Effect of Climate Smart Agriculture Innovations on Climate Resilience among Smallholder Farmers: Empirical Evidence from the Choke Mountain Watershed of the Blue Nile Highlands of Ethiopia

Abstract

Smallholder farmers’ capacities need to be strengthened to enable them to better withstand the upcoming impacts of climate change; these capacities not only include the responsive capacity, but also consider innovation, learning, and anticipation to be prepared for the projected impacts of a changing climate on the agriculture system. The objective of this paper is to examine the impact of climate smart agriculture (CSA) innovations on building climate resilience capacity in smallholder agriculture systems. A cross-sectional household survey was conducted among a multi-stage sample of 424 smallholder farmers selected from five agroecosystems of the Upper Blue Nile Highlands in Ethiopia. The study used an endogenous switching regression (ESR) model to examine the impact of CSA innovations on building climate resilience capacity among smallholder farmers. The true average adoption effects of climate resilience capacity under actual and counterfactual conditions showed that different CSA innovations have different effects on the climate resilience capacity of households. Except for SWC adopters, all CSA innovations significantly increased the climate resilience capacity of households. However, improved variety, crop residue management, and SWC have more profound effects on the non-adopters than adopters, =if non-adopters had adopted these CSA innovations. Strong absorptive, adaptive, and transformative capacities through strong disaster and early warning systems, climate-resilient infrastructure, a strong public agricultural extension system, a strong informal safety net, and social networks build a climate-resilient agriculture system among smallholder farmers. Thus, scaling up of CSA innovations may expand the benefit of CSA innovation on building the climate resilience capacities of households. Thus, strong risk management, disaster mitigation and early warning systems, adaptive strategies, information and training, informal safety nets, social networks, and infrastructure use may build the climate resilience capacity of smallholder farmers by facilitating the adoption of CSA innovation. Therefore, policies that strengthen good governance, social cohesion, disaster communication and early warning systems, input supply of drought-resistant varieties, climate smart extension service, and climate-resilient infrastructure are necessary.

1. Introduction

Climate change significantly impacts smallholder farmers’ livelihoods because of heavy or erratic rainfall, temperature rise, sudden hailstorms, repeated droughts, and floods that will worsen in the future [1]. Even though Africa has contributed the least to greenhouse gas emissions, key development sectors have suffered widespread loss and damage due to anthropogenic climate change [2]. Climate projections show that the drought changes over East Africa follow a “dry gets drier and wet gets wetter” trend [3]. Hence, climate change is straining Africa’s agriculture, forestry, fisheries, and aquaculture [4].

Smallholder farmers are the primary victim of the adverse effects of climate change, as they lose food, water, and livelihood security [5]. Smallholder farmers’ capacities need to be strengthened to withstand the climate change–related stresses, shocks and impacts; these capacities include the responsive capacity to already known threats and should also consider innovation, learning, and anticipation for the projected impacts of a changing climate on the agriculture system [6,7].

Climate resilience is the ability of agriculture systems to absorb and recover from climatic shocks and stresses while positively adapting and transforming their structures and means of living in the face of long-term change and uncertainty [6,8,9,10]. It is a combination of an agriculture system’s absorptive, adaptive, and transformative capacities, which can be delimited based on the responses to the level of climatic shocks and stresses [8]. Hence, in this study, we defined climate resilience as a smallholder agriculture system’s capacity to persist, incrementally change, or transform in the face of persistent climate change stresses and shocks [8]. Hence, climate resilience building involves intervention that promotes absorptive, adaptive, and transformative capacities.

Absorptive capacity is similar to coping capacity, which refers to the ability of a social-ecological system, such as a smallholder agriculture system, to manage and recover from adverse climate change conditions using available skills and resources. Food security of the household will be primarily affected by climate risk shocks and stresses such as drought, so the absorptive capacity of the household toward food insecurity should be strengthened [11]. Thus, absorptive capacity of the household should include all available resources in the socio-ecological system [12]. However, adaptive capacity won’t be an option once the household has used all of its absorptive capacity [13]. Adaptive capacity includes the various adjustments and strategies that households undergo in order to maintain the sustainability of their livelihood [13]. This capacity is the ability to design and implement effective adaptation strategies or react to evolving hazards and stresses to reduce the likelihood of their occurrence and the magnitude of harmful outcomes resulting from climate-related hazards [14]. However, as the intensity of stress and shock increases beyond their adaptive capacity, households will need to resort to applying transformative capacity in order to survive. Transformative capacity deals with the ability of a social system to adapt to, anticipate, and absorb climate extremes and disasters by adopting transforming policies that change the institutional rules of the game [8].

Climate-resilient agriculture safeguards food security by enhancing smallholder farmers’ productivity and transforms the current system to withstand current and future climate change effects on smallholder farmers’ livelihoods [15]. Climate-resilient households are thus more active in anticipating, resisting, coping with, and recovering from the shock impacts of climate change and maintaining or improving their living standards [16]. Hence, building a climate-resilient agriculture system is a priority that policymakers should not overlook when facing the challenge of future and current climate change risks [16,17].

Several climate smart agriculture (CSA) innovations can deliver climate resilience outcomes [18,19]. For instance, adoption of drought-resistant, early maturing, and high-yield improved varieties [20], crop residue management, crop rotation, compost, agroforestry, as well as soil and water conservation structures may lessen the effect of drought through water management [21,22,23,24].

In Ethiopia, the impact of climate change is manifested mainly through drought and food insecurity. Since the 1970s, meteorological droughts and agriculture have resulted in chronic food insecurity [25]. Historical and more recent climate-related events such as the 2008/2009 and 2011 food security crises in the Horn of Africa as well as the 2015/2016 El Niño effect have highlighted the impact of droughts and floods on food production, access to markets, and income from agricultural activities [26,27,28].

Ethiopian agriculture is characterized by rain-fed subsistence farming, practiced on too small a land size to be viable, with a low yield, and exposed to climate change risk due to its reliance on timely and sufficient rainfall [29,30]. The overreliance on rain-fed smallholder agriculture, widespread poverty, and land degradation increase Ethiopia’s vulnerability to climate change and variability [31]. Hence, Ethiopia is one of the countries most vulnerable to climate change and with the least capacity to respond [30,32,33].

The Choke mountain watershed is located in Ethiopia’s Blue Nile Highlands and comprises six distinct agroecosystem zones [34]. Agriculture is the main economic activity and source of livelihood. A wheat-maize-teff-dominated mixed crop–livestock production system characterizes the farming system. The Ethiopian ard (or maresha), an ancient plough, is used for tillage, leading to high rates of on-field erosion, particularly on steep slopes [35]. Overgrazing and deforestation have also contributed to erosion, while soil fertility decline, livestock feed shortages (open grazing), and fuel wood demands continue to exert significant pressure on the resource base [36]. Land degradation–induced climate change risks pose significant challenges for Ethiopia’s Blue Nile Highlands [37]. Consequently, low agricultural productivity, severe land degradation, and climate change and variability threaten the livelihood of smallholder agriculture households [35].

Although several pieces of literature on climate resilience are found globally, their approach to conceptualizing and measuring climate resilience differ [7,38,39,40]. Tambo [41] used the climate resilience index to evaluate the climate resilience of Ghanaian districts. Most empirical studies on resilience to climate change defined resilience as the other side of vulnerability [42,43], sustainability to community-based institutions [44], adaptive capacity [45], and societal transformative capacity [46,47]. Most of the literature supports this definition, for example, “… the ability of a system to bounce back or return to equilibrium following disturbance …” [48]. However, we need to transform our definition of resilience in the face of climate change to embrace the ability of a system not simply to bounce back but also to adapt and to transform [13].

