Impact of Farm Size on Farmers’ Recycling of Pesticide Packaging Waste: Evidence from Rural China

Abstract

Scale management has become an essential form of modern agricultural production. However, it is still unclear how farm size influences farmers’ pesticide packaging waste recycling behavior (FPPWRB). Based on the data from the China Rural Revitalization Survey 2020, this study quantitatively explores the impact of farm size on FPPWRB. This study found that (1) the ratio for FPPWRB is low, with only about 41.7% of the sample farmers expressing participation in recycling. (2) The empirical results show that for every 1% increase in farm size, the probability of FPPWRB increases by 3.59%. (3) Farmers in urban suburbs or younger farmers are more inclined to FPPWRB. (4) Farm size can improve FPPWRB by enhancing farmers’ environmental cognition levels. The research in this study provides insights into the improper disposal of pesticide packaging waste and offers references for formulating policies related to the resource utilization of pesticide packaging waste. Thus, the findings of this study can help provide a reference for the introduction of policies to manage pesticide packaging waste, which, in turn, can help promote sustainable agricultural development.

1. Introduction

Pesticides play an important role in achieving the global goal of “zero hunger”. However, they also pose environmental threats that must not be ignored. About 3.5 million tons of pesticides are applied globally each year [1]. Asia has the highest pesticide usage globally, accounting for 38%, while Europe has the lowest usage, accounting for only 13% [2]. Between 1990 and 2021, pesticide usage in Oceania saw the fastest growth, increasing by 206%, while the growth rates in the Americas, Africa, Asia, and Europe were 191%, 175%, 67%, and 1%, respectively [3]. Oceania applies a relatively low level of pesticides per hectare of farmland, which totals 1.66 kg/hectare. Still, when standardized by the agricultural production value (0.83 kg/USD1000), the pesticide application per capita is higher (1.30 kg/person) [4]. In contrast, the Americas have reached higher levels in all three indicators, with 3.01 kg/hectare, 1.49 kg/USD1000, and 1.23 kg/person, respectively [5]. The massive generation of pesticide packaging waste (PPW) accompanies the abuse of pesticides. Approximately 10 billion pieces of this waste are generated annually worldwide. However, there is limited research on the subsequent disposal of PPW and its potential threat to sustainable development [6].

The improper disposal of PPW hinders the comprehensive green transformation of agriculture and rural areas. Improper handling and disposal of this waste may bring hazards. Firstly, the presence of residual pesticides means improper disposal can exacerbate the harmful effects of pesticide overuse [4,7]. Li et al. [8] found that the misuse of pesticides can lead to pesticide resistance in pests, creating a vicious cycle that threatens biodiversity; Zhao et al. [9] pointed out that 20 types of pesticides have been detected in the serum, urine, and cerebrospinal fluid of urban populations in China. Second, the haphazard discarding of plastic pesticide packaging greatly harms soil and water. Studies show that white plastic pollution and microplastic contamination are significant driving factors of soil compaction and water body damage [10,11]. Third, the burning of PPW pollutes the air. Open burning releases large amounts of carcinogens, such as dioxins, polycyclic aromatic hydrocarbons, and particulate matter, into the atmosphere, causing severe environmental pollution that threatens human health [5]. Therefore, promoting the recycling of PPW is the key to achieving sustainable development in agriculture and rural areas.

Differences in farm size are considered a key factor in understanding the differences in farm management performance [12]. According to economies of scale, expanding the scale of operations within a specific period makes better use of production facilities and resources, thereby reducing average costs and increasing operating profits. In farm management, studies have found that large-scale farms have higher production efficiency compared to small-scale farms [13]. At the same time, studies have shown that large-scale farms have lower transaction costs compared to small-scale farms [14]. Existing studies have verified that farm size positively affects agricultural production. However, how it does so, i.e., the quantitative relationship between farm size and how farmers recycle PPW, is still unclear.

According to statistics from the Ministry of Agriculture and Rural Affairs of China, approximately 3.5 billion pieces of PPW are generated annually in China. Therefore, this study aims to identify the causal relationship between farm size and farmers’ recycling of PPW. Based on existing research, the marginal contribution of this study is as follows: (1) As the world’s largest developing country, sustainable agriculture is the foundation of China’s sustainable development [15]. This study is an empirical analysis based on rural big data in China, using the 2020 China Rural Revitalization Comprehensive Survey (CRRS 2020) data as the research sample. It aims to provide empirical evidence for sustainable farm management in other developing countries. (2) In discussing the relationship between farm size and farmers’ pro-environmental behaviors, existing studies have downplayed or neglected endogeneity, thus failing to pinpoint the causal relationship between farm size and farmers’ pro-environmental behaviors. This study introduces the average farm size of other farmers in the same village as an instrumental variable to address the endogeneity issue. It then examines the causal relationship between farm size and FPPWRB. (3) This study further examines the mechanism through which farm size influences FPPWRB, providing a theoretical foundation and empirical support for the design of targeted policies.

