countries in Asia and Africa revealed that adopting green pest control technologies not only reduces the use of agricultural chemicals and increases crop yields but also improves soil quality and farm hygiene [19]. At the Research Farm of the Water Management Project, Mahatma Phule Agricultural University in Rahuri (Maharashtra), India, tests on the effect of drip irrigation with urea fertilizer showed 20% to 40% of nitrogen savings compared to two fertilizations with furrow irrigation [20]. An analysis of survey data from agricultural producers in Uganda found that adopting green pest control technologies increased the net coffee income by 118%, with the estimated rural income multiplier being 1.27 [21].

From an economic perspective, adopting green production technologies can reduce fertilizer use by 11.97–20.98% and increase agricultural operating income by 8.15–9.07%. RFA decreases farmers’ reliance on chemical inputs, thus lowering agricultural chemical input costs [22]. Moreover, RFA can effectively promote the standardization of crop production, enhance the safety of agricultural products, increase the market competitiveness of agricultural products, and achieve a premium for green agricultural products, thereby increasing farmers’ income [23,24]. The application of soil testing and formulated fertilization technologies can increase agricultural product yields and have considerable economic benefits [25]. Green inputs can drive agricultural economic growth, achieving the goal of increasing production without increasing pollution [26]. Based on a 30-year review of pesticide reduction and chemical control in sub-Saharan Africa, it is urgent to change smallholder farmers’ improper use of agricultural chemicals. Promoting green production technologies that reduce chemical use is a key path to improving farmers’ economic gains [27]. An empirical study of onion farmers in the Philippines showed that these technologies not only significantly increased crop yields but also reduced agricultural chemical use by 25% to 65% [28]. However, some scholars argue that although RFA can reduce inputs, the current inadequate green agricultural product certification system in China has led to insufficient premiums for green production, making it difficult for farmers to directly gain additional profits from RFA [29], which hinders farmers’ income growth goals.

The existing literature reveals several key points: First, many conclusions about the economic and environmental effects of RFA are based on natural science experiments. There is a relative lack of empirical research at the micro-farmer level using econometric models. Because the internal and external environments of natural science experiments are strictly controlled, it is difficult to effectively reveal the impact of individual farmer heterogeneity in real-world settings. Therefore, to address the aforementioned shortcomings, this study employs an endogenous switching regression model (ESRM), which quantify the impact of different farmer characteristics on the economic and environmental effects of RFA based on the perspective of micro-level rice farmers. This approach aims to better reflect farmers’ decision-making behavior in real-world environments. Second, existing studies have a narrow scope in evaluating the effects of green agricultural production behaviors. They typically focus on the impact on crop yield or fertilizer use alone, lacking a dual evaluation system for both economic and environmental effects. Therefore, after reanalyzing the economic effects generated by farmers’ RFA, this study further explores the potential environmental effects that may be realized. Third, research from a micro-farmer perspective that measures agricultural environmental effects mainly focuses on the actual amount or cost of fertilizers and pesticides, making it difficult to effectively identify the extent to which farmers’ behaviors impact the ecological environment.

Given the above, this study explores the potential economic and environmental impacts of RFA among rice farmers in Jiangxi Province, China. Using data from a 2022 survey of 1345 rice farmers across 30 counties, we apply an ESRM to quantify the dual impacts of RFA. Additionally, an undesirable output DEA-SBM model is employed to assess the environmental effects, providing a comprehensive evaluation that considers both economic and ecological outcomes. This study focuses on rice farmers in Jiangxi Province, a major rice-producing region in China. Although the findings are specific to Chinese rice production, the methodological approach and insights into green agricultural practices may have broader applications. However, regional differences in farming practices and policies may limit the generalizability of the results to other regions or crops. Future research should include case studies from different geographical areas or agricultural systems to validate and extend these findings.

This paper is divided into six section: after the introduction, the theoretical framework is described in the second section; the third sections details the data sources, variable selection, and model setup; the fourth sections presents the empirical analysis results; the fifth section provides conclusions and policy recommendations; and the sixth section discusses research limitations and prospects.

2. Theoretical Framework

2.1. Economic Effects of Farmers’ RFA

As rational economic agents, maximizing household income is the key intrinsic motivation for farmers to adopt agricultural input-reduction production practices. Therefore, the positive expectations of the economic benefits of RFA drive farmers to implement these behaviors. This paper begins by exploring the economic effects on rice farmers, analyzing the potential economic effects of RFA from three aspects: cost savings, yield increases, and product premiums.

Firstly, adopting green agricultural production methods can help farmers avoid over-reliance on chemical inputs [30]. By reducing the amount and frequency of fertilizer applications, the input cost of fertilizers can be lowered [31]. Additionally, compared to traditional fertilization methods, farmers who use soil formulators and precision fertilizer appliances can improve fertilization efficiency. This method not only reduces the amount of fertilizer used but also allows farmers to gain a price advantage in the market through bulk purchasing fertilizers, thereby reducing production costs [32]. However, it is essential to consider that while cost savings are evident, farmers may face initial expenses related to the purchase of new equipment and learning about new technologies when transitioning to green production methods, which could temporarily increase production costs [33,34]. Therefore, the short-term changes in costs may not decline as directly as expected, and this dynamic deserves further study.

Secondly, according to the law of diminishing marginal returns on factor inputs, the relationship between fertilizer input and marginal output in agricultural production shows an “inverted U-shaped” trend. When the amount of fertilizer input reaches its maximum limit, the marginal output decreases as the application amount increases. Thus, under the current agricultural production technology conditions or farming habits, blindly increasing fertilizer input does not significantly enhance yield. Moreover, studies have shown that compared to small-scale farmers, large-scale farmers use less fertilizer but achieve higher yields [35]. This indicates that scientific and appropriate fertilization methods can increase agricultural product yields. Based on this point, this study argues that RFA is not merely a cost-saving strategy but also an important means of optimizing the allocation of agricultural production resources. Through more scientific fertilization methods, farmers can reduce resource waste while optimizing yield [36,37]. This effect has short-term economic significance and could have profound implications for farmers’ long-term production efficiency and sustainable development capabilities.

