Green pesticide practices and sustainability: empirical insights into agricultural services in China

ABSTRACT

In a world grappling with the overuse of pesticides, China emerges as a key consumer and producer, facing substantial challenges in environmental and agricultural sustainability. This study investigates how agricultural services can foster eco-friendly pesticide use among farmers. It leverages data from the ‘China Land Economic Survey’ (2020-2021) by Nanjing Agricultural University, covering 52 villages and 2600 households. Using a multi-dimensional fixed effects model and propensity score matching, the research addresses biases inherent in the data. Results show that agricultural business services significantly boost green pesticide adoption, outperforming agricultural technology and input sectors. These services increase the use of safer pesticides and the recycling of their containers by 11.6% and 6%, respectively. Farmer satisfaction and regular internet use also play a role in reinforcing this trend. Based on these insights, it recommends enhancing support for low-toxicity pesticides and recycling programmes, creating tailored policies for diverse agricultural needs, and implementing resource management and incentive mechanisms to promote sustainable practices. These strategies aim to bolster sustainable agriculture, balancing environmental health with economic viability.

KEYWORDS:

• Agricultural socialization services
• low-toxicity and low-residue pesticides
• pesticide packaging recycling
• heterogeneity
• multidimensional fixed effects

JEL CODES:

• Q12
• Q16
• Q18
• Q180
• Q280
• Q120

1. Introduction

Pesticides are fundamental to global food security, safety, and ecological health. As the primary consumer and producer of chemical pesticides, China’s usage per hectare is 1.5–4 times the global average (Lai, 2017). This intensive use leads to significant environmental challenges (Li et al., 2023). Overuse, especially of highly toxic pesticides, causes severe environmental and externalities issues (Zhao, 2015). These residues, persisting in soil, threaten soil health, water quality, and crop safety (Zhou et al., 2023). Over 10% of China’s farmland suffers from pesticide contamination (Zhao et al., 2018), highlighting the urgency to reduce highly toxic pesticide use. This challenge is not unique to China; developing countries like Vietnam (Hoi et al., 2016), India (Jaacks et al., 2022), Bangladesh (Biswas et al., 2021), and Thailand (Panuwet et al., 2012)also endeavour to mitigate similar issues. Numerous studies have investigated the effects of pesticides in developing countries, focusing on various aspects such as the impact on human health in Chinese agriculture (Kesavachandran et al., 2009; Koh & Jeyaratnam, 1996), the resulting environmental unsustainability (Bhandari et al., 2021), and the decline in agricultural productivity (Qaim & Zilberman, 2003).

Farmers increasingly turn to efficient, low-toxicity pesticides, including bio-pesticides and chemically advanced variants, to reduce environmental pollution, mitigate health risks, and improve agricultural product quality (Feng et al., 2021). However, environmental challenges extend beyond pesticide use to the disposal of non-biodegradable packaging, laden with residues, necessitating comprehensive recycling and disposal strategies (Moore et al., 2014).

In agricultural governance, the quality and yield of products, coupled with farmers’ market leverage, are critical. These outcomes are shaped by resource constraints and information asymmetries. Agricultural socialization services are emerging to address these challenges, promoting green pesticide use and recycling practices.

In China, these services offer a unique approach by integrating small-scale farmers into modern agriculture (Wang & Huan, 2023). They involve diverse market-based entities, providing a range of support from input provision to technology training. These entities, including cooperatives and agricultural firms, are scalable and pivotal in enhancing market interactions and disseminating insights, aiding farmers’ decision-making (Qing et al., 2023). The significant adoption of these services, with over 1.04 million organizations reaching 89 million small farmers by 2022, underscores their impact.

In Western contexts, these services parallel ‘agricultural extension services.’ For instance, in Ghana, extension services promote soil and water conservation (Danso-Abbeam, 2022), while in Bangladesh, they have boosted rice yields and efficiency (Biswas et al., 2021). India has launched extensive advisory services, encouraging small farmers to adopt new technologies (Varshney et al., 2022). In China, agricultural socialization services, encompassing resource sharing, collective management, and green practices (Huan et al., 2022), differ from the extension services’ focus on technology transfer and individual skill enhancement (Danso-Abbeam et al., 2018). This distinction highlights the complementary nature of these two service types in advancing agricultural sustainability and efficiency.

The unique aspect of agricultural socialization services, particularly in China, is their involvement not just between industries or supply chain stages, but also within the production domain itself, especially in operational aspects (Steinke et al., 2021; Yang & Zhang, 2023). Practices like outsourcing irrigation, drone fertilization, and pest control services for grain and cash crops represent a shift of activities traditionally internal to the production phase, addressing internal resource allocation inefficiencies (Qiu et al., 2023). While such services are less common in other countries, their potential for broader application is evident.

