Multiple pathways of institutional environment driving the reduction of veterinary antibiotic overuse: Evidence from dairy farmers in Heilongjiang Province, China

Abstract

Reducing veterinary antibiotic overuse is a critical strategy for protecting the ecological environment, ensuring food safety, safeguarding public health, and maintaining biosecurity. Currently, not only formal government institutions but also informal institutions have become critical in reducing the overuse of antibiotics by farmers; however, the literature on this topic remains limited. To advance research in this field, this study draws on micro-survey data from 141 dairy farmers with more than 100 cows in Heilongjiang Province, one of China's major dairy-producing regions. The institutional environment (IE) is conceptualized along three dimensions—regulative environment (RE), normative environment (NE), and cognitive environment (CE)—and the study employs Partial Least Squares Structural Equation Modeling (PLS-SEM) and Fuzzy-Set Qualitative Comparative Analysis (fsQCA) to investigate the mechanisms and pathways through which the IE and farmers' sense of responsibility (SR) influence the reduction of veterinary antibiotic overuse (RVAU). The results indicate that: (1) IE has a significant positive effect on RVAU by dairy farmers, and by enhancing their SR, it further promotes RVAU. (2) RE, NE, CE, and SR interact in multiple configurations, resulting in two key pathways that drive a high RVAU level: the “NE-SR” linkage type and the “RE-CE-SR” driving type. Among these, the “NE-SR” linkage type is slightly more applicable than the “RE-CE-SR” driving type, with high SR playing a crucial role in enhancing RVAU levels. There are two conditional configurations that lead to a non-high level of RVAU, namely the “NE-CE-SR” restriction type and the “RE-NE-SR” restriction type. (3) The pathways for improving RVAU vary across different farm scales. NE and SR generally impact the RVAU level across different farm scales, while RE has a significant impact on medium-scale farmers. Based on these findings, it is necessary to improve the control mechanism, optimize institutional coordination, innovate promotional strategies, and pay attention to scale differences to effectively promote farmers' reduction of antibiotic overuse.

1. Introduction

Veterinary antibiotics are a critical input in livestock production, playing a vital role in disease prevention, control, and yield improvement [1], [2]. However, improper use of veterinary antibiotics, such as excessive medication, failure to follow the withdrawal period, and use of prohibited antibiotics, has contributed to the rise of antimicrobial resistance (AMR) and a range of quality and safety issues in livestock and poultry products [3], [4]. AMR has been identified as one of the top ten global public health threats [5]. The extensive use of antibiotics in animals was blamed as a cause for the emergence and spread of antibiotic resistance genes [6]. AMR exerts its impact across an interconnected system, simultaneously compromising human, animal, and environmental health [7]. Because most antibiotics used in livestock farming are excreted unmetabolized, animal feces contain a large amount of antibiotics and drug-resistant bacteria, which ultimately enter the soil and water environment through various pathways, posing a serious threat to both human health and ecosystem security [2], [8]. Humans are exposed to antibiotic-resistant microorganisms primarily through contaminated food and drinks, direct contact with animals, or environmental exposure [9]. AMR affects not only humans but also animals, as resistant infections spread among livestock and poultry, leading to severe implications for veterinary medicine, as treatment failures in animals can lead to economic losses for the farmers and the food industry [10]. In addition, AMR compromises the treatability of common infections, contributing to elevated morbidity and mortality, which in turn exacerbates the burden on healthcare systems [11].

As one of the world's largest producers and consumers of antibiotics, China accounts for 52% of its antibiotic use toward livestock farming [12]. Farmer's overuse of livestock antibiotics, which exceeds the recommended dose standard of antibiotic guidelines or prescriptions, is remarkable in livestock and poultry production [13]. In response to the challenge of AMR caused by the abuse of antibiotics in animal husbandry, China has formulated a series of policies and regulations. Since 2018, the Ministry of Agriculture and Rural Affairs has implemented supervision, sampling, and risk monitoring programs for veterinary drugs. In 2021, the “National Action Plan for Reducing the Use of Antimicrobial Agents for Animal Use (2021-2025)” was released. In 2025, the No. 1 central document once again emphasized promoting the reduction of veterinary antibiotics use(RVAU). Nevertheless, the issue of antibiotic resistance has not been effectively resolved, and improper veterinary use still exists among some farmers [14].

Relying solely on formal government regulations has proven insufficient for effectively promoting RVAU. According to neoinstitutional theory, the institutional environment(IE) consists of the regulatory environment(RE), the normative environment(NE), and the cognitive environment(CE) [15]. The current government RE for RVAU has not achieved the expected results, mainly due to the neglect or underestimation of informal institutions such as the NE and CE. In rural areas where formal institutions are often inadequate, informal institutions may play a more crucial role in regulating farmers' behaviors [16], [17]. Farmers are the decision-makers in the use of veterinary antibiotics and one of the key actors in achieving veterinary antibiotic reduction targets. Their production decisions are deeply influenced by their institutional environment and tend to choose production methods that are compatible with it [18]. In reality, formal and informal institutions do not function independently; rather, they interact dynamically to influence and constrain farmers' behaviors at different levels [19]. Therefore, to achieve effective management of veterinary antibiotics, it is imperative to embrace the One Health framework and establish a composite institutional system that integrates and coordinates the RE, NE, and CE. By doing so, we can safeguard the health of humans, animals, and the environment at the source of livestock production.

Currently, scholars have conducted extensive research on the relationship between IE and the use behavior of veterinary antibiotics by farmers. Existing literature indicates that the impact of the institutional environment on the use behavior of veterinary antibiotics by farmers can be divided into direct and indirect pathways. Firstly, in terms of direct impact, regulatory policy tools have played a fundamental constraining role, such as the antibiotic prescription policy and withdrawal period policy, which can significantly reduce the excessive use of antibiotics by farmers [20]. The Chinese government does not directly subsidize farmers' standard use of antimicrobials [21]. However, indirect subsidies for compulsory immunization, genetic improvement of livestock, and the safe disposal of dead livestock can reduce disease incidence in livestock and, consequently, reduce antimicrobial overuse [22]. Furthermore, penalties imposed by the government, along with public awareness campaigns, are effective in reducing antibiotic use among farmers [23], [24]. The normative environment and the cognitive environment have also played an important role. The values of farmers have a significant impact on their rational use of veterinary antibiotics [25]. The farm veterinarian, the press, and peers were found to be the main sources of antibiotic information [26]. Among them, the peer effect had an inhibitory impact on the overuse of antibiotics by farmers involved in the food traceability system [27]. Secondly, in terms of indirect pathways, external interventions, such as norms and standards, and economic support, successfully improved antimicrobial use in animal husbandry by enhancing the veterinarian-farmer relationship, knowledge and perception, as well as motivation and confidence [28]. The food traceability system has a negative impact on the overuse of antibiotics by farmers through the mechanism of liability traceability for antibiotic residues, social reputation maintenance, and biosafety enhancement [27]. Policy advocacy promotes standardized use of antibiotics among meat duck farmers by increasing their public opinion pressure perception and moral responsibility [29]. The theory of planned behavior posits that external institutional factors translate into behavioral change via the mediation of internal psychological processes [30]. Insufficient SR among farmers will exacerbate the difficulty of addressing antibiotic resistance issues [31]. Therefore, in the process of IE affecting RVAU among farmers, their own SR also plays an important role. The interactions among RE, NE, CE, and SR form complex combinations that may be functionally equivalent, ultimately resulting in multiple pathways that promote RVAU.

