Livestock Farmers’ Intentions to Adopt Climate-Smart Agricultural Practices in Kenya’s Arid and Semi-Arid Lands: What Role Do Behavioural Factors Play?

Abstract

Pastoral livelihoods in Sub-Saharan Africa are under an increasing threat from climate change with arid and semi-arid lands (ASALs) being especially vulnerable. Climate-smart agriculture (CSA) is widely promoted as a strategy for enhancing resilience among smallholder livestock farmers by improving productivity, increasing farmers’ incomes and strengthening adaptive capacity. However, CSA adoption rates among pastoralists remains low. While existing studies emphasise socio-economic and institutional factors, this study explores the often-overlooked behavioural dimensions, attitudes, beliefs, and perceptions, which critically influence adaptation decisions. Guided by the theory of planned behaviour (TPB), this study investigates the behavioural drivers of CSA adoption among 737 livestock farmers in Kenya’s ASALs. Using ordered probit regression and structural equation modelling–confirmatory factor analysis (SEM-CFA), the results reveal that attitudes and perceived behavioural control are significant predictors of farmer intention to adopt CSA practices, with perceived behavioural control being the most influential predictor. Farmers with a positive attitude and confidence in their ability to implement CSA practices are more likely to adopt them. The study findings suggest that efforts to promote CSA adoption should prioritise transforming attitudes and building practical confidence by increasing exposure to demonstration farms and implementing awareness-raising initiatives within pastoral communities.

1. Introduction

Climate change is a global phenomenon that poses a serious threat to agriculture and food security. The rapid and uncertain weather changes, such as rising temperatures, unpredictable rainfall, prolonged droughts, and extreme weather events, continue to exacerbate the vulnerability of the agricultural systems globally [1]. While the impacts of climate change are global in nature, Sub-Saharan Africa (SSA) is particularly vulnerable due to its high dependence on rain-fed agriculture and the limited adaptive capacity of smallholder farmers [2,3]. In Kenya, this vulnerability is especially pronounced in arid and semi-arid lands (ASALs), which cover approximately 75% of the total country’s land area. This region is home to about 75% of total national livestock population, which is a key source of livelihood, supporting over 10 million people [4]. While the country’s livestock sector has a promising outlook driven by a rising demand for animal products, its sustainability is increasingly threatened by climate-related challenges, including recurrent droughts, erratic rainfall, and land degradation [5,6,7,8].

To address these challenges, climate-smart agriculture (CSA) has been championed by the Food and Agriculture Organization (FAO) and various development agencies as a key strategy for enhancing the resilience of agricultural systems [9,10,11]. CSA offers a holistic approach to addressing the challenges of climate change by simultaneously targeting three interlinked goals: sustainably increasing agricultural productivity and incomes; enhancing resilience to climate variability; and reducing greenhouse gas emissions [10]. While some investments have been made by the government and development partners over time through various policies and programmes [12,13,14,15], the uptake of CSA, particularly in ASAL regions, remains limited [2,5,8,16]. This highlights a disconnect between the promotion of CSA and the lived realities of pastoral communities, underscoring the need for a deeper analysis of adoption drivers and barriers.

The existing body of research on CSA adoption primarily concentrates on cropping systems [17,18,19,20], often with limited attention to pastoral livestock systems—despite their heightened vulnerability to climate change. This crop-centric focus fails to capture the distinct decision-making behaviours that differentiate livestock producers from crop farmers [2,8,21]. Additionally, a substantial body of literature [22,23,24,25] has mainly examined and identified a range of socio-demographic and socio-economic factors that significantly influence CSA adoption. While these extrinsic factors are important for shaping the adoption and scaling of climate-smart agricultural practices, they often overlook the behavioural processes that equally influence farmers’ adaptation decisions.

The existing evidence suggests that behavioural factors, such as attitudes, beliefs, and perceptions, could influence how farmers interpret climate change risks, assess adaptation strategies, and ultimately make behavioural decisions towards adaptation and mitigation [26,27,28]. Furthermore, the literature shows that farmers’ cognitive and emotional responses to climate change are largely shaped by their knowledge and perception of the adaptation strategies, their perceived ability to implement them, investment capacity and trust in the expected outcomes [3,26]. Despite their importance, these behavioural dimensions remain underexplored, particularly within pastoral communities in ASALs.

To address these gaps, this paper employs the theory of planned behaviour (TPB) to examine the behavioural factors that influence livestock farmers’ intentions to adopt CSA practices in Kenya. Specifically, the study aims to (1) assess livestock farmers’ intentions to adopt CSA practices (2) determine the influence of behavioural factors on livestock farmers’ intentions to adopt CSA practices, and (3) examine the interaction effects among these behavioural factors in shaping farmers’ intentions. Our study addresses the often-overlooked behavioural dimensions of CSA adoption, contributing to a better understanding of factors shaping adaptive decisions in climate-vulnerable livestock systems.

2. Materials and Methods

2.1. Study Area and Data Sampling

The study was conducted in three arid and semi-arid counties in Kenya–Kajiado (2°5′53.2″ N; 36°46′55.2″ E), Laikipia (0°21′38.2″ N; 36°46′55.2″ E), and TaitaTaveta (3°18′57.7″ S; 38°29′6″ E) (Figure 1). The selection of these counties was based on their dependence on livestock-based livelihoods, levels of vulnerability to climate change, and agro-ecological variability.

Kajiado, predominantly a semi-arid area, is highly drought-prone with limited crop farming potential, making pastoralism the main livelihood. It ranks among the top ASAL counties in livestock population but suffers frequent droughts, leading to over 70% livestock mortality and widespread crop failure. This exposes almost 80% of its households to food insecurity. Laikipia’s ecological zones range from dry lowlands to highland areas with mixed crop–livestock and wildlife systems. About 80% of its land is low in agricultural potential, and 43% of its population lives in poverty. Taita Taveta has relatively more humid highland zones suitable for mixed farming, alongside drier lowlands prone to drought. Despite this ecological diversity, 57% of its residents live in poverty. The county also grapples with rising human–wildlife conflict due to proximity to Tsavo National Park, compounding climate change-related stress [29,30,31].

A multistage sampling method was adopted in the study. Using the power sample size estimation formula [32], a sample size of 738 livestock–farming households were selected.