The scientific management of PPW plays a crucial role in ensuring food security in China. China needs to feed over 1.4 billion people annually, and ensuring food security requires the rational use of pesticides [15]. According to statistics, China uses over 1 million tons of pesticides annually, resulting in the generation of approximately 3.5 billion PPW items [16]. Previous research has primarily focused on how to scientifically and rationally use pesticides. For example, Qiao et al. [17], Liu and Huang [18], and Gong et al. [19] discussed the pesticide usage behavior of Chinese farmers and its influencing factors. However, there is limited research focusing on the FPPWRB. Therefore, the marginal contribution of this paper lies in the following points: (1) As the world’s largest developing country, sustainable agriculture is the foundation of China’s sustainable development [15]. Therefore, the scientific management of PPW in China is crucial for both China’s and global sustainable development. Based on the large-scale survey data of Chinese farmers, discussing their FPPWRB helps to clarify the micro-mechanisms underlying recycling behavior. (2) Scaled operations are key to ensuring agricultural sustainability. This paper aims to examine the quantitative impact and mechanism between scaled operations and FPPWRB using econometric models, providing valuable insights into the development of policies on managing PPW. (3) In discussing the relationship between farm size and farmers’ pro-environmental behaviors, existing studies have downplayed or neglected endogeneity, thus failing to pinpoint the causal relationship between farm size and farmers’ pro-environmental behaviors. This study introduces the average farm size of other farmers in the same village as an instrumental variable to address the endogeneity issue.

The structure of the remaining sections of this paper is as follows: (1) Section 2 discusses the theoretical relationship between farm size and FPPWRB. (2) Section 3 discusses the data and research methods used in this study. (3) Section 4 examines the causal relationship between farm size and FPPWRB. (4) Section 5 summarizes the findings of this study and provides policy recommendations.

2. Theoretical Analysis and Research Hypotheses

2.1. The Impact of Farm Size on FPPWRB

The core concept of the economies of scale theory is that as the scale of production increases, the per-unit production cost decreases. Large-scale production can spread fixed costs and improve production efficiency. Subsequently, Sharma et al. [20] pointed out that the larger the production scale alongside the amount of prepaid capital, the more efficiently the entire production system operates. He et al. [21] and Wang et al. [22] found that land scale has a positive impact on straw return. Based on data from South Africa [23], Europe [24], and Uruguay [25], studies in the field of manufacturing indicate that the size of a firm is conducive to technological innovation.

Firm size promotes the adoption of technology [14,15]. A large firm can reduce the risks of market development and financing and is a manifestation of increased risk tolerance for technological innovation and the adoption of new technologies. The same is true in the agricultural sector, where small farms face more risks when adopting new technologies. For example, Li et al. [4] studied the relationship between agricultural technological change and farm size, finding that large-scale farms are more likely to adopt new technologies because they are more able to bear the risks and costs associated with technology adoption. Similarly, Deng et al. [26], Zheng et al. [27], and Yu et al. [28] found that large-scale farms are more proactive in adopting labor-saving mechanized technologies.

Hypothesis 1. 

Farm size positively influences farmers’ recycling of pesticide packaging waste.

2.2. Farm Size, Hazard Awareness, and FPPWRB

Hazard awareness, which refers to people’s awareness of the possible negative impact of a certain behavior or substance, significantly influences the recycling of PPW. Studies indicate that the greater the awareness of potential harm, the more likely individuals are to take preventive actions [20]. In the field of environmental protection, hazard awareness is widely regarded as one of the key factors promoting environmentally responsible behavior [29]. Deng et al. [30] believe that the public’s awareness of environmental pollution hazards directly influences their willingness to engage in environmental protection behaviors. Similarly, Sikor et al. [5] found that there is a significant correlation between farmers’ awareness of pesticide pollution hazards and their waste disposal behavior. In addition, Song and Ye [31] found that when farmers have a deeper understanding of the harmful effects of pesticide waste, they are more likely to adopt recycling and reuse measures. Therefore, hazard awareness is important in promoting the recycling and reuse of PPW. Farmers who recognize the harm generated by pesticide waste are more likely to adopt recycling measures to reduce environmental pollution.