Finally, RFA can promote the standardization of crop production, improve the quality and safety of agricultural products [38], effectively avoid excessive fertilizer residues, increase market competitiveness, and achieve a premium for green agricultural products, thereby increasing farmers’ income [25,39]. As the quality of agricultural products improves, farmers gain stronger market competitiveness and a price premium advantage in the trading market [40]. Thus, appropriate RFA provides a foundation for stable crop income, improves the commercial value of crops, and increases farmers’ economic returns [41]. It is evident that RFA is not just a choice of an agricultural production technique but also a market strategy. By enhancing the quality and safety of agricultural products, farmers can achieve greater price premiums in a market that increasingly values food safety. This premium not only boosts current farmer incomes but also potentially drives the development of agricultural branding, promoting a shift from traditional production to modern, high-value agriculture. Thus, the following research hypothesis is proposed:

H1. 

Farmers’ RFA significantly increases the net profit per hectare (NPH) in rice production.

2.2. Environmental Effects of Farmers’ Fertilizer Reduction Behavior

Considering agricultural economic growth, it is also necessary to address the environmental issues arising from agricultural production processes. In recent years, scholars have begun to focus on the positive effects of green production technologies on the ecological environment. They have gradually incorporated agricultural pollution (such as agricultural carbon emissions, pesticide residues, and excessive fertilizer use) as undesirable outputs of agricultural production, putting them into the framework for evaluating agricultural productivity [42,43]. By analyzing the relationship between factor inputs, economic outputs, and the ecological environment in agricultural production, the environmental effects of agriculture can be measured. Agricultural green productivity (AGP) is a key indicator of environmental effects, representing the ability to achieve maximum agricultural output with minimal resource consumption and environmental pollution under a certain combination of agricultural input factors [2,44]. In the context of increasingly severe global environmental challenges, the importance of green production technologies is becoming more prominent. Agricultural pollution can be integrated as an undesirable output into the analysis framework of agricultural production capacity, which represents a critical shift in modern agricultural research. This approach not only aids in comprehensively assessing the environmental impact of agricultural production but also provides policymakers with more scientific evidence, thereby promoting the sustainable development of agriculture.

The environmental effects of farmers’ RFA are specifically reflected in the following ways: RFA can reduce agricultural NPS pollution. Unreasonable fertilizer application methods are a major cause of agricultural NPS pollution, affecting water, soil, and air. In rice production, the excessive and inefficient use of chemicals leads to numerous fertilizer residue in the soil, disrupting the soil’s pH balance, causing soil acidification and compaction, and hindering sustainable agricultural development [45]. Excessive fertilizer use leads to phosphorus accumulation in the topsoil and excessive heavy metals, which, in eutrophic water environments, impede crop survival and severely impact the agricultural ecological environment. Soil respiration produces carbon dioxide, and over-fertilization increases the emissions of carbon dioxide, nitrogen oxides, and other gasses, thus leading to atmospheric pollution [46]. RFA alleviates these environmental pollution issues and improves agricultural environmental efficiency. It is clear that RFA, as a green production method, not only directly reduces fertilizer usage, thereby lowering the risk of agricultural NPS pollution, but more importantly, it embodies a sustainable agricultural management philosophy. This practice is a proactive response to current environmental issues and a crucial step in guiding agricultural production from high input and high pollution toward low input and low pollution, which is of significant importance for achieving a green transformation in agriculture. Thus, the following research hypothesis is proposed:

H2. 

Farmers’ RFA significantly enhances AGP.

In summary, this study aims to explore the economic and environmental effects of farmers’ RFA by examining two key aspects: enhancing farmers’ economic returns and reducing agricultural NPS pollution. The specific research framework is illustrated in Figure 1.

3. Data Source, Variable Selection, and Model Setting

3.1. Data Source

The sample data for this study were collected through a micro-survey conducted by the research team among rice farmers in Jiangxi Province, China, between November 2022 and August 2023. The selection of the sample and regional focus was guided by several key factors. First, China is a global leader in rice production and consumption, with its rice planting area and output accounting for 20% and 33% of the world’s total, respectively. Previous research has shown that nitrogen fertilizer use in rice production contributes approximately 50% of the total carbon emissions, largely due to improper application [11]. As a major agricultural province and a key rice-producing region, Jiangxi has long promoted agricultural supply-side structural reforms. These reforms have resulted in a sustained reduction in fertilizer use, achieving an 18.57% decrease in 2022 compared to 2016. Therefore, focusing on rice farmers in Jiangxi to assess the economic and environmental impacts of RFA is both representative and highly relevant.

To ensure the scientific rigor of our sampling process, we employed a stratified random sampling method. First, the research team categorized 100 counties (cities and districts) in Jiangxi Province into three groups—high, medium, and low—based on per capita GDP. From each group, 10 counties were randomly selected. In each selected county, two townships were randomly chosen, and within each township, two villages were randomly selected. Finally, 5 to 15 rice farming households were randomly chosen from each village. This multi-stage sampling approach ensured that our sample accurately represented rice farmers across different economic regions of Jiangxi Province, thereby increasing the statistical validity of our findings. The survey data were coded hierarchically, starting with the county, followed by the township, and then the village. This hierarchical coding enabled a systematic analysis across various geographic levels. The survey used a structured questionnaire consisting exclusively of closed-ended questions to collect quantitative data on the economic and environmental impacts of RFA. The complete questionnaire is available in the appendix for reference.

The sample size was determined based on the total number of rice farming households in Jiangxi Province. Using Cochran’s formula for sample size determination [47], a 95% confidence level and a 5% margin of error were applied. Given the heterogeneity of the population, an initial target of 1500 households was set to account for potential non-responses or invalid surveys. Of the 1500 questionnaires distributed, 1410 were returned, resulting in a 94% response rate. After excluding incomplete or invalid responses, 1345 valid samples were retained for analysis.