This paper aims to analyze the impact of diverse, interconnected services within the agricultural supply chain on sustainable production. By doing so, it illuminates the benefits of agricultural socialization services and offers insights for developing countries to foster green and sustainable agricultural practices. Small-scale farmers often prioritize yield over the quality of agricultural products, facing challenges such as information asymmetry and market risks, which diminish their bargaining power (Nordin et al., 2022). This scenario renders price incentives for quality less effective in influencing their adoption of green pesticides, reduction efforts, and packaging recycling practices.

Agricultural socialization services offer a viable solution to these challenges. These services address technical skill gaps, labour shortages, financial constraints, weak risk resilience, and high transaction costs (Davide et al., 2010). They facilitate a shift in pesticide usage, improve management practices, reduce information asymmetry, encourage green technology adoption, and promote green production (Ruzzante et al., 2021).

Moreover, these services enhance farmers’ income and market bargaining power through quality premiums, leading to increased adoption of safer, residue-free pesticides, and improved pesticide application and packaging disposal practices. Investments in green practices by these organizations also generate economies of scale, improving agricultural product quality. However, existing research predominantly focuses on how these services decrease fertilizer and pesticide usage, overlooking their role in fostering the use of safer, less polluting pesticides and recycling behaviours. The government and related departments are now increasingly focusing on proper pesticide packaging recycling. This paper concentrates on green pesticide practices, such as the use of low-toxicity pesticides and recycling packaging, thereby addressing research gaps.

Furthermore, while existing studies recognize the impact of various socialized services on green production, they often neglect the role of service providers. These organizations, varying extensively in characteristics and nature (Abugre et al., 2023), significantly influence farmers’ production behaviours. For example, agricultural input department sales staff, who may lack pesticide expertise, often prioritize sales goals over promoting green application techniques (Xin et al., 2021). As service suppliers, these organizations play a pivotal role in shaping farmers’ production behaviours (Zhang et al., 2023).

Current research primarily focuses on the role of agricultural socialization services in reducing fertilizer and pesticide usage, with insufficient attention to the use and recycling of efficient, low-toxic, and eco-friendly pesticides (Li et al., 2023; Lin et al., 2022). This paper addresses this gap by characterizing green pesticide practices, including the use of efficient, low-toxicity pesticides with minimal residue and their packaging recycling.

Moreover, the literature lacks an in-depth analysis of the entities behind agricultural socialization services. The diversity in the types and characteristics of these organizations is significant (Li et al., 2023). For instance, input sales staff may prioritize sales over educating on green pesticide applications (Xin et al., 2021), impacting how these services influence farmers’ practices (Zhang et al., 2023).

This study examines the effects of agricultural socialization services on farmers’ green pesticide application behaviours, considering the diversity of service providers and information acquisition. It categorizes service organizations into agricultural operators, technical departments, and input companies, reflecting market-oriented, public welfare-oriented, and technology dissemination-oriented approaches, respectively. The study explores the effectiveness of different service organization types in promoting green pesticide practices among farmers.

First, it analyzes the heterogeneity of service providers. Previous research has highlighted the influence of service satisfaction on farmers’ behaviours (Amrutha & Geetha, 2021), but often overlooks how this satisfaction translates into responsive actions. This study delves into the nuances of service satisfaction and its correlation with farmers’ adoption of green practices, considering the technical training, information dissemination, and policy benefits provided by these organizations.

Second, the study investigates the heterogeneity in information acquisition. Asymmetric information among farmers can impact the effectiveness of agricultural socialization services. This analysis encompasses the roles of social networks and internet usage in shaping farmers’ information access (Bentley et al., 2019; Stern, 2008), examining variations in green pesticide application behaviours under different information acquisition conditions.

Integrating these perspectives, this study offers a comprehensive understanding of agricultural socialization services’ impact on farmers’ green pesticide behaviours, contributing to a deeper comprehension of farmers’ behavioural patterns in green agricultural practices and aiding in the formulation of relevant policies.

2. Material and methods

2.1. Data source

This research utilizes data from the ‘China Land Economics Survey’ (CLES) conducted by Nanjing Agricultural University, comprising panel data at household and village levels for 2020-2021. The China Land Economic Survey (CLES), initiated by the Humanities and Social Sciences Department of Nanjing Agricultural University in 2020 with assistance from the Jin Shan-bao Institute for Modern Agricultural Development, has established a significant data resource. Employing Probability Proportional to Size (PPS) sampling, the study selected 26 survey counties from 13 prefecture-level cities in Jiangsu Province. Within each county, 2 towns and subsequently 1 administrative village were randomly chosen, with 50 households sampled per village, totalling 2600 households across the province. The study processed variables related to agricultural socialization services, green pesticide application behaviour, and other factors. After addressing missing values and outliers, and controlling for fixed effects of region-time and service organization-time, the final analysis encompassed a sample of 1195 households.