The power sample size estimation formula [33] used was as follows:

$$n={\displaystyle \frac{2{(Z_a+Z_{1-\beta })}^2{\sigma }^2}{{∆}^2}}$$

(1)

where:

$n$ = the required sample size;

$Z_a$= z-value corresponding to the significance level (α = 0.05) = 1.96;

$Z_{1-\beta }$ = z-value corresponding to desired power (80%) of the study = 0.8416;

$\sigma$ = the estimated standard deviation = 1.2;

$∆$= minimum detectable effect size = 0.175.

$$n={\displaystyle \frac{2{(1.96+0.8416)}^2{1.2}^2}{{0.175}^2}}=737.99\approx 738$$

The standard deviation (σ = 1.2) was derived from prior studies [34,35] involving similar constructs among smallholder farmers in Sub-Saharan Africa, where behavioural intention scores typically range between 1 and 5. The study adopted this conservative estimate based on the variability observed in intention scales in comparable TPB-based agricultural studies. This was due to the absence of local pilot data. The effect size (Δ = 0.175) represents the smallest meaningful change in the behavioural intention score that the study aimed to detect. Hence, this value was chosen in alignment with Cohen’s [32] criteria for a small-to-moderate effect in social science research and supported by previous research on CSA adoption based on TPB [19,36].

One observation was excluded from the analysis due to incomplete information, resulting in a final sample size of 737. The distribution across the three counties was as follows: Kajiado (n = 248), Laikipia (n = 231), and Taita Taveta (n = 258).

2.2. Theoretical Framework: Theory of Planned Behaviour (TPB)

The study followed the theory of planned behaviour (TPB), a socio-psychological model developed by [37]. TPB posits that individual behaviour is driven by behavioural intention, which in turn is shaped by three key constructs: attitude toward the behaviour, subjective norms, and perceived behavioural control [38]. The theory provides a robust and widely accepted framework for understanding and predicting human behaviour, especially in decision-making processes.

In this study, TPB was applied to assess livestock producing households’ intentions to adopt nine CSA practices grouped into three domains: nutrition and health management, breeding management, and risk reduction management. In this context, behavioural intention refers to the willingness of the farmers to adopt CSA practices. Attitude reflects the degree to which a farmer favourably or unfavourably evaluates a behavioural intention to adopt CSA practice(s), based on the perceived benefits associated with the adoption. Subjective norm, refers to a farmer’s perception of social influence exerted by significant others, such as family, friends, neighbours, and community, to take up or not to take up the CSA practice(s). Lastly, perceived behavioural control is the farmer’s belief in their capacity to implement CSA practice(s) shaped by past experiences and perceived capability [39].

There is a large and growing body of empirical research demonstrating the utility of the TPB in understanding the household’s decision behaviour in the context of smallholder agricultural systems. Studies by [20,35,36] have demonstrated that farmers’ attitudes towards innovation, their perceived ease or difficulty of adoption, and social influence are key predictors of behavioural intentions to adopting sustainable and climate resilient farming practices. In addition, Sarkar et al. [19] highlighted the effectiveness of the expanded TPB in explaining the key determinants of farmers’ willingness to adopt sustainable agricultural practices. Ultimately, the literature underscores the explanatory power of the TPB, highlighting the potential influence of its core elements—attitudes, subjective norms, and perceived behavioural control—on farmers’ adoption behaviour.

Although TPB has been extensively used in agricultural studies to explain CSA adoption [19,35,36], other competing theoretical frameworks, such as the Technology Acceptance Model (TAM) and the Value-Belief-Norm (VBN) theory, also exist. Technology adoption studies frequently employ the Technology Acceptance Model (TAM), which places a strong emphasis on perceived utility and usability. Compared to TAM and VBN, TPB offers a more comprehensive behavioural framework, particularly well-suited to the complex and context-dependent nature of CSA adoption. While TAM primarily emphasises perceived usefulness and ease of use, it often fails to account for broader socio-cultural and structural factors. In contrast, TPB incorporates social norms and perceived behavioural control—key determinants in agricultural decision-making, where choices are shaped by community expectations, institutional support, and resource availability. Similarly, although VBN captures moral obligations and environmental values, it tends to underrepresent self-interested and instrumental motivations that often drive CSA adoption [40]. Empirical comparisons of the TPB and VBN in a study by [40] on Chinese rice systems demonstrate that TPB explains a larger percentage of behaviour variability in adaptation contexts (42.1% vs. 28.4%), especially for self-interested behaviours like CSA adoption.

TPB’s advantage lies in its flexibility to incorporate additional constructs, such as socio-economic status, institutional access, and past experience, while preserving a distinct behavioural structure [34]. This makes it well-suited for contexts like Kenya’s ASALs, where farmers’ decisions are shaped by both internal and external factors like access to extension services, markets, and credit availability. Guided by the study objectives and the theory of planned behaviour, a set of hypotheses was formulated and tested to examine the determinants of livestock farmers’ intentions to adopt CSA practices. The conceptual framework (Figure 2) outlines these hypothesised relationships. Specifically, the study tested: H₀₁: livestock farmers’ attitudes towards CSA practices have no significant effect on farmers’ intentions to adopt CSA practices; H₀₂: livestock farmers’ subjective norms have no significant effect on farmers’ intentions to adopt CSA practices; H₀₃: livestock farmers perceived behavioural control has no significant effect on farmers’ intentions to adopt CSA practices. In addition, the interactions among these TPB constructs were also tested: H₀₄: attitudes and perceived behavioural control are not interconnected to influence intentions to adopt CSA practices; H₀₅: attitudes and subjective norms are not interconnected to influence intentions to adopt CSA practices; H₀₆: perceived behavioural control and subjective norms are not interconnected to influence intentions to adopt CSA practices.

2.3. Estimation Strategy

The analysis was conducted in two main stages. First, descriptive analysis was used to summarise the socio-economic characteristics of sampled households, assess their intention to adopt CSA practices, and examine the distribution of TPB constructs. This provided a foundational understanding of the context and variation within the data. In the second stage, we employed a dual modelling approach comprising an ordered probit model and structural equation modelling (SEM). The ordered probit model was used to estimate the effects of socio-economic and behavioural factors on adoption intentions. The SEM, integrated with confirmatory factor analysis (CFA), was applied to examine the interaction effects among TPB constructs in shaping farmers’ behavioural intentions. This combined approach provides deeper insight into the behavioural mechanisms underlying CSA adoption decisions.

2.3.1. Assessing Household Characteristics

Household characteristics were assessed to understand their potential influence on CSA adoption. The descriptive analysis employed frequency distributions, means, and standard deviations to summarise socio-economic characteristics and TPB constructs. Kruskal–Wallis test was used to assess variations in adoption across the counties.

