Mitigating the inhibitory effect of information incompleteness on biopesticide adoption: The moderating role of social trust among smallholder rice farmers in China

1 Introduction

Excessive and inefficient chemical pesticide use drives substantial resource waste, with large volumes leaching into soil, water, and the atmosphere, causing a series of environmental problems (Sun et al., 2018; Aniah et al., 2021). Globally, the average pesticide application rate across croplands is 1.8 kg/ha [Food and Agriculture Organization of the United Nations (FAO), 2022], representing a 50% increase from the 1990s. This trend highlights persistent high usage intensity even amid regional efforts to optimize application efficiency [Food and Agriculture Organization of the United Nations (FAO), 2022]. Yet China's chemical pesticide application rate (2.13 kg/ha) remains notably higher than that of developed economies: 1.3 times the European Union (1.6 kg/ha) and 2.3 times the United States [Organisation for Economic Co-operation and Development (OECD), 2022]. To address this challenge, China has implemented policies that prioritize the scaled adoption of agricultural technologies focused on input reduction and efficiency enhancement, such as precision spraying and green pest control (China Ministry of Agriculture and Rural Development (MARD), 2023).

Biopesticides, as a core component of green pest management, can effectively control pests with minimal environmental impact. They degrade rapidly in ecosystems and present lower risks to non-target organisms (Abdollahzadeh et al., 2015; Constantine et al., 2020). Globally, biopesticides account for 5% of the total pesticide market, with penetration rates exceeding 20% in developed economies. This high penetration is largely driven by demand in the Americas, which accounts for 44% of global biopesticide use, and Europe, which accounts for 20% [China Statistics Bureau (CSB), 2022; Verified Market Reports, 2025]. China's biopesticide industry has experienced rapid expansion. By December 2023, the number of validly registered biopesticide products had exceeded 2,000, with an average annual growth rate of 7.55% over a 9-year period (Li et al., 2024). Unfortunately, a critical gap remains in China: biopesticide penetration is low, with fewer than 10% of farmers adopting this products (China Ministry of Agriculture and Rural Development (MARD), 2024). This discrepancy impedes both the sustainability of domestic agriculture and global efforts toward a green transition.

Existing research on the barriers to farmer adoption of biopesticides has primarily focused on two dimensions: external environmental constraints and farm-level internal factors. From an external perspective, three key challenges stand out. First, regarding market factors, incomplete market regulation, a clutter of brands, variable product quality, and relatively high prices collectively deter adoption (Shiferaw et al., 2015; Khan and Damalas, 2015; Guo and Zhao, 2017; Launio et al., 2020). Second, technical limitations further impede uptake, including slower efficacy onset, moderate pest-control efficacy, and narrow pest target ranges for management (Launio et al., 2020; Mkenda et al., 2020). Third, policy gaps exacerbate these barriers: insufficient technical training and limited extension services for biopesticide application fail to support farmer adoption (Sharafi et al., 2018; Wuepper et al., 2018; Yan et al., 2023b). From a farm-level internal perspective, individual constraints are dominant: farmers' cognitive biases toward biopesticides and limited proficiency in application directly constrain adoption decisions (Ngowi et al., 2007; Mkenda et al., 2020; Bakker et al., 2021; Aniah et al., 2021). This aligns with broader insights into smallholder decision-making under uncertainty, where limited knowledge of new agricultural inputs constrains adaptive choices (Mayorga et al., 2025). Ultimately, these external and internal constraints converge on a core issue: farmers' information incompleteness about biopesticides, which acts as the underlying driver of low adoption rates.

Farmers' information incompleteness about biopesticides primarily manifests in two interrelated challenges: the wide diversity of biopesticide products and the technical complexity of their application. For one, product diversity within biopesticide categories creates significant information barriers. Even under a single biopesticide label, the number of distinct formulations is substantial: statistics indicate 1,631 formulations of abamectin, 185 of Bacillus thuringiensis, and 114 of matrine (China Pesticide Information Network (CPIN), 2023). Critically, even for the same active ingredient, variations in dosage forms, application standards, and packaging labels further complicate farmers' ability to access comprehensive product information—ultimately hindering their willingness to adopt biopesticides. For another, the technical demands of biopesticide use exacerbate information gaps. Biopesticide efficacy against pests and diseases is highly sensitive to external environmental conditions, with performance tightly linked to both their pharmacological properties and proper application techniques (Tang and Luo, 2021; Huang et al., 2022). Many farmers lack proficiency in these specialized techniques, which directly undermines biopesticide effectiveness (Shiferaw et al., 2015). This poor performance, in turn, fosters the misperception that biopesticides are ineffective, further lowering adoption rates. Notably, information incompleteness is a well-documented barrier to agricultural technology adoption more broadly. Studies consistently show it leads farmers to abandon, reduce, or delay the uptake of new practices (e.g., biopesticides) (Wang and Liu, 1996; Zhang et al., 2012; Aniah et al., 2021; Xie et al., 2024).

Existing literature identifies three key interventions that improve farmers' access to biopesticide-related information: government publicity and guidance (Huang et al., 2021; Yan et al., 2023a), policy regulation (Lefebvre et al., 2015), and technical training (Wuepper et al., 2018). Yet this body of work overlooks a critical alternative channel: social trust derived from blood ties, kinship, geographic proximity, and industrial connections, which also serve as important sources of biopesticide information for farmers. Notably, individual economic activities are “embedded” in external interpersonal relationships (Granovetter, 1985), a framework aligning with analyses of rural agricultural decision-making in related studies. Among these contexts, social trust is the foundation of all social activities and a core component of social capital (Luhmann, 2000; Christian and Pierre-Guillaume, 2013). In traditional rural communities, the “private circles” shaped by social trust facilitate the sharing and dissemination of biopesticide information within local communities (Lin et al., 2019), which enhances information transparency and accessibility to a certain extent and mitigates the negative impacts of information constraints (Tao et al., 2022), directly promoting farmers' adoption of biopesticides. In addition, social trust further broadens individuals' information access, reduces information acquisition costs and alleviates information incompleteness (Ji and Zhang, 2021; Xie et al., 2024), all of which further drive the adoption of biopesticides by farmers.