Figure 1 portrays the agricultural landscape in Jiangsu Province, China. In 2021, the province was ranked tenth nationally in terms of total crop sowing area, amounting to 7,514.45 thousand hectares. Jiangsu’s agricultural sector is diverse, catering to staple grains such as rice and wheat, as well as high-value economic crops including honey peaches, loquats, and bayberries. This diversity is supported by an extensive network of 68,000 entities engaged in agricultural socialization services, comprising specialized farmer cooperatives, supply and marketing cooperatives, agricultural enterprises, and family farms. This indicates a well-developed system of agricultural socialization services in the province.

Figure 1. Research area (Jiangsu Province, China).

In terms of pesticide usage, Jiangsu reported consumption of 65,000 tons in 2020, ranking eighth in China. A concerning factor is the average DDT concentration in the soil, recorded at approximately 100μg/kg, surpassing the national standard of 50μg/kg (Ma et al., 2020). This underscores the need for careful attention to pesticide usage in the region.

Furthermore, the internal economic disparities within Jiangsu mirror the broader regional differences observed across Western, Central, and Eastern China (Li et al., 2023). Despite the research being confined to Jiangsu Province, the diversity in agricultural products, significant regional variations, and the spectrum of entities involved in agricultural socialization services render this province-based survey highly valuable. The advanced and systematic nature of Jiangsu’s agricultural socialization services significantly contributes to the study, offering insights applicable not only to other provinces in China but also to nations with emerging agricultural service systems. Consequently, the choice of Jiangsu Province as a research sample is both logical and representative, offering a microcosmic view of broader agricultural trends and practices.

2.2. Variable design and descriptive statistics

2.2.1. Dependent variable

The study measures green pesticide application behaviour through two indicators: the use of high-efficiency, low-toxicity, and low-residue pesticides (coded as 1 for usage and 0 for non-usage), and pesticide packaging recycling behaviour (coded as 1 for recycling and 0 for non-recycling). Missing values are assigned a 0 to mitigate sample selection bias.

2.2.2. Key independent variable

The survey identified a range of segmented services available in the market. This study focuses on services relevant to pesticide application as indicated in the responses. The variable for agricultural socialization services is coded as 1 if farmers have received any services such as improved seed, soil testing, crop cultivation management, pest control techniques, mechanized production, energy-efficient agricultural facilities, or disaster prevention and mitigation, and 0 otherwise.

2.2.3. Control variable

The analysis considers variables like individual farmer characteristics (age, gender, agricultural labour days, agricultural training experience), human capital (education level, health condition), household endowments (wealth, number of agricultural labourers, non-agricultural income share), and land endowments (contracted land area, pesticide expenses). Notably, social capital is divided into weak and strong ties networks. The number of contacts in a mobile phone measures the weak ties network, while the ability to borrow 50,000 yuan in times of difficulty measures the strong ties network. Detailed variables and descriptive statistics are presented in Table 1.

Table 1. Definitions and Descriptive Statistics of Variables.

Variable type	Variable Definition	Variable Description and Assignment	Mean value	standard deviation
Dependent Variable	High efficiency, low toxicity, and low residue pesticide usage	Usage in contracted land for agricultural purposes: Yes = 1; No = 0	0.285	0.452
	Pesticide packaging recycling	Recycling or proper disposal in contracted land: Yes = 1; No = 0	0.127	0.333
independent variable	Agricultural socialization services	Services received: Accepted = 1; Not accepted = 0	0.213	0.409
personal characteristics	Age	Age in years	61.575	11.388
	Age ^2	Age squared term	3921.148	1323.326
	gender	Male = 1; Female = 0	0.715	0.452
	Days in agricultural labour	Number of days engaged in agricultural labour within one year	73.670	98.436
	Agricultural training	Received any agricultural training: Yes = 1; No = 0	0.319	0.466
human capital	Education level	Years in school	7.032	3.965
	Health condition	Health status (1 = loss of labour ability to 5 = excellent)	3.974	1.070
Social Capital	Weak network	Number of contacts on your phone	95.091	328.565
	Strong Network	Number of people who can lend 50,000 yuan in difficulty	5.400	19.521
Family endowment	Family wealth level	Satisfaction with family life prosperity(1 = very dissatisfied to 5 = very satisfied)	3.622	0.917
	Family farm labour force	Number of labourers in the family farm	1.401	1.027
	Non-agricultural income proportion	Household non-agricultural income/total household income	0.653	0.387
Land endowment	Contracted land area	Total contracted land area (acres)	0.439	1.910
	Pesticide purchase cost	Logarithm of pesticide purchase cost for contracted land (Chinese RMB)	4.556	0.405
Internet usage	Smartphone usage	Number of smartphones	2.485	1.675
	Computer network usage	Number of computers with internet access	0.713	0.937

