Introduction

Dairy livestock is crucial for enhancing food and nutritional security by increasing dietary diversity, creating employment and improving individual well-being within households1. It enhances food and micronutrient security by supplying high-quality protein, essential vitamins (A, B2, B12, and D), fats, and minerals (calcium, phosphorus, zinc, and iodine), significantly contributing to dietary diversity and caloric intake that are vital for human health2,3,4. Consuming dairy products like milk, yogurt, and cheese enhances dietary diversity, reducing malnutrition risks, combating deficiencies such as iron-deficiency anemia, and stunting in children. However, dairy product consumption remains low due to the limited adoption of modern dairy technologies. This low adoption rate is attributed to farmers’ lack of knowledge, unfavorable attitudes, and inadequate practices regarding these innovations5. Insufficient awareness and technical expertise prevents farmers from utilizing advanced dairy farming techniques, leading to suboptimal yields and reduced market supply6. Additionally, socio-cultural barriers and resistance to change further impede the diffusion of improved dairy practices, exacerbating challenges in the sector7. Livestock-based interventions enhance food and micronutrient security by improving productivity, access to quality dairy, and income1. These initiatives reduce malnutrition risks and support maternal and child nutrition while promoting poverty alleviation8,9. Strengthening livestock value chains ensures sustainable food access, dietary diversity, and economic resilience10,11.

Strategies such as dairy intensification are promising for improving human nutrition by strengthening household food security12.Moreover, enhancing dairy cattle genetics and management practices can significantly increase the benefits of small-scale dairy farms, leading to improved livelihoods and self-sufficiency in livestock production13. Dairy farmers need access to information on innovative production techniques, disease control, breeding methods, and other management aspects14. Policies supporting livestock development programs, including animal husbandry training, and ensuring land tenure security, can strengthen household food and nutritional security15. Preserving livestock genetic resources is crucial for maintaining food security, nutrition, and sustainability in the face of climate change impacts on livestock production15,16. Farmers’ KAP regarding livestock interventions significantly influence FNS. Training enhances awareness, improving adoption of dairy hubs, veterinary services, and feed management1. Training enhances awareness and improves the adoption of dairy hubs, veterinary services, and feed management1.

Positive attitudes drive sustainable practices, reducing malnutrition risks and enhancing resilience9,11. Capacity building fosters long-term benefits17 .Targeted interventions, improved management practices, and support for farmers can greatly enhance food security and nutrition outcomes within the dairy sector.

The governance of Bangladesh has actively engaged in initiatives to improve livestock and dairy production through the livestock and dairy development project (LDDP). This project promotes private sector investment and sustainable development of livestock value chains, addressing issues like food safety, environmental pollution, climate change, and animal welfare. These efforts represent a concerted attempt to enhance dairy and livestock production in Bangladesh, focusing on livelihood improvement, market access, and sustainability, crucial for FNS in the country. The LDDP provide livestock-based interventions including formation of producer groups, provides training, capacity building, mobile veterinary clinics support, establishment of village milk collection centers, and dairy hubs aimed at improving overall productivity, sustainability, and market access in the livestock and dairy sector. Despite efforts to disseminate dairy technologies, adoption remains low due to a lack of farmers’ KAP. The LDDP interventions and studies on FNS-related KAP assess and explore people’s KAP related to nutrition, diet, foods, and the adoption of dairy farming. The KAP studies serve two main purposes: (1) collecting key information during situational analyses to inform the design of nutrition interventions, and (2) evaluating nutrition education interventions. The findings of the situational analysis will aid in planning a nutrition intervention to address identified nutrition problems. However, a significant research gap exists concerning the low adoption of DLI in Bangladesh, despite governance efforts to promote them. The causal pathways and mechanisms linking dairy interventions to food and nutritional security remain unexplored. Further investigation is needed into the specific knowledge, attitudes, and practices affecting dairy intervention adoption, and their impact on food safety. KAP surveys provide crucial information for exploring food safety factors and potential intervention strategies for dairy development. Therefore, assessing the KAP in the delta region of Bangladesh is essential to examine its relationship with the adoption of DLI. This research seeks to address the challenges in technology adoption and answers the following queries:

How do farmers’ influence the adoption of DLI in Bangladesh?

To that extent, does the adoption of DLI improve household FNS in the Delta region of Bangladesh?

What are the causal pathways through which dairy interventions impact FNS?

These questions ensure a structured investigation into the relationships between KAP, dairy technology adoption, and FNS while addressing gaps in causal pathways and intervention strategies. Because the farmers’ KAP understanding can guide policymakers and researchers in developing technologies and policies that better dairy production performance, leading to positive impacts on farmers’ livelihoods.‌.

Conceptual framework of the causal relationship between KAP, LDDP interventions and FNS

The conceptual framework illustrates (Figure-1) the causal relationship between the KAP, LDDP interventions, and FNS. Various livestock interventions, including PG, VMCC, MVC, DH, and CBT influence the adoption of dairy technologies. The adoption process is driven by farmers’ knowledge, attitudes, and practices5. Knowledge enhances awareness of improved technologies, whereas attitudes determine the willingness to adopt new practices, and proper management practices ensure effective implementation. The adoption of dairy livestock interventions strengthens FNS by improving dairy production and income, leading to household food security and better nutrition11. However, gaps in KAP, such as limited awareness, resistance to change, and inadequate management skills, hinder its widespread adoption7. Effective training and extension services play a critical role in overcoming these barriers and ensuring the sustainability of dairy innovations18. Addressing the KAP gaps is essential for enhancing dairy productivity and achieving long-term food and nutritional security.