Against this backdrop, this study draws on survey data from 992 rice farmers in Hubei Province, China, to examine the impact of information incompleteness on farmers' biopesticide adoption behavior from both theoretical and empirical lenses. It further investigates how the interaction between information incompleteness and social trust shapes such adoption behavior. The marginal contributions of this study are threefold. First, existing studies on farmers' biopesticides adoption have primarily focused on cognitive or perceptual factors, while overlooking both the extent of farmers' knowledge of biopesticides and the inhibitory role of information incompleteness in driving adoption decisions. Second, in discussions of trust's influence on farmers' behavioral decisions, existing research has typically examined trust through a single level or dimension—failing to fully address trust's role in biopesticide adoption, crucially, neglecting the key function of both institutional trust and interpersonal trust in the effects of information incompleteness on farmers' adoption of biopesticides. Third, this study specifically explores whether a synergistic effect exists between social trust and government promotion policies in the context of information incompleteness. It further analyzes the conditions under which social trust can fully reduce the inhibitory effect of information incompleteness on biopesticide adoption.

2 Theoretical framework and hypotheses

2.1 Impact of information incompleteness on farmers' adoption of biopesticide

Knowledge is a critical determinant of an individual's information processing, handling, and behavioral decision-making (Guo and Zhao, 2017; Aniah et al., 2021). For farmers to adopt a new technology, they must first be aware of its existence, understand its operational principles, and recognize its potential benefits (Wuepper et al., 2018). Thus, the extent of farmers' knowledge about biopesticides directly influences their adoption of such technologies.

On one hand, biopesticides are knowledge-intensive technologies that demand substantial skills and knowledge from farmers (Tang and Luo, 2021). In contexts of information incompleteness, farmers using biopesticides may encounter issues such as unscientific ratios of biopesticides to baits, mismatches between biopesticide selection and target pests, non-standard application procedures, and irrational operations (Fusar Poli et al., 2025). These problems can result in poor pest and disease control, or even outcomes that contradict expected control effects (Gao and Niu, 2019). In contrast, farmers with greater knowledge—equipped with more information on biopesticide control efficacy, application skills, and related details—are more likely to adopt biopesticides (Guo and Zhao, 2017; Mkenda et al., 2020).

On the other hand, subjective risk stemming from information incompleteness also exerts a significant influence on farmers' biopesticide adoption decisions (Bagheri et al., 2019; Ayat et al., 2020). Farmers often lack confidence in biopesticide efficacy and cannot accurately estimate the output levels or input costs associated with biopesticide (Tang and Luo, 2021). This can lead farmers to underestimate marginal output benefits or overestimate marginal costs, ultimately reducing biopesticide adoption rates (Su et al., 2022; Huang et al., 2022). Thus, uncertainty surrounding expected net returns may further influence farmers' adoption of biopesticide.

Based on this analysis, the following hypothesis is proposed:

H1. Information incompleteness has a negative effect on farmers' adoption of biopesticides.

2.2 Mechanisms of social trust in moderating the impact of information incompleteness on farmers' biopesticides adoption

Social trust serves as foundational social capital that underpins transactions and cooperative behaviors in agricultural contexts, including farmers' decisions to adopt new technologies (Hirsch, 1977; Yuan et al., 2022). For biopesticide adoption, information incompleteness frequently generates uncertainty regarding product efficacy, application techniques, and costs. In this context, social trust reflects the degree to which farmers are willing to rely on external advice or institutional support to mitigate such uncertainty (Luhmann, 2000; Xie et al., 2024). This aligns with frameworks emphasized in farm-level decision-making research, which highlight how social factors buffer smallholders against risks in technology adoption (Mayorga et al., 2025). As documented in existing literature, social trust can be generally divided into two interrelated dimensions: institutional trust and interpersonal trust. These two dimensions function through separate mechanisms to moderate the adverse impact of information incompleteness on farmers' biopesticide adoption behavior (Porta et al., 1997; Feng et al., 2016).

2.2.1 Institutional trust: Alleviating information gaps via formal support

Institutional trust refers to farmers' confidence in the credibility and competence of formal organizations, particularly those involved in agricultural policy, technology extension, and regulation (Tyler and Degoey, 1996; Norris, 2017). For biopesticide adoption, this trust primarily targets government departments and agricultural extension agencies actors that play a critical role in reducing information incompleteness, analogous to how formal regulations lower transaction costs for agrifood firms (Fiankor et al., 2025).

First, agricultural extension agents act as a key link between biopesticide technology developers and on-farm practice (Batoa et al., 2019). These agents integrate modern agricultural science, technical knowledge, and local agroecological insights to provide tailored guidance to farmers (Krishnan and Patnam, 2014). When farmers hold high trust in extension agents, they are more likely to engage in in-depth communication about biopesticide traits (e.g., target pests, application timelines), efficacy results, and troubleshooting (Wuepper et al., 2020). Such interactions directly reduce the degree of information incompleteness, helping farmers master technical operations that are critical for avoiding improper use and ensuring biopesticide effectiveness (Wang and Lu, 2015; Nyangau et al., 2020). Second, government policies and services further intervene to guide farmers' green production behavior (Ma et al., 2024). Formal initiatives including technical training programs, subsidy support and quality certification systems reduce farmers' information search costs for biopesticides and mitigate their risk perceptions of unproven efficacy and high input costs (Han et al., 2024; Fusar Poli et al., 2025). By standardizing biopesticide promotion processes and providing transparent information on costs and benefits, these policies improve the efficiency of technology diffusion (Mao et al., 2022; Han et al., 2024). Trust in government departments strengthens farmers' willingness to rely on such support, indirectly alleviating the risk of improper biopesticide use and reinforcing adoption intentions (Liu et al., 2016).

2.2.2 Interpersonal trust: Facilitating information sharing via peer networks

Interpersonal trust denotes farmers' general confidence in the reliability of peers, neighbors, and experienced growers (Simpson, 2007; Dinesen and Bekkers, 2017). This dimension of trust complements institutional trust by leveraging informal social interactions to fill information gaps, a pattern consistent with farm-level decision-making under uncertainty where social interactions often shape technology adoption choices (Chen et al., 2021; Mao et al., 2022).