The survey methodology involved face-to-face interviews conducted by trained enumerators. Before the survey, all enumerators received extensive training to ensure consistency in data collection. The training covered both the technical aspects of the survey and the ethical considerations for fieldwork. A pre-test of the questionnaire was conducted to identify potential issues, and adjustments were made to improve clarity and reliability. During the survey, a supervisor was present in the field to monitor enumerator performance and ensure adherence to the protocol. Regular meetings were held with the survey team to address any challenges encountered, and real-time data validation was performed to detect inconsistencies or missing data. After data collection, a comprehensive data cleaning process was undertaken, including cross-checking responses and removing outliers or invalid entries. The final dataset underwent consistency checks and statistical validation to ensure accuracy and reliability. These quality control measures were implemented to ensure the high reliability of the data and to minimize bias. The selection of survey sites and sample households was guided by established methods from previous agricultural studies [48], ensuring both scientific rigor and practical relevance. The distribution of sample counties is illustrated in Figure 2, which was generated using ArcGIS 10.8.

3.2. Model Setting

3.2.1. Endogenous Switching Regression Model

Considering that whether rice farmers practice RFA is a non-random event, this can lead to sample selection bias. Additionally, non-randomized controlled trials cannot obtain information about the “counterfactual”, making it difficult to identify causal relationships. Without correction, biased estimates may result. Therefore, considering the probability of rice farmers adopting RFA, estimating the treatment effects of this behavior on economic and environmental effects becomes crucial. In view of this, we employ an ESRM to overcome sample selection bias caused by unobservable variables [49]. This model can fit the effect decision equations and “counterfactual” equations for both adopters and non-adopters of RFA, compensating for the shortcomings of previous research methods.

First, we establish the decision equation for rice farmers’ RFA as follows:

$$D_i^{\ast }=\gamma Z_{ij}+u_i,D_i=\left\{\begin{array}{c}1,D_i^{\ast }>0\\ 0,D_i^{\ast }\le 0\end{array}\right.$$

(1)

$D_i^{\ast }$ represents the unobservable variable for the decision to adopt RFA. $Z_{ij}$ represents the observable factors influencing the adoption of RFA. $\gamma$ represents the parameters to be estimated. $u_i$ represents the random disturbance term. The impact effect equation for adopting RFA is given by the following:

$$Y_{1i}=\sum _{j=1}^n{\beta }_{1j}{X}_{ij}+{\epsilon }_{1i};\hspace{0.33em}D=1$$

(2)

$$Y_{0i}=\sum _{j=1}^n{\beta }_{0j}{X}_{ij}+{\epsilon }_{0i};\hspace{0.33em}D=0$$

(3)

$Y_{1i}$ and $Y_{0i}$ represent the economic and environmental effects produced by rice farmers who have and have not adopted RFA, respectively. $X_{ij}$ is a vector of exogenous variables that may impact the outcomes employed. At the same time, the variables $Z_{ij}$ in selection Equation (1) and variables $X_{ij}$ in outcome Equations (2) and (3) are allowed to overlap. In our case, selection Equation (1) is estimated based on all explanatory variables specified in outcome Equations (2) and (3) plus one instrumental variable. The valid instrumental variable is required to influence the choice of adopting RFA but does not impact the outcomes. Therefore, the average adoption of RFA by other farmers in the village, excluding the farmer in question, is selected as the identifiable variable. It should be noted that due to the “herd effect,” farmers’ decisions to adopt RFA are influenced by other farmers, meeting the requirement of correlation. Meanwhile, other farmers’ decisions to adopt RFA do not directly affect the NPH and AGP of the farmer in question, meeting the requirement for exogeneity. ${\beta }_i$ is the parameter to be estimated; ${\epsilon }_i$ is the random disturbance term.

The conditional expectations of rice farmers under the two scenarios of whether they adopt RFA are as follows:

$$E(Y_{1i}\left|D_i=1)={\beta }_{1j}{X}_{ij}+{\sigma }_{{\mu }_{1}v}\right.{\lambda }_{1i}$$

(4)

$$E(Y_{0i}\left|D_i=0)={\beta }_{0j}{X}_{ij}+{\sigma }_{{\mu }_{0}v}\right.{\lambda }_{0i}$$

(5)

${\lambda }_{1i}$ and ${\lambda }_{0i}$ are unobservable variables contributing to selection bias. ${\sigma }_{{\mu }_0}$ and ${\sigma }_{{\mu }_1}$ represent the covariances between the error terms of the selection equation and the outcome equation. If these covariances are significant, this indicates the “selection bias” issue between the two sets of equations, ${\sigma }_{{\mu }_{1}v}=cov({\mu }_1,v)$ and ${\sigma }_{{\mu }_{0}v}=cov({\mu }_0,v)$, where $v$ is a random error term with an expected value of zero.

Next, we estimate the expected values of the economic and environmental effects in the two “counterfactual” scenarios: the effects for rice farmers who have adopted RFA when they do not adopt it, and the effects for rice farmers who have not adopted RFA when they adopt it.

$$E(Y_{0i}\left|D_i=1)={\beta }_{0j}{X}_{ij}+{\sigma }_{{\mu }_{0}v}\right.{\lambda }_{1i}$$

(6)

$$E(Y_{1i}\left|D_i=0)={\beta }_{1j}{X}_{ij}+{\sigma }_{{\mu }_{1}v}\right.{\lambda }_{0i}$$

(7)

The average treatment effect on the treated (ATT) for the economic and environmental effects of rice farmers who have adopted RFA (treatment group) is represented by the difference between Equations (4) and (6):

$$ATT=E(Y_{1i}\left|D_i=1)-E(Y_{0i}\left|D_i=1)=({\beta }_{1j}-{\beta }_{0j})X_{ij}+({\sigma }_{{\mu }_{1}v}-{\sigma }_{{\mu }_{0}v}){\lambda }_{1i}\right.\right.$$

(8)

The average treatment effect on the untreated (ATU) for the economic and environmental effects of rice farmers who have not adopted RFA (control group) is represented by the difference between Equations (5) and (7):

$$ATU=E(Y_{1i}\left|D_i=0)-E(Y_{0i}\left|D_i=0)=({\beta }_{1j}-{\beta }_{0j})X_{ij}+({\sigma }_{{\mu }_{1}v}-{\sigma }_{{\mu }_{0}v}){\lambda }_{0i}\right.\right.$$

(9)

3.2.2. Propensity Score Matching Method

The core idea of the Propensity Score Matching (PSM) [50] method is to treat farmers who engage in RFA as the treatment group and those who do not as the control group. Each farmer in the treatment group is then matched with a sample from the control group who has the same or similar propensity score, ensuring that they are distributed similarly across the characteristic variables. In other words, farmers in the control group who have the same or similar endowment characteristics as those in the treatment group are matched, and their agricultural production effects (economic effects/environmental effects) are compared. The treatment effects are calculated based on these matched samples, which can effectively reduce estimation bias caused by sample selection bias.