Fig. 1
figure 1

Conceptual framework of causal relationship among KAP, DLI and FNS.

Full size image

Ensuring FNS requires an integrated approach linking dairy livestock interventions (DLI) with farmers’ KAP5. Strengthening extension services, capacity building, and market access enhances technology adoption, boosting dairy productivity and household income19. Improved milk production directly enhances nutritional intake, reducing food insecurity11. To ensure food and nutrition security among livestock farmers, effective linkages between stakeholders and services in the livestock sector are essential. Interlinking livestock producer groups, village milk collection centres, mobile veterinary clinics, DHs, and CBT can enhance food security and nutrition outcomes. This interconnected system promotes sustainable livestock production, improves livelihoods, and contributes to overall food security in rural communities.

Materials and methods

Study areas

This study was conducted under the LDDP, financed by the World Bank (Project ID: P161246). It was carried out in various districts of Bangladesh, namely: Sirajganj, Pabna, Munshiganj, Satkhira, and Chittagong, selected based on dairy cattle density and milk production. At least two Upazilas were purposively chosen from each district, considering their concentration of livestock farming: Shahjadpur and Belkuchi (Sirajganj), Bera and Santhia (Pabna), Sreenagar and Sirajdikhan (Munshiganj), Tala and Kaligonj (Satkhira), and Satkania and Potiya (Chittagong). The LDDP aims to enhance agricultural productivity and market access for smallholder dairy livestock farmers and small and medium-scale agro-entrepreneurs. Through regular education and training sessions on whole-farming systems (basic husbandry, nutrition, reproduction, and calf rearing) via farmer discussion groups, dairy farmers under the LDDP are benefiting areas.

Sampling technique

To ensure statistical reliability, a stratified random sampling technique was applied. The study population was stratified based on key characteristics, and a proportional random selection was conducted within each stratum. The following formula was used to determine the sample size\(\:n=\frac{{Z}^{2}.p.q.N}{{Z}^{2}p.q+(N-1){e}^{2}}\)

Where,

n = Sample size.

N = Population size.

e = Precision rate or amount of admissible error in the estimate.

p = Proportion of defectiveness or success for the indicator.

q = 1-p.

z = Standard normal variable at the given level of significance.

In the sampling estimate, given values are:

N = total households covered by the LDDP)

e = 0.06 (6% significance level/admissible error margin).

p = 0.5.

q = factor q (1-p) = 0.5.

z = 1.96 (value of the standard normal variable at 92% confidence level).

Using the given values, the estimated sample size was determined. Based on this calculation, a total of 400 respondents was selected, ensuring statistical acceptability at a 95% confidence level. The total sample of 400 was proportionally distributed across five Upazilas in five districts of Northern Bangladesh, ensuring representation from different livestock farming communities. The selection process accounted for variations in socio-economic conditions, farming practices, and geographical differences within the region. This methodological approach strengthens the study’s validity and ensures that the findings accurately reflect the food security and sovereignty dynamics among livestock farming communities in Bangladesh.

Table 1 Detailing the distribution of the 400 samples across the five districts.
Full size table

Table 1 presents the proportional distribution of the 400 samples across the five districts, ensuring representative coverage of key dairy farming regions. Pabna had the highest share (12.5%), while Munshiganj had the lowest share (7.5%), ensuring diverse socioeconomic and geographical representation.

Data collection

The primary data were collected using a structured survey questionnaire administered through Qualtrics e-survey software. Before conducting the survey, researchers identified potential interventions for dairy farmers provided by the LDDP. A structured questionnaire was developed to assess the KAP of dairy livestock farmers regarding the adoption of LDDP interventions. The questionnaire was pre-tested to ensure clarity and reliability, and necessary modifications were made to refine it. The finalized questionnaire was digitized and deployed using Qualtrics e-survey software to collect cross-sectional data. Secondary data were gathered from LDDP reports, government publications, research articles, and relevant institutional databases. These sources provided background information and contextual insights to complement the primary data findings. This comprehensive data collection approach ensures a robust analysis of dairy farmers’ adoption of LDDP interventions and their impact on food security in Northern Bangladesh.

Analytical techniques

Multistep analytical techniques were used in this study to assess farmers’ KAP regarding the implementation of interventions in dairy farming to ensure food security and nutrition security in Bangladesh. This research was conducted in five stages. First, a literature review was conducted to identify key areas of intervention related to dairy farming. Second, an estimate-talk-estimate approach was used to elicit experts’ opinions on the perception of key statements related to interventions in dairy farming for food security, as proposed by the LDDP. Third, an online survey was conducted in Bangladesh from December 2023 to May 2024, using a closed-ended questionnaire with a total of 400 cross-sectional data points. The data distribution was checked using a histogram with a density curve and the Shapiro-Wilk test. The Mann-Whitney U-test and Kruskal-Wallis test were used to compare KAP scores, with post hoc analysis using Dunn’s test where appropriate. Pairwise correlations among these measures were assessed using Spearman’s rho (r) test. A p-value < 0.05 was considered significant for all tests, except for Dunn’s test, where a p-value < 0.025 was considered significant. Finally, the study employed the Partial Least Square Structural Equation Modeling (PLS-SEM) technique using SmartPLS03 with STATA version 17. PLS-SEM is well-suited for analyzing farmers’ KAP towards livestock interventions, as it effectively handles complex relationships, small sample sizes, and non-normal data20. It enables the simultaneous assessment of direct and indirect effects, improving predictive accuracy and robustness in food security research21. The survey instrument received ethics approval from the Bangladesh Agricultural University’s ethics committee.