Recently, the idea of climate resilience as absorptive, adaptive, and transformative capacities has been gaining momentum [49,50,51]. Some of the literature has tried to assess building the climate resilience farming effect of push–pull technology (PPT) [15]. Yet, there is a dearth of literature on the concept of it absorptive, adaptive, and transformative capacities of climate resilience among smallholder farmers.

Hence, there is a pressing need to understand which CSA innovations have successfully built smallholder climate resilience capacity and how these capacities were built among smallholder agriculture households [52]. Therefore, this study aims to investigate the effect of CSA innovations in building the climate resilience (absorptive, adaptive, and transformative) capacity of smallholder farmers in the Upper Blue Nile Highlands of Ethiopia.

This study adopted an integrated social-ecological understanding of resilience for the analytical framework of climate-resilient agriculture [6,9,10,53]. Hence, according to the climate-resilient agriculture framework, adopting CSA innovations affects risk management, informal safety nets, disaster mitigation and early warning systems (DMEWS), adaptation strategies, wealth and income, food security, information and training, social networking, and infrastructure. These are subcomponents of the major component of climate resilience capacity, such as the smallholder agriculture system’s absorptive, adaptive, and transformative capacities. These absorptive, adaptive, and transformative capacities influence the climate resilience capacity of the smallholder agriculture household. Absorptive capacity also influences adaptive capacity as well as transformative capacity (Figure 1).

2. Methods

2.1. Study Area

The Choke mountain watershed is located between 9°38′00″ and 10°55′24″ north latitude and 37°07′00″ to 38°17′00″ east longitude. It has an elevation of 2100 to 4113 m above sea level, and the total land surface area of the watershed is approximately 15,950 km² with average annual rainfall of 200 to 2200 mm as well as average annual temperatures of 11.5 °C to 27.5 °C. The watershed has a slope gradient from flat to steep, and eight dominant soil types are found: Alisols, Andosols, Cambisols, Leptosols, Luvisols, Nitosols, Phaeozems, and Vertisols. The climate of the watershed ranges from the hot, arid climate of the Abay (Blue Nile) gorge to the cold and moist climate of the peak of the Choke Mountain [34].

2.2. Data Source, Sampling Design, and Data Collection

The sample size determination was calculated based on a finite population sample size calculation [54]. Because there had been no prior research on CSA innovation in the study area, the proportion of smallholder farmers who perceived CSA innovations as important innovations was assumed to be half of the population. We used the following formula:

$$\begin{array}{l}{n}_1=\frac{Z_{1-\alpha /2}^2}{d^2}P\left(1-P\right)\\ =\frac{{(1.96)}^2{(0.5)}^2}{{(0.05)}^2}\\ =385\end{array}$$

where n1 is sample size; Z_α/2 = 1.96 for 95 percent confidence interval; P is the proportion of the population who said CSA innovations are important for climate change adaptation; P = 0.5; and d is the error margin, taking d = 0.05. The study also assumed a 10 percent non-response rate, which equates to 39 households. The sample size then becomes 424 smallholder households.

A multi-stage sampling technique was used to randomly select 424 households from the five districts. The selection of the districts was through purposive sampling taking into consideration the agroecosystem zones they represent. The sampling frame was a one-to-five mobilization register obtained from the kebele extension officers. Second, one kebele from each woreda were randomly selected. The selected kebeles are Gelegele from Dejen, Enebi from Awobel, Limichim from Basoliben, Debere kelemu from Machakel, and Yeted from Sinan. In the second stage, a systematic random sampling technique was employed to select households from each of the five kebeles using a sampling frame of a one-to-five community mobilization group register. Finally, 424 households were randomly drawn from the sample kebeles on the basis of a probability proportional to size (PPS) sampling procedure (

Table 1. Sample woredas/districts and kebeles.

District/Woreda	Kebele	Population of HHs	Sample Size	Agroecosystem Zone (AESZ)
Dejen	Gelgele	7475	77	AESZ1: Lowland agroecosystem
Awabel	Enebi	5416	55	AESZ2: Midland with black soil
Basoliben	Limichim	10,147	104	AESZ3: Midland with brown soil
Machakel	Debre Kelemu	6207	63	AESZ4: Midland with sloping land
Sinan	Yeted	9533	125	AESZ5: The hilly and mountainous highland
Total		38,779	424

).

2.3. Measuring Climate Resilience

Resilience, a latent variable, is a dynamic multidimensional concept. In this study, the resilience tool proposed by the FAO is adapted to measure farm households’ resilience to climate change–induced shocks [55]. Adopting from the FAO’s analytical framework, climate resilience explains why one household is more resilient to climate change while another household is not. This analytical framework explains the interaction between shocks and their effects on households, with resilience accounting for the difference in outcomes between two similar households exposed to the same shock.

The resilience framework tool consists of three major components and 10 subcomponents. The three major components include absorptive capacity (ABP), adaptive capacity (ADP), and transformative capacity (TRAN). Further, these three major components were subdivided into the 10 subcomponents that include risk management (RM), social safety net (SSN), disaster mitigation and early warning system (DMEWS), adaptation strategies (ADPS), wealth (W), food security (FS), information and training (IT), social network (SN), and use of infrastructure (INFRA). Each of the 10 subcomponents has a specific set of variables collected from households that can convey climate resilience capacities of smallholder farmers. Hence, a climate resilience capacity index was constructed based on these indicators, as climate resilience is a function of these nine indicators that can be combined to give absorptive, adaptive, and transformative capacities of smallholder farmers:

CRi_j = f (RM_j, SSN_j, DMEWS_j, ADPS_j, W_j, FS_j, IT_j, SNj, INFRA_i)

(1)

where CRI_j is the climate resilience capacity index of household j, and RM_j, SSN_j, DMEWS_j, ADPS_j, W_j, FS_j, IT_j, SN_j, and INFRA_i are risk management (RM), social safety net (SSN), disaster mitigation and early warning system (DMEWS), adaptation strategies (ADPS), wealth (W), food security (FS), information and training (IT), social network (SN), and use of infrastructure (INFRA) subcomponent values, respectively, of household j for j = 1, …, n. Also, ABP_j = f(RM_j, SSN_j, DMEWS_j), ADP_j = f(ADPS_j, W_j, FS_j), and TRN_j = f(ES_j, SN_j, INFRA_i). To estimate CRi_i, it is necessary to estimate separately RM_j, SSN_j, DMEWS_j, ADPS_j, W_j, FS_j, IT_j, SN_j, and INFRA_i which are they latent variables (variables that are not measured or cannot be directly observed in a given survey rather than constructed from the set of indicators).

In this study, a two-stage procedure was used to estimate households’ CRi. During the first stage, nine resilience blocks were estimated using principal component analysis (PCA) based on 42 indicators. Likerat scale items were used to measure the different analytical constructs. A Cronbach’s alpha value of 0.90 for construct validity was obtained. In this study, PCA was used both for data reduction and identification of the dominant factors that explain a household’s resilience to climate change risks.

In order to obtain principal components, the study used a Kaiser Criterion of extracting factors with eigenvalues greater than 1. A varimax rotation technique was used for producing these constructs. Thus, we achieved the heaviest loading of principal component expressed in terms of the variables as an index for each household that captured the largest amount of information.

The resilience blocks include risk management (RM), social safety net (SSN), disaster mitigation and early warning system (DMEWS), adaptation strategies (ADPS), wealth (W), food security (FS), extension services (ES), social network (SN), and presence and use of infrastructure (INFRA). Based on the factor loading of each indicator, the resilience index of each individual household was computed using the PCA following Equation (2) as follows:

$$C_i={{\displaystyle \sum }}_{k=1}{f}_{ki}\left(\frac{\left(x_{ki}-\overline{x}\right)}{{\delta }_k}\right)$$

(2)

where f_ki is the component loading generated by PCA for the k^th variable of the i^th subcomponent; $x_{ki}$ is the observed value for the k^th variable of the i^th subcomponent; $\overline{x}$ and ${\delta }_k$ are the mean and standard deviations, respectively, of the k^th variable of the i^th subcomponent overall on households, and the value k varies according to the number of variables in the subcomponent.