2.3.2. Assessing Farmers’ Intentions to Adopt CSA Practices

Livestock farmers intentions to adopt CSA practices were measured using a 5-point Likert scale ranging from (1) strongly disagree to (5) strongly agree. An average intention score was calculated for each household based on their responses. Intention levels were categorised into three equal groups using Stata version 17.0 tertile command, which divides a continuous variable into three equal categories based on its distribution. Accordingly, the average household intentions were categorised into three levels; low (<3.35), medium (3.35 ≤ x < 3.5) and high (≥3.5). The intentions levels were analysed using Kruskal–Wallis test to determine whether there were any statistically differences among them. In addition, descriptive statistics to compared sampled livestock farmers across the three intention levels, with the Kruskal–Wallis test used to examine differences among these groups.

2.3.3. Assessing TPB Constructs

TPB constructs—attitude, subjective norms, and perceived behavioural control—were measured using multi-item Likert scale questions. A total of 29 initial items—9 items for attitude, 13 for subjective norms, 5 for perceived behavioural control, and 4 for adoption intention were subjected to confirmatory factor analysis. Their reliability and validity were assessed and only items that met established thresholds for eigenvalue, factor loadings, composite reliability, and average variance extracted were reatained. Descriptive statistics of the retained constructs were then analysed across farmers with low, medium, and high adoption intentions. This approach allowed an initial exploration of how these constructs vary with intention levels and also formed a foundation for the subsequent ordered probit and structural equation modeling. By summarising the mean scores and distributions of each construct across intention categories, the analysis offered insights into the behavioural dynamics that drive or hinder farmers’ readiness to adopt CSA practices.

2.3.4. Determinants of Livestock Farmers’ Intentions to Adopt CSA Practices

In modelling the key factors influencing livestock farmers’ intention to adopt CSA practices, a dual analytical approach was employed, combining ordered probit regression and structural equation modelling (SEM).

• Ordered Probit Regression

The ordered probit model was used to examine how observable socio-economic and institutional factors, as well as constructs from TPB, influence farmers’ adoption intentions. In the model, the outcome variable, farmers’ intentions are categorised into three ordered levels—low, medium, and high—which reflects an increasing degree of willingness to adopt CSA practices. This outcome variable was regressed against a set of explanatory variables, such as socio-economic characteristics and TPB constructs. The relationship is expressed as (Equation (2)):

$${INT}_i^{*}={\beta G}_i+{\beta Z}_i+{\epsilon }_i$$

(2)

where ${INT}_i^{*}$ is the farmers’ intention categorised into 3 (low, medium, high), β is a vector of parameters to be estimated, G is a vector of TPB constructs, Z is a vector of other socioeconomic and institutional variables, and ε is the error term.

The observed response categories (INT) that are tied to the latent variable satisfy the following model (Equation (3)):

$${INT}_i=\left\{\begin{array}{c}0, if {INT}_i^{*} \le A_1\\ 1, if A_1\le {INT}_i^{*}\\ 2, if {INT}_i^{*}>A_2\end{array}\right.\le A_2$$

(3)

where $A_1$ and $A_2$ are unknown cutoff parameters to be estimated with β.

The probability that consumer i will belong in group j is given by (Equation (4)):

$$Prob \left({INT}_i=j\right)=W\left(A_j- {\beta G}_i\right)-W\left(A_{j-1}- {\beta G}_i\right)$$

(4)

The marginal effect of the rth explanatory variable is calculated as follows (Equation (5)):

$${\displaystyle \frac{\mathsf{\partial }Prob\left({INT}_i=j\right) }{\mathsf{\partial }{G}_{ir}}}= W^{\prime }\left(A_j- {\beta G}_i\right)-W^{\prime }\left(A_{j-1}- {\beta G}_i\right){\beta }_r$$

(5)

where W is the standard normal CDF.

• Structural Equation Model–Confirmatory Factor Analysis

SEM was used to complement the ordered probit model and to provide a deeper understanding of the behavioural mechanisms while bringing out the predictive power and the interrelationships of the TPB constructs. CFA ensures that latent constructs are reliably and validly measured, strengthening the accuracy of the analysis. SEM allows for the simultaneous examination of multiple relationships between observed and latent variables, capturing complex interactions among behavioural factors. This method also accounts for measurement errors, providing more robust and nuanced insights into how attitudes, subjective norms, and perceived behavioural control influence farmers’ intentions to adopt CSA practices. Overall, SEM with CFA delivers a comprehensive and rigorous framework ideal for studying multifaceted decision-making processes.

The equation used is expressed as follows (Equation (6)):

$$INT={\beta }_1{ATT}_i+{\beta }_2{SBN}_i+{\beta }_3{PBC}_i+\mu$$

(6)

where $INT$ is the farmers’ adaption intention level, ${ATT}_i$ is the attitude, $SBN$ is the subjective norm, ${PBC}_i$ is the perceived behavioural control. Here, $\beta$ is an empirically estimated path coefficient for each of the constructs; and $\mu$ represents the disturbance term. This analysis was done for all categories of CSA—nutrition, breeding, and risk management categories.

3. Results and Discussion

3.1. Livestock Farming Household’s Characteristics

Table 1. Descriptive statistics of livestock farmers.

Variable	Variable Measurement	Counties			Pooled n = 737	K-Wallis Test
		Kajiado n = 248	T/Taveta n = 258	Laikipia n = 231
		Mean/% Frequency	Mean/% Frequency	Mean/% Frequency	Mean/% Frequency
Continuous Variables
Age of household head	Years	46.6	52.8	46.5	48.01 (14.14)	0.0000
Household size	Number	5.5	5.0	5.0	5.2 (2.1)	0.0000
Household TLU	Number	20.6217	16.6632	12.3142	16.6 (46.3)	0.0000
Years of L/S household farming expr	Years	21.69	18.09	19.93	19.4 (13.5)	0.0002
Categorical variables
Gender of household head	Female	17.7	24.0	26.4	22.66	0.0626
	Male	82.2	75.9	73.6	77.34
Education level of Household head	None	48.8	19.4	32.5	32.02	0.0005
	Primary	21.4	53.1	35.5	36.64
	Secondary	18.2	17.1	23.8	19.81
	Tertiary	11.7	10.5	8.2	11.53
Source of livelihood	Livestock rearing	89.1	36.1	66.7	63.5
	Other sources of livelihood	10.9	63.9	33.3	36.5
Types of livestock production system	Agro-pastoralism	8.06	46.5	18.2	24.7	0.0001
	Pastoralism	91.9	53.5	81.8	75.3
Land tenure	Own with title teed	65.3	58.5	15.2	53.7	0.0001
	Own without title teed	34.7	41.5	84.8	46.3
Access to extension services	Yes	51.2	63.2	51.5	67.6	0.0087
Access to credit	Yes	63.3	48.5	48.0	53.3	0.0006
Collective action participation	Yes	69.4	76.7	54.5	67.4	0.0001

Comparison of means based on Kruskal–Wallis rank sum test; (standard deviation values in parentheses).

shows the descriptive statistics of the socio-demographic characteristics of livestock farming households across the three ASAL counties—Kajiado, Taita Taveta, and Laikipia. These socio-demographic profile of farmers in the study area reveals important nuances that could influence CSA adoption patterns. Kruskal–Wallis test was undertaken to assess differences among the three counties.