When using the PSM method to estimate the treatment effect of RFA on farmers’ production effects, the explanatory variable is “whether fertilizer use is reduced,” which is a binary discrete variable (“0–1”). Therefore, this study uses a binary Logit model to estimate the conditional probability of farmers entering the treatment group, known as the “propensity score”:

$$P\left(x_i\right)=Pr(D_i=1|x_i)$$

(10)

In this context, $P\left(x_i\right)$ represents the propensity score value, $D_i$ is a dummy variable, $D_i=1$ indicates that the farmer engages in RFA, and $D_i=0$ indicates that the farmer does not engage in RFA. $x_i$ represents a set of characteristic variables.

After matching samples based on the propensity score, the average treatment effect of the treated group (ATT) using RFA on agricultural production effects is calculated by comparing the differences in effects between the treatment group and the control group. The ATT is expressed as follows:

$$\begin{array}{cc}\hfill ATT& =E\left[y_1-y_0|D_i=1\right]\hfill \\ & =E\left\{E[y_1-y_0|D_i=1,P\left(x_i\right)]\right\}\hfill \\ & =E\left\{E\left[y_1\right|D_i=1,P\left(x_i\right)]-E[y_0|D_i=0,P(x_i\left)\right]|D=1\right\}\hfill \end{array}$$

(11)

$y_1$ represents the effect under the treatment group condition, and $y_0$ represents the effect under the control group condition. However, in reality, we can only observe $y_1$, while $y_0$ is unobservable. Therefore, the Propensity Score Matching method is used to construct a counterfactual framework for analysis.

3.3. Variable Selection

3.3.1. Dependent Variables

(1) Economic Effect. From a “cost–benefit” perspective, this paper measures the economic effect of rice planting using the average NPH for 2022. The calculation is based on the sales price of rice multiplied by the average yield per hectare minus the average cost per hectare. The main costs include land rent, labor costs, agricultural input (fertilizers, pesticides, seeds, and herbicides), machinery costs, and other costs (such as electricity).

(2) Environmental Effect. The AGP measurement model accounts for input–output slack variables, effectively addressing the bias caused by slackness and the radial and angular selection of inputs and outputs. It not only achieves the required efficiency but also provides improvement targets and degrees for input factors and the distance from undesirable outputs in decision-making units. Therefore, this paper uses the DEA—SBM model based on undesirable outputs to measure the AGP, which represents the agricultural environmental effect. The input indicators include land input (hectares), labor input (workdays), the cost of agricultural inputs (fertilizers, pesticides, seeds, and herbicides) (CNY), and the cost of machinery (CNY). The output indicators include desirable and undesirable outputs; the desirable output is presented as the total rice yield (in kilograms). The undesirable outputs include the emissions of nitrogen and phosphorus (in kilograms) during agricultural production. According to the nutrient balance approach [51], the emissions of nitrogen and phosphorus during rice production equal the pure content of nitrogen and phosphorus in fertilizers minus their content in rice grains5. If the calculated index is positive, it indicates a negative impact of agricultural production on the environment; if the index is negative, it indicates a positive impact. The specific input–output descriptions and statistics are shown in

Table 1. Input–output indicators for measuring agricultural environmental effects.

Typology	Indicator	Description of Indicators
Input indicators	Land input	Crops sown (hectares)
	Labor input	Including own-account and hired labor hours (hours)
	Seed input	Total cost of seeds purchased (CNY)
	Fertilizer input	Total cost of fertilizer purchased, including compost and fertilizer (CNY)
	Pesticides input	Total cost of pesticides purchased (CNY)
	Herbicides input	Total cost of herbicides purchased (CNY)
	Mechanical input	Total cost including owned and hired machinery (CNY)
Expected outputs	Output	Total crop production (kilogram)
Non-expected outputs	Nitrogen emissions	Nitrogen emissions from agricultural production (kilogram)
	Phosphorus emissions	Phosphorus emissions from agricultural production (kilogram)

.

3.3.2. Independent Variables

Currently, studies on RFA typically use the actual amount of fertilizer input for measurement [52]. Some studies identify the degree of fertilizer reduction by measuring fertilizer use efficiency [53]. With the continuous attention from the government and academia on the issue of agricultural NPS pollution, the methods for measuring fertilizer reduction have also been improved. Drawing on the measurement methods of fertilizer reduction effects presented in [54,55], this paper first calculates the optimal amount of fertilizer input in rice production by constructing a Cobb–Douglas production function. Then, by comparing the actual amount of fertilizer input by farmers with the optimal application amount, it determines whether the farmers have implemented RFA. Farmers whose actual fertilizer application amount is at or below the optimal level are assigned a value of 1, indicating that they implement RFA; otherwise, they are assigned a value of 0, indicating that they have not implemented RFA.

3.3.3. Control Variables

To reduce estimation bias caused by omitted variables, this research controls the following variables: ① individual characteristics, including age, education level, risk preference, training frequency, whether they are a village cadre, and work experience [56]; ② household characteristics, including the number of laborers and total household income [57]; and ③ agricultural production characteristics, including the land management scale, degree of land parcel consolidation, soil fertility, and distance of the land plots [58,59]. A descriptive statistical analysis of the variables is shown in

Table 2. Variable definitions and descriptive analyses.