Econometric model for analysis

The PLS-SEM is a statistical technique that combines the strengths of two methods: PLS and SEM. The PLS is a multivariate method that is useful for analyzing relationships between many predictor variables and a small number of response variables. The SEM, on the other hand, is a method for analyzing relationships between a set of observed variables and a set of latent variables. The PLS-SEM is a causal-predictive approach that emphasizes prediction in estimating statistical models. It is designed to test the causal relationships between constructs with multiple measurement items. In PLS-SEM, the basic equations involve estimating the latent variables and their relationships. Consider a structural model with latent variables X, Y, Z. The equations for these latent variables in the context of PLS-SEM can be represented as follows.


Measurement model equations:

For the latent variable X; \({{\upchi\:}}_{i}={\lambda\:}_{\chi\:i}{\xi\:}_{i}+{\epsilon}_{\chi\:i}\)

For the latent variable Y; \({\text{y}}_{i}={\lambda\:}_{yi}{\eta\:}_{i}+{\epsilon}_{yi}\)

For the latent variable Z; \({\text{z}}_{i}={\lambda\:}_{zi}{\zeta\:}_{i}+{\epsilon}_{zi}\)

Here, \({{{\upchi}}_{i}, \:\text{y}}_{i},\:and\:{\text{z}}_{i}\:represent\:the\:observed\:indicators\:for\:variables\:X,\:Y,\:and\:Z\)\(\:respectively.\) \(\:{\lambda\:}_{\chi\:i},{\lambda\:}_{yi},\:and\:{\lambda\:}_{zi}\:are\:the\:factor\:loading\), \(\:{\xi\:}_{i},\:{\eta\:}_{i}\) and \(\:{\zeta\:}_{i}\)are the latent variables, and \({\epsilon}_{\chi\:i},\:{\epsilon}_{yi}\:and\:{\epsilon}_{zi}\:are\) the measurement errors.


Structural Model Equations:

$${\xi\:}_{i}={\beta}_{xy}{y}_{i}+{\beta}_{xz}{z}_{i}+{\zeta\:}_{x}+{\epsilon}_{\chi\:}$$
$${\eta\:}_{i}={\beta}_{yz}{x}_{i}+{\beta}_{yz}{z}_{i}+{\zeta\:}_{y}+{\epsilon}_{y}$$
$${\zeta\:}_{i}={\beta}_{zx}{x}_{i}+{\beta}_{zy}{y}_{i}+{\zeta\:}_{z}+{\epsilon}_{z}$$

Here, \({\beta}_{xy},\:{\beta}_{zx},\:{\beta}_{yz}\), \({\beta}_{xz}\), \({\beta}_{yz},\:\)and \({\beta}_{zy}\) represent the path coefficients, \({\zeta\:}_{x},\:{\zeta\:}_{y},\:and\:{\zeta\:}_{z}\) are the intercepts, and \({\zeta\:}_{x},\:{\zeta\:}_{y}\:and\:{\zeta\:}_{z}\:are\) the structural errors.

These equations represent the fundamental relationships in a PLS-SEM, where latent variables are estimated based on their observed indicators and their relationships are determined through structural equations.

The key latent variables hypothesis was created against the LDDP’s interventions

The interventions of the LDDP in Bangladesh, including LPGs, VMCCs, MVCs, DHs, and CBT, impact FNS. Key latent variables such as technology adoption, big data decision-making, small farm competition, real-time decision-making, land fragmentation, risk attitude, processor linkages, institutional framework, and precise condition detection influence these interventions’ effectiveness and outcomes. This brief discusses generating latent variables and highlights effective interventions in increasing food and nutrition security in rural areas.

Livestock producer groups (LPGs)

LPGs are farmer collectives that manage cattle, dairy, and poultry together. These groups pool resources and share knowledge to boost productivity and profitability. In Bangladesh and other developing countries, LPGs play a crucial role in ensuring food and nutrition security, as well as enhancing overall food production, nutrition, and economic benefits. The LDDP project has successfully organized 5,000 LPGs in Bangladesh. This initiative has helped farmers improve market access, share knowledge and skills, increase bargaining power, access credit and extension services, improve animal health and welfare, and adopt climate-resilient practices. As a result of these interventions, researchers have generated the following hypothesis.

Hypothesis 1

H0: \({\beta}_{LPGs}=0.\:\)The null hypothesis states that there is no statistically significant relationship (\({\beta}_{LPGs}=0)\) between LPGs and the FNS.

H1: \({\beta}_{LPGs}\ne\:0.\:\)The alternative hypothesis states that there is evidence to suggest that LPGs play a significant role to ensure FNS in livestock farming households.


Hypothesis 1 suggests that there is a significant correlation between knowledge and the adoption of DLI. It proposes that farmers with a strong grasp of CBT, livestock producer groups, mobile veterinary clinics, village milk collection centers, and DHs are more inclined to adopt these practices. This, in turn, can lead to enhanced food and nutrition security within livestock farming households. And the latent variable LPG was developed based on PG1- PG7 Likert scale observation indicators in Table 3.

Village milk collection centers (VMCCs)

The hypothesis suggests that knowledge has a significant impact on the adoption of DLI, specifically focusing on VMCCs to improve food and nutrition security among livestock farmers. VMCCs are important platforms for smallholder dairy farmers to sell their milk, enhance milk quality and safety, and access extension services and training. The LDDP project’s intervention has shown that establishing a system for milk collection and marketing through VMCCs helps dairy farmers obtain better prices for their milk, increasing their income and contributing to food and nutrition security. In addition, VMCCs have played a crucial role in empowering women farmers by increasing their income, creating employment opportunities, and reducing gender disparities through improved confidence, knowledge, and decision-making abilities. By examining the relationship between knowledge and the adoption of DLI, particularly through VMCCs, the hypothesis suggests that a deeper understanding and awareness among farmers can lead to more effective utilization of these interventions, ultimately enhancing food and nutrition security outcomes in rural areas.