Previous studies have laid a solid foundation for this research, yet several areas remain open for further exploration. Firstly, existing research on RVAU mainly focuses on livestock such as hogs and meat ducks [23], [29]. However, there is a notable lack of studies examining RVAU within the context of dairy farming. Secondly, existing research mainly focuses on analyzing only one dimension of RVAU, and their evaluation criteria are single. Thirdly, most prior studies have approached the influence of IE on the farmers' pro-environmental behaviors from a single dimensional perspective, failing to consider the combined effects of multidimensional institutional environments and their optimal pathways for promoting RVAU. In view of this, this study is based on field survey data from Heilongjiang Province, a major dairy industry province in China. It analyzes the impact of the IE in the three dimensions of “regulation-norm-cognition” on RVAU among dairy farmers, and further explores the effective path of the combination of different institutional factors and responsibility awareness driving RVAU from a configurational perspective. The marginal contribution of this article mainly lies in: firstly, in terms of research objects, this research focuses on dairy farmers. The dairy industry is a representative industry of food safety, and dairy products are an important component of a balanced diet. Their quality and safety are related to public health. Therefore, promoting dairy farmers to reduce the use of veterinary antibiotics is of significant practical importance. Secondly, in terms of research content, this article establishes an evaluation system for the level of dairy farmers' RVAU. Based on the reduction guidance principles in the National Action Plan for Reducing the Use of Antimicrobial Agents in Animals (2021–2025), the system covers five key aspects: breeding management, disease prevention and control, standardized and prudent antibiotic use, scientific antibiotic use, and application of substitutes. Thirdly, in terms of a research perspective, this article breaks the single-dimensional IE and examines the combined effects of formal RE, informal NE, and CE. It constructs a multidimensional “regulation–norm–cognition” analytical framework, identifies configurational pathways through which IE and farmers' SR jointly facilitate RVAU, and further analyzes distinct implementation pathways for farmers operating at different scales. The analytical framework used in the current study is shown in Fig. 1.

Fig. 1.

The analytical framework used in the study.

2. Materials and methods

2.1. Study sites, sampling, and participants

The data for this study were sourced from a micro-survey conducted by the research team in July and August 2024 across 10 cities in Heilongjiang Province, China. The research team conducted a small-scale pre-survey before the formal investigation and revised the questionnaire content based on feedback. The reasons for selecting Heilongjiang Province as the study area are as follows. First, Heilongjiang ranks among the top in China in terms of dairy cow inventory and raw milk output, and its dairy supply has a significant influence on the national dairy market and food safety. Second, the province features both large-scale modernized farms and small- to medium-scale household operations, providing a solid basis for comparing differences in the reduction of antibiotic overuse across farms of varying sizes. Finally, Heilongjiang has implemented veterinary antibiotic reduction initiatives, and its institutional environment for dairy farming is relatively well-developed. Therefore, choosing Heilongjiang Province as the study area is both representative and of broad applicability.

This survey targets dairy farmers in Heilongjiang Province who own 100 or more dairy cows. The spatial distribution of dairy farmers in Heilongjiang Province is uneven, and using simple random sampling may lead to sample dispersion in non-breeding areas and sparse breeding areas, resulting in insufficient sample representativeness. Therefore, the research adopted the principle of combining key sampling with random sampling. Firstly, based on data from the dairy industry association's big data platform and statistical information from regional agricultural and rural affairs departments, 10 cities with relatively developed dairy farming, including Harbin, Qiqihar, Mudanjiang, Jiamusi, Daqing, Jixi, Heihe, Hegang, Shuangyashan, and Suihua, were selected as the research areas. Secondly, based on the proportion of dairy farms with over 100 heads in each region to the total amount in the province, the basic sample quota for each city is determined. Considering that core breeding areas such as Daqing and Harbin have richer heterogeneity in terms of breeding scale and medication behavior, in order to ensure that the sample can comprehensively cover the production practices of different types of farmers, the sample size of these areas will be appropriately increased. Finally, random sampling will be conducted on eligible farmers in each survey city. Among them, Daqing had the largest number of sample dairy farmers, with 45, followed by Harbin(25), Qiqihar(21), and Jixi(18). According to the relevant data from the “China Animal Husbandry and Veterinary Yearbook (2023)”, there were 291 dairy farmers with a herd size of over 100 cows in Heilongjiang Province in 2022. A total of 150 questionnaires were collected during the survey, representing 51.55% of the large-scale dairy farmers in the province, thus providing strong explanatory power. After excluding invalid questionnaires lacking key data, 141 valid questionnaires were obtained, yielding an effective rate of 94%.

Table 1 presents the basic characteristics of the surveyed farmers. From the perspective of gender, the majority of respondents were male, comprising 73.76% of the sample. Regarding age, the majority were middle-aged, with a proportion of 68.09%. A larger proportion of farmers lived in urban areas, accounting for 52.48%. Farmers with a junior college or vocational college degree are the largest group, accounting for 27.66%. Regarding the duration of farming, most farmers have some farming experience, and the proportion of those with a farming duration of more than 5 years is 78.72%.In terms of the scale of farming, there are more large-scale farmers, accounting for 53.19%. The organizational degree of the farms is relatively high, with 65.25% of the farms operating in the form of cooperatives or companies.

Table 1.

Basic characteristics of interviewed dairy farmers.

Variable	Category	Sample size(household)	Percentage(%)	Variable	Category	Sample size(household)	Percentage(%)
Gender	Male	104	73.76	Degree of education	Junior high school and below	31	21.99
	Female	37	26.24		Senior high school or technical secondary school	36	25.53
Age	≤30	12	8.51		Junior college or vocational college	39	27.66
	31–40	53	37.59		Undergraduate and above	35	24.82
	41–50	43	30.50	Breeding years	≤5 years	30	21.28
	>50	33	23.40		6–10 years	39	27.66
Permanent residence	Town	74	52.48		11–15 years	32	22.70
	Village	67	47.52		>15 years	40	28.37
Management form	Cooperative	37	26.24	Breeding scale	100–499 heads	26	18.44
	Company	55	39.01		500–999 heads	40	28.37
	Individual operation	49	34.75		≥1000 heads	75	53.19

2.2. Variable selection

2.2.1. Explained variable

The explained variable was the reduction of veterinary antibiotic overuse(RVAU). Based on the National Action Plan for Reducing the Use of Veterinary Antibiotics (2021–2025) and the guidelines for reducing the overuse of veterinary antibiotics, this paper constructs an evaluation system for the level of RVAU by dairy farmers consisting of 5 primary indicators (breeding management, disease prevention and control, standardized antibiotic use, scientific and prudent antibiotic use, and application of substitutes), 11 secondary indicators, and a total of 16 questions, as shown in Table 2.