In the second stage, the climate resilience capacity index is derived from a weighted average of the interacting components estimated in the first stage using the standardized value of the variables that created the factor loadings in the first stage. The CRi was then obtained from a weighted average of the 10 subcomponents using Equation (3):

$$\mathrm{CRi}=\frac{{{\displaystyle \sum }}_{\mathrm{i}=1}^{11}{\mathrm{w}}_{\mathrm{i}}{\mathrm{C}}_{\mathrm{i}}}{{{\displaystyle \sum }}_{\mathrm{i}=1}^{11}{\mathrm{w}}_{\mathrm{i}}}$$

(3)

where CRI is the climate resilience capacity index and w_i is the weight of the i^th component.

Since this is the first time that this type of analysis has been carried out in the study area, exploratory factor analysis (EFA) and the regression analysis were generated based on factorial punctuations (construct s index), including the factorial punctuation (index) of the dependent variable. Hence, the resilience index calculated for each household was considered as a dependent variable for further regression analysis.

Following the above argument, this study used 16 indicators that collectively form risk management, informal safety nets, and disaster mitigation and early warning systems subcomponents that make up the absorptive capacity of households; 12 indicators for adaptive strategies, wealth, and food security subcomponents that make up the adaptive capacity of households; and 14 indicators for information and training, social network, and infrastructure subcomponents that make up the transformative capacity of households (

Table 2. Climate resilience capacities indicators.

Major Components	Subcomponents	Indicators	Literature
Absorptive capacity	Risk management	Decrease quantity of meal	[56]
		Decrease diversity of meal	[56]
		Borrow grain	[57]
		Seek NGO help	[57]
		Borrow cash	[57]
		Sales of livestock	[57]
		Labor work	[57]
	Informal safety net	Social insurance	[58]
		Informal support	[58]
		Friend support	[58]
	Disaster mitigation and early warning system (DMEWS)	Social communication	[59]
		Mobile phone	[59]
		Soil fertility Parcel 1	[50,60]
		Soil fertility Parcel 2	[50,60]
		Slope of Parcel 1	[50,61]
		Slope of Parcel 2	[50,61]
Adaptive capacity	Adaptation strategies	Late sowing of crop	[62,63]
		Use of drought-resistant variety	[64]
		Use of crop residual for farm	[65]
		Use of water-harvesting technologies	[66]
	Wealth and income	Durable asset	[42]
		Farm size	[42]
		Livestock holding	[42]
		Number of plots	[42]
		Farm income	[42]
	Food security	Household Dietary Diversity Score (HDDS)	[67]
		Food consumption score (FCS)	[68]
		Crop diversity	[69]
Transformative capacity	Information and training	Access to extension service	[70]
		Extension service on IO	[70]
		Extension service on CC	[70]
		Extension service on productivity	[70]
		FTC training	[70]
	Social network	Participation in watershed management	[71]
		Membership in mahber	[72]
		Participation in Debo/Wenfel	[73]
	Infrastructure use	Access to health post	[74]
		Access to market	[75]
		Access to FTC	[70,76]
		Access to all-weather road	[77]
		Access to water and sanitation	[78]
		Access to school	[79]

).

Endogenous Switching Regression (ESR)

The endogenous switching regression (ESR) estimation model assumes that farmers that adopted CSA innovations may have systematically different characteristics from the farmers who did not adopt, and they may have decided to adopt based on expected outcomes they obtained from the adoption. Unobservable characteristics of farmers and their farm may affect both the adoption decision and its outcome, resulting in inconsistent estimates. For example, if only the wealthy, informed, skilled, or motivated farmers choose to adopt and the analysis fails to control for these factors, then upward bias will be incurred. Thus, the study accounts for the endogeneity of the adoption decision, i.e., controlling the effect of factors on the adoption decision and its outcome simultaneously, by estimating a simultaneous equations model of adoption of CSA innovation and its impact with ESR by the full information maximum likelihood (FIML) estimation method.

For the model to be identified, it is important to use instrumental variables as selection instruments, not only those automatically generated by the non-linearity of the selection model of adoption, but also other variables that directly affect adoption of CSA innovations but not the adoption’s impact. The study established the admissibility of these instruments by performing a simple falsification test: if a variable is a valid selection instrument, it will affect the adoption decision but it will not affect the climate resilience capacity [80].

$$U_{1i}=X_i{\beta }_1+{ϵ}_{1i}$$

(4)

$$U_{2i}=X_i{\beta }_2+{ϵ}_{2i}$$

(5)

$$G_i^{*}=\partial \left(U_{1i}-U_{2i}\right)+Z_i\alpha +u_i$$

(6)

Here, $G_i^{*}$ is a latent variable that determines the utility obtained whether the household i adopted a CSA innovation or not; $U_{ji}$ is the outcome variable value of a household i who adopted CSA innovation and j referes to the two regiems 1 and 2; and $Z_i$ is a vector of characteristics that influences the decision to adopt the innovation but not the outcome variable value. $X_i$ is a vector of household characteristics that is thought to influence the decision to adopt the innovation; β₁, β₂, and γ are vectors of parameters; and $u_i$, ${ϵ}_{1i}$, and ${ϵ}_{2i}$ are the error terms.

The regression model coefficient of adoption is $\alpha$, which measures the impact of adopting the innovation should be random. But in the case of adoption of CSA innovations, farmers freely choose the particular CSA innovation they want to adopt by their own consent. Hence, there is the problem of self-selection, which leads to selection bias. The decision to adopt a given innovation is likely to be affected by unobservable characteristics that may be correlated with the outcome variables (the climate resilience capacity index). Finally, the error terms in Equations (4)–(6) are assumed to have a trivariate normal distribution, with zero mean and covariance matrix, i.e., ($\upsilon$, ε₁, ε₂) ∼ N(0, $\mathsf{\Sigma }$):

$${{\displaystyle \sum }}^{}=\begin{array}{ccc}{\sigma }_{\upsilon }^2& {\sigma }_{\upsilon 1}& {\sigma }_{\upsilon 2}\\ {\sigma }_{1\upsilon }& {\sigma }_1^2& .\\ {\sigma }_{2\upsilon }& .& {\sigma }_2^2\end{array}$$

where ${\sigma }_{\upsilon }^2$ is the variance of the error term in the CSA adoption in Equation (6), which can be assumed to be equal to 1, since the coefficients are estimable only up to a scale factor; ${\sigma }_1^2$ and ${\sigma }_2^2$ are the variances of the error terms in the outcome variable functions (5) and (6); and ${\sigma }_{1\upsilon }$ and ${\sigma }_{2\upsilon }$ represent the covariance of $\upsilon$_i and ε_1i and ε_2i. Since Equations (4) and (5) are not observed simultaneously, the covariance between ε_1i and ε_2i is not defined (reported as dots in the covariance matrix). An important implication of the error structure is that because the error term of the selection Equation (6) $u_i$ is correlated with the error terms of the outcome variable functions (4) and (5) (ε_1i and ε_2i), the expected values of ε_1i and ε_2i conditional on the sample selection are nonzero:

$$\mathrm{E}({\epsilon }_{1i}|{\mathrm{G}}_{\mathrm{i}}=1)={\sigma }_{1\upsilon }\frac{\varphi \left(Z_i\alpha \right)}{\mathsf{\Phi }\left(Z_i\alpha \right)} ={\sigma }_{1\upsilon }{\lambda }_{1i}$$

$$\mathrm{E}({\epsilon }_{2i}|{\mathrm{G}}_{\mathrm{i}}=0)=-{\sigma }_{2\upsilon }\frac{\varphi \left(Z_i\alpha \right)}{1-\mathsf{\Phi }\left(Z_i\alpha \right)}={\sigma }_{2\upsilon }{\lambda }_{2i}$$