Variable Name	Meaning and Assignment	Reduction		Non-Reduction		Discrepancy
		Average Value	Standard Deviation	Average Value	Standard Deviation
RFA	Fertilizer reduction application or not? 0 = no; 1 = yes	n = 1078		n = 267		-
Economic effects	Farmer’s net profit per hectare (NPH) from rice production in 2022 (CNY)	9870.65	191.70	7521.47	119.96	2349.26 ***
Environmental effects	The DEA—SBM model, based on non-expected output, measures AGP with values between 0 and 1	0.51	0.02	0.27	0.01	0.24 ***
Age	Actual age of farm householder (years)	51.54	0.49	55.01	0.28	3.47 ***
Education level	Educational attainment of farmers; 1 = primary school and below, 2 = junior high school, 3 = senior high school/vocation secondary school/technical school/vocational high school, 4 = post-secondary school, 5 = bachelor’s degree and above	1.87	0.05	1.87	0.03	0.01
Risk preference	1 = risk aversion, 2 = risk neutrality, 3 = risk appetite	1.85	0.05	1.50	0.02	0.35 ***
Training frequency	Frequency of farmers’ participation in agricultural technology training; 1 = never, 2 = occasionally, 3 = often	1.43	0.04	1.32	0.02	0.12 ***
Whether they are a village cadre	Is the farmer a village cadre? 0 = no; 1 = yes	0.27	0.03	0.29	0.01	0.02
Work experience	1 = always farming, 2 = part-time farming, 3 = labor or business, 4 = other	0.27	0.03	0.29	0.01	0.02
Total household income	Gross farm household income (ten thousand)	9.03	0.95	5.95	0.19	3.07 ***
The number of laborers	Total number of agricultural laborers in farming households (person)	1.83	0.04	1.84	0.02	0.01
Land management scale	Total actual operating scale of rice production (hectares)	5.16	6.18	6.57	7.84	1.41 ***
Degree of land parcel consolidation	1 = very scattered, 2 = more scattered, 3 = partially contiguous, 4 = all contiguous	3.11	0.05	2.43	0.03	0.69 ***
Soil fertility	Soil fertility status of largest plots; 1 = poor, 2 = medium, 3 = good, 4 = excellent	3.21	0.04	2.87	0.03	0.25 ***
Irrigation conditions	Irrigation conditions of largest plots; 1 = poor, 2 = medium, 3 = good, 4 = excellent	3.01	0.04	2.86	0.03	0.15 ***
Distance of the land plots	Distance of largest plot from farmer’s house (kilometer)	0.94	0.03	1.79	0.13	0.84 ***

Note: ***, **, and * indicate that the differences between the mean values of the variables for the fertilizer reduction group and the non-reduction group are statistically significant at the 1%, 5%, and 10% levels, respectively.

.

To compare the economic and environmental outcomes between the fertilizer reduction group and the non-reduction group, we conducted an independent sample t-test. This test evaluates whether there is a statistically significant difference between the means of two independent groups, making it appropriate for comparing the average outcomes of farmers who practice fertilizer reduction and those who do not. The results, presented in Table 2, indicate that the average NPH of the fertilizer reduction group (CNY 10,162.78) is significantly higher than that of the non-reduction group (CNY 7567.85). Similarly, the average AGP of the fertilizer reduction group (0.51) is also significantly higher than that of the non-reduction group (0.27).

4. Results and Discussion

4.1. Simultaneous Estimation Results and Analysis

This study utilizes the Stata 15.0 software to conduct the following analyses. To test the potential economic and environmental effects of farmers’ RFA, a simultaneous estimation was performed using the decision-making function of farmers’ RFA, along with the NPH and AGP function of rice. The test results are shown in

Table 3. Impact of farmers’ RFA on NPH of rice production.

Variable Name	RFA	Outcome Equation
	Selection Equation	NPH of Rice		AGP
		Reduction	Non-Reduction	Reduction	Non-Reduction
Age	−0.027 ***	−0.002	0.002	−0.004	−0.001
	(0.0067)	(0.003)	(0.003)	(0.002)	(0.001)
Education level	−0.039	0.013	0.029	0.015	0.004
	(0.064)	(0.034)	(0.027)	(0.025)	(0.005)
Risk preference	0.267 ***	0.043	0.001	0.029	−0.012 **
	(0.061)	(0.031)	(0.031)	(0.023)	(0.006)
Training frequency	0.083	−0.024	−0.002	0.083 ***	0.013 *
	(0.074)	(0.038)	(0.038)	(0.028)	(0.007)
Whether they are a village cadre	−0.184 *	0.016	0.063	−0.056	0.019 **
	(0.110)	(0.060)	(0.049)	(0.044)	(0.009)
Work experience	−0.048	0.003	0.110 ***	−0.007	−0.004
	(0.064)	(0.030)	(0.032)	(0.022)	(0.006)
Total household income	0.020 ***	−0.001	0.010 **	0.0002	0.001
	(0.007)	(0.002)	(0.004)	(0.002)	(0.001)
The number of laborers	0.033	0.024	−0.020	0.133 ***	0.019***
	(0.078)	(0.040)	(0.033)	(0.029)	(0.006)
Land management scale	0.009	0.005	−0.009 **	0.006 *	−0.000
	(0.009)	(0.005)	(0.004)	(0.003)	(0.001)
Degree of land parcel consolidation	0.494 ***	0.073 **	−0.029	0.011	0.001
	(0.057)	(0.035)	(0.023)	(0.026)	(0.005)
Soil fertility	0.104	0.085 **	0.029	−0.048	0.001
	(0.0662)	(0.040)	(0.027)	(0.030)	(0.005)
Irrigation conditions	0.0210	−0.042	0.001	−0.006	−0.001
	(0.069)	(0.039)	(0.029)	(0.029)	(0.006)
Distance of land plots	−0.357 ***	0.006	−0.004	0.011	0.001
	(0.063)	(0.047)	(0.005)	(0.034)	(0.001)
Identifying variables	3.843 ***
	(−0.316)
Constant term	−2.253 ***	8.617 ***	8.389 ***	0.287	0.267 ***
	(0.513)	(0.282)	(0.215)	(0.208)	(0.042)
ρ0	-	−0.239 ***	-	0.288 **	-
		(0.082)		(0.118)
ρ1	-		0.181 *	-	0.104
			(0.104)	-	(0.134)
Wald test	-	29.47 ***		29.66 **
LR test	-	chi2(2) = 9.29 Prob > chi2 = 0.009		chi2(2) = 6.32 Prob > chi2 = 0.042
Sample volume	1345	267	1078	267	1078

Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively, with robust standard errors in parentheses. Also, to attenuate the heteroskedasticity of the data in the model, the NPH of rice in the model was treated as logarithmic.