Table 2.

Scale for RVAU among dairy farmers.

Second-order latent variable	First-order latent variable	Title	Title assignment
Breeding management (BM)	Feeding model	BM1 Your farm has adopted a reasonable grouping management system	Strongly disagree = 1—strongly agree = 5
		BM2 The types of environmental monitoring practices implemented on your farm	1 type or below = 1, 2 types = 2, 3 types = 3, 4 types = 4, 5 or more types = 5
		BM3 Your farm maintains a comprehensive digital management system for cattle-related information	Strongly disagree = 1—strongly agree = 5
	Nutrition	BM4 Your farm applies a scientifically formulated feed and dietary regime	Strongly disagree = 1—strongly agree = 5
Disease prevention and control (DPC)	Epidemic prevention conditions	DPC1 The number of conditions available in your farm's quarantine area	1 or below = 1, 2 = 2, 3 = 3, 4 = 4, 5 or more =5
	Protective measures	DPC2 The degree of standardization of disinfection measures on your farm	Strongly not standardized = 1—strongly standardized = 5
Standardized antibiotics use (SAU)	Safe use	SAU1 There are no quality issues with the veterinary antibiotics being used.	Strongly disagree = 1—strongly agree = 5
	Regulations on the Use of Veterinary Antibiotics	SAU2 Your adherence to usage instructions when administering veterinary antibiotics	Never = 1, rarely = 2, occasionally = 3, often = 4, always = 5
		SAU3 Your compliance with withdrawal period regulations	Never = 1, rarely = 2, occasionally = 3, often = 4, always = 5
Scientific and prudent antibiotics use (SPAU)	Precise use of antibiotics	SPAU1 Your adherence to veterinary prescriptions for the use of prescription antibiotics	Never = 1, rarely = 2, occasionally = 3, often = 4, always = 5
	Cautious combination therapy	SPAU2 During the breeding process, a single antibiotic is used when effective, rather than multiple antibiotics simultaneously	Strongly disagree = 1—strongly agree = 5
	Graded and classified use of antibiotics	SPAU3 During the breeding process, general-grade antibiotics are used when sufficient, instead of higher-grade antibiotics	Strongly disagree = 1—strongly agree = 5
		SPAU4 During the breeding process, narrow-spectrum antibiotics are used when effective, instead of broad-spectrum antibiotics.	Strongly disagree = 1—strongly agree = 5
	“Reduce antibiotics use”	SPAU5 Your farm is increasing the use of precision antibiotic treatments for individual cows while reducing the use of group-based preventive antibiotics.	Strongly disagree = 1—strongly agree = 5
Application of substitutes (AS)	Green veterinary drugs replace antibiotics	AS1 The number of types of alternatives to antibiotics used on your farm	2 types or below = 1, 3–4 types = 2, 5–6 types = 3, 7–8 types = 4, 9 or more types = 5
		AS2 The proportion of antibiotic alternatives in relation to the total amount of antibiotics used on your farm	5% or below = 1, 5%–10%(including 10%) = 2, 10%–15%(including 15%) = 3, 15%–20%(including 20%) = 4, 20%or more = 5

Note: Some questions in the Animal Drug Reduction Level Scale are summarized from specific questions in the questionnaire.

The specific questions in the BMB2 questionnaire are: which of the following are included in the environmental monitoring of breeding farms, including temperature and humidity monitoring, carbon dioxide concentration monitoring, noise monitoring, ammonia monitoring, hydrogen sulfide monitoring, pressure monitoring inside and outside the breeding community, light intensity monitoring, wind speed and direction monitoring? Formulate scale questions based on the species selected by the farmers.

The specific questions in the DPB1 questionnaire are: What are the isolation walls, green buffer zones, entrance warning signs, environmental monitoring, and disinfection facilities (or equipment) in the isolation area of the breeding farm？

The specific questions in the DPB2 questionnaire are: the frequency of disinfection of the surrounding environment, factory area, and roads by the breeding farm; The disinfection frequency of feeding tools, feeding troughs, and feed beds in the breeding farm; The disinfection frequency of daily utensils (veterinary utensils, midwifery utensils, breeding utensils, milking equipment, milk tanks, etc.) in the breeding farm; The disinfection frequency of drinking water in aquaculture farms is considered very non-standard if all disinfection measures do not meet the standards. If one standard is met, it is considered relatively non-standard; if two standards are met, it is considered average; if three standards are met, it is considered relatively standardized; and if all four standards are met, it is considered very standardized.

The specific questions in the ASB1 questionnaire are: Which probiotics do farms add to their feed, such as yeast, lactic acid bacteria, and Bacillus; What enzyme preparations should be added to the feed in the breeding farm: amylase, lysozyme, lipase, cellulase, xylanase, phytase, pectinase, protease; Which of the following acidifiers are added to feed in aquaculture farms: formic acid, acetic acid, citric acid, fumaric acid, Corydalis yanhusuo, phosphoric acid; Which of the following antimicrobial peptides are added to feed in aquaculture farms: fluoxetine ZY4; Which of the following traditional Chinese medicines are added to animal feed in breeding farms: pain relieving and anti-inflammatory drugs (extract of Paeonia lactiflora saponins, honeysuckle, purple clover, astragalus polysaccharide, double Ding injection, Chaihu injection, Houttuynia cordata injection), diarrhea treating drugs (astragalus, cortex, three needles, yellow cypress, Huangqin, Duxie Kang), growth promoting drugs (Jianqu, hawthorn appetizing powder, white peony, earthworm powder, sesame leaves, seabuckthorn), deworming drugs (Xianhe grass root buds, Bai Bu, Lei Wan, pomegranate peel, betel nut, pumpkin seeds), immune enhancing drugs (Codonopsis pilosula, Epimedium, Chuanxinlian, Yunzhi polysaccharide, shiitake mushroom polysaccharide, goji berry polysaccharide). Form scale questions based on the species selected by the farmers.

2.2.2. Core explanatory variable

The core explanatory variable was institutional environment (IE). The measurement of the IE is based on Scott's New Institutionalism theory, including three primary indicators: regulative environment (RE), normative environment (NE), and cognitive environment (CE), with 8 secondary indicators and a total of 21 questions. Among them, the RE refers to the scale proposed by Lu et al. [32], which is divided into incentive regulation, guiding regulation, and constraining regulation, reflecting the mandatory institutional factors that can promote or restrict the behavior of farmers; the NE refers to the scales proposed by Wang et al. [33]and Wang et al. [34], which are divided into values, social norms, and morality, reflecting the institutional factors that are socially accepted and considered natural; the CE refers to the scale proposed by Chen [35], which is divided into common understanding and behavioral template, reflecting farmers' understanding of the external environment and their perception of the behaviors of other members in the social network, as shown in Table 3.

Table 3.

Institutional environment scale.