where $\varphi$ (.) is the standard normal probability density function, $\mathsf{\Phi }$(.) the standard normal cumulative density function, λ_1i = $\frac{\varphi \left(Z_i\alpha \right)}{\mathsf{\Phi }\left(Z_i\alpha \right)}$, and λ_2i = − $\frac{\varphi \left(Z_i\alpha \right)}{1-\mathsf{\Phi }\left(Z_i\alpha \right)}$. If the estimated covariance ${\sigma }_{1\upsilon }$ and ${\sigma }_{2\upsilon }$ are statistically significant, then the decision to adopt and the outcome variable are correlated, that is, evidence of endogenous switching was found, which rejects the null hypothesis of the absence of sample selectivity bias. This model is defined as a “switching regression model with endogenous switching” [81,82,83]. An efficient method to estimate endogenous switching regression models is the full information maximum likelihood estimation [83,84]. The logarithmic likelihood function given the previous assumptions regarding the distribution of the error terms is

$$\ln {\mathrm{L}}_{\mathrm{i}}={{\displaystyle \sum }}_{i=1}^N{A}_i[\ln \varphi (\frac{{ϵ}_{1i}}{{\sigma }_1})-\ln {\sigma }_1+\ln \mathsf{\Phi }({\theta }_{1i})]+(1-{\mathrm{A}}_{\mathrm{i}})[\ln \varphi (\frac{{ϵ}_{2i}}{{\sigma }_2})-\ln {\sigma }_2+\ln \left(1-\mathsf{\Phi }\left({\theta }_{2i}\right)\right)]$$

where θ_ji = $\frac{Z_i\alpha +\frac{{\epsilon }_{ij}}{{\sigma }_j}{\rho }_j}{\sqrt{(1-{\rho }_j^2})}$, j = 1, 2, with ρ_j denoting the correlation coefficient between the error term $u_i$ of the CSA innovation adoption Equation (6) and the error term ε_ji of Equations (4) and (5), respectively.

The ESR model can be used to compare the expected outcome variable of the farm households that adopt a particular innovation (a) with respect to the farm households that did not adopt (b), and to investigate the expected outcome variable result in the counterfactual hypothetical cases (c) that the adopted farm households did not adopt, and (d) that the non-adopted farm household adopted (

Table 3. Conditional Expectations, Treatment, and Heterogeneity Effects.

	Adoption Decision		Treatment Effect
	To Adopt	Not to Adopt
Adopters	Y11 = $\mathrm{E}(U_{1i}$|Gi = 1)	Y21 = $\mathrm{E}(U_{2i}$|Gi = 1)	ATT = Y11 − Y21
Non-adopters	Y10 = $\mathrm{E}(U_{1i}$|Gi = 0)	Y20 = $\mathrm{E}(U_{2i}$|Gi = 0)	ATU = Y10 − Y20
Heterogeneity effects	H1 = Y11 − Y10	H2 = Y21 − Y20	TH = ATT − ATU

Note: (Y11) and (Y20) represent observed expected CSA innovation adoption outcome variable; (Y10) and (Y21) represent counterfactual expected outcome variable G_i = 1 if farm households adopted CSA innovation; G_i = 0 if farm households did not adopt; Y_1i: the outcome variable if farm households adopted; Y2i: the outcome variable if farm households did not adopt; TT: the effect of adopting the innovation on the farm households that adopted the innovation; TU: the effect of adoption of the innovation on the untreated, i.e., farm households that did not adopt; Hi: the effect of base heterogeneity for farm households that adopted the innovation (i = 1), and did not adopt the innovation (i = 2); TH = (TT − TU), i.e., transitional heterogeneity [80].

).

$$\mathrm{E}(U_{1i}|{\mathrm{G}}_{\mathrm{i}}=1)=X_{1i}{\beta }_1+{\sigma }_{1\upsilon } {\mathsf{\lambda }}_{1\mathrm{i}}$$

(7a)

$$\mathrm{E}(U_{2i}|{\mathrm{G}}_{\mathrm{i}}=0)=X_{2i}{\beta }_2+{\sigma }_{2\upsilon } {\mathsf{\lambda }}_{2\mathrm{i}}$$

(7b)

$$\mathrm{E}(U_{2i}|{\mathrm{G}}_{\mathrm{i}}=1)=X_{1i}{\beta }_2+{\sigma }_{2\upsilon } {\mathsf{\lambda }}_{1\mathrm{i}}$$

(7c)

$$\mathrm{E}(U_{1i}|{\mathrm{G}}_{\mathrm{i}}=0)=X_{2i}{\beta }_1+{\sigma }_{1\upsilon } {\mathsf{\lambda }}_{2\mathrm{i}}$$

(7d)

Cases (a) and (b) along the diagonal of Table 2 represent the actual expectations observed in the sample. Cases (c) and (d) represent the counterfactual expected outcome variable. In addition, the effect of the treatment “to adopt” on the treated (ATT) as the difference between (a) and (c) were calculated as

$$\mathrm{ATT}=\mathrm{E}({\mathrm{U}}_{1\mathrm{i}}|{\mathrm{G}}_{\mathrm{i}}=1)-\mathrm{E}({\mathrm{U}}_{2\mathrm{i}}|{\mathrm{A}}_{\mathrm{i}}=1)={\mathrm{X}}_{1\mathrm{i}}({\mathsf{\beta }}_1-{\mathsf{\beta }}_2)+({\mathsf{\sigma }}_{1\mathsf{\upsilon }}-{\mathsf{\sigma }}_{2\mathsf{\upsilon }}){\mathsf{\lambda }}_{1\mathrm{i}}$$

(8)

which represents the effect of adoption of CSA innovations on the outcome variable result of the farm households that actually adopted a particular CSA technology [85,86]. Similarly, the effect of the treatment on the untreated (TU) for the farm households that actually did not adopt was calculated as the difference between (d) and (b):

$$\mathrm{TU}=\mathrm{E}({\mathrm{U}}_{1\mathrm{i}}|{\mathrm{A}}_{\mathrm{i}}=0)-\mathrm{E}({\mathrm{U}}_{2\mathrm{i}}|{\mathrm{A}}_{\mathrm{i}}=0)={\mathrm{X}}_{2\mathrm{i}}({\mathsf{\beta }}_1-{\mathsf{\beta }}_2)+({\mathsf{\sigma }}_{1\mathsf{\upsilon }}-{\mathsf{\sigma }}_{2\mathsf{\upsilon }}){\mathsf{\lambda }}_{2\mathrm{i}}$$

(9)

The expected outcomes described in Equations (7a)–(7d) also can be used to calculate the heterogeneity effects. For example, farm households that adopted may be better than farm households that did not adopt regardless of the fact that they decided to adopt but because of unobservable characteristics, such as their assets. The effect of base heterogeneity [87] for the group of farm households that decided to adopt as the difference between (a) and (d) can be calculated as:

$$\mathrm{BH}1=\mathrm{E}({\mathrm{U}}_{1\mathrm{i}}|{\mathrm{A}}_{\mathrm{i}}=1)-\mathrm{E}({\mathrm{U}}_{1\mathrm{i}}|{\mathrm{A}}_{\mathrm{i}}=0)=({\mathrm{X}}_{1\mathrm{i}}-{\mathrm{X}}_{2\mathrm{i}}){\mathsf{\beta }}_{1\mathrm{i}}+{\mathsf{\sigma }}_{1\mathsf{\upsilon }} ({\mathsf{\lambda }}_{1\mathrm{i}}-{\mathsf{\lambda }}_{2\mathrm{i}})$$

(10)

Similarly, for the group of farm households that decided not to adopt, the effect of base heterogeneity is the difference between (c) and (b):

$$\mathrm{BH}2=\mathrm{E}({\mathrm{U}}_{2\mathrm{i}}|{\mathrm{A}}_{\mathrm{i}}=1)-\mathrm{E}({\mathrm{U}}_{2\mathrm{i}}|{\mathrm{A}}_{\mathrm{i}}=0)=({\mathrm{X}}_{1\mathrm{i}}-{\mathrm{X}}_{2\mathrm{i}}){\mathsf{\beta }}_{2\mathrm{i}}+{\mathsf{\sigma }}_{2\mathsf{\upsilon }} ({\mathsf{\lambda }}_{1\mathrm{i}}-{\mathsf{\lambda }}_{2\mathrm{i}})$$

(11)

Finally, the transitional heterogeneity (TH) was investigated, that is, whether the effect of adopting the innovation is larger or smaller for farm households that actually adopted the innovation or for farm households that actually did not adopt, but in the counterfactual case if they did adopt, this is the difference between Equations (8) and (9) (i.e., TT and TU) (Table 3).