Hypothesis 2

H0: \({\beta}_{VMCCs}=0.\:\)The null hypothesis states that there is no statistically significant relationship (\({\beta}_{VMCCs}=0)\) between VMCCs and the FNS.

H1: \({\beta}_{VMCCs}\ne\:0.\:\)The alternative hypothesis states that there is evidence to suggest that VMCCs play a significant role in ensuring FNS in livestock farming households.

The latent variable VMCC was developed based on V1-V5 Likert scale observation indicators in Table 3.

Hypothesis 2 proposes that attitude significantly affects the adoption of DLI. Farmers’ attitudes towards CBT, LPGs, MVCs, VMCCs, and DHs can impact their willingness to adopt these practices. Positive attitudes are expected to promote adoption, while negative attitudes may hinder it. The statistical analysis revealed a significant link between attitude and the adoption of DLI, highlighting the need to understand and address farmers’ attitudes to enhance FNS outcomes in Bangladesh.

Mobile veterinary clinic (MVC)

The MVCs positively impact food and nutrition security in rural areas by providing essential animal health services like vaccinations, deworming, and treatment for common illnesses and injuries. LDDP introduced MVCs in Upazila level in Bangladesh, offering various services and training on animal care and husbandry. MVCs are crucial for disease prevention and control, enhancing food security in rural areas22.

Hypothesis 3

H0: \({\beta}_{MVCs}=0.\:\)The null hypothesis states that there is no statistically significant relationship (\({\beta}_{MVCs}=0)\) between MVCs and the FNS.

H1: \({\beta}_{MVCs}\ne\:0.\:\)The alternative hypothesis states that there is evidence to suggest that MVCs play a significant role in ensuring FNS in livestock farming households.

The latent variable MVC was developed based on M1-M5 Likert scale observation indicators.

Hypothesis 3 states that practices have a statistically significant impact on the adoption of dairy interventions. This suggests that farmers’ practices in adopting dairy interventions can influence their adoption. Farmers who engage in certain practices related to DLI are more likely to adopt these interventions, ultimately improving food and nutrition security in livestock farming households.

Dairy hub (DH)

DHs play a crucial role in improving food and nutrition security in livestock farming households. They achieve this by increasing milk yields, enhancing household resilience, providing market access, generating income, promoting sustainable practices, and fostering community development23. DHs also give smallholder farmers access to markets, inputs, and services, leading to better economic well-being. LDDP project developed dairy hub though its intervention, they provide various services that can improve the availability and quality of dairy products, leading to increased food security, household income, and leading to better nutrition and food security for the household.

Hypothesis 4

H0: \({\beta}_{DHs}=0.\:\)The null hypothesis states that there is a no statistically significant relationship (\({\beta}_{DHs}=0)\) between DHs and the FNS.

H1: \({\beta}_{DHs}\ne\:0.\:\)The alternative hypothesis states that there is evidence to suggest that DHs play a significant role to ensure FNS in livestock farming households.

And the latent variable DH was developed based on D1-D6 Likert scale observation indicators.

Hypothesis 4 suggests that DHs have a significant impact on food and nutrition security in livestock farming households. DHs improve food security by increasing milk yields, enhancing household resilience, providing market access, generating income, promoting sustainable practices, and fostering community development. Statistical analysis will determine if the p-value associated with Hypothesis 4 is less than 0.05, thereby supporting the alternative hypothesis of a significant relationship between DHs and food and nutrition security.

Capacity Building training (CBT)

The LDDP in Bangladesh offers CBT activities to enhance the skills and knowledge of dairy farmers, aiming to improve productivity, profitability, and food security. This includes training in animal nutrition, breeding, veterinary care, and milk production. This can help to increase their participation in the sector and improve their food and nutrition security in livestock farming households24,25. The researcher (s) have been drowning the hypothesis below.

Hypothesis 5

H0: \({\beta}_{CBs}=0.\:\)The null hypothesis states that there is no statistically significant relationship (\({\beta}_{CBs}=0)\) between CBT and the FNS.

H1: \({\beta}_{CBs}\ne\:0.\:\)The alternative hypothesis states that there is evidence to suggest that CBT play a significant role in ensuring FNS in livestock farming households.

The latent variable CB was developed based on C1- C7 Likert scale observation indicators in Table 3.

Hypothesis 5 suggests that CBT has a significant impact on food and nutrition security in livestock farming households. However, the results do not support this hypothesis as the path coefficient for CBT to food and nutrition security is not statistically significant. This suggests that CBT may not directly improve food and nutrition security in the context of DLI in Bangladesh.


Key latent variables hypothesis created against the KAP.

Knowledge: Farmers’ understanding of the benefits of CBT, LPGs, MVCs, VMCCs, and DHs can influence their willingness to adopt these practices. The hypothesis is that there is a statistically significant impact of knowledge on the adoption of LDDP interventions26,27,28,29.

Hypothesis 1

H0: \({\beta}_{Knowledge}=0\), The null hypothesis states that there is no statistically significant impact (\({\beta}_{Knowledge}=0)\) of knowledge on the adoption of dairy interventions.

H1: \({\beta}_{Knowledge}\ne\:0\), The alternative hypothesis states that there is a statistically significant impact (\({\beta}_{Knowledge}\ne\:0)\) of knowledge on the adoption of dairy interventions.