Farm size has a positive impact on the awareness of the harmful effects of PPW. A large farm size often comes with more resources and better educational conditions, allowing farm owners to access more knowledge and information about the hazards of pesticides. The theory of economies of scale suggests that large-scale operations can improve the efficiency of information acquisition and technology learning [32]. Wang et al. [33] pointed out that large-scale farm owners are more likely to participate in agricultural training and environmental education programs, thereby increasing their awareness of the hazards of pesticides. Yan et al. [34] found that large-scale farms have significant advantages in information dissemination and technology application, making large-scale farmers more likely to understand and recognize the potential hazards of pesticides. Additionally, Zhang et al. [35] emphasized that large-scale farm owners, due to their operational scale, are more motivated and capable of acquiring knowledge related to environmental protection. Therefore, large-scale farm owners, with their resources and educational advantages, are better able to recognize the hazards of PPW, making them more sensitive and proactive in hazard awareness compared to small-scale farm owners.

Hypothesis 2. 

Farm size positively influences the awareness of pesticide packaging hazards, further promoting the recycling and reuse of pesticide packaging waste.

3. Data, Variables, and Model

3.1. Data

This study employs data from the China Rural Revitalization Survey, which was conducted by the Rural Development Institute, Chinese Academy of Social Sciences. According to the introduction on its official website, the survey has the following characteristics: (1) The survey covers a wide range of topics, including rural population and labor force, rural industrial structure, rural governance, and comprehensive rural reforms. (2) The survey covers a large geographic area. It includes 50 county-level units from both economically developed and underdeveloped provinces in China, with data collected from 300 village surveys and over 3800 farmer surveys. (3) The sample is representative. The sampling procedure is as follows: (a) sample provinces are randomly selected based on their economic development level, regional location, and agricultural development, covering eastern, central, western, and northeastern regions; (b) sample counties are randomly selected within each province based on per capita GDP at the county level using an equal interval random sampling method; (c) sample townships (or towns) and villages are randomly selected based on economic development levels; and (d) sample households from the household register are randomly selected provided by the village committees.

After excluding questionnaires with significant missing data and responses that deviated from reality, this study used a total of 2663 valid farmer surveys.

3.2. Variables

3.2.1. Dependent Variable

The dependent variable in this study is “FPPWRB”: farmers’ pesticide packaging waste recycling behavior. In the CRRS (China Rural Revitalization Survey), farmers were asked about their methods for handling pesticide packaging waste. In this study, households that answered “recycle to a designated point” or “recycle to the agricultural supply market” are defined as households with FPPWRB (coded as 1); conversely, households that did not report recycling are defined as households without FPPWRB (coded as 0).

3.2.2. Key Variable

The key variable in this study is farm size. Most studies use the actual total cultivated area to represent the scale of land management [36]. Therefore, this study uses the land area managed by the farm as the measure of farm size.

3.2.3. Mediator Variables

The main mediator variable in this study is the awareness of the hazards of PPW. Zhang et al. [36] found that larger farms are more likely to adopt environmental protection measures due to their stronger resource integration capabilities and greater environmental awareness, providing theoretical support for using the awareness of the hazards of PPW as a mediator variable.

3.2.4. Control Variables

This study introduces control variables in the empirical model to mitigate the impact of omitted variables on the estimation results. Drawing on the studies of Huang and Elahi [37], Wang et al. [38], and Chen et al. [39], the research includes the characteristics of the household’s head (e.g., gender, age, marital status), household characteristics (e.g., level of organization, per capita income, number of household laborers), and community characteristics (e.g., village topography and distance from the village committee to the township government) as control variables. The variable definitions and descriptive statistics for this study are shown in

Table 1. Descriptive statistical analysis.

Variables	Definition	Mean	S.D.
FPPWRB	Whether pesticide packaging waste is recycled: 1 = Yes; 0 = No	0.417	0.493
farm size	Farm area under farm management (mu)	25.586	80.887
gender	Gender: 1 = Male; 0 = Female	0.944	0.229
age	Farmer’s age (years)	54.548	10.750
marriage	Marital status: 1 = Married; 0 = Unmarried	0.930	0.256
education	Whether the farmer received education at the high school level or above: 1 = Yes; 0 = No	0.145	0.352
party	Whether the farmer is a member of the Communist Party of China: 1 = Yes; 0 = No	0.221	0.415
nonfarm work	Whether the farmer engages in nonfarm work: 1 = Yes; 0 = No	0.093	0.291
leader	Whether the farmer is a village official: 1 = Yes; 0 = No	0.020	0.138
farm cooperation	Whether the farmer joined an agricultural cooperative: 1 = Yes; 0 = No	0.235	0.424
income	Per capita income of the household	9.361	1.103
labor	Total number of laborers in the household	2.919	1.284
machine	Whether mechanized services are used for pesticide application: 1 = Yes; 0 = No	0.201	0.401
outsource	Whether the household outsources pesticide spraying services: 1 = Yes; 0 = No	0.069	0.254
Terrain 1	Whether the village is located in a plain: 1 = Yes; 0 = No	0.465	0.499
Terrain 2	Whether the village is located in a hilly area: 1 = Yes; 0 = No	0.212	0.409
recovery point	Whether the village has a pesticide packaging waste recycling point: 1 = Yes; 0 = No	0.395	0.489
distance	Distance from the village committee to the county government	23.314	16.241

.