Second-order latent variable	First-order latent variable	Title	Title assignment
Regulatory environment (RE)	Incentive regulation	IR1 The extent of government subsidies provided to farmers implementing RVAU	Strongly small = 1—strongly large = 5
		IR2 The extent of government recognition and honorary awards granted to farmers implementing RVAU	Strongly small = 1—strongly large = 5
	Guiding regulation	GR1 The diversity of channels used by the government to disseminate information on RVAU	2 or below = 1, 3–4 = 2, 5–6 = 3, 7–8 = 4, 9 or more = 5
		GR2 The frequency of government-provided training and technical guidance on RVAU	1 time or less = 1, 2 times = 2, 3 times = 3, 4 times = 4, 5 times or more = 5
		GR3 The number of RVAU-compliant livestock demonstration farms established by the government in your county	1 or less = 1, 2 = 2, 3 = 3, 4 times = 4, 5 times or more = 5
	Constrained regulation	CR1 The frequency of government monitoring of antibiotic residues and antibiotics resistance	Little monitoring = 1, once a month = 2, once a quarter = 3, once every half year = 4, once a year = 5
		CR2 The frequency of government inspections of livestock farms' medication records	Rarely checked = 1, once a month = 2, once a quarter = 3, once every half year = 4, once a year = 5
		CR3 The strength of the regulatory effects of government-formulated veterinary drug laws and regulations on your antibiotic use	Strongly small = 1—strongly large = 5
		CR4 The government's enforcement of regulations against the misuse of antibiotics	Strongly small = 1—strongly large = 5
		CR5 The severity of government penalties for improper antibiotic use	Strongly small = 1—strongly large = 5
Normative environment (NE)	Values	VN1 Reducing antibiotic use contributes to mitigating antimicrobial resistance	Strongly disagree = 1—strongly agree = 5
		VN2 Reducing antibiotic use decreases the level of antibiotic residues in milk	Strongly disagree = 1—strongly agree = 5
		VN3 Reducing antibiotic use helps mitigate environmental pollution	Strongly disagree = 1—strongly agree = 5
	Social norm	SN1 The strength of the influence of industry association requirements on your antibiotic use	Strongly small = 1—strongly large = 5
		SN2 The strength of the influence of industry association production requirements on your antibiotic use	Strongly small = 1—strongly large = 5
	Morality	MN1 Reducing the dosage of antibiotics is in line with social moral standards.	Strongly disagree = 1—strongly agree = 5
Cognitive environment (CE)	Common understanding	UC1 The degree to which your peers are concerned with reducing antibiotic dosage	Strongly not concerned = 1—strongly concerned = 5
		UC2 The degree to which your friends and relatives are concerned with reducing antibiotic dosage	Strongly not concerned = 1—strongly concerned = 5
	Behavior template	BC1 The majority of the farms in your vicinity or that you are familiar with generally adopt RVAU	Strongly disagree = 1—strongly agree = 5
		BC2 Adopting RVAU can gain recognition from peers	Strongly disagree = 1—strongly agree = 5
		BC3 Adopting RVAU can gain recognition from friends and relatives	Strongly disagree = 1—strongly agree = 5

2.2.3. Mediating variable

The mediating variable was sense of responsibility (SR). Regarding the measurement of the SR, we refer to the scale proposed by Yu et al. [36] It measures the sense of responsibility of farmers from 3 primary indicators (responsibility cognition, responsibility attribution, and responsibility assumption), 5 secondary indicators, and a total of 5 questions, as shown in Table 4.

Table 4.

Responsibility awareness scale.

Second-order latent variable	First-order latent variable	Title	Title assignment
Sense of responsibility (SR)	Responsibility cognition	SR1 RVAU is an important concern for you.	Strongly disagree = 1—strongly agree = 5
	Responsibility attribution	SR2 You consider yourself responsible for RVAU
	Responsibility assumption	SR3 You are willing to devote time and effort to RVAU.
		SR4 You are willing to comply with the requirements of RVAU.
		SR5 You would voluntarily suggest to others to reduce the dosage of veterinary antibiotics.

This paper conducts a descriptive statistical analysis on 141 sample data. Among them, the six latent variables, including RE, NE, CE, SR, and RVAU, together with 42 measurable variables, are analyzed. The specific results of the mean and standard deviation of the measurable variables are presented in Table 5.

Table 5.

Descriptive statistics of each variable.

Latent variable	Item code	Mean value	Standard deviation	Latent variable	Item code	Mean value	Standard deviation
Regulatory environment (RE)	IR1	2.652	0.652	Sense of responsibility (SR)	SR1	4.099	0.861
	IR2	2.851	0.704		SR2	4.128	0.815
	GR1	2.965	0.846		SR3	4.106	0.805
	GR2	2.922	0.685		SR4	4.333	0.839
	GR3	2.972	0.780		SR5	4.057	0.913
	CR1	3.028	0.833	Reduction of veterinary antibiotic use (RVAU)	BM1	4.241	0.807
	CR2	2.993	0.700		BM2	3.277	0.835
	CR3	3.511	0.579		BM3	4.220	0.876
	CR4	3.496	0.514		BM4	4.213	0.770
	CR5	3.468	0.578		DPC1	3.355	0.706
Normative environment (NE)	VN1	4.177	0.917		DPC2	4.142	0.768
	VN2	4.014	0.960		SAU1	4.234	0.701
	VN3	4.142	0.880		SAU2	4.099	0.727
	SN1	4.035	0.846		SAU3	3.957	0.807
	SN2	4.504	0.813		SPAU1	3.993	0.767
	MN1	4.248	0.818		SPAU2	4.241	0.780
Cognitive environment (CE)	UC1	3.504	0.759		SPAU3	4.213	0.788
	UC2	3.475	0.749		SPAU4	4.128	0.858
	BC1	3.773	0.688		SPAU5	4.149	0.833
	BC2	3.582	0.764		AS1	3.184	0.912
	BC3	3.489	0.768		AS2	3.376	0.855

2.3. Model selection

2.3.1. Partial least squares structural equation model (PLS-SEM)

PLS-SEM can effectively address the issue of collinearity among variables [37]. Therefore, in this study, PLS-SEM was employed to analyze the direct influence and mechanism of different dimensions of IE and SR on RVAU of dairy farmers.

$$\eta =\mathit{\beta \eta }+\mathrm{\Gamma }\xi +\zeta$$ (1)

$$Y={\mathrm{\Lambda }}_y\eta +\epsilon$$ (2)

$$X={\mathrm{\Lambda }}_x\xi +\delta$$ (3)

In the formula:$\xi$represents IE；$\eta$represents RVAU；$\beta$and $\mathrm{\Gamma }$are the path coefficients between latent variables；$\zeta$is the residual；$X$and $Y$are the observed variables of the latent variables of IE and RVAU,${\mathrm{\Lambda }}_x$and ${\mathrm{\Lambda }}_y$are the coefficient matrices between IE $\xi$and RVAU $\eta$,$\delta$and $\epsilon$are the error terms.