2.3. Econometrics models

2.3.1. Multi-dimensional fixed effects model

To address omitted variable bias and endogeneity, this study employs a multi-dimensional fixed effects model for baseline regression (Duflo, 2004). Acknowledging disparities in economic development and service organization within Jiangsu Province, the model includes regional-time and service organization-time interactive fixed effects. These account for time-varying economic factors and policy impacts across different regions and service organizations, isolating macro and meso-level influences on the analysis. The econometric model is specified as follows: (1) $Y_{it}={\beta }_0+{\beta }_1{X}_{it}+{\beta }_2{Z}_{it}+{\gamma }_{at}+{\delta }_{zt}+{\epsilon }_{it}$(1) In the model, ${}^{\mathrm{\prime }}{i}^{\mathrm{\prime }}$denotes the household, ${}^{\mathrm{\prime }}{t}^{\mathrm{\prime }}$ the time, ${}^{\mathrm{\prime }}{a}^{\mathrm{\prime }}$ the region,${}^{\mathrm{\prime }}{z}^{\mathrm{\prime }}$ the service organization. $Y_{it}$ represents the green pesticide application behaviour of household$i$ in the year $t$, $X_{it}$ indicates the level of specialized services received by household $i$ in the year $t$, and $Z_{it}$ includes control variables such as individual characteristics, human capital, social capital, family, and land endowments. ${\gamma }_{at}$ and ${\delta }_{zt}$ represent regional-time and service organization-time fixed effects, respectively, and ${\epsilon }_{it}$ is the error term.

2.3.2. Propensity score matching model

The study employs Propensity Score Matching (PSM) to control for observable heterogeneity and address endogeneity from self-selection in non-randomized experiments (Rosenbaum & Rubin, 1983). PSM is used to counteract the self-selection bias in farmers’ adoption of agricultural socialization services. The sample is divided into a treatment group (farmers receiving the services) and a control group (farmers not receiving the services). The propensity score is estimated using: (2) $P=P_r(D_i=1|X_i)=E(D_i=0|X_i)$(2) In this model, $D_i=1$ indicates farmers receiving services, $D_i=0$ those not receiving services, and $X_i$ represents the characteristic variables of the sample farmers. The Average Treatment Effect (ATT) on farmers’ green pesticide application behaviour is calculated as: (3) $ATT=E(Y_{1i}|D_i=1)-E(Y_{0i}|D_i=1)=E(Y_{1i}-Y_{0i}|D_i=1)$(3) $Y_{1i}$, $Y_{0i}$ represent the green pesticide application behaviour of farmers with and without receiving agricultural socialization services, respectively. The nearest neighbour, calliper, and kernel matching methods are used for analysis.

3. Results

3.1. Impact of agricultural socialization services on green pesticide practices

Table 2 presents the effects of agricultural socialization services on the use of low-toxicity and low-residue pesticides by contracted farmers. The study employs comparative analysis to determine the most appropriate model. Model 1, without covariates but controlling for fixed time, regional, and service organization effects, indicates a positive impact of agricultural socialization services on high-efficiency, low-toxicity, and low-residue pesticide use, significant at the 1% level.

Table 2. Impact of Agricultural Services on Low-Toxicity Pesticide Use by Farmers

variable	High efficiency, low toxicity, and low residue pesticide usage
	model1	model2	model3
Agricultural socialization services	0.171***	0.196***	0.193***
	(0.027)	(0.031)	(0.032)
Age		0.005	0.006
		(0.008)	(0.008)
Age^2		−0.000	−0.000
		(0.001)	(0.000)
Gender		−0.014	−0.008
		(0.034)	(0.034)
Days in agricultural labour		0.000	0.000
		(0.000)	(0.000)
Agricultural training		−0.022	−0.021
		(0.032)	(0.032)
Education level		−0.002	−0.002
		(0.004)	(0.004)
Health condition		0.001	−0.002
		(0.015)	(0.016)
Weak network		0.000***	0.000***
		(0.000)	(0.000)
Strong network		0.001	0.001
		(0.000)	(0.000)
Family wealth level		0.0117	0.010
		(0.015)	(0.015)
Family farm labour force		0.075***	0.078***
		(0.013)	(0.013)
Non-agricultural income proportion		−0.001	0.002
		(0.035)	(0.036)
Contracted land area		0.012***	0.014***
		(0.004)	(0.004)
Pesticide purchase cost		−0.143***	−0.147***
		(0.042)	(0.042)
Time fixed effect	YES	YES	NO
Regional fixed effect	YES	YES	NO
Service organization fixed effect	YES	YES	NO
Region-time fixed effect	NO	NO	YES
Service organization-time fixed effect	NO	NO	YES
Constant (_cons)	0.205***	0.522*	0.528*
	(0.023)	(0.315)	(0.319)
N	1276	1199	1195
R^2	0.159	0.216	0.221

Note: *, **, *** indicate significance at the 10%, 5%, and 1% levels, respectively. Standard errors in parentheses are heteroscedasticity-robust.

Model 2, adding control variables such as farmer characteristics, human capital, social capital, family assets, and land assets, maintains that agricultural socialization services significantly enhance the usage of such pesticides. This effect remains statistically significant at the 1% level, even after adjusting for fixed effects over time, regionally, and from service organizations.