And the latent variable Knowledge was developed based on K1- K5 Likert scale observation indicators in Table 3.

Attitude

Farmers’ attitudes towards capacity building training, livestock producer groups, mobile veterinary clinics, village milk collection centers, and DHs can impact their adoption. Positive attitudes encourage adoption, while negative attitudes hinder it. Positive attitudes towards new technologies and practices increase the likelihood of adoption28,29,30 .

Hypothesis 2

H0: \({\beta}_{attitude}=0.\:\)The null hypothesis states that there is a statistically significant relationship (\({\beta}_{attitude}=0)\) between attitudes and the adoption of dairy interventions.

H1: \({\beta}_{attitude}\ne\:0.\:\)The alternative hypothesis states that there is evidence to suggest that attitudes play a significant role in the adoption of dairy interventions.

And the latent variable Attitude was developed based on A1- A5 Likert scale observation indicators in Table 3.

Practices

Farmers’ adoption of these interventions can be influenced by their knowledge and attitudes. Those with knowledge and positive attitudes are more likely to adopt them, which enhances food and nutrition security in livestock farming households. The hypothesis predicts a significant impact of practice on the adoption of dairy interventions27.

Hypothesis 3

‌Ho: \({\beta}_{practices}=0\:.\) The null hypothesis states that there is no statistically significant impact (\({\beta}_{practices}=0)\) of practices on the adoption of dairy interventions.

‌H1: \(practices\ne\:0).\:\text{T}\text{h}\text{e}\:\text{a}\text{l}\text{t}\text{a}\text{n}\text{a}\text{t}\text{i}\text{v}\text{e}\:\text{h}\text{y}\text{p}\text{o}\text{t}\text{h}\text{e}\text{s}\text{i}\text{s}\:\text{s}\text{t}\text{r}\text{a}\text{t}\text{e}\text{s}\:\text{t}\text{h}\text{a}\text{t}\:\text{t}\)here is a statistically significant impact (\({\beta}_{practices}\ne\:0)\) of practices on the adoption of dairy interventions.

The latent variable Practice was developed based on P1- P5 Likert scale observation indicators.

While practice does have a statistically significant impact on dairy intervention adoption, it is just one of several factors that affect adoption. The degree of influence varies depending on the intervention and context. Similarly, the positive relationship between knowledge, awareness, and adoption intention is one of many influencing factors that also vary depending on the context and intervention.

Hypothesis 4

H0: \({\beta}_{knowledge,}-{\beta}_{awarness}=0\:.\) The null hypothesis states that there is no statistically significant relationship \({\beta}_{knowledge,}-{\beta}_{awarness}=0\:\) between knowledge and awareness and the intention to adopt dairy farm interventions.

H1:\({\beta}_{knowledge,}-{\beta}_{awarness}\ne\:0\:.\)

\(\:At\:least\:one\:of\:the\:{\beta}_{knowledge,\:awarness}=0\:.\:The\:alternative\:hypothesis\:states\:that\:t\)here is a statistically significant relationship \({\beta}_{knowledge,}-{\beta}_{awarness}\ne\:0\:\)between knowledge and awareness and the intention to adopt dairy farm interventions.

Results and discussion

Dairy farming household socio-demographic profiles

Table 2 provides a socio-demographic profile of dairy farmers in Bangladesh. The majority (68%) fall within the age bracket of 27 to 49 years, with 38.5% having completed high school education. Most households are medium-sized (77%, with 4 to 9 members), and 57.5% own small to marginal landholdings. About 46.5% of farmers are in the medium-income bracket. Herd sizes mainly range from 4 to 7 animals for 58.5% of farmers. Social participation is moderate for 35% of respondents. The primary information sources are newspapers and local networks, relied upon by 90.5% of farmers. Regarding risk orientation, 71.5% exhibit a medium level of risk tolerance. This data offers insights into the diversity and needs of dairy farming households in Bangladesh, emphasizing the importance of considering these socio-demographic factors when designing interventions to enhance food and nutrition security in the agricultural sector.

Table 2 Socio-demographic profile of the dairy farmers.
Full size table

The measurement model

Composite reliability (CR) and Cronbach’s α were used to measure the internal consistency of the model, while factor loading, and the average variance extracted (AVE) were used to measure the convergent validity of the model. Given the controversial nature of model fit in PLS-SEM21, the Consistent-PLS program, following31, was employed to test the hypotheses. The first step involves evaluating the outer model, and assessing how well the items load on the hypothesized constructs. The reflective outer model assessment includes examining indicator reliability, reliability of latent variables, internal consistency (Cronbach alpha and composite reliability), construct validity (loading and cross-loading), convergent validity (AVE), and discriminant validity (Fornell-Larcker criterion, cross-loading, HTMT criterion) (Hair et al.,2014). Key latent variables of LDDP interventions such as LPGs, VMCCs, MVCs, DHs, and CBT were determined (Table 3), alongside farmer perceptions towards the adoption of interventions like KAP for ensuring FNS. Latent variables, unobserved and inferred from observed variables, explain the relationships between them. Structural equation modelling (SEM), a statistical technique, estimates the relationships between latent and observed variables. To establish convergent validity, the factor loading of the indicator, CR, and the AVE were considered. AVE values exceeding 0.50 are considered adequate for convergent validity30,32,33. The study found AVE to be more than 0.50 for VMCC (0.809), MVC (0.731), DH (0.680), CBT (0.621), Knowledge (0.555), Attitude (0.752), and Practices (0.521), except for LPG.