2.4. Data Analysis

The data were subjected to descriptive analysis in order to obtain frequencies and cross-tabulations that showed the data’s relationships. Mean, standard deviation, and percentage were utilized depending on the nature of the variable and the need for presentation. T-tests and chi-square tests were used to determine whether variations in the CSA innovation adoption on climate resilience were statistically significant. This study used the ESR model to estimate the effect of CSA innovations on climate resilience. The dependent variables for the ESR are the adoption of CSA innovations such as improved variety, crop residue management, crop rotation, compost, row planting, soil and water conservation (SWC), and agroforestry [33,42,88,89,90,91,92]. The independent variables were selected based on an extensive literature review on socio-demographic, economic, institutional, and environmental factors of climate resilience.

3. Results

3.1. Climate Resilience Capacities

This paper used a climate resilience index (CRi) as a proxy measurement of household climate resilience capacity. Mean comparison of adopters’ against non-adopters’ CRi of improved variety, crop residue management, crop rotation, compost, row planting, SWC, and agroforestry was conducted.

3.1.1. Comparison of Absorptive Capacity of the Adoption of CSA Innovations

Table 4. Comparison of absorptive capacity of the adoption of CSA innovations.

Indicators	Category	Improved Variety	Crop Residue Management	Crop Rotation	Compost	Row Planting	SWC	Agroforestry
Decrease quantity of meal	Adopter	0.027	−0.012	−0.02	−0.01	−0.02	0.01	−0.05
	Non-adopter	−0.012	0.011	0.012	0.01	0.07	−0.01	0.01
Decrease diversity of meal	Adopter	0.06	−0.01	−0.03	0.00	−0.01	0.01	−0.08
	Non-adopter	−0.03	0.01	0.02	0.01	0.04	−0.01	0.02
Borrow grain	Adopter	0.02	−0.07	0.04	0.00	0.03	0.02	−0.03
	Non-adopter	−0.01	0.06	−0.03	0.01	−0.10	−0.02	0.01
Seek NGO help	Adopter	−0.017	−0.042	0.001	−0.011	0.005	0.002	−0.055
	Non-adopter	0.007	0.036	−0.001	0.021	−0.016	−0.003	0.014
Borrow cash	Adopter	0.06	0.05	−0.04	0.02	0.01	−0.01	−0.07
	Non-adopter	−0.03	−0.04	0.02	−0.04	−0.01	0.01	0.02
Sales of livestock	Adopter	0.07	0.02	−0.02	0.01	−0.03	0.00	−0.05
	Non-adopter	−0.03	−0.02	0.01	−0.02	0.10	0.00	0.01
Labor work	Adopter	−0.01	0.08	−0.12	−0.01	−0.04	0.00	−0.01
	Non-adopter	0.00	−0.07	0.07	0.01	0.11	0.00	0.00
Risk management	Adopter	0.032	0.001	−0.025	0.001	−0.008	0.005	−0.049
	Non-adopter	−0.014	−0.001	0.014	−0.001	0.026	−0.005	0.013
		0.045 (0.036)	0.002 (0.033)	−0.039 (0.034)	0.001 (0.035)	−0.034 (0.039)	0.01 (0.03)	−0.061 (0.041)
Friend support	Adopter	−0.012	0.048	−0.065	0.009	−0.02	−0.004	−0.05
	Non-adopter	0.005	−0.041	0.037	−0.017	0.064	0.004	0.013
Informal social insurance	Adopter	0.027	0.035	−0.034	0.024	0.007	0.036	0.042
	Non-adopter	−0.003	−0.018	0.03	−0.026	0.004	−0.025	−0.003
Informal aid	Adopter	0.073	0.101	−0.036	0.018	0.004	−0.014	0.06
	Non-adopter	−0.032	−0.086	0.021	−0.036	−0.013	0.014	−0.016
Informal safety net	Adopter	0.03	0.06	−0.044	0.017	−0.003	0.007	0.017
	Non-adopter	−0.01	−0.048	0.029	−0.026	0.018	−0.003	−0.002
		0.04(0.031)	0.109 *** (0.028)	−0.074 ** (0.03)	0.043 (0.03)	−0.022 (0.034)	0.01 (0.03)	0.019 (0.036)
Social communication	Adopter	0.02	0.049	0.003	0.007	0.001	0.017	0.025
	Non-adopter	−0.009	−0.042	−0.002	−0.014	−0.002	−0.018	−0.007
Mobile phone communication	Adopter	0.035	0.017	0.022	0.001	0.000	0.005	0.013
	Non-adopter	−0.017	−0.016	−0.014	−0.005	−0.004	−0.007	−0.005
Soil fertility Parcel 1	Adopter	0.022	0.007	−0.001	0.022	0.003	−0.032	−0.026
	Non-adopter	−0.01	−0.007	0.001	−0.043	−0.007	0.032	0.007
Soil Fertility Parcel 2	Adopter	0.011	0.001	0.023	0.012	−0.006	−0.035	−0.05
	Non-adopter	−0.005	−0.001	−0.014	−0.023	0.02	0.036	0.013
Slope of Parcel 1	Adopter	0.132	0.088	0.049	0.033	0.002	−0.087	0.106
	Non-adopter	−0.058	−0.076	−0.028	−0.063	−0.005	0.089	−0.028
Slope of Parcel 2	Adopter	0.098	0.053	0.036	0.005	0.001	−0.038	0.076
	Non-adopter	−0.043	−0.045	−0.021	−0.009	−0.003	0.039	−0.02
Disaster mitigation and early warning system (DMEWS)	Adopter	0.052	0.036	0.022	0.013	0.000	−0.028	0.024
	Non-adopter	−0.024	−0.031	−0.013	−0.026	0.000	0.029	−0.007
		0.076 *** (0.025)	0.067 *** (0.023)	0.035 (0.024)	0.04 (0.025)	0 (0.027)	−0.06 ** (0.02)	0.03 (0.029)
Absorptive capacity index	Adopter	0.035	0.035	−0.024	0.01	−0.005	−0.001	−0.005
	Non-adopter	−0.014	−0.028	0.016	−0.017	0.018	0.003	0.003
		0.049 ** (0.022)	0.062 *** (0.02)	−0.04 * (0.021)	0.027 (0.021)	−0.022 (0.024)	−0.004 (0.02)	−0.008 (0.025)

Standard errors in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01.

shows that adopters of improved variety (p = 0.002) and crop residue management (p = 0.004) households significantly enhanced the disaster and early warning systems of their livelihoods through better mobile phone and social communication than non-adopter households.