3.3. Methods

3.3.1. Baseline Model

To examine the impact of farm size on FPPWRB, a probit regression model was constructed as follows:

$${pesticide\_packaging}_i={\alpha }_0+{\beta }_1{Lnland\_area}_i+{\beta }_2{Controls}_i+{\gamma }_i+{\epsilon }_i$$

(1)

In Equation (1), i represents the household; $pesticide\_packaging$ denotes the binary dummy variable for the household’s FPPWRB; $land\_area$ represents the land area operated by the household; Controls refers to a set of control variables; ${\gamma }_i$ indicates provincial fixed effects; and ${\epsilon }_i$ is the random disturbance term.

3.3.2. Mediation Effect Model

By using the awareness of pesticide packaging hazards as a mediating variable, a mediation effect model was constructed to explore the transmission mechanism of the land management scale’s influence on FPPWRB. The specific equations are given as follows:

$${cognition}_i={\alpha }_0+{\beta }_1{lnland\_area}_i+{\beta }_2{Controls}_i+{\gamma }_i+{\epsilon }_i$$

(2)

$${pesticide\_packaging}_i={\alpha }_0+{{\beta }_{1}cognition}_i{+\beta }_2{lnland\_area}_i+{\beta }_3{Controls}_i+{\gamma }_i+{\epsilon }_i$$

(3)

In Equations (2) and (3), $cognition$ represents the awareness of the hazards of pesticide packaging waste. Since $cognition$ is a type of ordinal data, Equation (2) is specified as an ordered probit model.

3.3.3. Multicollinearity Test Method

To eliminate the influence of multicollinearity on empirical results, this study employed a variance inflation factor (VIF) test to assess whether the model suffers from multicollinearity. Following Chen [40], we constructed Equation (4) to calculate the variance inflation factor (VIF).

$$\mathrm{V}\mathrm{I}\mathrm{F}={\displaystyle \frac{1}{1-{\mathrm{R}}^2}}$$

(4)

4. Results

4.1. Results of Multicollinearity Test

Table 2. Multicollinearity—VIF test.

Variable	VIF	1/VIF
Terrain 1	1.83	0.543755
machine	1.65	0.603058
Terrain 2	1.54	0.647147
outsource	1.46	0.684195
Ln (farm size)	1.41	0.702131
distance	1.22	0.793182
age	1.18	0.844405
labor	1.14	0.873237
nonfarm work	1.14	0.876215
recovery point	1.12	0.885026
party	1.12	0.889887
marriage	1.09	0.895756
education	1.09	0.917843
Ln(income)	1.08	0.918074
leader	1.05	0.923301
gender	1.04	0.950771
farm cooperation	1.04	0.960651
mean VIF	1.25

presents the test results. The VIF values of all variables, as well as the model’s mean VIF, are less than two, indicating that the empirical estimates are not significantly affected by multicollinearity [41]. Additionally, the correlation coefficients between the core explanatory variables and control variables are all below 0.3, which preliminarily rules out multicollinearity among the independent variables and facilitates more accurate results in the subsequent regression analysis. However, it is worth noting that correlation analysis only measures the pairwise relationships between variables without accounting for the influence of other factors. Therefore, more accurate estimation results should be further addressed and adjusted in the subsequent multivariate regression analysis.

4.2. Estimated Results for the Impact of Farm Size on FPPWRB

We used a stepwise approach to add variables to mitigate the impact of omitted variables on the empirical results and examined the relationship between farm size and FPPWRB.

Table 3. Regression results–probit model.