. The independence LR test for the two-stage equations reject the null hypothesis of mutual independence between the selection equation and the outcome equation at the 10% level. The Wald test rejects the null hypothesis that the behavior equation and the outcome equation are independent at the 5% significance level, indicating that the model has strong significance. Additionally, the error term correlation coefficient ρ1 was significant at the 5% statistical significance level, suggesting that unobservable factors influence farmers’ decisions on RFA, indicating the presence of selection bias, which validates the use of the ESRM [60]. Moreover, the estimated values of correlation coefficient ρ1 in both sets of simultaneous equations were significantly positive, indicating that farmers who adopted RFA had higher NPH and AGP than the average of the sample farmers, which is consistent with research hypotheses 1 and 2.

Regarding the selection equation, factors such as the farmer’s age, risk preference, whether they are a village cadre, total household income, degree of land parcel consolidation, and distance of land plots significantly influenced the farmer’s RFA behavior. In terms of the outcome equation, from the economic effect perspective, the land management scale had a significant positive impact on the NPH of rice production, indicating the presence of scale effects in rice production. The training frequency and the total number of laborers significantly positively impacted the NPH of farmers who have practiced RFA, suggesting that participation in training enhances farmers’ ability to acquire production and market information, thereby improving agricultural production benefits. Additionally, a higher number of agricultural laborers in a household where income mainly comes from agricultural production increases the likelihood of being exposed to and understanding green pest control technologies, resulting in greater economic effects. From the environmental effect perspective, the frequency of training participation and the number of agricultural laborers had a significant positive impact on the AGP of farmers, further highlighting the importance of agricultural training and labor input in agricultural production. The land management scale significantly positively impacted the AGP of farmers who have practiced RFA, indicating the presence of scale effects in green agricultural production. Soil fertility had a significant positive impact on the AGP of farmers who have practiced RFA. For land with higher soil fertility, farmers tend to apply less fertilizer, resulting in higher environmental effects.

4.2. An Analysis of the Economic Effects of Farmers’ RFA

To reflect the impact of farmers’ RFA on their NPH and AGP, the expected values of the NPH for non-reduction households and reduction households, adjusted for sample bias, were calculated using Equations (4) and (5). Additionally, the NPH and AGP values of the reduction group and non-reduction group under two “counterfactual” scenarios were calculated using Equations (6) and (7). Finally, the treatment effects of farmers’ RFA on the NPH, AGP, ATT, and ATU were calculated using Equations (8) and (9). The estimated average treatment effects of RFA on the NPH are shown in

Table 4. Mean treatment effect of farmers’ RFA on NPH.

Groups	Fertilizer Reduction	No Fertilizer Reduction	ATT	ATU
Fertilizer reduction	9.129	8.586	0.543 ***	-
No fertilizer reduction	8.964	8.744	-	0.220 ***

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

. The ATT value is 0.543, indicating that if farmers adopting fertilizer reduction technology did not adopt this technology, their NPH from rice production would decrease. The ATU value is 0.220, indicating that, for farmers who have not implemented fertilizer reduction, adopting this technology would increase their NPH from rice production.

To more clearly reflect the economic effects of farmers’ RFA, the probability density distribution graphs of the NPH for reduction and non-reduction households were described separately. As shown in Figure 3 (left), if farmers who reduced fertilizer use chose not to reduce it, the probability density distribution curve of their NPH would shift significantly to the left. In the “counterfactual” scenario in which reduction households do not reduce fertilizer, their NPH would significantly decrease by 5.94%. Conversely, the ATU in Figure 3 (right) indicates that in the “counterfactual” scenario in which non-reduction households choose to reduce fertilizer, the probability density distribution curve of their NPH would shift to the right, suggesting that if non-reduction farmers choose to practice RFA, their NPH would increase by 2.52%. Hence, sample farmers who practice fertilizer reduction behavior could increase their NPH by 2.52–5.94%. Based on the above analysis, it can be concluded that H1 is validated.

4.3. An Analysis of the Environmental Effects of Farmers’ RFA

Table 5. The mean treatment effects on the impact of farmers’ RFA on AGP.

Groups	Fertilizer Reduction	No fertilizer Reduction	ATT	ATU
Fertilizer reduction	0.508	0.285	0.223 ***	-
No fertilizer reductio	0.389	0.269	-	0.120 ***

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

presents the average treatment effects of farmers’ RFA on AGP. The results show that adopting RFA has a significant positive treatment effect on farmers’ AGP. The ATT value is 0.223, indicating that if farmers who adopted RFA did not adopt this technology, their AGP would decrease. The ATU value is 0.120, indicating that for farmers who have not implemented fertilizer reduction, adopting this behavior would increase their AGP.

To more clearly reflect the environmental effects of farmers’ RFA, the probability density distribution graphs of AGP for reduction and non-reduction households were described separately. As shown in Figure 4 (left), if farmers who reduced fertilizer use chose not to reduce it, the probability density distribution curve of their AGP would shift significantly to the left. In the “counterfactual” scenario where reduction households do not reduce fertilizer, their AGP would significantly decrease by 43.90%. Conversely, the ATU in Figure 4 (right) indicates that in the “counterfactual” scenario in which non-reduction households choose to reduce fertilizer, the probability density distribution curve of their AGP would shift significantly to the right, suggesting that if non-reduction farmers choose RFA, their AGP would increase by 44.61%. Therefore, the RFA of sample farmers could increase their AGP by 43.90–44.61%. Based on the above analysis, it can be concluded that H2 is validated.

4.4. Robustness Test

To ensure the reliability of the above conclusions, robustness tests were conducted by replacing the estimation methods. The OLS model, Tobit model, and Propensity Score Matching (PSM) method were used to examine the economic and environmental effects of farmers’ RFA. The specific test results are shown in

Table 6. Results of OLS and Tobit regressions on economic effects of farmers’ RFA.

Variable	Economic Effects (OLS)	Environmental Effects (Tobit)
RFA	0.340 ***	0.234 ***
	(0.038)	(0.014)
Control variable	control	control
Constant term	8.491 ***	0.224 ***
	(0.183)	(0.050)
R2/Pseudo R2	0.073	−1.022
Sample volume	1345	1345

Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively, with robust standard errors in parentheses.

. Table 6 presents the regression results using the OLS model and Tobit model. The results show that RFA has a significant positive impact on the NPH and AGP of rice farmers.