The results show that the average age of household heads was 48 years, with Kajiado farmers being younger (46.6 years) than those in Taita Taveta (52.8 years). Differences in age suggest varying perspectives on agricultural innovation. Although a higher mean age of farmers may be an indication of more experience in livestock farming, it can also increase the risk averseness in the adoption of agricultural technologies [25,41]. This resonates with existing evidence that younger farmers embrace innovation and adaptive practices more readily than older ones. This could be due to their education levels and greater access to information [42].

The average household size was five persons, with significant variations across counties. Larger households may offer increased labour capacity, which is particularly beneficial for labour-intensive CSA practices. However, greater household size can also elevate resource demands, potentially straining financial resources and influencing decision-making priorities. Thus, household size may exert a dual influence—serving both as a labour asset and a financial constraint in the adoption of CSA practices.

The study considered the Tropical Livestock Unit (TLU), which is used to aggregate and compare different types of livestock based on their body weight or metabolic needs. The mean TLU for all the farming households was at 16.6 but with variations across the counties. It was highest in Kajiado (20.6) and lowest in Laikipia (12.3). A high herd size may indicate greater financial investment, potentially influencing farmers perceived value in adopting risk management-oriented CSA practices. Livestock farming experience also varied significantly (p = 0.0002), with Kajiado farmers reporting slightly longer experience than Taita–Taveta and Laikipia.

Regarding gender, majority of the household heads were male (77%) with differences across counties weakly significant (p = 0.0626). There was a notably higher proportion of female-headed households in Laikipia (26.4%) and Taita–Taveta (24.0%) compared to Kajiado (17.7%). Given that women play a critical role in household adaptation strategies but often lack equal access to key resources—such as land, credit, and extension services—targeted support for female farmers could enhance the effectiveness and equity of CSA interventions in these regions.

One-third of the respondents do not have formal education, with the highest being in Kajiado (48%). Slightly over one-third had secondary education with more than half of the respondents in Taita Taveta falling into this category. Overall, the results show that the levels of education are generally low in the study areas. Research [22,43,44,45] suggests that education may contribute to a better understanding of the benefits associated with CSA and enhance confidence in farmers’ adoption [22,42].

The main livelihood for most farmers was livestock rearing (64%). This was dominant in Kajiado (89.1%) compared to Laikipia (66.7%) and Taita–Taveta (36.1%) and with statistically significant differences. Likewise, pastoralism dominated in Kajiado (91.9%) and Laikipia (81.8%); however, agro-pastoralism was more common among the farmers in Taita-Taveta County. These differences could shape risk tolerance and the need to invest in livestock-related CSA. Borges et al. [46] suggested that farmers’ direct dependence on their livelihoods increases their likelihood of investing in its sustainability.

In terms of land tenure, almost half of the farmers owned lands without title deeds (46.3%), and notably, the majority (84.8%) of them were from Laikipia County. This finding has important implications for CSA adoption as land tenure security can influence the willingness of the farmer to invest in long-term CSA practices. Secure land rights have been widely associated with higher levels of investment in sustainable land management practices, which are the central pillar of CSA adoption.

Nearly half (45.2%) of the sampled households had received extension services in the year preceding the survey (2021). Extension is an important source of information on climate change and promotes adaptation and coping strategies [22,41]. It was also noted that more than two-thirds (67%) of the households had membership in community collective action. Participation in collective action was high across all the counties with Taita-Taveta County recording the highest participation of around 77%. This practice has been seen to enhance peer learning among the farmers, access to extension services through group training and therefore positively influence the adoption of the climate smart agricultural practices [18,27].

3.2. Categories of CSA Practices and Their Adoption Intentions

Table 2. Farmers intentions to adopt CSA practices.

CSA Categories	CSA Practices	Kajiado n = 248		T/Taveta n = 258		Laikipia n = 231		K-Wallis
		Mean	Std Dev	Mean	Std Dev	Mean	Std Dev
Nutrition and health management	Fodder conservation	3.489	0.348	3.224	0.547	3.119	0.503	0.0001
	Water harvesting and storage	3.484	0.364	3.395	0.443	3.416	0.371	0.0938
	Regular vaccination of livestock against key diseases	3.511	0.266	3.469	0.414	3.437	0.307	0.0201
Total Nutrition and Health Management		3.495	0.242	3.363	0.357	3.324	0.266	0.266
Breeding management	Rearing animals that are resilient to climate change, and choosing adaptive varieties	3.549	0.245	3.422	0.430	3.439	0.390	0.0053
	Selecting and maintaining adaptive breeds within the current herd	3.475	0.317	3.350	0.449	3.373	0.391	0.0058
Total Breeding Management		3.493	0.233	3.409	0.377	3.405	0.276	0.276
Risk reduction management	Maintaining a diverse range of livestock	3.487	0.290	3.434	0.410	3.468	0.354	0.2933
	Selling livestock at the beginning of drought to reduce losses (destocking)	3.411	0.524	3.346	0.433	3.325	0.490	0.0270
	Diversifying into crop production activities	3.445	0.395	3.368	0.452	3.102	0.483	0.0001
	Livestock insurance scheme	3.323	0.496	3.000	0.608	3.039	0.495	0.0001
Total Risk Reduction Management		3.416	3.416	0.288	3.287	0.326	3.233	0.288

presents the mean intention scores for various CSA practices, categorised under nutrition and health management, breeding management, and risk reduction management, across the three study counties. The intention scores, measured on a five-point Likert scale (1 = strongly disagree to 5 = strongly agree), reflect farmers’ willingness to adopt specific CSA practices. Results of the Kruskal–Wallis test are also reported to verify whether the differences in mean intention scores across counties are statistically significant. The analysis provides insights into county-specific patterns in CSA adoption intentions, which may be influenced by many factors including local agro-ecological, socio-economic, and institutional contexts.

The results indicate varying levels of farmers’ intention to adopt different CSA practices across the three counties. Overall, the mean intention scores are high, suggesting that farmers demonstrate a strong inclination toward implementing CSA measures, although with notable differences between practices and counties.