	(1)	(2)	(3)	(4)	(5)
Ln (farm size)	0.1227 ***	0.1037 ***	0.0928 ***	0.1029 ***	0.0359 ***
	(5.21)	(4.24)	(3.64)	(3.93)	(3.96)
gender		0.0735	0.0789	0.0807	0.0282
		(0.65)	(0.69)	(0.70)	(0.70)
age		−0.0054 **	−0.0048 *	−0.0057 **	−0.0020 **
		(−2.12)	(−1.82)	(−2.15)	(−2.16)
marriage		0.0486	0.0276	0.0201	0.0070
		(0.48)	(0.27)	(0.19)	(0.19)
education		−0.0443	−0.0464	−0.0710	−0.0247
		(−0.59)	(−0.61)	(−0.93)	(−0.93)
party		0.0117	0.0131	0.0034	0.0012
		(0.18)	(0.20)	(0.05)	(0.05)
nonfarm work		−0.2593 ***	−0.2631 ***	−0.2620 ***	−0.0913 ***
		(−2.71)	(−2.74)	(−2.71)	(−2.72)
leader		−0.1523	−0.1554	−0.0627	−0.0219
		(−0.79)	(−0.81)	(−0.32)	(−0.32)
farm cooperation			−0.0193	0.0007	0.0002
			(−0.31)	(0.01)	(0.01)
Ln (income)			0.0131	0.0059	0.0021
			(0.54)	(0.24)	(0.24)
labor			0.0198	0.0193	0.0067
			(0.91)	(0.88)	(0.88)
machine			0.0931	0.0620	0.0216
			(1.04)	(0.69)	(0.69)
outsource			0.0215	0.0397	0.0139
			(0.18)	(0.33)	(0.33)
Terrain 1				0.2601 ***	0.0907 ***
				(3.40)	(3.42)
Terrain 2				0.3418 ***	0.1192 ***
				(4.22)	(4.26)
recovery point				0.2545 ***	0.0888 ***
				(4.25)	(4.29)
distance				−0.0036 **	−0.0013 **
				(−1.98)	(−1.98)
_cons	−0.5929 ***	−0.3455	−0.5650 *	−0.8621 **
	(−4.89)	(−1.51)	(−1.70)	(−2.51)
prov effect	Yes	Yes	Yes	Yes	Yes
N	2663	2663	2663	2663	2663
Log likelihood	−1666.8552	−1661.1485	−1659.7635	−1628.6105	−1628.6105
χ2	284.62 ***	296.04 ***	298.81 ***	361.11 ***	361.11 ***

Note: *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. The T-values are in parentheses. The fifth column reports the estimated marginal effects.

reports the estimation results.

Model (1) incorporates farm size Ln (farm size) as the only core explanatory variable for FPPWRB. Model (2) adds a series of farm-owner characteristics—sex, age, marriage, education, party affiliation, nonfarm work, and leader—as control variables on top of Model (1). Model (3) incorporates a set of farm characteristics—farm cooperation, Ln(income), labor, machine, and outsourcing—as control variables. Model (4) continues by introducing Terrain 1, Terrain 2, the recovery point, and distance, estimating a nonlinear probit model. Since the coefficients of Model (4) cannot directly explain the quantitative relationship between two variables, marginal effects were estimated; the results are presented in the fifth column of Table 3. In terms of model selection, the χ² results support the validity of using a probit model, making the probit estimation appropriate.

From the regression results in columns (4) and (5), the estimated coefficient of Ln (farm size) is 0.1029, and this is positively significant at the 1% level. This indicates that the core explanatory variable (farm size) has a statistically significant positive effect on the dependent variable (FPPWRB) at the 1% level. As shown in column (5), the marginal effect of the land management scale on FPPWRB is 0.0359, and this effect is significant at the 1% level. In other words, the higher the farm size, the more likely FPPWRB is to increase by 3.59%. According to the regression results, farm size is significantly and positively associated with FPPWRB in all models. The empirical results of this study are consistent with the theoretical analysis presented earlier and the findings of Yan et al. [34], Zhang et al. [35], and Zhang et al. [42], implying that as farm size increases, farmers are more likely to recycle and reuse pesticide packaging waste. This finding aligns with the studies of Han et al. [43], Ren et al. [15], and Zheng and Luo [44], as a larger farm size implies that farmers have more resources and capacity to manage waste. Therefore, hypothesis 1 (i.e., farm size positively influences household willingness to participate in waste sorting) is confirmed.

Another interesting finding from this study is that geographical location can influence FPPWRB. In Model (4), the coefficients for Terrain 1 and Terrain 2 are significantly positive (at the 1% statistical level), indicating that farmers in plain and hilly areas are more inclined to recycle pesticide packaging waste than those in mountainous regions. This finding suggests that establishing a universal resource recycling system requires increased attention to mountainous regions (geographically disadvantaged areas).

4.3. Endogeneity Test for Farm Size and FPPWRB

Endogeneity analysis verifies whether the relationship between the explanatory variable (farm size) and the dependent variable (FPPWRB) is accurately captured. If endogeneity is present, modeling results may lead to false causal relationships and misleading policy recommendations. Farm size serves as the explanatory variable, while FPPWRB is the dependent variable. FPPWRB may be influenced by other unobserved factors, which could simultaneously affect farm size. Such unobservable factors can include economic factors, policy factors, etc. Therefore, there may be an endogeneity issue between farm size and FPPWRB. If there are issues such as omitted variables or unobservable factors, endogeneity analysis can help identify them and explore how to incorporate them into the model. This can help eliminate model errors and more accurately measure the relationship.