Adopters of SWC households, on the other hand, have significantly worse disaster mitigation and early warning systems than non-adopters to deal with climate change–related floods because they live in mountainous areas where mobile and social communication are difficult (p = 0.015). Furthermore, farmers who practice crop residue management have a significantly better informal safety net in the form of community support from friends and neighbors in the event of a disaster (p < 0.001). Crop rotation adopters (p = 0.012), on the other hand, have a shaky informal safety net of community support. Thus, improved variety and crop residue management have a high capacity to absorb the effects of climate change–related disasters such as drought and flooding, whereas crop rotation adopters have a low capacity to withstand climate change–related disasters.

3.1.2. Comparison of Adaptive Capacity of the Adoption of CSA Innovations

Table 5. Comparison of adaptive capacity of the adoption of CSA innovations.

Indicators	Category	Improved Variety	Crop Residue Management	Crop Rotation	Compost	Row Planting	SWC	Agroforestry
Late sowing of crop	Adopter	−0.047	0.011	0.004	−0.009	−0.010	0.015	−0.054
	Non-adopter	0.021	−0.009	−0.002	0.018	0.033	−0.015	0.014
Use of drought-resistant variety	Adopter	0.005	0.041	−0.035	−0.030	−0.038	0.014	−0.097
	Non-adopter	−0.002	−0.035	0.021	0.058	0.122	−0.014	0.025
Use of crop residual for farm	Adopter	0.001	0.011	−0.007	−0.033	−0.002	0.025	−0.084
	Non-adopter	−0.001	−0.009	0.004	0.063	0.007	−0.025	0.022
Use of water-harvesting technologies	Adopter	−0.012	0.000	−0.012	−0.004	0.001	0.009	0.000
	Non-adopter	0.005	0.000	0.007	0.007	−0.002	−0.009	0.000
Adaptive strategies	Adopter	−0.014	0.016	−0.013	−0.018	−0.013	0.016	−0.059
	Non-adopter	0.006	−0.014	0.007	0.036	0.040	−0.016	0.015
		−0.02 (0.03)	0.03 (0.03)	−0.02 (0.03)	−0.06 (0.03)	−0.05 (0.04)	0.03 (0.03)	−0.07 * (0.04)
Durable asset	Adopter	0.104	0.065	0.094	0.080	0.038	0.072	0.164
	Non-adopter	−0.045	−0.056	−0.055	−0.155	−0.122	−0.074	−0.043
Total farm size	Adopter	0.146	0.115	0.115	0.052	−0.005	0.035	0.178
	Non-adopter	−0.065	−0.099	−0.066	−0.101	0.017	−0.036	−0.046
Livestock holding	Adopter	0.210	0.082	0.105	0.095	0.046	0.004	0.195
	Non-adopter	−0.092	−0.070	−0.060	−0.183	−0.147	−0.004	−0.051
Number of plots	Adopter	0.169	0.020	0.051	0.043	0.036	−0.025	0.159
	Non-adopter	−0.074	−0.017	−0.030	−0.082	−0.115	0.026	−0.041
Farm income	Adopter	0.242	0.131	0.108	0.094	0.041	−0.037	0.221
	Non-adopter	−0.106	−0.112	−0.062	−0.182	−0.131	0.038	−0.057
Wealth and Income	Adopter	0.175	0.083	0.094	0.072	0.031	0.009	0.184
	Non-adopter	−0.076	−0.071	−0.055	−0.141	−0.100	−0.010	−0.048
		0.25 *** (0.04)	0.15 *** (0.03)	0.15 *** (0.03)	0.21 *** (0.03)	0.13 *** (0.04)	0.02 (0.03)	0.23 *** (0.04)
Household Dietary Diversity Score (HDDS)	Adopter	0.196	0.114	0.061	0.100	0.084	−0.053	0.303
	Non-adopter	−0.086	−0.098	−0.036	−0.194	−0.266	0.055	−0.079
Food consumption score (FCS)	Adopter	0.157	0.258	−0.047	0.014	0.040	−0.177	0.048
	Non-adopter	−0.069	−0.221	0.027	−0.026	−0.128	0.181	−0.012
Crop diversity	Adopter	−0.015	0.008	0.014	0.001	0.000	0.009	−0.005
	Non-adopter	0.007	−0.007	−0.007	−0.003	−0.001	−0.009	0.002
Food security	Adopter	0.113	0.127	0.009	0.039	0.041	−0.073	0.115
	Non-adopter	−0.050	−0.109	−0.005	−0.074	−0.132	0.075	−0.030
		0.16 *** (0.04)	0.24 *** (0.04)	0.02 (0.04)	0.11 *** (0.04)	0.17 *** (0.04)	−0.2 *** (0.04)	0.15 *** (0.04)
Adaptive capacity index	Adopter	0.079	0.068	0.025	0.025	0.016	−0.013	0.063
	Non-adopter	−0.035	−0.058	−0.015	−0.048	−0.051	0.013	−0.017
		0.11 *** (0.02)	0.13 *** (0.02)	0.04 * (0.02)	0.07 *** (0.02)	0.07 *** (0.02)	−0.025 (0.02)	0.08 *** (0.03)

Standard errors in parentheses * p < 0.1, *** p < 0.01.

shows that adopters of improved variety, crop residue management, compost, row planting, and agroforestry have significantly increased the wealth, income, and food security of households with a greater number of plots, livestock holdings, and farm income (p < 0.001).

3.1.3. Comparison of Transformative Capacity of the Adoption of CSA Innovations

Table 6. Comparison of transformative capacity of the adoption of CSA innovations.

Indicators	Category	Improved Variety	Crop Residue Management	Crop Rotation	Compost	Row Planting	SWC	Agroforestry
Access to extension service	Adopter	0.11	−0.08	0.09	0.05	0.06	0.05	0.04
	Non-adopter	−0.05	0.07	−0.05	−0.10	−0.18	−0.05	−0.01
Extension service on farm input	Adopter	0.48	0.26	0.37	0.32	0.35	0.26	0.35
	Non-adopter	0.20	0.31	0.24	0.22	0.09	0.32	0.27
Extension service on climate change	Adopter	0.01	0.03	0.03	0.04	0.03	0.02	0.07
	Non-adopter	0.00	−0.02	−0.02	−0.08	−0.10	−0.02	−0.02
Extension service on productivity	Adopter	0.13	0.06	0.04	0.06	0.05	−0.03	0.13
	Non-adopter	−0.06	−0.05	−0.03	−0.12	−0.15	0.04	−0.03
Farmers’ Training Center (FTC) training	Adopter	0.01	−0.01	0.00	0.01	0.00	0.00	0.01
	Non-adopter	0.00	0.00	0.00	−0.01	0.00	0.00	0.00
Information and Training	Adopter	0.146	0.052	0.106	0.097	0.097	0.057	0.118
	Non-adopter	0.018	0.063	0.03	−0.019	−0.069	0.058	0.042
	Diff	0.13 *** (0.03)	−0.01 (0.03)	0.08 ** (0.038)	0.12 *** (0.03)	0.165 *** (0.033)	0 (0.029)	0.076 * (0.036)
Participation in watershed management	Adopter	0.031	0.029	0.03	0.03	0.033	0.029	0.036
	Non-adopter	0.05	−0.03	0.05	0.03	0.12	−0.04	0.01
Membership in mahber	Adopter	0.04	−0.03	0.09	−0.09	0.00	−0.10	−0.02
	Non-adopter	−0.02	0.03	−0.05	0.18	−0.01	0.11	0.00
Participation in debo/wenfel	Adopter	−0.01	0.00	−0.01	0.04	0.02	0.05	0.04
	Non-adopter	0.01	0.00	0.01	−0.07	−0.08	−0.05	−0.01
Social Networks	Adopter	−0.029	0.003	−0.003	−0.024	−0.003	−0.005	−0.002
	Non-adopter	0.013	−0.003	0.002	0.045	0.009	0.005	0.001
	Diff	−0.04 (0.04)	0.005 (0.04)	−0.01 (0.04)	−0.07 * (0.03)	−0.013 (0.043)	−0.01 (0.036)	−0.003 (0.045)
Access to health post	Adopter	−0.004	0.01	0.03	−0.01	−0.02	0.02	−0.02
	Non-adopter	0.002	−0.01	−0.02	0.02	0.05	−0.03	0.01
Access to market	Adopter	−0.03	−0.02	0.02	0.04	0.01	0.05	−0.01
	Non-adopter	0.02	0.01	−0.01	−0.07	−0.02	−0.05	0.00
Access to FTC	Adopter	0.05	0.02	−0.01	−0.004	0.01	−0.06	0.06
	Non-adopter	−0.02	−0.02	0.004	0.01	−0.04	0.06	−0.02
Access to all-weather road	Adopter	0.03	−0.05	−0.03	−0.01	0.04	−0.06	−0.02
	Non-adopter	−0.01	0.04	0.02	0.01	−0.12	0.06	0.01
Access to water and sanitation	Adopter	0.04	−0.03	−0.01	0.01	0.02	−0.02	0.02
	Non-adopter	−0.02	0.02	0.01	−0.02	−0.06	0.02	−0.01
Access to school	Adopter	−0.01	0.01	0.01	0.01	0.00	0.00	0.03
	Non-adopter	0.01	−0.01	−0.01	−0.01	−0.01	0.00	−0.01
Infrastructure use	Adopter	0.011	−0.008	0.003	0.005	0.010	−0.010	0.009
	Non-adopter	−0.005	0.007	−0.002	−0.011	−0.032	0.010	−0.003
	Diff	0.02 (0.02)	−0.02 (0.02)	0.004 (0.02)	0.02 (0.02)	0.042 * (0.025)	−0.02 (0.021)	0.012 (0.026)
Transformative capacity index	Adopter	0.039	0.016	0.034	0.024	0.033	0.015	0.040
	Non-adopter	0.010	0.022	0.010	0.009	−0.027	0.024	0.014
		0.03 (0.02)	−0.005 (0.02)	0.02 (0.02)	0.014 (0.021)	0.06 *** (0.022)	−0.009 (0.02)	0.027 (0.024)