In nutrition and health management practices, farmers in all three counties reported a high intention to engage in fodder conservation, with the highest mean in Kajiado (3.489) and the lowest in Laikipia (3.119). This indicates statistically significant differences between counties (p = 0.0001), suggesting that factors such as forage availability, extension services, and drought experience may contribute to influence these disparities. Likewise, regular vaccination recorded high means across counties, with significant differences (p = 0.0201), perhaps reflecting varying levels of veterinary service access and livestock disease prevalence. On the contrary, water harvesting and storage had relatively smaller variation (p = 0.0938), implying comparable levels of intention across counties, possibly due to the universal recognition of the importance of water resource to the survival of livestock.

The intention to adopt practices such as rearing climate-resilient and adaptive livestock breeds was generally high, with Kajiado farmers showing the greatest inclination (mean = 3.549). The high means suggest that changing to more resilient breeds is widely recognised as a key adaptation strategy to climate change. The significant differences (p = 0.0053) may be attributed to regional differences in breed availability, market demand, and extension services.

Practices under the risk reduction category showed more variability. Keeping a variety of livestock was common across all counties with no significant difference (p = 0.2933). This implies that this practice could be viewed as a widely accepted risk mitigation approach. Conversely, the practice of selling of livestock at the onset of drought and diversification into crop production showed significant regional differences. Diversification of livelihoods into crop farming was markedly lower in Laikipia (3.102), likely reflecting farmers’ reluctance to adopt this practice due to prevalence of human wildlife conflict in the county. Notably adoption intention of livestock insurance scheme showed one of the lowest intentions overall with highly significant differences (p = 0.0001), likely due to limited awareness, affordability constraints, or low trust in insurance providers.

Overall, the highest levels of intention across the three CSA domains was observed in nutrition and health management as well as breeding management practices. Risk reduction practices such as livestock insurance had comparatively lower scores. The significant Kruskal–Wallis results for many practices highlight the role of local conditions, institutional support, and socio-economic factors in shaping farmers’ intentions.

Livestock farmers’ intentions to adopt each CSA practice were assessed and the average intention score was computed for each household. These average scores ranged from a minimum of 2.475 to a maximum of 4.075, and to facilitate analysis, the intention scores were categorised into three levels—low (<3.35; n = 248), medium (≥3.35 to <3.5; n = 294), and high (≥3.5; n = 195)—using Stata version 17.0 tertile command. Significant differences in the levels of intentions were observed across all categories of CSA practices, as indicated by the K-Wallis test with a p-value of 0.0001 for all comparisons (

Table 3. Distribution of farmers’ level of intentions across counties.

Levels of Intention	Kajiado n = 248	T/Taveta n = 258	Laikipia n = 231	K-Wallis
	% Frequency	% Frequency	% Frequency
Low	22.6	36.4	41.0	0.0001
Medium	32.8	38.2	28.9	0.0007
High	49.7	28.5	21.7	0.0001

).

3.3. Descriptive Analysis of TPB Elements

This section presents the descriptive analysis of the TPB constructs—attitude subjective norms and perceived behavioural control—across the levels of farmers’ intentions to adopt CSA practices (See

Table 4. Descriptive test statistics of TPB constructs based on the level of intentions.

TPB Constructs	Statements (Ordinal Items Ranging from (1) Strongly Disagree to (5) Strongly Agree)	Levels of Intentions			Pooled n = 737	K-Wallis Test
		Low n = 248	Medium n = 294	High n = 195
		Mean	Mean	Mean	Mean
Attitude towards CSA Practices (ATT)		3.659	3.788	4.154	3.834
ATT1	If I adopt CSA practices it will increase the production of my livestock	3.626	3.762	4.091	3.796	0.0001
ATT2	If I adopt CSA practices it will increase the survival of my livestock during dry season	3.660	3.786	4.156	3.834	0.0001
ATT3	Adoption of CSA practices on my farm is wise choice	3.691	3.817	4.216	3.872	0.0001
Subjective Norms		3.092	3.267	3.529	3.271
SBN1	My friends expect me to take up the CSA practices	3.141	3.355	3.514	3.319	0.0001
SBN2	My neighbours expect me to take up the CSA practices	3.083	3.275	3.443	3.250	0.0003
SBN3	My village elders expect me to take up the CSA practices	3.076	3.176	3.486	3.218	0.0001
SBN4	My local chiefs, community expect me to take up the CSA practices	3.061	3.224	3.589	3.258	0.0001
SBN5	My community expect me to take up the CSA practices	3.099	3.303	3.616	3.309	0.0001
Perceived Behavioural Control		3.401	3.712	3.862	3.651
PBC1	I have necessary resources which allows me to adopt the CSA practices on my farm	3.485	3.683	3.838	3.674	0.0001
PBC2	I can adopt the CSA practices on my farm without any problem	3.317	3.741	3.886	3.627	0.0001
Intentions		2.974	3.46	3.85	3.407
INT1	My regular use of this CSA practices is so far good	3.316	3.942	4.361	3.825	0.0001
INT2	I intend to continue using/surely adopt this CSA practices within the next 12 months	3.269	3.872	4.124	3.721	0.0001
INT3	I will be confidently recommend/inspire others to adopt the CSA	1.706	2.063	2.534	2.139	0.0001
INT4	I will be happy to discuss about the adoption CSA practices with other	3.606	3.963	4.381	3.941	0.0001

ATT—Attitude; SBN—Subjective Norm; PBC—Perceived Behavioural Control; INT—Intentions to adopt CSA.

). Understanding how these behavioural factors differ among farmers with low, medium, and high adoption intentions offers insights into the underlying behavioural dynamics influencing CSA uptake among the livestock farmers.

The descriptive analysis of TPB constructs across the levels of intention showed significant differences (p < 0.01 for all items). Livestock farmers with higher adoption intentions reported more favourable attitudes toward CSA practices, particularly in terms of perceived benefits, such as improved livestock productivity and resilience during dry seasons. The mean attitude scores increased consistently from the low- to high-intention level categories, demonstrating a positive relationship between perceived benefits and behavioural intent to adopt CSA practices. Likewise, subjective norms, which reflect social influence from different people, were stronger among farmers with medium and high levels of intention. Additionally, perceived behavioural control displayed an upward trend, with household in higher intention categories expressing greater confidence in their access to resources and ability to implement CSA practices without difficulty. However, a notable finding was the consistently lower mean score across all intention categories for the statement “I will confidently recommend/inspire others to adopt the CSA practices.” Even though farmers may demonstrate a personal willingness to adopt CSA, they may not be confident enough to promote them within their social networks. The hesitation is linked to limited hands-on experiences and lack of success stories or demonstrable outcomes and probably underlying uncertainty about the long-term viability of CSA practices in their specific contexts.

These findings reveal the relevance of all the TPB constructs in explaining behavioural intentions in the context of CSA adoption. They also play a critical role in shaping decision making around climate adaptation efforts among the farmers. Hence the need to strengthen all these dimensions promoting uptake of CSA practices among livestock farmers.