Model 3, encompassing all control variables and accounting for fixed effects of regional, temporal, and service organization factors, shows the best fit (R² = 0.221) and reaffirms the positive impact. This model, with an estimated coefficient of 0.193 significant at the 1% level, confirms that agricultural socialization services notably encourage the adoption of these pesticides on contracted lands. The analysis of control variables indicates that factors like weak social networks, the number of family agricultural labourers, and land assets positively influence this usage, all significant at the 1% level. Conversely, the cost of purchasing pesticides has a significant negative impact at the same level.

Table 3 displays the results from a series of regression models analyzing the influence of agricultural socialization services on pesticide packaging recycling practices among farmers. In Model 1, which controls only for time, region, and service organization effects, these services are found to positively affect packaging recycling, with statistical significance at the 5% level. Model 2 extends the analysis by including control variables relating to personal characteristics, human and social capital, and family and land assets. This model confirms the positive effect observed in Model 1. Model 3, the most comprehensive, includes all control variables while also accounting for region-time and service organization-time fixed effects. This model provides the best fit (R² = 0.110) and continues to affirm the positive impact of agricultural socialization services on recycling practices.

Table 3. Impact of Agricultural Services on Pesticide Packaging Recycling in Contracted Land.

Variable	Pesticide packaging recycling
	model1	model2	model3
Agricultural socialization service	0.063**	0.071**	0.071**
	(0.026)	(0.032)	(0.032)
Age		0.005	0.006
		(0.007)	(0.007)
Age ^2		−0.000	−0.000
		(0.000)	(0.000)
Gender		0.0194	0.024
		(0.030)	(0.030)
Days in agricultural labour		−0.000	−0.000
		(0.000)	(0.000)
Agricultural training		0.024	0.024
		(0.026)	(0.027)
Educational level		−0.001	−0.001
		(0.004)	(0.004)
Health condition		0.010	0.010
		(0.013)	(0.013)
Weak network		−0.000	0.000
		(0.000)	(0.000)
Strong network		0.001	0.001
		(0.000)	(0.000)
Family wealth level		0.016	0.016
		(0.013)	(0.013)
Household agricultural labour force		0.062***	0.064***
		(0.012)	(0.012)
Non-agricultural income proportion		−0.023	−0.020
		(0.030)	(0.031)
Contracted land area		0.008**	0.009**
		(0.004)	(0.004)
Pesticide purchase cost		−0.063	−0.065*
		(0.040)	(0.040)
Time fixed effect	YES	YES	NO
Regional fixed effect	YES	YES	NO
Service organization fixed effect	YES	YES	NO
Region-time fixed effect	NO	NO	YES
Service organization-time fixed effect	NO	NO	YES
Constant (_cons)	0.128***	0.113	0.106
	(0.0224)	(0.287)	(0.287)
N	1276	1199	1195
R^2	0.052	0.105	0.110

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

Specifically, Model 3 estimates a coefficient of 0.071, significant at the 5% level, suggesting that agricultural socialization services substantially encourage pesticide packaging recycling on contracted lands. The analysis of control variables indicates that the number of family agricultural labourers positively affects recycling, with a significant impact at the 1% level. The area of contracted land is also positively related to recycling activities, with a significant effect at the 5% level. In contrast, the cost of purchasing pesticides is found to have a negative impact on recycling, with significance at the 10% level.

3.1. Service subjects and information acquisition heterogeneity analysis results

Table 4 presents the findings from an analysis exploring the heterogeneity of agricultural socialized service organizations. These entities include agricultural technology departments (e.g. government agricultural extension departments, university research and education institutions, technical R&D units in agricultural academies, specialized agricultural technology service organizations), agricultural management entities (like cooperatives, leading enterprises, and large-scale farmers in farming and breeding), and agricultural input enterprises.

Table 4. Heterogeneity of Agricultural Socialization Service Organizations.

	High efficiency, low toxicity, and low residue pesticide usage			Pesticide packaging recycling
	(1)	(2)	(3)	(4)	(5)	(6)
	Agricultural Technology Department	Agricultural Business Entity	Agricultural Means of Production Enterprises	Agricultural Technology Department	Agricultural Business Entity	Agricultural Means of Production Enterprises
Agricultural socialization service	0.192***	0.283**	0.100	0.065*	0.244***	−0.087
	(0.034)	(0.113)	(0.234)	(0.036)	(0.080)	(0.191)
control variable	YES	YES	YES	YES	YES	YES
Region-time fixed effect	YES	YES	YES	YES	YES	YES
Service organization-time fixed effect	YES	YES	YES	YES	YES	YES
Constant (_cons)	0.718**	0.661	−1.165	0.187	0.820	−0.565
	(0.361)	(1.234)	(1.065)	(0.326)	(1.128)	(0.833)
N	963	127	91	963	127	91
R^2	0.249	0.298	0.240	0.125	0.298	0.245

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

The results indicate that services from agricultural technology departments and agricultural management entities significantly promote farmers’ use of high-efficiency, low-toxicity, and low-residue pesticides, and positively influence the recycling of pesticide packaging. The estimated coefficients for these effects are 0.192 and 0.283 for pesticide use, and 0.065 and 0.244 for packaging recycling, respectively. However, the services rendered by agricultural input enterprises do not show a significant impact on farmers’ pesticide application behaviour.