First, neighborhood communication drives the spread of context-specific biopesticide information among village groups. Peers share practical knowledge including biopesticide product characteristics, optimal application methods, and dosage adjustments for local pest pressures, which may not be accessible through formal channels (Krishnan and Patnam, 2014; Fusar Poli et al., 2025). Higher interpersonal trust between neighbors reduces uncertainty about the credibility of this shared information (Zeng et al., 2022), lowering individual risk perceptions and encouraging farmers to align their biopesticide adoption decisions with peers who have demonstrated successful use. Second, large-scale farmers gain priority access to green technology adoption resources: demonstration base construction, early access to new biopesticides and technical support (Feng et al., 2016; Zheng et al., 2021; Deepak et al., 2022). When trust is high, knowledge spillovers of on-farm efficacy results, cost-saving practices and operational tips from these growers are more likely to be adopted by surrounding farmers (Gorman et al., 2019). Farmers who trust large-scale growers tend to actively learn their biopesticide use patterns, which reduces learning costs related to information incompleteness and enhances adoption enthusiasm (Yan et al., 2023b).

Both institutional and interpersonal trust operate by reducing uncertainty, lowering information search costs, and facilitating knowledge acquisition—all of which mitigate the negative impact of information incompleteness on farmers' biopesticide adoption. Institutional trust leverages formal support to address structural information gaps, while interpersonal trust uses informal channels to fill context-specific knowledge voids. Together, these two dimensions of social trust strengthen farmers' ability to navigate information incompleteness when considering biopesticide adoption.

Based on this reasoning, the following hypotheses are proposed:

H2. Social trust exerts a positive moderating effect, mitigating the negative impact of information incompleteness on farmers' biopesticide adoption.

H2a. Institutional trust exerts a positive moderating effect, mitigating the negative impact of information incompleteness on farmers' biopesticide adoption.

H2b. Interpersonal trust exerts a positive moderating effect, mitigating the negative impact of information incompleteness on farmers' biopesticide adoption.

3 Methodology and data sources

3.1 Data sources

The data for this study was collected from a micro-level farmer survey conducted by the research team in July 2023 across eight counties (cities) in Hubei Province. Hubei Province was selected as the survey area mainly based on the following three considerations: first, industry representativeness. Hubei Province is an important rice-producing region in China, with rice planting areas accounting for 7.60% and 7.68% of the national rice planting area in 2021 and 2022, respectively, both ranking fifth nationwide [National Bureau of Statistics of China (NBSC), 2022, 2023]. It is typical in the national rice production pattern and can provide samples with national significance for research. Second, problem focus. The phenomenon of unreasonable pesticide application in rice production is relatively common, and existing research has shown that the overuse of pesticides in rice production in China is as high as 57% (Zhang et al., 2015). As a major rice-producing province, Hubei Province's pesticide use behavior, pest and disease control methods, and adoption of biopesticides have strong practical relevance and policy reference value in rice-growing areas. Third, scientific sample coverage. The survey samples covered major agricultural ecological types in Hubei Province, including the Jianghan Plain, the central Hubei hills, and the northern Hubei uplands. The selected counties (cities) included rice-producing areas, green high-yield demonstration zones, and traditional agricultural areas, taking into account natural conditions, economic development levels, and cropping systems, which is conducive to enhancing the robustness of the empirical conclusions of the samples.

Sample selection followed a multi-stage random sampling protocol to ensure representativeness of Hubei's rice farming population. The process proceeded in three sequential steps: (1) County level sampling: eight counties were randomly selected from Hubei Province, including Xiangzhou, Nanzhang, Jianli, Shishou, Jingshan, Zhongxiang, Qianjiang, and Xiantao—covering diverse agroecological conditions and rice farming systems to avoid geographic bias. (2) Township and village sampling: within each selected county, four townships were randomly sampled; within each township, three villages were further randomly selected to capture local variations in farming practices and resource access. (3) Farmer level sampling: between 10 and 12 rice-growing households were randomly selected from each sampled village, focusing exclusively on households engaged in rice cultivation to align with the study's research focus. This sampling process initially yielded 1,047 completed questionnaires. To ensure data quality, responses were screened for validity: questionnaires with missing key information and outliers were excluded. After this screening, 992 valid samples were retained, resulting in a sample validity rate of 94.75%.

Three measures were implemented to safeguard data comprehensiveness, authenticity, and validity. First, experienced researchers were recruited for the survey, and all received professional training on questionnaire administration—including standardized approaches to asking technical questions (e.g., about biopesticide application methods) and minimizing interviewer bias. Second, a pre-survey was conducted with 50 rice farmers in a non-sampled county of Hubei; feedback from this pre-survey was used to refine questionnaire content and structure, ensuring the instrument was contextually appropriate for rural rice farmers. Third, all formal data collection relied on face-to-face interviews; with farmers' explicit consent, conversations were recorded to allow post-survey cross-verification of responses, reducing recall bias and ensuring accuracy for critical variables.

3.2 Variables setting

3.2.1 Dependent variable

The dependent variable in this study is farmers' adoption behavior of biopesticides. It is derived from the survey question: “Have you used biopesticides?” Following standard binary variable coding practices, the response is assigned a value of 1 if the farmer reported using biopesticides and 0 otherwise. To assess the robustness of the main findings, this study employs biopesticide use intensity as an alternative dependent variable. This continuous measure is defined as the share of biopesticide expenditure in farmers' total pesticide expenditure for rice production (Tang and Luo, 2021; Yan et al., 2023a). Unlike the binary adoption indicator, biopesticide use intensity quantifies the degree to which biopesticides are integrated into farmers' pest management strategies, capturing not just the presence of adoption but also its relative importance in overall pesticide input decisions—an approach that enhances the reliability of results by addressing potential limitations of binary measures in capturing nuanced input use patterns.

3.2.2 Independent variable

The independent variable of this study is information incompleteness, a core concept in information economics. It referring to the state where market participants lack comprehensive knowledge of the technical and environmental conditions related to a specific economic activity, and the market mechanism struggles to independently generate and efficiently allocate such critical information (Zhang et al., 2012). Combined with the technical attributes of biopesticides and rice farmers' production practices, this study specifically defines information incompleteness as gaps in farmers' mastery of biopesticide-related technical and application knowledge, following the research logic on agricultural technology information gaps from Ayat et al. (2020); Wang and Liu (1996).