Therefore, to determine whether variables or causal relationships are included or not in the model and to enhance the credibility of the research results, instrumental variables were used to conduct an endogeneity test on the model.

Table 4. Endogeneity regression results.

	Phase 1	Phase 2
size_ratio	1.7759 ***
	(12.90)
Ln (farm size)		0.1936 *
		(1.81)
_cons	0.8313 ***	−1.0164 ***
	(3.18)	(−2.63)
control variables	Yes	Yes
prov effect	Yes	Yes
F-value of Phase 1	90.37 ***
Wald chi2	322.14 ***
Wald endogeneity value	0.3809
N	2663	2663
R-squared	0.471	0.125

Note: *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. The T-values are in parentheses.

shows the endogeneity regression results. The instrumental variable for farm size is “the proportion of other large-scale households (i.e., households with land size greater than the sample mean) in the village (excluding the household itself)” [34,45]. In selecting the instrumental variable, the first criterion is validity, meaning that the instrumental variable should be correlated with the explanatory variable and capable of influencing its value. In this case, “the proportion of other large-scale households in the village (excluding the household itself)” reflects the farm size situation within a village, making it a potentially valid instrumental variable. Second, the instrumental variable is available. In some cases, obtaining raw data on farm size may not be easy. By using “the proportion of other large-scale households in the village (excluding the household itself)” as an instrumental variable, existing data sources or survey results can be leveraged, making it easier to obtain data and conduct the study. Additionally, using “the proportion of other large-scale households in the village (excluding the household itself)” as an instrumental variable provides a clearer explanation of the determinants of farm size. For example, if this instrumental variable is found to have a significant relationship with the explanatory variable, it can allow for further exploration of how other large-scale households influence farm size. The regression results in Table 4 show that the first-stage estimation of the instrumental variable test yields an F-value of 90.37 (p-value = 0.000), which is significant at the 1% level. This indicates that there is no weak instrument problem. In the second stage, the Wald test rejects the null hypothesis that the model has no endogenous variables at the 1% significance level, indicating that endogeneity is indeed a concern. From the estimation results using the instrumental variable method, the direction of the coefficient for farm size and FPPWRB is consistent with the baseline regression results. This suggests that even after addressing endogeneity with the instrumental variable method, farm size still has a significant positive effect on FPPWRB, further confirming H1.

4.4. Robustness Check for the Impact of Farm Size on FPPWRB

To strengthen the reliability of the research conclusions and results, we conducted robustness checks by changing the measurement method of the core explanatory variable and employing alternative estimation models.

Given that parameter estimation using the probit model may be subject to variability due to different distributional assumptions, this study used the logit model to test for robustness. Logit and probit models are commonly used binary choice models for dealing with binary response variable problems. However, compared to the probit model, the logit model imposes less restrictive distributional assumptions on the explanatory variables, making it more effective in handling heteroskedasticity and outliers. In addition, the coefficients of the logit model can be interpreted as odds ratios, enhancing the intuitive interpretability of the model’s results. Therefore, the robustness check using the logit model aims to confirm the reliability of the research findings and ensure that the results are not solely influenced by the choice of a specific model or data-handling approach. As shown in

Table 5. Robustness check.

	(1)	(2)	(3)
	Logit	Irrigable Area	Maximum Plot Size
Ln (farm size)	0.1685 ***
	(3.89)
Ln (farm size2)		0.0671 ***
		(3.06)
Ln (farm size3)			0.0889 **
			(2.31)
_cons	−1.4562 **	−0.7731 **	−0.9415 ***
	(−2.55)	(−2.24)	(−2.66)
control variables	Yes	Yes	Yes
prov effect	Yes	Yes	Yes
N	2663	2565	2545
pseudo R2	0.0999	0.0965	0.0974

Note: *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. The T-values are in parentheses.

, after the probit model is replaced by the logit model, the coefficients of the core explanatory variable are still significantly positive, which verifies the fact that the robustness test of the replacement model is reasonable.

Additionally, to ensure the robustness of the core explanatory variable, we incorporated the “irrigable area” and “maximum plot size” as alternative measures. By replacing the original core explanatory variable (farm size) with these two new indicators, Models (2) and (3) were constructed. As shown in Table 5, regardless of whether the core explanatory variable was replaced by the “irrigable area” or “maximum plot size,” the coefficients remained significantly positive. This further confirms the positive impact of farm size on FPPWRB, validating the effectiveness of the robustness check through variable substitution. Therefore, the robustness test conducted by replacing the estimation model and the measurement method of the core explanatory variable not only strengthened the credibility of the research conclusions but also ensured the robustness of the research results, showing that the conclusions remain consistent and reliable under different model frameworks and measurement indicators. The results of the robustness check are shown in Table 5.