2.3.2. Fuzzy sets qualitative comparative analysis (fsQCA)

Due to the limitations of the PLS-SEM method in addressing complex causal relationships, fsQCA was employed to further investigate the interaction between variables and explore the configuration paths that promote RVAU. fsQCA is a qualitative comparative analysis (QCA) method that uncovers asymmetric, complex causal relationships between antecedent variables and outcomes through the construction of fuzzy set data [38]. The fsQCA method can avoid endogeneity issues such as reverse causality, omitted variable bias, and sample selection bias from the source [39]. In the QCA analysis process, consistency and coverage are used to assess the validity of the results. A consistency value greater than 0.75 indicates that the condition variables or configurations are sufficient conditions for the outcome. Coverage measures the extent to which the set of relationships that passed consistency tests explains the results.

3. Result

3.1. Common method bias test, reliability test, and validity test

In order to further verify the reliability of the data, this study conducted a common method bias test using Harman's single factor test method. The exploratory factor analysis results showed that the first single factor explained 31.30% of the total variance, which was below the critical value of 40%, indicating that there was no significant common method bias in the variables.

This study employed the KMO value and Bartlett's sphericity test to assess the suitability of the data for factor analysis. The results showed a KMO value of 0.919, exceeding the threshold of 0.7, indicating that the data were suitable for factor analysis. The probability P value of Bartlett's sphericity test statistic was 0.000, which was less than the 0.05 significance level, further confirming the appropriateness of the selected variables for factor analysis. Reliability was assessed using Cronbach's α coefficient and composite reliability (CR). The results indicated that both the Cronbach's α coefficient and CR for all variables were greater than 0.6, demonstrating good internal consistency and high reliability of the questionnaire data. Validity analysis includes two aspects: convergent validity and discriminant validity. Convergent validity was assessed based on factor loadings and average variance extracted (AVE). The results revealed that the factor loadings of all observed variables exceeded the 0.6 threshold, and the AVE values for each latent variable were above the 0.5 criterion, indicating strong convergent validity (See Table 6). Discriminant validity was tested using the Fornell-Larcker criterion, based on the fact that the square root of the AVE of latent variables is greater than the correlation coefficient between latent variables. As shown in Table 7, the square roots of the AVE for the 5 latent variables in this study were greater than the correlation coefficients between latent variables [40], indicating that the measurement model exhibited good discriminant validity.

Table 6.

Reliability and validity test results.

Latent variable	Observed variable	Factor loading	AVE	CR	Cronbach's α
RE	IR1	0.658	0.530	0.907	0.901
	IR2	0.643
	GR1	0.766
	GR2	0.694
	GR3	0.733
	CR1	0.703
	CR2	0.697
	CR3	0.819
	CR4	0.805
	CR5	0.738
NE	VN1	0.857	0.581	0.874	0.853
	VN2	0.743
	VN3	0.864
	SN1	0.659
	SN2	0.616
	MN1	0.801
CE	UC1	0.665	0.569	0.822	0.810
	UC2	0.747
	BC1	0.766
	BC2	0.851
	BC3	0.731
SR	SR1	0.710	0.627	0.857	0.850
	SR2	0.873
	SR3	0.819
	SR4	0.754
	SR5	0.793
RVAU	BM1	0.877	0.549	0.954	0.944
	BM2	0.659
	BM3	0.802
	BM4	0.820
	DPC1	0.606
	DPC2	0.650
	SAU1	0.789
	SAU2	0.629
	SAU3	0.625
	SPAU1	0.637
	SPAU2	0.837
	SPAU3	0.873
	SPAU4	0.817
	SPAU5	0.831
	AS1	0.608
	AS2	0.693

Table 7.

Correlation coefficient of latent factor AVE value and correlation coefficient between factors.

Latent variable	RVAU	RE	NE	CE	SR
RVAU	0.741
RE	0.568	0.728
NE	0.673	0.461	0.762
CE	0.659	0.470	0.573	0.754
SR	0.736	0.583	0.632	0.679	0.792

3.2. PLS-SEM analysis

In this study, the goodness-of-fit (GOF) value was calculated as 0.822, indicating that the model exhibits a good fit [41]. The R² values for SR and RVAU were 0.570 and 0.613, respectively, suggesting that the model has a moderate or above explanatory power. Therefore, the structural model of this study can be considered relatively robust. The results of the hypothesis tests are presented in Table 8.

Table 8.

Path coefficient and significance level of structural model.

Path	Path coefficient	Standard deviation	t	P
(1) RE → RVAU	0.150	0.068	2.208	0.014**
(2) NE → RVAU	0.274	0.137	2.006	0.022**
(3) CE → RVAU	0.202	0.104	1.950	0.026**
(4) RE → SR	0.268	0.067	4.026	0.000***
(5) NE → SR	0.286	0.105	2.721	0.003**
(6) CE → SR	0.389	0.081	4.824	0.000***
(7) SR → RVAU	0.338	0.140	2.419	0.008**

Note: *, ** and *** indicate significant impacts at 10%, 5% and 1% levels respectively.

(1) Based on the PLS-SEM model analysis, lines (1)–(3) show that the RVAU of dairy farmers is affected by the external institutional environment. Line (1) represents the effect of RE on RVAU. According to the estimation results, RE (β = 0.155, P = 0.039) has a significant positive impact on RVAU. Line (2) shows the effect of NE on RVAU. The results indicate that NE (β = 0.273, P = 0.030) has a positive impact on RVAU. Line (3) represents the effect of CE on RVAU. The result indicates that CE (β = 0.223, P = 0.031) has a significant positive impact on RVAU.

(2) Lines (4)–(6) show the influence of IE on the SR. Line (4) shows the effect of RE on SR. The RE (β = 0.288, P = 0.000) has a significant positive impact on the SR of dairy farmers. Line (5) represents the effect of NE on SR. The results show that NE (β = 0.229, P = 0.033) has a significant positive impact on SR, meaning SR increases with the increase in NE. Line (6) shows the effect of CE on SR. According to the estimation results, the CE (β = 0.408, P = 0.000) has a significant positive impact on SR.

(3) As can be seen from column (7) of Table 8, the effect of SR on RVAU is positively significant, indicating that farmers with a stronger SR demonstrate greater self-discipline and are more likely to actively implement RVAU practices in their production process. Thus, farmers' SR plays a mediating role in the process through which the RE, NE, and CE influence RVAU. The robustness of the mediating effect of SR was tested using the Bootstrap method, and the results are presented in Table 9. The results indicated that SR played a mediating role in the relationships among RE, NE, CE, and RVAU of farmers. This suggests that the IE promotes RVAU of farmers through the cultivation of SR.

Table 9.

The intermediary effect test of sense of responsibility.

Path	Effect value	Standard deviation	P	95% confidence interval
RE → SR → RVB	0.091	0.042	0.015**	[0.031, 0.165]
NE → SR → RVB	0.097	0.056	0.044**	[0.023, 0.207]
CE → SR → RVB	0.131	0.066	0.024**	[0.043, 0.259]

Note: *, ** and *** indicate significant impacts at 10%, 5% and 1% levels respectively.

3.3. fsQCA analysis

3.3.1. Analysis of the necessity of individual conditions

Based on the actual situation of the questionnaire, this article calibrated the data according to three anchors: complete subordination (0.95), intersection point (0.5), and complete non-subordination (0.05). Following this calibration, a necessity analysis was conducted. A condition is considered necessary for the outcome if its consistency with the result exceeds 0.9 [42]. The results, as shown in Table 10, indicate that no condition variables have a consistency greater than 0.9. This suggests that the level of RVAU is not determined by a single factor, but rather by a combination of conditions that together form a configuration path.