Based on the technical process of biopesticide use and the decision-making needs of rice farmers, and referring to the indicator design approach of Wang and Liu (1996) and Ayat et al. (2020), this study measures the degree of information incompleteness from three complementary dimensions: experiential information, operational information, and control information.

3.2.2.1 Experiential information

Captures farmers' knowledge of contextual precautions for biopesticide application, specifically focusing on three key environmental and timing factors that influence efficacy: the appropriate temperature range for application, humidity requirements, and optimal timing of pest control when using biopesticides.

3.2.2.2 Operational information

Reflects farmers' understanding of biopesticide mixing and proportioning protocols, measured through three critical compatibility questions: whether biopesticides can be mixed with alkaline agents, whether they can be mixed with fungicides, and the maximum number of biopesticide types that can be co-applied safely.

3.2.2.3 Control information

Assesses farmers' knowledge of optimal control periods for key rice pests and diseases when using biopesticides. This dimension targets four major rice pests and pathogens: Chilo suppressalis, Scirpophaga incertulas, Brown rice planthopper and Rice blast.

The specific measurement of information incompleteness follows three steps. First, targeted survey questions are designed for each of the three dimensions (experiential information, operational information, and control information) to capture knowledge gaps in each dimension. Second, each question's responses are coded using a binary scale: a correct answer (indicating no knowledge gap in the corresponding area) is scored 1, while an incorrect answer (indicating a knowledge gap) is scored 0. Third, the scores of all questions for each farmer are summed to measure their degree of information incompleteness—a lower total score indicates larger knowledge gaps and a higher degree of information incompleteness, while a higher total score indicates more sufficient knowledge mastery and a lower degree of information incompleteness. The specific measurement methods are described below.

Mi=1-Ni10(1)

In Equation 1, M_*i* denotes the degree of information incompleteness of the ith farmer. N_*i* denotes the total score of the ith farmer who got the correct questions, and the degree of the farmer's score characterizes the degree of information mastery on biopesticides. If the value of N_*i* is smaller, it indicates a higher level of information incompleteness among farmers. Table 1 reports the measure of information incompleteness and descriptive statistical analysis.

Table 1. The measure of information incompleteness and descriptive statistical analysis.

3.2.3 Instrumental variable

Following the approach of Deng et al. (2019) and Gao and Feng (2025), this study selects the degree of incompleteness of information about neighbors in the same village as the instrumental variable. Specifically, it refers to the average degree of incompleteness information of other farmers in the village, excluding the farmer himself/herself. The rationality lies in that same-village neighbors share the same village-level production and information environment, with access to consistent technology promotion channels and resources, thus meeting the correlation requirement between the instrumental variable and the core independent variable. Their information status is influenced by village-level common factors, independent of individual farmers' adoption decisions and heterogeneous characteristics, which conforms to the exogeneity assumption and can effectively address the endogeneity issues in this study.

3.2.4 Moderating variable

3.2.4.1 Social trust

Following the measurement approaches of Porta et al. (1997) and Daskalopoulou (2019), social trust in this study is categorized into two distinct dimensions: institutional trust and interpersonal trust, with each dimension operationalized via a weighting approach.

3.2.4.2 Institutional trust

Institutional trust captures farmers' trust in formal rural support and governance entities (Fuglsang and Jagd, 2015; Ji and Zhang, 2021). It is characterized by two key indicators relevant to agricultural production contexts: (1) the degree of trust that farmers hold in staff of agricultural technology departments (who provide professional technical guidance on agricultural practices); (2) the degree of trust that farmers hold in village cadres (who coordinate local agricultural policy implementation and rural public affairs). To construct the institutional trust variable, we first collect farmers' trust scores for each of these two indicators and then compute the composite measure via a weighting approach.

3.2.4.3 Interpersonal trust

Interpersonal trust reflects farmers' trust in individuals within their local social networks (Lewicki et al., 2006; Chen et al., 2021; Tao et al., 2022). It is measured using two contextually appropriate indicators: (1) the degree of trust that farmers hold in large-scale planting households within the village (who often serve as peer models for agricultural technology adoption); (2) the degree of trust that farmers hold in their neighbors (who represent daily social interaction partners and informal information sources). Similar to institutional trust, interpersonal trust is operationalized by first gathering farmers' trust scores for each indicator (via the same consistent rating scale) and then aggregating these scores using a weighting approach.

3.2.5 Control variable

Following the studies of Aniah et al. (2021) and Yan et al. (2023b), this study selects control variables across five dimensions: household head characteristics, family characteristics, cropland characteristics, external environment, and village characteristics. Details are presented in Table 2.

Table 2. Definition of variables, mean, and standard.

^*1 hectare = 15 acres.

3.3 Model construction

3.3.1 Benchmark model construction

The explanatory variable in this paper is farmers' biopesticides adoption behavior, which is a binary discrete variable. Therefore, a binary Logit model is used to discuss the effect of information incompleteness on farmers' biopesticides adoption behavior. The specific model is as follows:

Yi=β0+β1Mi+βiXi+υi(2)

In Equation 2, Y_*i* denotes the ith farmer's biopesticide adoption behavior; M_*i* denotes the degree of information incompleteness of the ith farmer; X_*i* denotes the control variable, β₁ andβ_*i* are the regression coefficient of the explanatory variables; β₀ is the intercept of the regression; υ_*i* is the random error term.

3.3.2 Construction of endogeneity model

When estimating the impact of information incompleteness on farmers' adoption of biopesticides, it is necessary to focus on the estimation bias caused by endogeneity issues to ensure the reliability of causal inference in empirical results. The endogeneity issue in this study may have two aspects: first is the bidirectional causal relationship. Biopesticides are inherently characterized by slow efficacy, high technical application barriers, uncertain product quality and market returns, coupled with the initial state of information incompleteness, which elevates farmers' risk perceptions and trial-and-error costs. This prompts farmers to persist in chemical pesticide application, thereby lowering the probability of biopesticide adoption. Conversely, a persistently low adoption rate of biopesticides impedes the exertion of village-level demonstration effects and neighborhood learning externalities, leads to insufficient accumulation of practical experience, and dampens the investment enthusiasm of investors and promotion institutions. This further results in a shortage of key information including technical guidance, reliable data and market signals, thus exacerbating the state of information incompleteness. Estimation solely with the binary Logit model fails to disentangle the mutual interference of such bidirectional causality, giving rise to biased estimation results. Second is the omitted variable bias. Despite the selection of control variables covering farmers' individual characteristics, household characteristics, agricultural production attributes and external environmental factors, farmers' biopesticide adoption behavior is still affected by certain unobservable factors. These unmodeled unobservable factors may exert a simultaneous influence on both information incompleteness and biopesticide adoption behavior, leading to omitted variable bias and hindering the accurate identification of core causal relationships.