Table 7. PSM regression results of economic and environmental effects of farmers’ RFA.

Matching Method	Economic Effects (NPH)		Environmental Effects (AGP)
	ATT	T	ATT	T
Nearest neighbor matching	0.326 ***	6.82	0.246 ***	11.17
Kernel matching	0.322 ***	7.69	0.243 ***	11.40
Caliper matching	0.313 ***	6.77	0.246 ***	11.39
Average value	0.320	-	0.245	-

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

shows the impact results of farmers’ RFA on the NPH and AGP, estimated by nearest neighbor matching, kernel matching, and caliper matching. It can be found that the average treatment effect (ATT), estimated by the three matching methods, is significantly positive, indicating that farmers’ RFA can significantly improve their economic and environmental effects. The results of the OLS model, Tobit model, and the PSM method are consistent with the results of the ESRM, indicating that the model results are robust.

Table 7 presents the average results of the three matching methods’ ATT. It can be found that farmers’ RFA can significantly improve the economic and environmental effects of rice production. After excluding other factors, RFA can increase the NPH of rice farmers by an average of 0.32, representing an increase of 3.55%; RFA can increase the AGP of rice farmers by an average of 0.25, representing an increase of 48.33%. The results of the PSM test, whether it is the increase in net profit or the increase in AGP, are within the estimation range of the ESRM, further verifying the robustness of the research conclusions.

4.5. Analysis of Differences in Economic and Environmental Effects of RFA among Farmers of Different Management Scales

To further analyze the heterogeneous effects of farmers’ RFA on economic and environmental outcomes at different quantiles, this study employs quantile regression to examine the effects at the 25th, 50th, and 75th quantiles of economic and environmental outcomes. The previous research conclusions indicate that RFA has a significant positive impact on the NPH and AGP. Quantile regression provides a clearer understanding of the differential impacts of RFA on various levels of NPH and AGP. The test results are shown in

Table 8. Quantile regression of economic and environmental effects of farmers’ RFA.

Variable	Economic Effects (NPH)			Environmental Effects (AGP)
	0.25 Quartile	0.5 Quartile	0.75 Quartile	0.25 Quartile	0.5 Quartile	0.75 Quartile
RFA	0.395 ***	0.260 ***	0.173 ***	0.046 ***	0.187 ***	0.460 ***
	(0.047)	(0.038)	(0.034)	(0.140)	(0.055)	(0.028)
Control variable	control	control	control	control	control	control
Constant term	8.276 ***	8.787 ***	9.126 ***	0.144 ***	0.180 **	0.300 ***
	(0.254)	(0.202)	(0.156)	(0.038)	(0.081)	(0.060)
R2	0.062	0.043	0.038	0.048	0.048	0.228
Sample volume	1345	1345	1345	1345	1345	1345

Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively, with robust standard errors in parentheses.

. From an economic perspective, RFA has a significant positive impact across different levels of NPH. The impact is the highest at the lower net profit level (0.25 quantile) and decreases as the net profit level increases. This indicates that the economic effect of RFA diminishes as the profit from rice production increases, suggesting that for farmers with higher profits, the economic benefits of RFA have reached an optimal level. Thus, RFA should not be a continuous practice, but should aim to achieve the optimal application amount to maximize economic benefits. From an environmental perspective, RFA has a significant positive impact across different levels of AGP, with the impact increasing significantly as green productivity improves. This suggests that farmers with relatively low green productivity have a stronger inertia towards high fertilizer application, and the path dependency of agricultural production limits the environmental benefits of RFA. Therefore, for farmers with relatively low AGP, improving environmental outcomes will also rely on additional collaborative measures beyond fertilizer reduction.

The above analysis is based on regression results at specific quantiles. This study further describes the marginal contributions and trends in RFA on economic and environmental outcomes across all quantiles (as shown in Figure 5). The results in Figure 4 indicate that the coefficients of the impact of RFA on the NPH and AGP are generally greater than 0, confirming that RFA enhances both economic and environmental outcomes. These findings are consistent with the main conclusions presented earlier.

4.6. Further Discussion

The neighborhood effect refers to how the behaviors and characteristics of neighbors influence an individual’s socio-economic actions [61]. According to the social cognitive theory, external environments shape individuals’ cognition, intentions, and behaviors [62,63]. In China’s rural areas, which are built on close-knit social networks, farmers’ behaviors within a village tend to influence one another [64]. Farmers adjust their own behaviors by observing and learning from their neighbors, and they also adopt new technologies from highly regarded neighbors in their community [65]. In the context of green production practices, a stronger neighborhood effect implies more frequent interaction and learning among farmers. Increased communication can raise farmers’ environmental awareness, making them more likely to adopt fertilizer reduction practices. Therefore, this paper proposes further discussion to analyze the moderating role of neighborhood influence on the relationship between farmers’ fertilizer reduction behaviors and their economic and environmental impacts. In Qiao and Li’s study [64], the questionnaire item “Can you obtain useful information from communicating with other farmers?” was used to measure the neighborhood effect, with the responses being assigned values on a Likert scale as follows: strongly disagree = 1, disagree = 2, neutral = 3, agree = 4, and strongly agree = 5. The specific test results are presented in

Table 9. Moderating effect of neighborhood effect.

Variable	Economic Effects (OLS)	Environmental Effects (Tobit)
RFA	0.845 ***	0.468 ***
	(0.285)	(0.123)
Neighborhood effect	0.040 **	0.003
	(0.018)	(0.004)
RFA * neighborhood effect	−0.119 *	−0.050 *
	(0.061)	(0.026)
Control variable	control	control
Constant term	8.364 **	0.209 ***
	(0.186)	(0.052)
R2/Pseudo R2	0.078	−1.032
Sample volume	1345	1345

Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively, with robust standard errors in parentheses.