Table 10.

Necessary condition analysis.

Antecedent variable	High level of RVAU		Non high level of RVAU
	Consistency	Coverage	Consistency	Coverage
RE	0.723	0.795	0.451	0.461
∼RE	0.510	0.450	0.799	0.729
NE	0.811	0.830	0.519	0.494
∼NE	0.506	0.530	0.822	0.802
CE	0.757	0.765	0.554	0.521
∼CE	0.526	0.560	0.751	0.742
SR	0.825	0.796	0.456	0.410
∼SR	0.388	0.434	0.773	0.804

3.3.2. Analysis of the sufficiency of the conditional configuration

The consistency and coverage of the calibrated values were analyzed. The consistency threshold, PRI threshold, and frequency threshold are set as 0.80, 0.70, and 1, respectively, resulting in complex solutions, intermediate solutions, and simplified solutions. The intermediate solution was used as the primary reference, with the nested relationship between the intermediate and simplified solutions serving as an auxiliary basis for comparison. Two antecedent configurations were identified as promoting a high level of RVAU, and two antecedent configurations were found to trigger a non-high level of RVAU, as shown in Table 11. The overall consistency for both the high level of RVAU (0.870) and the non-high level of RVAU (0.926) exceeds 0.8, suggesting that the identified configurations are sufficient conditions for both outcomes. Additionally, the overall coverage for both levels of RVAU is 0.736 and 0.662, respectively, both above the 0.2 threshold, indicating that the configurations explain a substantial portion of the variance in RVAU levels. Based on the configuration analysis, two configurations promoting a high level of RVAU (P1, P2) were identified: P1, labeled the “NE-SR” linkage type, and P2, the “RE-CE-SR” driving type. Two configurations triggering a non-high level of RVAU (L1, L2) were also found: L1, termed the “NE-CE-SR” restrictive type, and L2, the “RE-NE-SR” restrictive type.

Table 11.

Results of sufficiency configuration analysis.

Conditional variable	High level of RVAU		Non high level of RVAU
	P1	P2	L1	L2
RE		●		⊗
NE	●		⊗	⊗
CE		●	⊗
SR	●	●	⊗	⊗
Consistency	0.891	0.912	0.937	0.932
Original coverage	0.702	0.560	0.603	0.601
Unique coverage	0.176	0.033	0.061	0.059
Overall consistency	0.870		0.926
Overall coverage	0.736		0.662

Note: ①_⊗ represents a low level of antecedent conditions, ● represents a high level of antecedent conditions. ②The large circle represents the core condition, the small circle represents the edge condition, and the space represents that the antecedent condition is optional for the occurrence of the result.

The configuration P1 indicates that a high NE and a high SR can lead to a high level of RVAU. This configuration accounts for 70.2% of cases, slightly surpassing configuration P2 in terms of its universality. After excluding common elements shared with other configurations, the proportion of cases where a single configuration leads to a high level of RVAU is 17.6%. The consistency of this configuration is 0.891, indicating that 89.1% of the farmers who meet the conditions of this configuration have achieved a high level of RVAU.

The configuration P2 indicates that a high RE, CE, and SR can lead to a high level of RVAU. This configuration accounts for 56.0% of the cases. After excluding shared components with other configurations, the proportion of cases in which this single configuration leads to a high level of RVAU is 3.3%. The consistency of this configuration is 0.912, meaning that 91.2% of farmers meeting the conditions of this configuration have achieved a high level of RVAU.

In the configuration for the non-high level of RVAU, configuration L1 indicates that the simultaneous absence of NE, CE, and SR will result in a non-high level of RVAU. Similarly, configuration L2 suggests that the simultaneous absence of RE, NE, and SR will also lead to a non-high level of RVAU.

An analysis of the distribution of each element in the condition configurations reveals that SR is crucial for achieving a high level of RVAU. Both pathways for the high level of RVAU identify SR as the core condition. Likewise, in the configurations for the non-high level of RVAU, both pathways are characterized by low SR as the core condition.

3.3.3. Robustness test

This study adopts the robustness test method proposed by Du et al. [43] Firstly, the consistency threshold is raised from “0.8” to “0.85”, and the resulting configuration paths are basically consistent. Secondly, the case frequency threshold is increased from “1” to “2”, and the resulting configuration is also basically consistent, indicating that the research conclusion is robust.

3.3.4. Conditional configurations of the high level of RVAU among different-sized farmers

Due to significant variations in resource endowments, value orientations, and other factors across dairy farmers of different sizes, the scale characteristics of the farmers may serve as important moderating factors influencing RVAU under the IE and SR. This, in turn, may lead to distinct patterns and effects of RVAU among farmers of different sizes. Based on existing research [44], this study categorizes the sample into small-scale (with a herd size of 100–499 heads), medium-scale (with a herd size of 500–999 heads), and large-scale (with a herd size of more than 999 heads), and investigates the impact of IE and SR on the high levels of RVAU across these groups. The results are presented in Table 12.

Table 12.

Configuration path of high veterinary drug reduction behavior levels among farmers of different scales.

Conditional variable	Small scale		Medium scale			Large scale
	G1	G2	G3	G4	G5	G6
RE	⊗	•	●	●
NE	●	●	●		●	●
CE		●		•	•
SR	●		●	●	●	●
Consistency	0.962	0.937	0.943	0.972	0.895	0.858
Original coverage	0.406	0.637	0.590	0.564	0.616	0.705
Unique coverage	0.052	0.283	0.059	0.033	0.085	0.705
Overall consistency	0.932		0.884			0.858
Overall coverage	0.689		0.708			0.705

Note: ①_⊗ represents a low level of antecedent conditions, ● represents a high level of antecedent conditions. ②The large circle represents the core condition, the small circle represents the edge condition, and the space represents that the antecedent condition is optional for the occurrence of the result.

From Table 12, it is evident that there are distinct pathways to achieving a high level of RVAU among farmers of different scales. For small-scale farmers, two configurations are identified. Configuration G1 shows that, in the absence of regulatory environment (RE) conditions, small-scale farmers can achieve high RVAU with the core conditions of normative environment (NE) and self-responsibility (SR). Configuration G2 suggests that, with NE and cognitive environment (CE) as core conditions and RE as a peripheral factor, small-scale farmers are also able to reach a high level of RVAU.

For medium-scale farmers, two configurations promote high RVAU. Configuration G3 reveals that the combination of RE, NE, and SR as core conditions enables medium-scale farmers to achieve high RVAU. Configurations G4 and G5 indicate a mutual substitution relationship between RE and NE. Specifically, in Configuration G4, high RVAU is achieved with RE as the core condition and SR and CE as peripheral factors, while in Configuration G5, the combination of RE and SR as core conditions, with CE as a peripheral factor, also leads to high RVAU.

For large-scale farmers, Configuration G6 highlights that the core conditions of NE and SR are sufficient to drive high RVAU.