Against this backdrop, this study selects appropriate instrumental variables for the core explanatory variables, as the instrumental variable method can effectively mitigate the endogeneity issues in the analysis (Wooldridge, 2015). Considering that the information incompleteness of farmers' biopesticides has a notable neighborhood diffusion effect. If the degree of information incompleteness of other rural farmers is low, the degree of information incompleteness of that farmer will also decrease. Therefore, this study draws on the practices of scholars Deng et al. (2019) and Gao and Feng (2025) and selects “Village neighbors' information incompleteness” as the instrumental variable. The specific model construction is as follows:

Mi=μ0+μ1Zi+μ2Xi+υ1(3)

Yi=δ0+δ1Mi+δ2Zi+δ3Xi+υ2(4)

In Equations 3, 4, Z_*i* is an instrumental variable that represents the degree of information incompleteness about neighbors in the same village; μ₁, μ₂, δ₁, δ₂ and δ₃ are the regression coefficient of the explanatory variables; μ₀ and δ₀ are the intercept of the regression; υ₁ and υ₂ are the random error term. Other variables are the same as those in Equation 2.

3.3.3 Mechanism of action model construction

Based on Equation 2, the variable of social trust is introduced to explore the impact of information incompleteness and social trust on farmers' adoption behavior of biopesticides. The specific model is as follows:

Yi=α0+α1Mi+α2Ti+α3Mi×Ti+αiXi+εi(5)

In Equation 5, T_*i* denotes the degree of trust of the ith farmer, M_*i*×T_*i* denotes the interaction term between information incompleteness and social trust of the ithfarmer, X_*i* denotes the control variable, α₁,α₂, α₃ and α_*i* are the regression coefficient of the explanatory variables; α₀ is the intercept of the regression; ε_*i* is the random error term. Other variables are the same as those in Equation 2.

4 Results and discussion

4.1 Baseline regression results

4.1.1 Impact of information incompleteness on farmers' adoption of biopesticides

This study employed the chi-square test to examine the statistical association between farmers' information incompleteness and their biopesticide adoption behavior. Farmers were grouped by the mean value of information incompleteness: values above the mean were categorized as “high information incompleteness” and those below as “low information incompleteness.” The specific results are shown in Table 3. Pearson's correlation test between information incompleteness and farmers' biopesticide adoption was statistically significant at the 1% level. Specifically, farmers in the low information incompleteness group exhibited a higher biopesticide adoption rate (88.57%) than those in the high information incompleteness group. This result indicates that lower levels of information incompleteness among farmers are associated with a higher likelihood of biopesticide adoption, confirming that information incompleteness hinders farmers' adoption of biopesticides.

Table 3. Analysis of information incompleteness and adoption of biopesticides by farmers.

^*** represent significant at 1% statistical levels, respectively.

To address potential multicollinearity among variables, we conduct a multicollinearity test using variance inflation factors (VIF) as the diagnostic metric. Results reveal VIF values for all variables are below 2, confirming the absence of severe multicollinearity. Table 4 presents benchmark regression results for the relationship between information incompleteness and farmers' biopesticide adoption. Model 1 estimates only information incompleteness's impact without control variables. Model 2 extends it with individual, household, and farmland characteristics to refine the effect. Model 3 adds regional fixed effects and shows better goodness of fit than Models 1–2. Model 4 builds on Model 3 to estimate marginal effects of all variables. Given sequential refinement, Model 4's superior explanatory power and actionable marginal effects, subsequent analysis focuses on its results.

Table 4. Benchmark regression results.

^*, ^**, ^*** represent significant at 10%, 5%, and 1% statistical levels, respectively.

The empirical results indicate that information incompleteness exerts a significant negative impact on farmers' biopesticide adoption behavior. Specifically, a 1-unit increase in the degree of information incompleteness reduces the probability of farmers adopting biopesticides by 26.20%. A plausible explanation for this finding is that farmers have a low level of mastery over biopesticide technical information. This lack of knowledge creates difficulties in core operational processes related to biopesticides, including application, dosage proportioning, and mixing. Such operational challenges directly undermine the pest and disease control effectiveness of biopesticides, which in turn discourages farmers from adopting the technology (Mkenda et al., 2020). These results provide empirical support for Hypothesis H1.

Among control variables, farmers' part-time employment exhibits a significantly positive effect on biopesticide adoption. Farmers with part-time jobs likely have greater exposure to new information and stronger motivation to acquire knowledge, prompting them to proactively learn about and adopt biopesticide (Huang et al., 2021). Social network exerts a significantly positive effect on adoption. Early in biopesticide diffusion—when its advantages are not widely known—farmers whose social networks include more biopesticide adopters can learn the technology from existing users, accelerating adoption (Constantine et al., 2020; Xie et al., 2024). Household income has a significantly negative effect on adoption, meaning lower income is associated with a higher likelihood of biopesticide use. Plausibly, low-income households rely more on agricultural production and adopt low-toxicity, high-efficiency biopesticides to produce high-quality rice, aiming for higher expected returns (Zhang and Luo, 2020). Information access shows a significantly positive effect on adoption, as farmers' decision-making on new technologies heavily depends on external information inputs (Mkenda et al., 2020). Village economic development level exerts a significantly positive effect. Villages with higher economic development have greater physical and human capital for new technology adoption, which notably promotes the diffusion and application of biopesticide technology (Tang and Luo, 2022).

4.1.2 Endogeneity test results

Table 5 reports the instrumental variables estimation results for Model 1 and Model 2. The first stage regression results show that the degree of information incompleteness of village level neighbors has a significant positive impact on the degree of information incompleteness of farmers themselves. The higher the information incompleteness of neighbors in the same village, the greater that of the surveyed farmers. Furthermore, the first-stage F-statistic exceeds 10, and combined with the second-stage Wald test results, the problem of weak instrumental variables can be effectively ruled out, indicating that the instrumental variable selected in this study has good validity. The second-stage regression results reveal that information incompleteness has a significantly negative impact on farmers' biopesticide adoption behavior, meaning it significantly suppresses the probability of farmers adopting biopesticides. The above findings demonstrate that after effectively addressing endogeneity issues, the benchmark regression conclusion remains robust, and Hypothesis H1 is thereby validated.