Table 3 Construct of key latent variables.
Full size table

Table 3 presents the key variables of the DLI and farmer perceptions towards the adoption of DLI for FNS. The variables include LPG, VMCCs, MVCs, DHs, CBT, Farmer’s KAP. The variables in Table 2 represent interventions implemented under the LDDP, such as LPG, VMCCs, MVCs, DHs, and CBT. These interventions aim to enhance agricultural productivity and market entry for smallholder farmers and agro-entrepreneurs. The factors loading and reliability measures indicate the strength and consistency of the relationships between the variables and their indicators. On the other hand, Table 3 presents variables related to farmers’ knowledge, attitudes, and practices in adopting dairy interventions for improving food and nutrition security. These variables represent farmers’ perceptions and behaviors towards adopting sustainable agricultural practices, livestock management strategies, and CBT for better food and nutrition outcomes. Overall, the tables provide a comprehensive view of the key variables involved in the LDDP and their impact on farmers’ perceptions and behaviors related to dairy interventions for enhancing food and nutrition security in Bangladesh. We evaluated indicator consistency using Cronbach’s alpha and composite reliability. Four items were dropped due to loading values below 0.70. Cronbach’s alpha ranged from 0.679 to 0.918, exceeding the threshold of 0.70. Composite reliability scores were mostly above 0.70. Factor loading values, indicating indicator consistency, ranged from − 1 to 1. KAP are vital latent variables influencing food and nutrition security. SEM helps identify key latent variables like KAP related to food production, distribution, and consumption. Farmers’ KAP towards CBT, LPGs, MVCs, VMCCs, and DHs can affect FNS in livestock farming households. The AVE for every construct exceeded the 0.50 cut-off, ranging from 0.68 to 0.94. Lastly, the VIF values, assessing multicollinearity, ranged from 1.308 to 2.936, indicating no multicollinearity and relatively stable model results.

Construct reliability and validity

The Fornell-Larcker criterion

Table 4 displays the results of the Fornell-Larcker criterion analysis for checking discriminant validity, revealing the correlation values between different constructs in the model. The diagonal features the square root of the AVE for each construct, while the off-diagonal values represent correlations between the constructs. To ensure discriminant validity, the square root of the AVE for each construct should exceed the correlations with other constructs. For instance, considering the first row and column for LPGs,” with an AVE value of 0.970, it surpasses all correlation values with other constructs, indicating good discriminant validity.

Table 4 Fornell-Larcker criterion analysis for checking discriminant validity.
Full size table

Table 5 presents the Fornell-Larcker Criterion Analysis for Checking Discriminant Validity. The diagonal values are the square roots of the AVE for each construct, while the values below the diagonal are correlation coefficients between the constructs. Comparing the square roots of the AVE with the correlations between the constructs, all AVE values exceed the correlation coefficients, indicating good discriminant validity within the model.

Table 5 Fornell-Larcker criterion analysis for checking discriminant validity.
Full size table

The Heterotrait–Monotrait ratio (HTMT)

The Heterotrait–Monotrait ratio (HTMT) is a criterion used to assess discriminant validity in structural equation modelling (SEM). The HTMT value is used to determine whether discriminant validity has been established between two constructs. According to the research, the HTMT ratio must be less than 1.00, and some authors suggest a threshold of 0.85 or 0.9017,34. Tables 6 and 7 display the HTMT values for different constructs related to DLI and FNS in Bangladesh. The values in the tables indicate the relationships between various factors, such as LPG, VMCCs, MVCs, DHs, CBT, KAP, and FNS. The HTMT values shown in these tables are below the threshold of 0.85, confirming the discriminant validity of the constructs. This means that the relationships between the different variables and constructs are distinct and show that the model proposed in the study has good discriminant validity. Overall, the HTMT analysis helps in understanding the relationships between key factors in DLI and their impact on food and nutrition security in Bangladesh.

Table 6 The Heterotrait–Monotrait ratio (HTMT).
Full size table
Table 7 The Heterotrait–Monotrait ratio (HTMT).
Full size table

DLI adoption is causally linked to FNS. Livestock ownership correlates with increased consumption of nutrient-rich animal-source foods (ASFs) like milk, meat, and eggs, enhancing children’s nutritional outcomes35. In Bangladesh, household ownership of dairy- and meat-producing animals is associated with elevated infant intakes of eggs, dairy, and meat36. Ethiopia’s study shows that adopters of dairy development interventions experience significantly higher incomes, potentially improving their FNS30. As livestock ownership is increasingly advocated in food security strategies, animals provide nutrient-dense foods, regular income, and other benefits. Although the impact of nutrition-sensitive livestock interventions on diet and nutritional status varies, studies indicate a rise in ASF consumption due to livestock interventions37. Overall, DLI adoption improves FNS by augmenting ASF consumption, increasing income, and providing additional benefits, albeit with contextual variations.

Hypothesis 5

‌Ho: \({\beta}_{intervention}=0\:.\) The null hypothesis states that there is no statistically significant causal relationship (\({\beta}_{intervention}=0)\) between DLI adoption and FNS.

H1: \({\beta}_{intervention}\ne\:0).\:\text{T}\text{h}\text{e}\:\text{a}\text{l}\text{t}\text{a}\text{n}\text{a}\text{t}\text{i}\text{v}\text{e}\:\text{h}\text{y}\text{p}\text{o}\text{t}\text{h}\text{e}\text{s}\text{i}\text{s}\:\text{s}\text{t}\text{r}\text{a}\text{t}\text{e}\text{s}\:\text{t}\text{h}\text{a}\text{t}\:\text{t}\)here is a statistically significant causal relationship between DLI adoption and FNS.