Table 5 presents the analysis of farmers’ satisfaction with agricultural socialization services and its impact on green pesticide practices. Satisfaction was assessed through the survey question ‘Are you satisfied?’, with a score of 5 indicating satisfaction and lower scores indicating less satisfaction. The experimental group’s samples were divided into two categories: satisfied and less satisfied, for comparative regression analysis against the control group.

Table 5. Heterogeneity in Satisfaction with Agricultural Socialization Services.

	High efficiency, low toxicity, and low residue pesticide usage		Pesticide packaging recycling
	(1)	(2)	(3)	(4)
	satisfied	Less satisfied	satisfied	Less satisfied
Agricultural socialization service	0.238***	0.152***	0.0882**	0.0570
	(0.0411)	(0.0540)	(0.0449)	(0.0481)
control variable	YES	YES	YES	YES
Region-time fixed effect	YES	YES	YES	YES
Service organization – time fixed effect	YES	YES	YES	YES
Constant (_cons)	0.334	0.910*	−0.111	0.386
	(0.429)	(0.510)	(0.394)	(0.449)
N	709	483	709	483
R^2	0.236	0.273	0.141	0.135

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

For farmers satisfied with the services, the estimated coefficient for the impact of agricultural socialization services on the use of high-efficiency, low-toxicity, and low-residue pesticides on contracted land is 0.238, significant at the 1% level. The coefficient for the effect on pesticide packaging recycling is 0.088, significant at the 5% level.

For farmers who are less satisfied, the estimated coefficient for the impact on pesticide usage is 0.152, significant at the 1% level. However, the coefficient for the effect on pesticide packaging recycling is 0.057, which is not statistically significant.

These findings suggest that agricultural socialization services are more effective in promoting green pesticide application behaviour among farmers who express satisfaction with these services compared to those who are less satisfied.

Table 6 delineates the results of analyzing the heterogeneity among farmers’ social network levels. The study classified farmers’ social capital into weak and strong tie networks, based on the number of contacts in their mobile phones and their ability to borrow money in times of difficulty, respectively. These two dimensions were integrated using the entropy method, setting the average value as a threshold. Farmers were then categorized as having higher or lower social networks based on this average.

Table 6. Social Network Heterogeneity.

	High efficiency, low toxicity, and low residue pesticide usage		Pesticide packaging recycling
	(1)	(2)	(3)	(4)
	Low level social networks	High level social network	Low level social networks	High level social network
Agricultural socialization service	0.202***	0.179***	0.0330	0.145***
	(0.0423)	(0.0493)	(0.0425)	(0.0517)
control variable	YES	YES	YES	YES
Region-time fixed effect	YES	YES	YES	YES
Service organization – time fixed effect	YES	YES	YES	YES
Constant (_cons)	0.437	1.100*	0.0254	0.407
	(0.382)	(0.606)	(0.357)	(0.573)
N	742	451	742	451
R^2	0.237	0.263	0.098	0.192

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

The analysis shows that for farmers with lower social networks, agricultural socialization services significantly increased the use of high-efficiency, low-toxicity, and low-residue pesticides on contracted land (estimated coefficient: 0.202, significant at the 1% level). For those with higher social networks, the coefficient was 0.179, also significant at the 1% level. However, the impact on pesticide packaging recycling was significant only for farmers with higher social networks.

Table 7 presents the findings from the analysis of farmers’ Internet usage levels. The study bifurcated Internet usage into two dimensions: smartphone usage and computer network usage (Esselaar et al., 2007). The number of smartphones and internet-accessible computers reported in the survey were used as measures. Similar to the social network analysis, the entropy method computed a comprehensive Internet usage score, with the average value determining higher or lower levels of Internet usage.

Table 7. Internet Usage Heterogeneity.

	High efficiency, low toxicity, and low residue pesticide usage		Pesticide packaging recycling
	(1)	(2)	(3)	(4)
	Low level Internet use	High level of Internet use	Low level Internet use	High level of Internet use
Agricultural socialization service	0.186***	0.190***	0.038	0.082*
	(0.052)	(0.040)	(0.050)	(0.043)
control variable	YES	YES	YES	YES
Region-time fixed effect	YES	YES	YES	YES
Service organization – time fixed effect	YES	YES	YES	YES
Constant (_cons)	−0.201	0.653	−0.606	0.200
	(0.655)	(0.431)	(0.461)	(0.383)
N	493	701	493	701
R^2	0.221	0.260	0.141	0.128

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

For farmers with lower Internet usage, the impact of agricultural socialization services on the use of high-efficiency, low-toxicity, and low-residue pesticides on contracted land was significant (estimated coefficient: 0.186, significant at the 1% level). For those with higher Internet usage, the coefficient was 0.19, also significant at the 1% level. However, the influence of these services on pesticide packaging recycling was not significant for farmers with lower levels of Internet usage. In contrast, for farmers with higher Internet usage, receiving agricultural socialization services significantly promoted pesticide packaging recycling, with an estimated coefficient of 0.082, significant at the 5% level.