Table 5. The results of endogeneity test and robustness test results.

^**, ^*** represent significant at 5%, 1% statistical levels, respectively.

4.1.3 Robustness test

This study employs the methods of replacing the independent variable and the dependent variable to test the robustness of the empirical results regarding the impact of information incompleteness on farmers' biopesticide adoption behavior. First, robustness testing is conducted by replacing the independent variable: “information incompleteness” is substituted with “information completeness,” and the empirical results are presented in Model 3 of Table 5. The results indicate that the degree of information completeness exerts a significantly positive impact on farmers' biopesticide adoption behavior. The more farmers understand biopesticides, the more inclined they are to adopt them. This finding indirectly confirms Hypothesis H1. Second, robustness testing is performed by replacing the dependent variable: “biopesticide adoption behavior” is replaced with “biopesticide adoption degree.” As shown in Model 4 of Table 5, information incompleteness has a significantly negative impact on farmers' biopesticide adoption degree at the 1% statistical level. This result is consistent with the benchmark empirical findings, further validating Hypothesis H1.

4.2 Mechanism of action testing

4.2.1 Information incompleteness, social trust, and farmers' biopesticide adoption behavior

To test the moderating effect of social trust on the relationship between information incompleteness and farmers' biopesticide adoption behavior, this study sequentially incorporated the interaction term of information incompleteness and social trust into the regression model.

Model 1 in Table 6 focuses on the combined impact of information incompleteness and institutional trust on farmers' biopesticide adoption. The results reveal that the interaction term “information incompleteness × Institutional trust” is significantly and positively associated with farmers' biopesticide adoption. Specifically, under conditions of information incompleteness, a higher level of institutional trust corresponds to greater enthusiasm among farmers for adopting biopesticides. This finding suggests that institutional trust has a positive moderating effect on the adoption of biopesticides in the presence of incomplete information. Model 2 in Table 6 examines the impact of information incompleteness and interpersonal trust on farmers' biopesticide adoption. The results show that the interaction term “information incompleteness × Interpersonal trust” exerts a significant positive effect on farmers' biopesticide adoption. This result indicates that interpersonal trust has a positive moderating effect on farmers' adoption behavior of biopesticides in terms of information incompleteness. In summary, social trust (encompassing both institutional trust and interpersonal trust) exerts a positive moderating effect on the relationship between information incompleteness and farmers' biopesticide adoption behavior. Based on these empirical results, Hypotheses H2, H2a, and H2b are all effectively validated.

Table 6. Mechanism test results.

^*, ^**, ^*** represent significant at 10%, 5%, and 1% statistical levels, respectively.

4.2.2 Robustness test

To assess the robustness of the benchmark regression results, this study conducted robustness checks by replacing the moderating variables and the dependent variables. The specific results are presented in Table 7.

Table 7. Robustness test.

^*, ^**, ^*** represent significance at the 10%, 5%, and 1% statistical levels, respectively. Standard errors are presented in parentheses.

4.2.2.1 Robustness test: Replacing moderating variables

This study measures institutional trust separately from the trust of agricultural technicians and village cadres, and measures interpersonal trust separately from the trust of large-scale growers and neighbors. Model 1 examines the interaction between information incompleteness and agricultural technicians trust. The results show this interaction term exerts a statistically significant positive impact on farmers' biopesticide adoption at the 1% level, indicating that the trust level of agricultural technicians has a positive moderating effect on the adoption rate of biopesticides due to incomplete information. Model 2 focuses on the interaction between information incompleteness and village cadres trust. This interaction term also exhibits a significant positive impact on biopesticide adoption, confirming that village cadres trust has a positive moderating effect on the adoption rate of biopesticides due to incomplete information. These findings further validate the positive moderating role of institutional trust in the relationship between information incompleteness and farmers' biopesticide adoption, consistent with the benchmark regression results.

Model 3 analyzes the interaction between information incompleteness and large growers' trust. The interaction term is significantly and positively associated with biopesticide adoption at the 10% level, suggesting that large growers' trust has a positive moderating effect on the adoption rate of biopesticides due to incomplete information. Model 4 investigates the interaction between information incompleteness and neighbor's trust. The results reveal this interaction term has a significant positive impact on adoption at the 1% level, indicating that neighbor's trust has a positive moderating effect on the adoption rate of biopesticides due to incomplete information. These results not only confirm the positive moderating role of interpersonal trust but also indirectly support the positive moderating effect of social trust in the relationship between information incompleteness and biopesticide adoption. In summary, these results are consistent with those presented earlier, validating hypotheses H2, H2a, and H2b.

4.2.2.2 Robustness test: Replacing dependent variables

A robustness check was conducted to validate the results by replacing the dependent variable. Specifically, the original dependent variable “biopesticide adoption behavior” is a binary indicator, and it was substituted with “biopesticide adoption intensity,” a continuous measure that reflects the extent of farmers' biopesticide use. Detailed results of check are reported in Table 7.

Model 5 focuses on the interaction between information incompleteness and institutional trust. The results show this interaction term exerts a statistically significant positive impact on farmers' biopesticide adoption intensity at the 1% level. Model 6 examines the interaction between information incompleteness and interpersonal trust, and the findings reveal this interaction term is also significantly and positively correlated with biopesticide adoption intensity. These results confirm that social trust and its two core dimensions (institutional trust and interpersonal trust) have a positive moderating effect on the impact of information incompleteness on farmers' adoption of biopesticides. This conclusion is highly consistent with the benchmark regression analysis results, providing additional empirical evidence to support hypotheses H2, H2a, and H2b.

4.2.3 Heterogeneity analysis

This study further discusses the heterogeneity of the impact of information incompleteness and social trust on farmers' adoption behavior of biopesticides across different levels of education, generations, cropland area, and government propaganda. The empirical results are presented in Tables 8, 9.

Table 8. Heterogeneity analysis (I).