Farmers’ KAP towards CBT, MVCs, VMCCs, and DHs play a crucial role in ensuring FNS in livestock farming households. By promoting positive KAP, these organizations can help farmers adopt practices that lead to improved food and nutrition security.

The structural model

The structural model examines the significance of the respective relationships (inner model) whereby path coefficient and t-values were investigated. Table 5 shows the path coefficient and its corresponding t-values for the direct relationships whereby t-value > 1.96 is equivalent to a significant relationship (p < 0.05). The relative importance of the exogenous constructs in predicting dependent constructs is shown in Table 8. Therefore, it can be inferred that two relationships were supported and two were not supported in this study. According to Falk and Miller, the R² -values should be equal to or greater than 0.10 for the variance explained of a particular endogenous construct to be deemed adequate. In this study, the R²-value of 61.3, meets the adequacy requirement of Falk and Miller and the variances are considered substantial in the intention to adopt good dairy farming and hygiene practices.

Table 8 Path coefficient in the structural model.
Full size table

Table 8 presents the path coefficients in the structural model. Path coefficients represent the strength and direction of the relationships between the latent variables in the study. The table shows the path coefficient, standard error, t-value, p-value, result, R2, f2, and VIF for each relationship. The path coefficient for knowledge is -0.14, with a t-value of 1.35, indicating a non-significant relationship (p > 0.05). The path coefficient for attitude is 0.42, with a t-value of 3.66, indicating a significant positive relationship (p < 0.05). The path coefficient for practices is 0.11, with a t-value of 0.70, indicating a non-significant relationship (p > 0.05). The path coefficient for DLIA is 0.29, with a t-value of 2.22, indicating a significant positive relationship (p < 0.05). The path coefficient for FNS is not provided in Table 8. Overall, the results indicate that attitude and DLI have significant positive effects on farmers’ intention to adopt dairy interventions. Knowledge and practices, on the other hand, do not show significant relationships in the structural model.

Hypothesis testing

Table 9 presents the results of hypotheses testing, including beta values, T statistics, and P values for each hypothesis related to DLI and their impact on food and nutrition security. The table provides insights into the relationships between variables and their significance in the study. For example, in hypothesis H1, which examines the association between LPGs and FNS, the beta value is 0.285, the T statistic is 3.594, and the P value is 0.001. This indicates a statistically significant positive association between LPG and FNS at the 0.05 level.

Table 9 Hypotheses, beta, T statistics and P values.
Full size table

The following are the three general hypotheses of this study:

Hypothesis 1

(H1). There is a statistically significant impact of knowledge on the adoption of DLI, α < 0.05.

Hypothesis 2

(H2). There is a statistically significant impact of attitude on the adoption of DLI, α < 0.05.

Hypothesis 3

(H3). There is a statistically significant impact of practice on the adoption of DLI, α < 0.05.

Similarly, in hypothesis H2, which focuses on VMCCs and FNS, the beta value is -0.241, the T statistic is 3.056, and the P value is 0.002. This suggests a statistically significant negative association between VMCC and FNS. Contrary to the negative association mentioned, Village VMCCs contribute to FNS. VMCCs provide a structured platform for rural dairy farmers, particularly women, to sell their milk directly to processors, thereby increasing income and improving livelihoods. This model has been successfully implemented in Bangladesh, resulting in increased milk production, improved nutrition, and enhanced food security for rural households38. In summary, Table 9 provides a comprehensive overview of the results of hypotheses testing, shedding light on the impact of different DLI on food and nutrition security in Bangladesh.

Discussions

This study provides new insights into the KAP of dairy farmers regarding the adoption of LDDP interventions to enhance food and nutrition security. DLI are pivotal strategies aimed at ameliorating food and nutrition security through the utilization of livestock. Such interventions often aim at augmenting the productivity and profitability of livestock operations while enhancing access to nutritious food derived from livestock products. Examples of these interventions include: (a) livestock health interventions such as vaccination and disease treatment, which can boost milk and meat production, thus improving food and nutrition security; (b) Livestock feed interventions, which ensure animals receive balanced diets, consequently enhancing productivity and contributing to food and nutrition security; (c) Livestock breeding interventions that improve the genetic potential of livestock through selective breeding, thereby increasing productivity and improving food and nutrition security; (d) Market-oriented livestock interventions, which promote market-oriented livestock production, increasing income and livelihood opportunities and improving food and nutrition security through increased access to nutritious food; (e) Women’s empowerment interventions, which support women’s involvement in livestock production and ownership, thereby enhancing gender equality and food and nutrition security. The adoption of DLI by farmers is influenced by their KAP. Knowledge is influenced by understanding the benefits and limitations of the interventions, technical understanding, and access to information and training. Attitudes are shaped by perceptions toward interventions, past experiences, cultural beliefs, and economic incentives. Practices are determined by existing routines, capacity, and resource availability.

Therefore, a multifaceted approach that addresses farmers’ knowledge, attitudes, and practices is crucial for promoting the adoption of DLI. The study measured knowledge using a test developed by30, attitude through a scale developed by7, and adoption of scientific dairy farming practices through an index developed by39. The results of correlation analysis showed a significant relationship between knowledge and DLI, indicating that adequate knowledge positively influences attitudes toward adoption. Furthermore, a positive and significant correlation between practice and the adoption of DLI was found, implying that good knowledge helps understand concepts and adopt good livestock practices.