3.2. Robustness test results

To validate the stability of its conclusions, the study implemented robustness tests by substituting the dependent variables. Specifically, variables related to the ‘use of high-efficiency, low-toxicity, and low-residue pesticides’ and ‘recycling of pesticide packaging’ on contracted land were replaced with equivalent variables for recipient land. The outcomes of these robustness checks are detailed in Table 8.

Table 8. Impact of Socialized Services on Green Practices in Receiving Areas.

variable	High efficiency, low toxicity, and low residue pesticide usage	Pesticide packaging recycling
	model1	model2
Agricultural socialization service	0.130***	0.099***
	(0.030)	(0.026)
Age	0.017***	0.008*
	(0.005)	(0.004)
Age ^2	−0.000***	−0.000***
	(0.000)	(0.000)
Gender	0.028	0.039**
	(0.024)	(0.019)
Days in agricultural labour	0.000***	0.000***
	(0.000)	(0.000)
Agricultural training	0.029	0.021
	(0.022)	(0.018)
Educational level	−0.003	−0.006**
	(0.003)	(0.002)
Health condition	0.000	0.013
	(0.012)	(0.009)
Weak network	0.000***	−0.000
	(0.000)	(0.000)
Strong network	−0.000	−0.000
	(0.000)	(0.000)
Family affluence	0.00201	0.000663
	(0.0109)	(0.009)
Household agricultural labour force	0.043***	0.023***
	(0.010)	(0.008)
Non-agricultural income proportion	−0.0288	−0.022
	(0.029)	(0.024)
Transferred land area	0.001***	0.001***
	(0.000)	(0.000)
Pesticide purchase cost	−0.243***	−0.106
	(0.082)	(0.080)
Constant (_cons)	0.680	0.251
	(0.419)	(0.393)
N	1194	1194
R^2	0.205	0.171

Note: *, **, and *** represent significance at the 10%, 5%, and 1% levels, respectively; values in parentheses indicate heteroscedasticity-robust standard errors.

The findings from these tests reinforce the study’s primary conclusions. The estimated coefficient for the impact of agricultural socialization services on the use of sustainable pesticides on recipient land is 0.13, significant at the 1% level. In parallel, the effect on recycling pesticide packaging on recipient land shows a coefficient of 0.099, also significant at the 1% level. The consistency of these results with the baseline regression findings bolsters the robustness and reliability of the study’s conclusions.

An in-depth analysis of control variables identified an inverted U-shaped relationship between age and the adoption of green pesticide practices on recipient land. Other factors, including weak connectivity, the number of family agricultural labourers, and land endowments, also consistently influence farmers’ adoption of green pesticide practices.

While robustness checks were additionally conducted for the heterogeneity analysis, space constraints preclude the inclusion of these results in the paper. Nonetheless, these checks further substantiate the study’s initial conclusions.

3.3. Endogeneity analysis results

To mitigate self-selection bias in the sample, this study employed propensity score matching (PSM) to derive more robust conclusions. The results, obtained using three distinct PSM methods – nearest neighbour, calliper, and kernel matching – are presented in Tables 9 and 10. These results consistently demonstrate that agricultural socialization services significantly boost the use of efficient, less toxic pesticides and enhance the recycling of pesticide packaging.

Table 9. Effect of Agricultural Services on ‘Sustainable’ Pesticide Usage.

Matching method	treatment group	control group	ATT	S.E.	T-value
the nearest neighbour matching method	0.397	0.296	0.101***	0.024	4.21
calliper matching method	0.396	0.271	0.125***	0.021	5.91
kernel matching method	0.396	0.274	0.122***	0.021	5.74

Table 10. The Impact of Agricultural Services on the Recycling of Pesticide Packaging.

Matching method	treatment group	control group	ATT	S.E.	T-value
the nearest neighbour matching method	0.200	0.142	0.057***	0.018	3.10
calliper matching method	0.200	0.138	0.062***	0.016	3.81
kernel matching method	0.200	0.139	0.061***	0.016	3.71

Note: The matching criterion for nearest neighbour matching is 3 units nearby; calliper matching is a one-to-four matching, with a calliper range of 0.06, and matching is conducted within this calliper range; kernel matching uses the default bandwidth of 0.06.