^*, ^**, ^*** represent significance at the 10%, 5%, and 1% statistical levels, respectively. Standard errors are presented in parentheses.

Table 9. Heterogeneity analysis (II).

^*, ^**, ^*** represent significance at the 10%, 5%, and 1% statistical levels, respectively. Standard errors are presented in parentheses.

4.2.3.1 Heterogeneity analysis: By farmer education level

This study further explores the heterogeneous impacts of information incompleteness and social trust on farmers' biopesticide adoption behavior across different education levels. To conduct this analysis, farmers are divided into two groups based on their education attainment: the low-education group (defined as those with primary school education or below) and the high-education group (defined as those with junior high school education or above). Empirical results for this heterogeneity analysis are presented in Table 8.

Model 1 examines the interaction between information incompleteness and institutional trust across the two education groups. For the high-education group, the interaction term “information incompleteness × institutional trust” exerts a statistically significant positive impact on farmers' biopesticide adoption. In contrast, this interaction term is statistically insignificant for the low-education group. Model 2 focuses on the interaction between information incompleteness and interpersonal trust across the same groups. For the low-education group, the interaction term “information incompleteness × interpersonal trust” shows a significant positive impact on biopesticide adoption, while this interaction term is not statistically significant for the high-education group.

The research results indicate that the moderating effect of social trust on the impact of information incompleteness on farmers' adoption behavior of biopesticides varies significantly at different educational levels. For the high-education group, farmers tend to place greater trust in biopesticide-related information provided by government-related organizations. This trust in institutional sources enables them to effectively reduce the negative impact of information incompleteness on biopesticide adoption. For the low-education group, farmers are more likely to trust biopesticide information shared by local peer groups. This reliance on interpersonal trust helps them alleviate the inhibitory effect of information incompleteness on their adoption of biopesticides.

4.2.3.2 Heterogeneity analysis: By farmer generation

This study explores intergenerational heterogeneity in the impacts of information incompleteness and social trust on farmers' biopesticide adoption behavior. Following the elderly classification criteria proposed by Yan et al. (2023a), farmers are divided into two groups based on age: the elderly group (aged above the sample mean) and the youth group (aged below the sample mean). Empirical results for this intergenerational analysis are presented in Table 8.

Model 3 examines the interaction between information incompleteness and institutional trust across the two generational groups. For the youth group, the interaction term “information incompleteness × institutional trust” exerts a statistically significant positive impact on farmers' biopesticide adoption. In contrast, this interaction term is statistically insignificant for the elderly group. Model 4 focuses on the interaction between information incompleteness and interpersonal trust across the same groups. For the elderly group, the interaction term “information incompleteness × interpersonal trust” shows a significant positive impact on biopesticide adoption, while this interaction term is not statistically significant for the youth group.

The research results indicate that the moderating effect of social trust on the impact of information incompleteness on farmers' adoption of biopesticides varies significantly across different generations of farmers. Young farmers place greater trust in biopesticide-related information from official channels, indicating they depend more on institutional trust to bridge information gaps (Huang et al., 2022). For elderly farmers, their longer residence in villages has fostered closer and more harmonious relationships with neighbors. As a result, they tend to trust biopesticide information shared by relatives, friends, and neighbors, meaning they rely more on interpersonal trust to address the constraints posed by information incompleteness (Kasse et al., 2020).

4.2.3.3 Heterogeneity analysis: By cropland area

This study further investigates heterogeneous impacts of information incompleteness and social trust on farmers' biopesticide adoption behavior across groups with different cropland areas. Based on the sample mean of cropland area, farmers are divided into two groups: smallholder farmers (cropland area above the sample mean) and large-scale farmers (cropland area below the sample mean). Empirical results for this analysis are presented in Table 9.

Model 1 examines the interaction between information incompleteness and institutional trust across the two cropland area groups. The interaction term “information incompleteness × institutional trust” exerts a statistically significant positive impact on farmers' biopesticide adoption. This impact is significant in the large-scale farmer group but statistically insignificant in the smallholder farmer group. Model 2 focuses on the interaction between information incompleteness and interpersonal trust across the same groups. The interaction term “information incompleteness × interpersonal trust” also shows a significant positive impact on biopesticide adoption, but this effect is only statistically significant in the smallholder farmer group and insignificant in the large-scale farmer group.

The research results indicate that the moderating effect of social trust on the impact of incomplete information on farmers' adoption of biopesticides varies significantly across different cropland area. For large-scale farmers, institutional trust is more effective in compensating for the deficiencies caused by information incompleteness. This is likely because large-scale farmers typically have more frequent interactions with official agricultural entities, making them more willing to trust biopesticide information from institutional sources (Constantine et al., 2020). For smallholder farmers, interpersonal trust plays a more critical role in alleviating the negative impact of information incompleteness on biopesticide adoption. Smallholder farmers often rely on local social networks (including neighbors, fellow smallholders, and village acquaintances) for agricultural information exchange, and trust in these peer sources helps reduce information gaps related to biopesticides (Behzad and Guilherme, 2025).

4.2.3.4 Heterogeneity analysis: By government biopesticide propaganda

This study explores heterogeneous impacts of information incompleteness and social trust on farmers' biopesticide adoption behavior, grouped by whether farmers have access to government biopesticide propaganda. Here, “government propaganda” is defined as the presence of government-led policy initiatives or outreach activities that promote biopesticide knowledge and use. Grouped regression analyses were conducted based on this indicator, with empirical results presented in Table 9.

Model 3 and Model 4 examine the interaction effects of information incompleteness with institutional trust and interpersonal trust, respectively, across the two groups. For the propaganda group—farmers who have received government biopesticide propaganda—both the interaction term “information incompleteness × institutional trust” and the interaction term “information incompleteness × interpersonal trust” exert statistically significant positive impacts on farmers' biopesticide adoption. In contrast, neither of these interaction terms is statistically significant in the non-propaganda group—farmers without access to such government outreach.

These results indicate that in the absence of complete information, government promotion enhances the positive regulatory effect of social trust on farmers' adoption of biopesticides. Specifically, the synergistic effect of “government propaganda + social trust” effectively reduces the information gap faced by farmers. Government propaganda likely provides a foundational framework of credibility for biopesticide-related information, making farmers more receptive to trust-based information sources and thereby promoting their adoption of biopesticides (Adere et al., 2024).