. The results indicate that the interaction term between RFA and the neighborhood effect has a significant negative impact on both the economic and environmental outcomes for farmers. This suggests a substitution relationship between fertilizer reduction behavior and neighborhood communication. One possible explanation is that farmers may require time to adapt to new technologies and realize their benefits. If they observe that their neighbors, who have adopted similar practices, have not yet experienced notable economic or environmental gains, they might delay their own adoption, leading to short-term negative impacts [66,67]. Additionally, farmers’ fertilization decisions may be influenced by their neighbors’ behaviors, particularly when there is considerable uncertainty regarding fertilizer reduction technologies. Research indicates that farmers’ perceptions of risk significantly influence their willingness to adopt green production technologies, which may ultimately lead to a decline in overall benefits. In certain contexts, the neighborhood effect can exacerbate this uncertainty, resulting in negative impacts on both economic and environmental outcomes [68]. Additionally, when farmers in neighboring plots implement similar fertilizer reduction measures, competitive effects may emerge. For instance, if neighboring farmers collectively reduce fertilizer usage, it could lead to an oversupply of agricultural products in the market, thereby lowering prices and negatively affecting individual farmers’ economic gains. Such detrimental competitive effects could inhibit economic benefits in some cases. Therefore, future research should further delineate the sources of the neighborhood effect and differentiate between various types of neighborhood influences to more accurately explain the heterogeneous outcomes associated with fertilizer reduction behavior.

5. Conclusions and Policy Recommendations

This study, based on a sample of 1345 rice farmers in Jiangxi Province, examines the economic and environmental effects of farmers’ RFA. The findings indicate that RFA has significant positive impacts on both economic and environmental outcomes. Further analysis revealed a substitution effect between farmers’ fertilizer reduction practices and the neighborhood effect on these outcomes.

From an economic perspective, RFA has a notably positive impact on the NPH, contributing to increased farmer income. This result is consistent with existing research, such as studies of apple farmers in Shanxi and Gansu, which demonstrated that fertilizer reduction technologies boost income by optimizing input use and improving crop quality [69]. Similarly, other studies have found that adopting green production technologies stabilizes yields and reduces input costs, thus enhancing farmers’ economic returns [70]. However, our findings also suggest a diminishing economic effect of RFA as the NPH increases, which aligns with the principle of diminishing marginal returns observed in agricultural economics [71].

In terms of economic impact, RFA significantly enhances AGP, thereby improving the agricultural environment. This is consistent with studies that highlight the environmental benefits of fertilizer reduction, including improved soil health, reduced greenhouse gas emissions, and decreased water pollution [67,72]. Research on sustainable agricultural practices suggests that these technologies not only mitigate environmental harm but also promote the long-term sustainability of farming systems by increasing fertilizer efficiency and maintaining productivity [67].

Regarding heterogeneous effects, the variation in income-enhancing and environmental-improving outcomes across different levels of NPH and AGP observed in this study is consistent with studies in the literature that emphasize the role of farm size and resource availability in shaping the effects of fertilizer reduction practices [71]. Larger farms, which typically have greater capital and access to advanced technology, tend to experience more substantial environmental benefits, while smaller farms may achieve more immediate economic gains. This suggests that policy interventions promoting RFA should be tailored to account for these differences in order to maximize both economic and environmental outcomes.

Based on the above research conclusions, the following policy recommendations are proposed to promote green production behaviors among farmers and effectively achieve fertilizer reduction: Firstly, RFA should be encouraged among low-profit farmers, and low-profit farmers should be encouraged and guided to adopt integrated green agricultural production technologies to stimulate income potential. The previous analysis confirmed that farmers’ RFA behavior can increase their income and improve their private profit objectives, primarily through cost savings. However, to achieve more significant income growth, it is necessary to accelerate the development of the green agricultural product market, enabling premium pricing for high-quality products. This, in turn, will lead to more pronounced income effects. Additionally, focus should be placed on improving the quality supervision system for green agricultural products within the market circulation mechanism. The certification of agricultural product quality should be promoted based on the quality evaluation results from monitoring departments, and differentiated pricing should be implemented for agricultural products of different quality levels. The development of green and organic agricultural products should be increased. This should be based on promoting green and organic agricultural products at a high quality and good prices. Secondly, the empirical analysis results indicate that farmers’ land management scale positively impacts their RFA behavior, which should create favorable land-scale conditions for fertilizer reduction technology promotion. The research also indicates that expanding farmers’ operational scale is crucial for achieving RFA [73,74]. In the context of the “three rights separation” reform in contracted land, actively promoting property rights system reform and further activating the rural land management rights market will accelerate land transfer and expand farmers’ operational scale. This will facilitate RFA through scale economies. Third, considering the significant economic effects of farmers’ RFA, policies should implement an externality internalization compensation system: RFA improves the environment and enhances social benefits, thus having the characteristics of public goods. Corresponding compensation policies should be developed for behaviors with positive externalities to ensure the effective implementation of RFA. Governments should provide technical, financial, and other compensations to farmers who actively adopt fertilizer reduction technologies.

6. Research Limitations and Prospects

This paper deepens the analysis in three key aspects compared to existing studies. First, there is a methodological distinction: while most studies on the economic and environmental effects of RFA are based on natural science experiments, few employ econometric models for empirical research at the micro-level of individual farmers. Natural science experiments often strictly control environmental factors, limiting their ability to capture the heterogeneity of individual farmers in real-world settings. By using an ESRM at the micro-level of rice farmers, this study quantifies the impact of various farmer characteristics on the economic and environmental effects of RFA, providing a more accurate reflection of decision making in real-world environments. Secondly, existing research typically focuses on a single aspect, such as crop yield or fertilizer use, without providing a dual evaluation of both the economic and environmental outcomes of green production practices. This study not only examines the economic effects of RFA but also explores the potential environmental impacts, creating a more comprehensive evaluation framework. Thirdly, this study improves the measurement of environmental effects. Current studies often assess agricultural environmental effects by examining fertilizer and pesticide usage or costs, making it difficult to accurately capture the broader ecological impacts of farmers’ behaviors. By constructing an undesirable output DEA-SBM model, this study provides a more precise measurement of the environmental effects of RFA, capturing the potential environmental consequences more effectively. However, this study has the following limitations: On the one hand, the sample scope is limited, and whether the research conclusions based on Jiangxi Province in China apply to other rice-producing regions remains to be verified. In the future, the survey scope should be expanded to enhance the generalizability of the conclusions. On the other hand, this study uses cross-sectional data, which need to account for the dynamic effects of RFA on economic and environmental outcomes over time. Future research should include follow-up surveys to form a long panel dataset, allowing for the multi-stage characteristics of farmers’ RFA to be explored.