In summary, the driving paths for improving RVAU levels differ across small, medium, and large-scale farmers. While both NE and SR consistently enhance RVAU levels across all three groups, RE plays a significant role primarily for medium-scale farmers.

4. Discussion

4.1. Discussion of empirical results

Antibiotic use on farms is contributing to the rise of resistant bacteria, which represents a major threat to human health [45]. Therefore, it is crucial to reduce the overuse of antibiotics. Many countries have introduced formal government interventions to curb the overuse of antimicrobial drugs [20], [46], [47]. Nevertheless, reliance on formal institutions alone remains insufficient. Informal institutions also play a crucial role in constraining and guiding farmers' use of veterinary antibiotics [48], [49]. Formal and informal institutions complement and reinforce each other in addressing the challenge of excessive antibiotic use. Consequently, investigating their joint influence on farmers' RVAU is vital for designing effective governance strategies, safeguarding consumer health, and promoting ecological sustainability.

First, this study confirms the effectiveness of RE, NE, and CE on RVAU by farmers. Consistent with Dong et al. [50], the finding demonstrates the effectiveness of the regulatory environment in reducing the use of veterinary drugs. The possible reasons for this are as follows. In terms of incentive regulation, the government leverages market mechanisms to increase farmers' expected income and reduce their perceived risks through subsidies, rewards, and similar strategies [51], [52]. This enhances farmers' willingness to adopt RVAU, ultimately improving its implementation. For restrictive regulations, the government enforces mandatory constraints on farmers' behaviors through the formulation and implementation of laws and regulations. To avoid economic penalties or reputation losses, farmers will produce according to institutional regulations [53] and strive to elevate their RVAU performance. For guided regulation, the government conducts demonstration projects and technical training on RVAU through channels such as the internet. This helps farmers understand the principles of reducing veterinary antibiotics use, enabling them to recognize the economic and environmental benefits of RVAU and thereby increasing their willingness to engage in such practices. Similar studies, such as Shi and Zhang [54] and Cobo et al. [25], argued the effectiveness of the normative environment—such as values and social norms—in reducing the use of agricultural inputs. The possible reasons for this are as follows. By guiding farmers' values toward RVAU, it is possible to make them realize the dangers of excessive use of veterinary antibiotics and other chemicals [25]. In the meantime, social norms can create informal constraints among farmers to avoid opportunistic behaviors [55], restrict illegal antibiotic use, and thereby influence farmers' decisions to reduce veterinary antibiotic usage. As the importance of reputation and social prestige grows within rural communities, farmers are increasingly mindful of their standing within the community [56]. To avoid social pressure and moral condemnation, farmers are likely to adjust their current unreasonable production practices and reduce their overuse of veterinary antibiotics. This reflects the transformation of farmers from “economic men” to “social men”, with their production decisions increasingly constrained by social evaluation systems. Meanwhile, Yan et al. [57] indicated that the cognitive environment significantly promotes farmers' adoption of pro-environmental behaviors. The possible explanations can be categorized into two main aspects. On one hand, influenced by institutions and technologies factors, the dairy farming industry has evolved from natural clustering to social clustering. Farmers tend to align their production behavior with that of their friends, family, and peers in order to gain social recognition. When friends, family, and peers pay attention to and approve of the reduction behavior, the likelihood of excessive use of antibiotics by farmers decreases. On the other hand, driven by risk aversion, farmers often replicate the successful experiences of others [58]. When those around them reduce the amount of veterinary drugs, farmers are more likely to adopt a herd mentality and follow suit.

Secondly, the study explores the influence mechanism through which the RE, NE, and CE influence RVAU. The results confirm that farmers' SR plays a mediating role in the process through which the RE, NE, and CE influence RVAU. Research in social psychology indicates that the behaviors and psychological motivations of farmers are influenced by the external social environment [59]. Through socialization, farmers absorb and internalize the values and norms of the external environment. As they gain experience, they gradually develop an awareness of their responsibilities, experience corresponding emotions, and take actions in line with these realizations, ultimately forming SR. [60] When farmers have a stronger SR for reducing RVAU, they are more likely to impose self-regulation on their own production behaviors. This process transforms their cognitive understanding into concrete actions to reduce veterinary antibiotic overuse. This also reflects a key psychological link in the transformation of institutional environmental efficacy, which holds significant practical value for reducing long-term regulatory costs and enhancing the sustainability of reduction policies.

Third, the study explores the configurational pathways of the IE and farmers' SR in promoting RVAU. Based on the configuration analysis, two configurations promoting a high level of RVAU (P1, P2) were identified: P1, labeled the “NE-SR” linkage type, and P2, the “RE-CE-SR” driving type. Configuration P1 indicates that a high NE and a high SR can lead to a high level of RVAU. This result shows that social values related to environmental protection and food safety, alongside regulatory measures from dairy enterprises and industry associations, as well as broader societal moral expectations, play a significant role in promoting RVAU among farmers. These factors, in conjunction with the linkage between the social environment and farmers' sense of responsibility, contribute to effective RVAU. This is in line with the current reality, where the government has implemented various measures to reduce the dosage of veterinary antibiotics, though improper use still persists. The possible reason is that farmers are not only self-interested economic agents but also socially norm-oriented individuals. The external environment, including social relationships and cultural traditions, can amplify internal motivations for RVAU, creating feelings of achievement and honor for reducing antibiotic use, and shame and guilt for overuse. These emotions, in turn, impact farmers' production decisions. If government policies focus exclusively on economic incentives without addressing the social dimensions of farmers' behaviors and meeting their broader needs, it will be challenging to sustain and effectively promote RVAU. When dairy processors raise their acquisition standards and industry associations develop more detailed conventions, reducing the use of veterinary antibiotics becomes a market rule. In this context, the media continues to expose cases of illegal use of veterinary antibiotics, amplifying the value of reduction and creating a public opinion supervision environment. Enable farmers to clearly perceive that excessive use of antibiotics may not only result in economic losses due to product rejection, but also lose reputation among peers. From this, re-examine one's medication behavior and encourage them to transform external regulatory requirements into an internal sense of responsibility, thereby promoting reduction behavior. At this time, farmers' RVAU is more influenced by the RE and their SR. The configuration P2 indicates that a high RE, CE, and SR can lead to a high level of RVAU. This result indicates that, under the personal constraints of the farmers' own SR, the RE established by the government, the attention and recognition from relatives and friends toward RVAU, as well as the exemplary behavior of peers in the CE, all interact to promote the implementation of RVAU by farmers. The underlying reason may lie in the differing functions of these two institutional environments. The RE, driven by the government, carries authority and enforceability, while the CE, which arises from mutual influence among individuals, is more easily accepted and internalized. Specifically, the agricultural authorities has established a legal framework and regulatory basis for reducing veterinary antibiotics use by formulating regulations and implementing action plans for antibiotic reduction; The environmental protection authorities has effectively blocked the environmental transmission path of antibiotic resistance by regularly monitoring antibiotic residues in aquaculture waste and establishing environmental standards for the prevention and control of drug-resistant bacteria; The public health authorities establishes a bacterial resistance monitoring network to systematically assess the risks of animal derived drug resistance to human health, providing scientific support for policy adjustments. The collaborative policy-making of agriculture, environmental protection, public health, and other authorities has significantly strengthened the authority and systematicity of regulating the environment. However, in order for these policies to be better implemented, the support of the social network of farmers, such as peers and relatives, is needed to achieve the best results. Leading enterprises play an important role in this process, transforming government antibiotics policies into replicable behavioral templates through technical guidance, demonstration farms, and other means, thereby shaping the behavioral reference for reduction and the social expectation that reduction can be recognized. Together, these two institutional environments complement each other, enhancing the overall effectiveness of the institutional framework. Farmers with a high SR for RVAU are thus more likely to achieve a high level of RVAU, benefiting from the combined influence of both environments. The governance paths implied by these two configurations are shown in the Fig. 2. Meanwhile, this study also examines the differentiated pathways for promoting RVAU across farms of different scales.