3.4. Determinants of Farmers’ Intentions to Adopt CSA Practices

To comprehensively examine the factors shaping farmers’ intentions to adopt CSA practices, this study employed two complementary analytical approaches: the ordered probit regression model and structural equation modelling (SEM). The ordered probit model quantifies the direct influence of individual socio-economic, institutional, and attitudinal variables on farmers’ intention levels, categorised as low, medium, or high. SEM, on the other hand, allows for simultaneous estimation of direct and indirect effects, capturing the interrelationships among the TPB constructs and their combined influence on intention. The subsequent subsections detail the findings from each approach.

3.4.1. Ordered Probit Results

In this section, we report the findings on the determinants of livestock farmers’ intention to adopt CSA practices using an ordered probit regression model. Based on this model, the dependent variable—the intention to adopt CSA level was categorised as low, medium, and high. Independent variables included behavioural constructs from the theory of planned behaviour, socio-economic characteristics, and institutional factors. In this model, positive coefficients indicate that an increase in the respective variable raises the likelihood of falling into a higher intention category, whereas negative coefficients indicate the reverse. Marginal effects presented for each predictor variable illustrate how one-unit change in an independent variable influences the probability of a farmer being in the low, medium, or high intention category (

Table 5. Ordered probit model results.

Variables	Coefficient	Std. Err.	Marginal Effects
			Low Intentions	Medium Intention	High Intention
TPB Constructs
Attitude	0.357 ***	0.069	−0.113 ***	0.016 ***	0.097 ***
Subjective norm	0.179 ***	0.050	−0.057 ***	0.008 ***	0.048 ***
Perceived behavioural control	0.262 ***	0.064	0.083 ***	0.012 ***	0.071 ***
Socio-economic Characteristics
Gender of household head	−0.118	0.116	0.037	−0.005	−0.032
Age of household head	−0.014 ***	0.004	0.004 ***	−0.000 ***	−0.004 ***
Household size	−0.007	0.023	0.002	−0.000	−0.001
Household Head’s Education Level (reference no formal education)
Primary	0.116	0.122	−0.038	0.009	0.030
Secondary	0.374 **	0.147	−0.120 **	0.016 **	0.103 **
Tertiary	0.432 **	0.172	−0.137 **	0.016 *	0.121 **
Household access to credit	−0.310 ***	0.096	0.098 ***	−0.014 **	−0.084 ***
Main Source of Livelihood (reference-livestock)
Other Sources of livelihood	−0.031	0.113	0.010	−0.001	−0.008
Collective action	−0.046	0.113	0.015	−0.002	−0.012
Household access to extension services	0.255 **	0.103	−0.081 **	0.012 **	0.069 **
Household TLU	0.000	0.001	−0.000	0.000	0.000
Years of livestock farming experience	0.008 *	0.004	−0.003 *	0.000 *	0.002 *
Livestock Production System (reference-agro-pastoralism)
Pastoralism	0.012	0.124	−0.004	0.001	0.003
Own land without title teed	−0.560 ***	0.097	0.184 ***	−0.031 ***
/cut1	1.787	0.470
/cut2	2.928	0.476
Pseudo R2	0.218
Likelihood ratio chi-square (χ2)	175.37
Prob > chi2	0.0000
Observation	737

*** p < 0.01; ** p < 0.05, * p < 0.1. Values indicate marginal effects for ordered probit regression that are significant.

).

The results in Table 5 show that attitude, subjective norms, and perceived behavioural control were statistically significant predictors of intention. The results corroborate with those of [46], whose findings reveal that all three TPB constructs were positive and significant in explaining the intentional behaviour of Brazilian cattle farmers to use improved natural grasslands. Similarly, the results of [42] confirm the importance of the three TPB constructs in predicting Iranian farmers’ intentions for safe us of chemical fertilizers.

A positive attitude toward the CSA significantly increased the likelihood of medium and high intention levels among the livestock farmers, while reducing the likelihood of low intention. Livestock farmers with favourable perceptions are more likely to consider adopting CSA practices. These findings are consistent with previous research [46,47,48,49] which identified a significant and positive relationship between attitude and behavioural intention in agricultural contexts. In contrast, Nguyen and Drakou [50] established that attitude was not a predictor of intentions to adopt sustainable agriculture among Vietnamese coffee farmers. This could arise from contextual differences in the farming systems, such as institutional support, cultural values, or the structure of farming systems, which play a great role in shaping how farmers process and act on attitudinal beliefs.

The findings also show that subjective norms have a positive and significant effects of farmers’ intention to take up CSA. Social influence from family members, friends, neighbours, co-farmers, and communities can shape adaptation decision making in the household. Consistent with previous studies [47,50,51,52,53], farmers are more likely to adopt CSA when they feel their decision is socially accepted or endorsed by their significant others. Contrary to the current study’s findings, Buyinza et al. [54] reported that subjective norm did not have a significant influence on smallholder farmers’ intention to adopt agroforestry practice within the Mt. Elgon region in Uganda. This could be attributed to the differences in social cohesion, varied strengths of communal structures, and differences in exposure to adaptation messages across the study contexts. Pastoral communities have strong social cohesion enabling decision making to be more collective and publicly visible—increasing social pressures. In contrast, in more individualised farming systems, social expectations may carry less influence over farmers’ behavioural intentions.

Perceived behavioural control has a significant and positive impact on livestock farmers’ intention to take up CSA practices. This finding suggests that the more they perceived to have the ability and capability to perform the CSA practice, the greater the intention towards CSA uptake. The studies by [46,47,50] also confirm that PBC has a positive and significant influence on intentional behaviour. In contrast, the findings of [49] report that PBC has no significant impact on intention. The possible explanation for this contradiction may be due to variations on how farmers conceptualise control whether related to physical resources, institutional support, or their confidence in carrying out the CSA practice(s).

The age of the household head had a negative and significant influence on the intention to adopt CSA practices, suggesting that older farmers are less receptive to change compared to their younger counterparts. This could be attributed to older farmers being entrenched in traditional practices, risk aversiveness, and low exposure to current CSA interventions. Similarly, access to credit was negatively associated with intention, indicating that farmers who accessed credit were less likely to intend to adopt CSA practices. This unexpected result may reflect a possibility of allocation of credit to non-agricultural uses or credit recipient being risk averse about investing in unfamiliar ventures such as CSA practices.