This study evaluated construct reliability and validity using Cronbach’s α, composite reliability (CR), factor loading, and average variance extracted (AVE). CR values exceeded 0.70, ensuring strong internal consistency20. AVE values surpassed 0.50 for most constructs, confirming convergent validity32. However, LPG had a lower AVE (0.373), indicating weaker reliability, which aligns with33 findings on similar AVE thresholds, reinforcing our results. Discriminant validity was confirmed using the Fornell-Larcker criterion, as the square root of AVEs exceeded inter-construct correlations21. The HTMT ratio, consistent with17, further validated construct distinctiveness. Multicollinearity was minimal (VIF: 1.308–2.936). These findings align with prior SEM-based agricultural adoption studies40, providing empirical insights into farmers’ perceptions of dairy interventions and strengthening methodological rigour in PLS-SEM applications. We conducted robustness checks to mitigate endogeneity in PLS-SEM using instrumental variable approaches, controlling for omitted variable bias, and applying the latent marker variable technique. Additionally, we tested for multicollinearity (VIF) and performed bootstrapping to ensure the stability and reliability of the results.

The hypotheses tested in this study supported these findings. The first hypothesis (H1) reveals a statistically significant influence of knowledge on DLI adoption, consistent with recent studies that emphasize the critical role of knowledge in technology adoption among farmers41,42 The second hypothesis (H2) found no statistically significant impact of attitude on the adoption of DLI, aligning with recent findings that attitude alone may not directly translate into adoption behavior without supportive external conditions43,44,45. The third hypothesis (H3) demonstrated a statistically significant impact of practice on the adoption of DLI, consistent with recent research highlighting that practical experience and resource availability significantly enhance the likelihood of adopting agricultural interventions5,46.

The study emphasizes the importance of comprehensive livestock interventions and their collective influence on food and nutrition security outcomes, particularly in rural and low-income communities where livestock production is vital for food, income, and employment. The findings underscore the significance of integrating various components such as LPGs, VMCCs, MVCs, DHs, and CBT to enhance livestock management practices, increase productivity, and improve the well-being of both livestock and farmers. Such interventions facilitate knowledge sharing, collective action, market access, and value addition, leading to higher incomes, improved food security, and better access to nutritious food. The discussion also refers to the work of47,48,49,50, which highlight the pivotal role of livestock in improving food security and income generation for farmers in developing countries like Bangladesh, further supporting the idea that dairy farming significantly contributes to ensuring food and nutrition security for rural households.This study offers novel insights into dairy farmers’ KAP regarding LDIs to enhance the FNS. This highlights the significant role of knowledge and practices in adopting LDIs, whereas attitudes showed no significant impact. This research underscores the importance of multifaceted approaches, integrating technical, market-oriented, and gender-focused strategies to improve livestock productivity, farmer livelihoods, and food security in rural communities. Methodologically, it strengthens the empirical rigor of PLS-SEM applications within agricultural adoption studies.

Conclusion and policy recommendations

Addressing FNS in Bangladesh through DLI requires a holistic approach that integrates farmers’ KAP. This study demonstrates that improved KAP significantly influences the adoption of sustainable dairy practices, leading to enhanced milk production, better input and fodder availability, and stronger market linkages. The findings reveal that interventions such as LPGs, VMCCs, MVCs, DHs, and CBT play a critical role in improving dietary diversity, household food security, and rural livelihoods. The study also emphasizes the need for targeted awareness campaigns, especially for older farmers, and advocates for a One Health approach fostering collaboration between veterinarians and public health officials to enhance livestock health and food security outcomes. Overall, this study reinforces the positive impact of DLI on food consumption, dietary diversity, and household food security. By improving farmers’ access to information, technology, and markets, the sector can contribute significantly to rural economic growth and national food security. To ensure the effective implementation of DLI and their contribution to FNS, the following policy recommendations are proposed;

  1. 1.

    Strengthen Extension Services: Targeted extension services that provide information and training on sustainable agricultural practices, livestock nutrition, and disease management can improve farmers’ knowledge and adoption of dairy interventions.

  2. 2.

    Support Capacity Building: Investing in capacity-building activities, including training programs on animal nutrition, breeding, and management, can enhance farmers’ skills and promote the adoption of innovative practices in dairy farming.

  3. 3.

    Facilitate Market Linkages: Establishing market connections for smallholder dairy farmers can improve their access to markets, increase income, and ultimately contribute to food security and nutrition outcomes.

  4. 4.

    Enhance Collaboration Between Stakeholders: Promoting collaboration between stakeholders in the livestock sector, such as livestock producer groups, village milk collection centers, and mobile veterinary clinics, can create synergies that benefit farmers and improve food and nutrition security.

  5. 5.

    Monitor and Evaluate Interventions: Regular monitoring and evaluation of DLI can help identify gaps and measure their impact on food and nutrition security, guiding future policy and program development.

These policy recommendations prioritize technology adoption, farmer education, and cross-sector collaboration to ensure the long-term sustainability of the dairy sector in Bangladesh. Strengthening farmers’ KAP through targeted interventions will not only boost income and food security at the household level but also contribute to national agricultural sustainability.

Limitations

The study, conducted in the Southern Delta region of Bangladesh with 400 farmers, delves into social capital and gender dynamics within smallholder farming. While it provides valuable insights, limitations include reliance on self-reported data, potential bias, and the cross-sectional nature of the analysis, allowing for associations but not causality. Despite these constraints, the research significantly contributes to understanding social capital dimensions in agricultural decision-making processes. Further studies should explore longitudinal trends in social capital, its causal impact on agricultural decisions, and regional variations across agro-ecological zones. Investigating social capital’s role in climate resilience and the impact of digital and financial inclusion on gendered decision-making is crucial. Research on government programs, cooperative models, and behavioral economics can provide deeper insights into peer influence on technology adoption and policy interventions to enhance social capital and gender equity in smallholder farming communities.