Standard errors in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01.

shows that households that adopted improved variety, crop rotation, compost, row planting, and agroforestry have a significantly higher information and training index (p < 0.001) due to higher public extension and FTC training scores for adopters of these CSA innovations, while compost adopter households have a marginally poor social network (p < 0.1) due to lower membership in mahber, which is the dominant social network structure in the study area. Row-planting adopters have a marginally significant higher score for infrastructure use than non-adopters, as the latter have better access to basic services such as markets, farmers’ training centers (FTC), all-weather roads, water and sanitation, and schools. Hence, higher information and training indexes through a strong public agricultural extension system as well as strong infrastructure use led to higher transformative capacity among row-planting adopters.

3.1.4. Comparison of Climate Resilience Capacity of the Adoption of CSA Innovations

Table 7. Comparison of climate resilience capacity of the adoption of CSA innovations.

CSA Innovations	Category	Absorptive Capacity Index	Adaptive Capacity Index	Transformative Capacity Index	Climate Resilience Capacity Index
Improved variety	Adopter	0.035	0.079	0.039	0.055
	Non-adopter	−0.014	−0.035	0.010	−0.016
	Diff	0.049 ** (0.022)	0.113 **** (0.022)	0.029 (0.021)	0.07 ** (0.015)
Crop residue management	Adopter	0.035	0.068	0.016	0.043
	Non-adopter	−0.028	−0.058	0.022	−0.026
	Diff	0.062 *** (0.02)	0.126 *** (0.02)	−0.005 (0.02)	0.07 *** (0.013)
Crop rotation	Adopter	−0.024	0.025	0.034	0.014
	Non-adopter	0.016	−0.015	0.010	0.002
	Diff	−0.04 * (0.021)	0.04 * (0.021)	0.024 (0.02)	0.012 (0.014)
Compost	Adopter	0.01	0.025	0.024	0.020
	Non-adopter	−0.017	−0.048	0.009	−0.022
	Diff	0.027 (0.021)	0.072 *** (0.021)	0.014 (0.021)	0.042 *** (0.02)
Row planting	Adopter	−0.005	0.016	0.033	0.016
	Non-adopter	0.018	−0.051	−0.027	−0.024
	Diff	−0.022 (0.024)	0.067 *** (0.024)	0.06 *** (0.022)	0.04 ** (0.016)
SWC	Adopter	−0.001	−0.013	0.015	−0.002
	Non-adopter	0.003	0.013	0.024	0.013
	Diff	−0.004 (0.02)	−0.025 (0.02)	−0.009 (0.02)	−0.015 (0.014)
Agroforestry	Adopter	−0.005	0.063	0.040	0.036
	Non-adopter	0.003	−0.017	0.014	−0.002
	Diff	−0.008 (0.025)	0.08 *** (0.025)	0.027 (0.024)	0.039 ** (0.017)

Standard errors in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01, **** p < 0.001.

shows that households with improved variety, crop residue management, compost, row planting, and agroforestry have a significantly higher climate resilience capacity index (p < 0.001).

3.2. Effect of CSA Innovations on Climate Resilience Capacity

Table 8. Adoption effect of CSA innovations on average expected climate resilience.

CSA Innovations	Households	Adoption Decision: to Adopt (A)	Adoption Decision: Not to Adopt (B)	Impact of CSA Innovation on Climate Resilience Capacity Index (CRi) (ESR) (C)	Heterogeneity Effect
Improved variety	Adopters	0.054	−0.132	ATT = 0.185 (0.01) ***	−0.040 (0.01) ***
	Non-adopters	0.123	−0.016	ATU = 0.139 (0.009) ***
Crop residue management	Adopters	0.043	0.055	ATT = −0.012 (0.01)	−0.041 (0.01) ***
	Non-adopters	0.041	−0.026	ATU = 0.066 (0.009) ***
Crop rotation	Adopters	0.013	−0.192	ATT = 0.205 (0.012) ***	0.018 (0.01) *
	Non-adopters	0.093	0.002	ATU = 0.091 (0.01) ***
Compost	Adopters	0.020	−0.040	ATT = 0.06 (0.008) ***	−0.007 (0.007)
	Non-adopters	0.074	−0.022	ATU = 0.096 (0.012) ***
Row planting	Adopters	0.015	−0.047	ATT = 0.062 (0.008) ***	0.035 (0.01) ***
	Non-adopters	0.029	−0.024	ATU = 0.053 (0.015) ***
SWC	Adopters	0.002	0.064	ATT = −0.062 (0.008) ***	−0.082 (0.01) ***
	Non-adopters	−0.217	0.013	ATU = −0.230 (0.01) ***
Agroforestry	Adopters	0.035	−0.127	ATT = 0.160 (0.02) ***	0.069 (0.01) ***
	Non-adopters	0.127	−0.002	ATU = 0.129 (0.008) ***

Standard errors in parentheses * p < 0.1, *** p < 0.01.

presents the true average adoption effects of climate resilience capacity under actual and counterfactual conditions. In this table, the climate resilience capacity index (CRi) of farm households who adopted the CSA innovation were compared with the outcome variables that would have been found if the households had not adopted. In order to determine the average adoption effects, the study compared Columns A and B of Table 8. Column C presents the impacts of the adoption of a CSA innovation on the climate resilience capacity, computed as the difference between Columns A and B.