Land tenure exhibited a negative and significant relationship, with farmers lacking formal title deeds being less likely to intend to adopt CSA practices compared with those with secure land ownership. This underscores the importance of land tenure security as an incentive to long-term investments such as CSA. This finding is consistent with [55], who reported a positive correlation between farmland tenure and the adoption of soil and water management technologies among rice farmers in Vietnam. In contrast, the education level of the household head was positively and significantly associated with intention. This implies that more educated farmers are more likely to plan to adopt CSA practices, possibly because they are better positioned to understand the benefits of CSA, have better access to information, and improved decision-making skills. These findings are supported by studies such as [22,44,45] that link education with high technology adoption and greater adaptive capacity. Additionally, farming experience was positively and significantly linked to intention, indicating that more experienced farmers are more open to adopting CSA practices. However, this contrasts with [56], who reported a negative relationship between livestock farming experience and the intention to adopt insect-based feeds among dairy farmers in Kenya. The discrepancy noted in the two studies could reflect the differences in the innovation type being promoted. Insect-based feed may be viewed as an unusual and culturally unaccepted practice among the older and more experienced farmers, unlike the CSA practices being promoted in the study, which aligned with the existing norms, practices, and traditions, hence gaining wider acceptance.

3.4.2. Structural Equation Model (SEM) Estimate Results

This section presents the results of the structural equation model estimation by testing the predictive power of TPB constructs. The results presented in this section reflect standardised path coefficients, model fit indices, and significance levels. These provide a comprehensive assessment of the behavioural determinants of CSA intention, which allows us to confirm or reject the relevant null hypotheses. This section also offers important insights into the behavioural influences that facilitate or impede CSA adoption by livestock farmers, thereby informing the design of more impactful, farmer-centred interventions.

Prior to conducting SEM analysis, the robustness of the measurement model was done, where a total of 29 initial items were subjected to exploratory and confirmatory factor analysis. These included nine items for attitude, thirteen for subjective norms, five for perceived behavioural control (PBC), and four for adoption intention (INT). Items were evaluated based on established thresholds specifically a minimum eigenvalue threshold of 1.0 and factor loadings above 0.70 [34]. Items that did not meet these criteria were excluded. The final model retained three attitude items, five subjective norm statements, two PBC, and four intention statements, as shown in

Table A1. Data adequacy and consistency of measurement of the constructs.

Constructs		No. of Items	L	CR	AVE	α	Determinant Value	KMO	Bartlett’s Test
									X2	DF	Sig
Attitude towards CSA Practices (ATT)		3		0.940	0.839	0.93	0.074	0.742	1916.25	3	***
ATT1	If I adopt CSA practices it will increase the production of my livestock		0.910
ATT2	If I adopt CSA practices it will increase the survival of my livestock during dry season		0.933
ATT3	Adoption of CSA practices on my farm is wise choice		0.904
Possible Societal Influence/Expectation Subjective Norms (SBN)		5		0.962	0.836	0.95	0.005	0.889	3919.61	10	***
SBN2b	My friends expect me to take up the CSA practices		0.890
SBN2c	My neighbours expect me to take up the CSA practices		0.928
SBN2d	My village elders expect me to take up the CSA practices		0.938
SBN2e	My local chiefs, community expects me to take up the CSA practices		0.925
SBN2f	My community expects me to take up the CSA practices		0.889
Perceived Behavioural Control (PBC)		2		0.812	0.684	0.63	0.784	0.500	178.82	1	***
PBC1	I have necessary resources which allow me to adopt the CSA practices on my farm		0.871
PBC5	I can adopt the CSA practices on my farm without any problem		0.780
Adoption Intentions (INT)		4		0.919	0.738	0.90	0.077	0.784	1883.44	6	***
INT1	My regular use of this CSA practices is so far good		0.874
INT2	I intend to continue using/surely adopt this CSA practices within the next 12 months		0.868
INT3	I will be confidently recommend/inspire others to adopt the CSA		0.857
INT4	I will be happy to discuss about the adoption CSA practices with others		0.838
Model		14				0.83	0.000	0.839	827.47	91	***

*** p < 0.01 Values indicate structural equation model estimates that are significant; L—Factor loadings; AVE—Average Variance Extracted; CR—Construct Reliability, α—Cronbach’s alpha; KMO—Keser Meyer Olkin; ATT—Attitude; SBN—Subjective Norm; PBC—Perceived Behavioural Control.

. Despite the smaller number of retained items for the PBC, the construct showed satisfactory psychometric performance with strong factor loadings (0.871 and 0.780), acceptable composite reliability (CR = 0.812), and average variance extracted (AVE = 0.684). While Cronbach’s alpha (α = 0.63) was marginal, it remains acceptable in cases of two-item constructs when CR and AVE exceed recommended thresholds, as shown in Table A1 [57]. Although some factor loadings were high (ATT2 = 0.933; SBN2d = 0.938), they were theoretically distinct and did not exhibit multicollinearity or redundancy. According to Hair et al. [57], high loadings are not problematic if items reflect separate but complementary facets of the latent construct (Table A1). Altogether, the retained constructs explained approximately 70% of the total variance. The measurement model demonstrated adequate construct validity and internal consistency, justifying its use in the subsequent structural model analysis.

Upon confirmation of the measurement model’s validity and reliability, the structural model was estimated. Based on the study’s conceptual framework, standardised parameter estimates were computed to assess the relationships between the endogenous variables. These estimates formed the basis for testing the main null hypotheses within each category of CSA practices.

To evaluate the model’s overall fit, several Goodness-of-Fit (GoF) indices were assessed and found to be within acceptable thresholds. The Chi-square to degrees of freedom ratio (CMIN/df) was 3.28, within acceptable model fit (<5). The comparative fit index (CFI) was 0.933, and the Tucker–Lewis Index (TLI) was 0.923, both exceeding the recommended acceptable cut-off value of 0.90, suggesting a good degree of fit between the research framework and the data. Additionally, the root mean square error of approximation (RMSEA) was 0.060, within the acceptable range of 0.05 to 0.08, and the standardised root mean square residual (SRMR) was 0.058 below the 0.09 cut-off. Collectively, these indicators confirm that the structural model demonstrates a good overall fit to the data (Table A1).

Table 6. Structural model results: influence of TPB constructs across CSA practice categories.

CSA Categories	Nutrition Management		Breeding Management		Risk Management
Intentions <-	Coef.	Std. Err.	Coef.	Std. Err.	Coef.	Std. Err.
ATT	0.164 ***	0.264	0.139 ***	0.024	0.082 ***	0.024
SBN	0.025	0.019	0.035 **	0.018	0.104 ***	0.019
PBC	0.259 ***	0.045	0.169 ***	0.040	0.207 ***	0.041

*** p < 0.01; ** p < 0.05 Values indicate structural equation model estimates that are significant; ATT—Attitude; SBN—Subjective Norm; PBC—Perceived Behavioural Control.

shows the SEM estimates highlighting the differential influence of TPB constructs on farmers’ intentions to adopt CSA practices across CSA categories: nutrition management, breeding management, and risk management.