4.5. Heterogeneity Analysis for the Impact of Farm Size on FPPWRB

To further investigate the heterogeneity of the effect of farm size on FPPWRB, we conducted two key heterogeneity analyses.

First, to examine whether the effect of farm size on FPPWRB varies by location, households were categorized into urban and non-urban suburbs, and regressions were conducted separately for each group. From the heterogeneous regression results in

Table 6. Heterogeneity regression results.

	(1)	(2)
	Urban Suburbs	Non-Urban Suburbs
Ln (farm size)	0.2890 ***	0.0723 **
	(3.95)	(2.53)
_cons	−1.1736	−1.0166 ***
	(−1.20)	(−2.67)
control variables	Yes	Yes
prov effect	Yes	Yes
N	519	2144
pseudo R2	0.1324	0.1096

Note: *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. The T-values are in parentheses.

, it can be observed that the effect of farm size on FPPWRB differs between urban and non-urban suburbs. Specifically, for farmers in urban suburbs, each unit increase in farm size results in a 0.2890 effect on FPPWRB, which is significantly greater than the 0.0723 effect observed in non-urban suburbs. The potential reasons for this discrepancy may include the following: (1) Urban suburbs are closer to markets, making it easier to access relevant technologies and beneficial information [46,47]. Therefore, farmers in urban suburbs are more likely to access technology and information related to waste management than those in non-urban suburbs, which may lead to greater motivation to participate in PPW recycling. (2) Urban suburbs are more likely to receive government support for the construction of recycling facilities. Therefore, from a regional perspective, the impact of farm size on FPPWRB may vary.

Second, farmers were divided into older and younger generations based on a 40-year-age threshold, and separate regressions were conducted. The regression results in

Table 7. Heterogeneity regression results—generational differences.

	(1)	(2)
	Over 40 Years Old	Under 40 Years Old
Ln (farm size)	0.0894 ***	0.1714 **
	(3.20)	(2.08)
_cons	−0.6507 *	−0.4711
	(−1.67)	(−0.39)
control variables	Yes	Yes
prov effect	Yes	Yes
N	2411	252
pseudo R2	0.0980	0.1847

Note: *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. The T-values are in parentheses.

show that the effect of farm size on FPPWRB is 0.1714 for younger farmers, compared to 0.0894 for older farmers. This indicates that each unit increase in farm size has a greater impact on FPPWRB among younger farmers than older farmers. This may be because younger farmers are more open to adopting new agricultural technologies and management practices. Additionally, they might place a greater emphasis on environmental protection, making them more inclined to recycle and reuse pesticide packaging waste.

4.6. Mediation Effect for the Impact of Farm Size on FPPWRB

To examine the impact of the hazard awareness of pesticide packaging on the relationship between farm size and FPPWRB, we further conducted a mediation effect analysis to explore the underlying mechanism.

As shown in

Table 8. Regression results of mediation effect.

	(1)	(2)	(3)
	Pesticide_Packaging	Cognition	Pesticide_Packaging
Ln (farm size)	0.1029 ***	0.1698 ***	0.0847 ***
	(3.93)	(7.68)	(3.17)
cognition			0.2425 ***
			(7.44)
_cons	−0.8621 **		−1.1927 ***
	(−2.51)		(−3.41)
control variables	Yes	Yes	Yes
prov effect	Yes	Yes	Yes
N	2663	2663	2663
pseudo R2	0.0998	0.0346	0.1152

Note: *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. The T-values are in parentheses.

, mediation effect analysis was conducted using three models to examine the relationships between farm size (Ln (farm size)), farmers’ environmental awareness (cognition), and the impact of farm size on FPPWRB (pesticide_packaging). This analysis aimed to determine whether farmers’ environmental awareness mediates the relationship between farm size and FPPWRB.

Model (1) examines the direct impact of farm size on FPPWRB. In this model, the coefficient of Ln (farm size) is 0.1029, which is statistically significant (p < 0.01). This indicates that a large farm size is associated with a higher probability of FPPWRB. The results for Model (1) indicate a positive relationship between farm size and FPPWRB, confirming the initial research hypothesis.

Model (2) analyzes the effect of farm size on farmers’ environmental awareness. In this model, the coefficient of Ln (farm size) is 0.1698, which is also statistically significant. This indicates that a larger farm size is associated with stronger environmental awareness among farmers. This finding suggests that farm size not only directly affects FPPWRB but may also indirectly influence it by enhancing farmers’ environmental awareness.