The Average Treatment Effect on the Treated (ATT) across all methods is significant at the 1% level, reinforcing the robustness of our findings. The average ATT values suggest that these services increase the likelihood of adopting sustainable pesticide usage and recycling practices by approximately 11.6% and 6%, respectively. This evidence strongly supports the effectiveness of agricultural socialization services in promoting environmentally sustainable practices in agriculture.

The efficacy of the propensity score matching (PSM) process is evaluated using kernel density plots. These plots are instrumental in comparing the distributions of the matched post-treatment group and the control group. A successful matching effect is indicated by minimal disparities between the groups in these plots.

Figure 2 presents the kernel density plots for the usage of low-toxicity and low-residue pesticides, specifically employing the nearest neighbour matching method. The visualization in Figure 2 shows a marked increase in the overlap between the matched post-treatment and control groups, indicative of a robust matching effect. This substantial overlap enhances the reliability of the analysis, suggesting that the conclusions derived are likely to be accurate and reflective of the true impact of agricultural socialization services.

Figure 2. Kernel Density Plot of Efficient, Low-Toxicity, Low-Residue Pesticide Use.

4. Discussion

This study provides a comprehensive analysis of the impact of agricultural socialization services on the adoption of green pesticide practices among farmers. The empirical evidence suggests a significant positive influence, leading to increased usage of low-toxic and low-residue pesticides and enhanced rates of pesticide packaging recycling on contracted land (Li et al., 2023). These findings affirm the pivotal role of agricultural socialization services in promoting sustainable farming practices.

In identifying the determinants affecting green pesticide adoption, our study highlights the positive effects of family agricultural labourers and land resources while noting the inverse relationship with pesticide costs (Teklewold et al., 2013). This nuanced understanding is crucial for developing targeted agricultural policies (Wilson & Tisdell, 2001). Our methodology’s strength lies in its large-scale data analysis and rigorous approach, including robustness checks and propensity score matching to address endogeneity issues (Li et al., 2023; Su et al., 2022).

Recognizing the limitations of a cross-sectional analysis, the study acknowledges the need for longitudinal research to capture long-term effects and underlying mechanisms. Future inquiries should explore these dimensions to provide a more dynamic understanding of the issues at hand.

The research’s implications are profound, demonstrating a substantial impact of agricultural socialization services on promoting sustainable pesticide usage and recycling practices, enhancing both environmental and agricultural sustainability (Akmukhanova et al., 2023). Such insights are vital for China and other developing nations aiming to mitigate environmental hazards associated with pesticide use and underscore the importance of green agricultural development (Li et al., 2023).

Our findings also contribute to the literature by addressing the gap concerning the effectiveness of agricultural services in fostering the adoption of efficient, low-toxicity pesticides and promoting recycling practices. This study delineates the positive influence of various service types, including mechanical outsourcing and agricultural extension services, in reducing overall pesticide usage (Lin et al., 2022); (Angeon et al., 2024; Hoi et al., 2016). Furthermore, it sheds light on the variations in service effectiveness and emphasizes the role of internet usage and social networks in disseminating green practices (Wang & Liu, 2021; Zhang et al., 2021).

5. Conclusions and policy recommendations

This study confirms the vital role of agricultural socialization services in enhancing sustainable agricultural practices, notably in increasing the use of low-toxicity pesticides and improving recycling rates of pesticide packaging. Recognizing the impact of family labour, land resources, and pesticide costs on these practices, this paper proposes the following policy recommendations:

5.1. Support for green pesticide adoption and packaging recycling

It is recommended to increase support for the adoption of low-toxicity pesticides and enhance recycling efforts of pesticide packaging. This can involve providing resources and incentives for farmers to shift to environmentally friendly alternatives and establishing recycling programmes.

5.2. Tailored policies for diverse agricultural needs

This paper suggests crafting policies that specifically address the unique needs of various agricultural stakeholders. This includes providing resources for the development and dissemination of green pesticides, technical assistance to agricultural operators, and incentives for broader participation in sustainable practices.

5.3. Resource management and incentive mechanisms

It is recommended to assist farmers in effectively managing their labour and land resources to promote sustainable agricultural practices. Introducing incentive mechanisms to encourage the adoption of green practices is also suggested, considering the influence of economic factors such as the cost of pesticides.

By implementing these recommendations, the agricultural sector can further contribute to the development of sustainable agricultural systems, ensuring environmentally responsible and economically viable farming practices.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The authors confirm that the data supporting of this study are available from the corresponding author (Dr. Yifeng Zhang) on request.

Additional information

Funding

Supported by 2023 Jiangsu Universities’ Major Project for Philosophy and Social Sciences Research (2023SJZD064) “Studyon the Mechanism of Effective Supply-Demand Matching for Jiangsu’s Agricultural Social Services Empowered by Digital Technology”, 2023 Jiangsu Provincial Social Science Fund Key Project (23EYA005), “Research on the Supply and Demand Matching ofJiangsu Digital Empowerment Agricultural Socialized Services Based on Rural Industry Revitalization”. Project Grantee: Dr. Zhang Yifeng.