The results indicate that attitude and perceived behavioural control significantly and positively contribute to adoption intentions across all the three CSA categories, with PBC being most influential predictor. PBC coefficient scores were highest in the nutrition management category (coef = 0.259, p < 0.01), risk management (coef = 0.207, p < 0.01), and breeding management (coef = 0.169, p < 0.01). This underscores the important role of farmers’ perceived capability and confidence, which significantly influences the adoption of CSA practices. Other studies [19,46,50] have also demonstrated that PCB is an important predictor in the adoption of agricultural innovations. Conversely, while subjective norms had a significant influence in breeding (coef = 0.035, p < 0.05) and risk management (coef = 0.104, p < 0.01), it did not influence intention in the nutrition management category (coef = 0.025, ns). This suggests that social norms had limited influence on nutrition-related decisions, likely because such choices are more individualised and depend on farmers’ resource availability. Breeding and risk management practices, on the other hand, have practices such as adoption of drought resistant breeds and participation in vaccination campaigns that are often communal in nature due to shared livestock systems among the pastoralists. Consequently, these practices are subject to a stronger social influence. Previous studies [49,54] also demonstrated that the influence of social pressure on behavioural intention is contingent on factors, such as visibility of practice, the strength of shared norms, and the extent to which collective knowledge is applied. These results suggest that to accelerate CSA adoption, interventions must then be tailored to the behavioural and social factors that are most pertinent to each CSA domain. For nutrition management, interventions that will increase farmers’ self-efficacy and access to resources are likely to be more effective while mobilising social networks and community leaders would be more critical for breeding and risk management categories.

In addition to examining the direct effects of the TPB constructs on adoption intentions, the study tested the following null hypotheses: H₀₄: attitudes and perceived behavioural control do not jointly influence intentions to adopt CSA practices; H₀₅: attitudes and subjective norms do not work together to influence intentions to adopt CSA practices; H₀₆: perceived behavioural control and subjective norms do not jointly influence intentions to adopt CSA practices. These hypotheses were used to assess whether the constructs interact in a manner that could jointly influence farmers’ behavioural intentions to adopt CSA practices.

Table 7. Interrelationships among TPB constructs and their influence on CSA adoption intentions.

CSA Categories	Nutrition Management	Breeding Management	Risk Management
Relationship	Correlation’s Value	Correlation’s Value	Correlation’s Value
ATT <-> PBC	0.135 ***	0.117 ***	0.116 ***
ATT <-> SBN	0.004	0.004	0.004
SBN <-> PBC	0.034	0.025	0.025

*** p < 0.01. Values indicate structural equation model estimates that are significant; ATT—Attitude; SBN—Subjective Norm; PBC—Perceived Behavioural Control.

reports the correlation estimates among these constructs across the CSA categories.

Although the analysis shows positive interrelationships between ATT and PBC for all CSA categories, nutrition (r = 0.135 ***), breeding (r = 0.117 ***), and risk management (r = 0.116 ***), they are relatively small in magnitude and suggest a modest but consistent association among the constructs. This implies that farmers with positive attitudes may elevate self-efficacy, reinforcing their confidence to adopt CSA. This is in line with previous studies [19,39] that described a relationship of mutual reinforcement between attitude derived from knowledge and perceived behavioural control in relation to intention.

Conversely, no significant correlations were found between subjective norms and the other constructs (ATT and PBC) in all categories, and therefore, we were unable to reject hypotheses H₀₅ and H₀₆. This suggests that perceived social expectations are largely independent of farmers’ beliefs. Despite this weak interrelationship of SBN with both ATT and PBC, the findings in Table 7 show that subjective norms do have a significant and direct effect on the breeding and risk management practices. This suggests that social influence may affect intention directly, rather than through interaction with ATT or PBC, indicating that subjective norms operate independently of both attitude and perceived behavioural control in this case. The weak and negligible correlation may also point to salient contextual factors specific to breeding or risk management particularly in pastoral communities, where these practices are often collective than individualised. In summary, although SBN does not correlate strongly with attitude and perceived behavioural control, it still plays a vital role in shaping intentions, particularly in breeding and risk management practices. ATT and PBC, on the other hand, both have significant direct effects on intention and exhibit modest but positive interrelationship. Leveraging this relationship can help develop more effective and farmer-centred behavioural change intervention to promote CSA adoption among livestock farmers.

4. Conclusions, Policy Implications, and Recommendation

This study applied the theory of planned behaviour within a structural equation modelling framework to examine the factors influencing livestock farmers’ intentions to adopt CSA practices in Kenya’s arid and semi-arid lands. By quantifying the influence of each TPB construct, the analyses identified the elements of the constructs that act as stronger motivators or barriers to adoption participation, thereby informing the design of targeted interventions.

The findings indicate that while all the three TPB constructs significantly influence livestock farmers intentions to adopt CSA, perceived behavioural control was the most influential. This highlights the crucial role of farmers believing in their capabilities and access to resources in shaping their willingness to adopt CSA practices. Consistent to prior research, these results underscore the importance of enhancing capacity and reducing structural barriers to facilitate farmers uptake of CSA practices. Moreover, this study reveals a moderate and significant relationship between farmers’ attitudes and perceived behavioural control. This insight supports the need for integrated interventions that simultaneously strengthens positive attitudes while enhancing perceived behavioural control among the farmers.

Attitude constructs had a consistently significant and a positive effect across all CSA categories, reinforcing the idea that favourable perceptions shaped by knowledge, experience, and emotional assessment are essential precursors to behavioural intention. Subjective norms, while statistically significant in breeding and risk management categories, did not influence intention within nutrition category. This suggests that the influence of social pressure may be context-specific depending on the relevance of particular practices within the community.

To enhance CSA adoption, there is a need to cultivate positive attitude among farmers and boost their sense of capability. This can be achieved by exposing them to tangible, contextualised success stories, on-farm demonstrations, collaborative farmer-to-farmer learning and engagement, and tailored extension services. Such approaches enhance positive perception and empower farmers by showing them tangible outcomes, which can be a powerful catalyst for behavioural change.

The findings of this study point to the need for policy and development practitioners to prioritise building farmers’ capacities by improving knowledge and addressing structural and resource-based constraints. Scaling up targeted extension services, facilitating access to information, and promoting locally adapted CSA technologies can collectively enhance farmers’ capacity to act on their intentions. Such efforts will foster an enabling environment that supports both intention and sustained adoption of CSA practices in ASAL regions.