Finally, Model (3) incorporates both farm size and farmers’ environmental awareness to assess their combined effect on FPPWRB. After including farmers’ environmental awareness as a mediating variable, the effect of farm size on FPPWRB decreased from 0.1029 to 0.0847 but remained statistically significant. Meanwhile, the coefficient for farmers’ environmental awareness is 0.2425, which is also highly significant. This suggests that farmers’ environmental awareness indeed plays a mediating role in the relationship between farm size and FPPWRB, thus confirming H2.

In summary, the analysis results of this series of models provide empirical evidence for our hypothesis—namely, the fact that farm size indirectly promotes FPPWRB by enhancing farmers’ environmental awareness. At the same time, farm size also has a direct positive effect. These findings are consistent with existing studies, which have found that larger farms are more likely to adopt environmentally friendly practices, partly due to their stronger resource integration capabilities and greater awareness of the environment. This analysis not only provides a new perspective on the impact of farm size on FPPWRB but also highlights the crucial importance of enhancing farmers’ environmental awareness when promoting pro-environment actions.

5. Conclusions

5.1. Findings

Based on the micro-survey data from CRRS 2020, this study covers a sample of 2775 rural households. Using the probit model, we thoroughly examined the impact of farm size on FPPWRB and its underlying mechanisms. The main findings are given as follows:

(1)

Farmers engage in limited recycling of PPW. Descriptive statistics show that only 41.7% of the sample farmers participated in pesticide packaging waste recycling. This indicates that PPW management has not been given adequate attention by farmers, presenting new challenges to the fragile agricultural production environment.

(2)

There is a significant positive correlation between farm size and FPPWRB. The empirical results show that for every 1% increase in farm size, the probability of farmers participating in pesticide packaging waste recycling increases by 3.59%. Namely, larger farms are more likely to engage in recycling. This may be because these farms have more resources and incentives to implement environmental protection measures, thereby reducing the environmental impact of agricultural production.

(3)

Farmers in urban suburbs or younger farmers are more inclined to FPPWRB. In the heterogeneity analysis, the coefficient of the farm size variable was 0.2980 for the urban suburbs group and 0.1714 for the younger group (under 40 years old), both of which are higher than those for the non-urban suburbs group and the middle-aged group (over 40 years old). Namely, farm size has a more positive impact on FPPWRB among younger farmers and those in urban suburbs, suggesting that these groups are more likely to adopt recycling measures because of their specific advantages.

(4)

Farm size can improve FPPWRB by enhancing farmers’ environmental cognition levels. The mediation effect estimation results show that farm size significantly and positively affects environmental awareness at the 1% level. Meanwhile, environmental awareness significantly and positively influences FPPWRB at the 1% level. Namely, farm size influences farmers’ perception of the hazards associated with pesticide packaging waste, which, in turn, affects FPPWRB. Therefore, enhancing farmers’ knowledge and awareness is a key approach to promoting FPPWRB.

5.2. Implications

First, the government and relevant authorities should strengthen public awareness campaigns and provide technical guidance to enhance farmers’ awareness of PPW recycling, especially among small-scale farmers. Collaboration with local villagers can be established by offering gift incentives to encourage active participation in PPW recycling. At the same time, recycling bins can be set up at the village level, and staff should be organized to visit the village regularly for collection and processing. Encouraging land transfer and large-scale farming through policy guidance and economic incentives can help integrate resources and improve recycling and utilization efficiency.

Second, differentiated support measures should be developed based on the characteristics of different farmers. For farmers with strong technical skills and higher education levels, market information and technical support should be provided. For farmers in regions with poor natural conditions, financial and technical assistance should be offered. For example, in hilly areas—where the terrain is complex and transportation inconvenient—innovative PPW recycling strategies can be developed. Incentive-based methods, such as the “pesticide bottle exchange for gifts” program, can be used to encourage recycling.

Third, media campaigns and policy interpretation can strengthen farmers’ awareness of environmental protection and economic benefits, continuously enhancing their overall understanding of PPW recycling and utilization. At the same time, this can improve the incentive mechanism to promote the importance of recycling and utilization from both environmental and economic perspectives.

5.3. Limitations

This study also has some limitations, and future research can further improve upon these aspects. (1) There may be a dynamic relationship between farm size and FPPWRB. Therefore, future research could construct panel data to explore the dynamic relationship between these two variables. (2) The environmental impact of FPPWRB has not been sufficiently explored. Therefore, future research could develop new datasets to assess the environmental impact of FPPWRB. (3) Based on the Chinese case, it was found that as farm size increased, farmers were more willing to engage in environmentally friendly actions. Therefore, future research could examine whether the conclusions of this study are applicable to other countries.