Fig. 2.

The “NE–SR” and “RE–CE–SR” configurational paths.

4.2. Limitations and future prospects of research

There are still certain limitations in the present study that need further elaboration. Firstly, the measurement of farmers' sense of responsibility has certain limitations. In this study, responsibility was assessed using subjective self-reported indicators from the survey, which are inherently influenced by respondents' attitudes, cognitive levels, and situational factors. Future research can combine behavioral experimental methods or conduct in-depth interviews to more comprehensively and objectively evaluate the true state of farmers' sense of responsibility. Secondly, the selection of variables is subject to constraints. While the institutional environment was conceptualized along three dimensions—regulative, normative, and cognitive—and responsibility awareness was introduced as a mediating variable to explain farmers' reduction behavior, other potentially important factors, such as market price fluctuations and technological availability, may have been overlooked. Subsequent research can include economic factors and technological conditions in the analysis. Again, although this study examined differentiated pathways among farms of varying scales, it did not further distinguish between different developmental stages or levels of organizational capacity. In the future, we can approach from the perspective of organizational capacity and expand the research scope to different regions for comparative analysis. Finally, the RVAU measurement used in this study reflects the behavioral response of farmers to the quantitative reduction of veterinary antibiotics, but cannot reflect the absolute changes in the dosage of veterinary antibiotics. Future research can further combine aquaculture ledger records and input-output data to calculate the actual intensity of veterinary antibiotic use, complementing the findings of this study. The exploration of these limitations and future prospects provides important guidance for further research on the treatment of veterinary antibiotics.

5. Conclusion and policy implications

5.1. Main conclusion

Within the framework of global public health governance, the use of animal-derived antibiotics has become a major challenge for human health, animal health, and the ecological environment. Consequently, developing effective institutional arrangements to reduce the excessive use of veterinary antibiotics by farmers has become a critical task for advancing the green transformation of animal husbandry and ensuring public safety.

This study is based on the micro-research data from dairy farmers in 10 cities of Heilongjiang Province, China, conducted in July and August of 2024. Using the structural equation model and fuzzy directional comparative analysis methods, it analyzed the influence mechanism of the RE, NE, and CE three-dimensional IE and SR on RVAU of dairy farmers. The findings reveal the configuration paths for improving the level of RVAU by farmers. The results are as follows: firstly, the regulatory environment (RE), normative environment (NE), and cognitive environment (CE) each have a significant positive impact on farmers' reduction in veterinary antibiotic use (RVAU). Furthermore, these three dimensions of the institutional environment (IE) significantly enhance farmers' sense of responsibility (SR), which in turn further promotes their RVAU. Farmers' sense of responsibility (SR) plays a mediating role in the relationship between IE and RVAU. Secondly, RE, NE, CE, and SR individually do not constitute necessary conditions for achieving a high level of RVAU. Integrated with the results of PLS-SEM, whether a given factor promotes dairy farmers' RVAU depends on its configurational effects in combination with other influencing factors. Two configurational paths lead to a high level of farmers' RVAU: the “NE-SR” linkage type and the “RE-CE-SR” driving type. Among these, the applicability of the “NE-SR” linkage type is slightly higher than that of the “RE-CE-SR” driving type. Both configurational paths include a high level of SR, indicating that SR plays a crucial role in achieving high levels of RVAU. In addition, two configurational paths lead to a non-high level of RVAU: the “NE-CE-SR” restrictive type and the “RE-NE-SR” restrictive type. Finally, the paths to improving RVAU vary by farmer scale. Both NE and SR universally impact the RVAU levels across all farmer scales, while RE significantly influences medium-scale farmers.

5.2. Policy implications

Based on the above research conclusions, several policy implications can be drawn: firstly, improve the control mechanism and strengthen demonstration and leadership. The agricultural authorities should develop full-chain standards for veterinary antibiotic control and collaborate with research institutions to promote antibiotic alternatives. Market supervision departments should guide dairy enterprises and industry associations to implement rigorous testing and traceability systems, effectively blocking drug-resistant bacteria from entering the human food chain through raw milk. Leading enterprises and cooperatives should provide training and on-site guidance on reduction practices, while establishing communication platforms to help farmers adopt effective production methods.

Secondly, optimize institutional collaboration and establish a long-term mechanism. In early policy stages, leverage the cognitive environment by encouraging large-scale farmers, village officials, and Party members to lead reduction initiatives through technical exchange and mutual assistance activities. In later stages, prioritize a mandatory regulatory environment: improve policies requiring harmless manure treatment to prevent antibiotic-resistant bacteria and residues from contaminating water and soil. Establish information-sharing between environmental and agricultural departments for rapid pollution source identification and precise governance.

Thirdly, innovate promotional strategies to promote internalization of responsibilities. Governments and grassroots organizations should use public advertisements and community activities to raise awareness that antibiotic-resistant bacteria threaten human health through dust, water, and other exposure routes. Emphasize farmers' responsibilities in reduction efforts and the social consequences of noncompliance, while encouraging their active participation as key agents in promoting veterinary antibiotic reduction.

Finally, pay attention to scale differences and provide precise empowerment for reduction. Implement differentiated strategies for farmers of different scales to collectively enhance reduction behaviors. For small-scale and large-scale farmers, strengthen reputation and high-quality, high-price mechanisms to incentivize responsibility fulfillment. For medium-scale farmers, intensify environmental regulation enforcement and drug withdrawal supervision, guiding upgrades to biosecurity conditions and manure treatment equipment to cut off the transmission of drug-resistant bacteria from both production sources and environmental endpoints.

CRediT authorship contribution statement

Bing Jiang: Writing – review & editing, Methodology, Investigation, Funding acquisition, Formal analysis. Meijia Li: Writing – original draft, Software, Investigation, Formal analysis, Data curation. Huixin Bai: Investigation, Data curation.

Funding

This work was funded by the National Natural Science Foundation of China (72203034), Natural Science Foundation of Heilongjiang Province (LH2021G002), and Philosophy and Social Sciences Planning of Heilongjiang Province (22JYB223).

Declaration of competing interest

The authors declare that they have no competing interests.

Data availability

The data that has been used is confidential.