4.3 Discussion

To explore whether a “threshold effect” exists between information incompleteness, social trust, and farmers' biopesticide adoption behavior at varying levels of social trust, this study uses Hansen's (2010) threshold regression model for cross-sectional data. The model aims to test for a “turning point” in the impact of information incompleteness and institutional trust (or interpersonal trust) on farmers' biopesticide adoption behavior. The basic form is as follows:

yi=θ1'xi+εi,qi≤γyi=θ2'xi+εi,qi>γ(6)

In Equations 6, 7, y_*i* and x_*i* represent the dependent variable and the independent variable, respectively. q_*i* is the threshold variable (institutional trust or interpersonal trust), and the sample is divided into two groups based on the corresponding threshold value γ; θ1' and θ2', respectively, represent the regression coefficients of two sets of samples; ε_*i* represents the random intervention group. Further define the dummy variable, d_*i*(y) = {q_*i* ≤ Y}, {·} as the indicator function, and let X_*i*(Y) = X_*i*d_*i*(y), then the threshold regression model is set as follows:

yi=θ1'+δn'xi(γ)+εi(7)

Finally, using the OLS model, estimate the corresponding threshold values and calculate Bootstrap using bootstrap method P-value, perform significance test on threshold regression model, specific steps can be found in Hansen (2010).

Table 10 presents the results of the threshold regression analysis. For Model 1, which uses institutional trust as the threshold variable, the sample is partitioned into two groups based on a threshold value of 3: one group with institutional trust ≤ 3 and another with institutional trust >3. In both groups, the interaction term “information incompleteness × institutional trust” exerts a positive impact on farmers' biopesticide adoption behavior, and this impact is statistically significant at the 1% level. Notably, the coefficient of this interaction term in the group with institutional trust >3 is significantly larger than that in the group with institutional trust ≤ 3. For Model 2, which takes interpersonal trust as the threshold variable, the sample is divided into two groups using a threshold of 3.5: one with interpersonal trust ≤ 3.5 and the other with interpersonal trust >3.5. In both groups, the interaction term “information incompleteness × interpersonal trust” shows no statistically significant impact on farmers' biopesticide adoption behavior. This finding implies that when farmers' institutional trust score exceeds three, the inhibitory effect of information incompleteness on biopesticide adoption is significantly mitigated. Specifically, farmers need to increase their trust in agricultural technology extension centers and village cadres to a score of 4 to effectively alleviate the negative impact of information incompleteness.

Table 10. Threshold regression results.

^**, ^*** represent significant at 5%, 1% statistical levels, respectively.

5 Conclusion

Using survey data from 992 rice growers in Hubei Province, this study investigates the impact of information incompleteness on farmers' biopesticide adoption behavior and further validates the mechanism through which information incompleteness and social trust jointly influence such adoption. The key empirical findings are as follows. First, information incompleteness exerts a statistically significant negative impact on farmers' biopesticide adoption: for each 1-unit increase in information incompleteness, farmers' propensity to adopt biopesticides decreases by 26.20%. Second, social trust has a positive moderating effect on the impact of incomplete information on the use of biopesticides; specifically, the interaction terms “information incompleteness × institutional trust” and “information incompleteness × interpersonal trust” both exhibit statistically significant positive moderating effects on farmers' biopesticide adoption. Third, heterogeneity analysis reveals that the interaction term “information incompleteness × institutional trust” exerts a stronger influence in the high-education group, youth group, large-scale households, and the group with access to government biopesticide propaganda; while the interaction term “information incompleteness × interpersonal trust” has a more pronounced effect in the low-education group, elderly group, small-scale farming households, and the group with government propaganda exposure. Fourth, when institutional trust scores exceed three, it effectively reduces the inhibitory effect of information incompleteness on farmers' biopesticide adoption.

Against this backdrop, targeted policy recommendations are proposed to advance the large-scale popularization and application of biopesticides. First, leverage the strengths of institutional actors including agricultural technicians, agricultural technology extension centers, village cadres and township officials. Through policy advocacy, government publicity, skill training and practical guidance, these entities can reduce the costs and risk perceptions of farmers in acquiring information on new agricultural technologies, alleviate the issue of information incompleteness, and thereby enhance farmers' cognitive level and practical application capabilities regarding biopesticides. Second, fully exploit the exemplary and leading role of large-scale farmers by providing targeted technical support and policy incentives to this group. Relying on the interpersonal trust networks in rural communities, the effective experience of green production technology application can be widely disseminated among smallholder farmers, so as to arouse their enthusiasm for biopesticide adoption. Third, implement differentiated promotion strategies based on farmers' heterogeneous characteristics: for highly educated, young and large-scale farmers, prioritize the cultivation and enhancement of institutional trust; for low-educated, elderly and smallholder farmers, focus on constructing localized interpersonal technical communication platforms to facilitate technology sharing. In addition, achieve full coverage of government publicity in rural areas and further strengthen the synergistic effect between government publicity and social trust, so that the two can complement and reinforce each other in promoting biopesticide adoption.

This study has several limitations that suggest promising future research directions. First, focusing on social trust's moderating role, the study only explores mitigating information incompleteness' negative impacts via trust cultivation, but lacks a balanced analysis of the trade-offs between direct information improvement measures and indirect trust-based regulatory interventions in cost-benefit, implementation efficiency and applicable scenarios. This may lead readers to a one-sided view that regulatory interventions are more advisable than direct investment in information improvement. Future research could build a comprehensive analytical framework integrating information improvement and trust regulation, empirically measure their trade-off relationship and optimal combination boundary, and explore the heterogeneous performance of this trade-off logic across different regions and planting scales, providing more refined theoretical support for biopesticide popularization. Second, findings' generalizability is limited by sample and crop scope. Based on 992 Hubei rice growers' cross-sectional data, the single-province sample may not apply to distinct agricultural regions, and excluding other crops (e.g., vegetables) leaves gaps in understanding unique information barriers and trust dynamics in green technology adoption.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Vodjo Nicodème Fassinou Hotegni, University of Abomey-Calavi, Benin

Reviewed by: Anthea Coggan, Commonwealth Scientific and Industrial Research Organisation (CSIRO), Australia

Xinjie Li, Northwest A&F University, China