The treatment effect (ATT) result indicated that adoption of improved variety has increased the climate resilience capacity of the adopters by 18.5%, while for the non-adopters, climate resilience capacity would have increased by 13.9% if they had adopted improved variety. However, the negative sign of treatment heterogeneity effect shows that adoption of improved variety is more pronounced for adopters than non-adopters, i.e., some characteristics of non-adopters have made the effect of adoption of improved variety more appropriate for non-adopters than actual adopters (p < 0.001). Hence, the climate resilience capacity of counterfactual adopters’ households would have increase by 4% if they had adopted improved variety, which supports the study by [20], who reported that drought-resistant improved variety has increased the climate resilience of smallholder farmers. Adoption of crop residue management has an insignificant effect on the climate resilience capacity of actual adopters (p = 0.26), while for the non-adopters of crop residue management, the climate resilience capacity would have increased by 6.6% had they adopted crop residue management. The negative effect of adoption of crop residue management shows that some characteristics of non-adopters have made the effect of adoption of crop residue management more appropriate for non-adopters than actual adopters (p < 0.001). The adoption treatment effect (ATT) of crop rotation indicated that adoption has increased the climate resilience capacity of the adopters by 20.5%, while for the non-adopters, climate resilience capacity would have increased by 9.1% if they had adopted crop rotation. Hence, the positive sign of treatment heterogeneity effect shows that adoption of crop rotation has a more pronounced effect for adopters than non-adopters (p < 0.001). This supports a similar study in China [93] that reported that by improving technical efficiency, crop rotation creates a stronger agricultural production system that is resilient to climate risks. Hence, the climate resilience capacity of actual as well as counterfactual adopters’ households would increase by 1.8%. The adoption treatment effect (ATT) of compost showed that adoption has increased the climate resilience capacity of the adopters by 6.0%, while for non-adopters, climate resilience capacity would have increased by 9.6% if they had adopted compost. The adoption treatment effect (ATT) of row planting indicated that adoption has increased the climate resilience capacity of the adopters by 6.2%, while for the non-adopters, climate resilience capacity would have increased by 5.3% if they had adopted row planting. Hence, the positive sign of treatment heterogeneity effect shows that adoption of row planting has a more pronounced effect for adopters than non-adopters (p < 0.001). Thus, the climate resilience capacity of actual as well as counterfactual adopters’ households would increase by 3.5%. The treatment effect (ATT) result indicated that adoption of SWC has reduced the climate resilience capacity of the adopters by 6.2% while for the non-adopters, climate resilience capacity would have reduced by 23.0% if they had adopted SWC. Hence, the climate resilience capacity of actual as well as counterfactual adopters’ households would have reduced by 8.2% if they had adopted SWC. The adoption treatment effect (ATT) of agroforestry indicated that adoption has increased the climate resilience capacity of the adopters by 16.0%, while for the non-adopters, climate resilience capacity would have increased by 12.9% if they had adopted agroforestry. Hence, the climate resilience capacity of actual as well as counterfactual adopters’ households would increase by 6.9%.

The literature on the impact of CSA innovations on climate resilience in Ethiopia is limited [42]. Hailemariam et al. [42] investigated the impact of the adoption of improved varieties, chemical fertilizer (a risk multiplier), and water management (a risk reducer) on agricultural revenue; however, they did not assess the capacity of smallholder farmers for climate resilience. Even though they evaluated the smallholder farmers’ potential for climate resilience through absorptive, adaptive, and transformative capacities, Asmamaw et al. [50] contrasted the agroecology capacity of the agriculture system rather than the effect of technology or innovations the farmers adopted.

4. Discussions

Climate change–induced hazards such as drought, floods, hailstorms, and erratic rainfall have been happening in Ethiopia. Such climate shocks disproportionately affect farmers with low adaptive capacities, with varying degrees of severity. The extent of the impact is further magnified when shocks hit households with different resilience capacities. Importantly, this study concurs with the finding by [94], who reported that information on the occurrence of a climate shock such as floods increases the climate resilience capacity of farmers in Ghana. Moreover, ref. [95] reported that mobile phone technologies can be used to improve inclusivity and local knowledge production for disaster risk mitigation systems in resilience building. Ref. [96] also support mobile phone usage during disaster preparedness as a factor for increased resilience, by improving mobile phone messaging to be used for communication during disasters as well as by establishing a redundant communication structure. Ref. [97] reported that social communication mediates the dissemination and interpretation of natural hazard risk messages in the community. Finally, ref. [98] report that community resilience through the interaction-based community of informal social networks is more visible in disaster response and recovery. Poor risk management strategies that focus on food consumption styles, borrowing grains and cashes, distress livestock sales, and labor have not helped household absorptive capacity, whereas disaster risk mitigation and early warning systems, as well as informal safety nets, enhance household absorptive capacity, thereby building the climate resilience capacity of CSA innovators.

Regarding adaptive capacity, this study concurs with the findings of [99], who reported that crop rotation households have significantly higher wealth and income than SWC adopter households (p < 0.001). Because of this, SWC adopter households have significantly lower food security because they consume less low-quality and low-quantity food. Therefore, among adopters of improved variety, crop residue management, compost, row planting, and agroforestry, households’ increased productivity through higher wealth, income, and food security, which includes more durable assets, a larger farm size, a larger livestock holding, a greater number of plots, farm income, and improved food security in terms of quantity and quality of food consumption, fosters strong adaptive capacity that is supported by [41,42,100,101]. This finding concurs with the study by [102], who reported that income plays a significant role in the household’s resilience building.

Furthermore, regarding the transformative capacity of farmers, this study shows that access to basic services is the main source of transformative capacity for smallholder farmers, which concurs with the study by Asmamaw [50]. However, a higher information and training index was offset by a lower social network index among adopters of compost, which showed the influence of social networks on transformative capacity among smallholder farmers and led to an insignificant difference in transformative capacity between adopters and non-adopters of compost, as the latter need more labor as a prerequisite for adoption. Similar studies also showed that access to extension services, farmers’ training centers, and infrastructure increases the transformative capacity of smallholder agriculture systems [70,76,103,104].

In general, improved variety, crop residue management, compost, row planting, and agroforestry adoption showed significant increases in climate resilience capacities. Similar results supported our finding that improved variety in the form of drought-resistant variety (DRV) adoption increased the climate resilience capacity of smallholder households [105,106]. Moreover, other studies also concur with our finding that adopters of crop residue management as a component of conservation have built climate resilience through mitigating the negative impacts of deviations in rainfall due to drought and rainfall decrease [107]. Similar findings have been observed by [108], who reported that compost alone or in combination with nitrogen and phosphorus (NP) fertilizer improved soil properties and crop productivity, which builds climate resilience. Studies also concur with our finding that row planting adopters increased their climate resilience. A study by [109] reported that row planting remained an essential adaptation strategy for sustainable food production. Similar studies on row planting by [110] concur with our finding that the mean yield of row-planted wheat was higher compared to conventional broadcast planting methods, which increases the climate resilience capacity of smallholder wheat farmers. Finally, studies by [110] find that reported maintenance and enhancement of locally evolved agroforestry systems, with high resilience and multiple benefits, can contribute to climate resilience.

Similar results also obtained using the ESR model. The finding concurs with the study by [111] who reported soil fertility management technologies increases climate resilience through increased net agricultural income, yield, and productivity. However, the negative sign of treatment heterogeneity effect shows that adoption of SWC is more pronounced for non-adopters than adopters, i.e., some characteristics of non-adopters have made the effect of adoption of SWC more appropriate for non-adopters than actual adopters (p < 0.001); this may be in line with the study result by [112], who reported that SWC increased crop yields and improved the resilience of the agroecosystem to environmental stress.

5. Conclusions

The objective of this paper is to examine the impact of climate smart agriculture (CSA) innovations on building climate resilience capacity in smallholder agricultural systems. A cross-sectional household survey was conducted among multi-stage sampled 424 smallholder farmers selected from five agroecosystems of the Upper Blue Nile Highlands in Ethiopia. This study used an endogenous switching regression (ESR) model to examine the impact of CSA innovations on building climate resilience capacity among smallholder farmers. Principal component analysis was used to generate an index of absorptive, adaptive, and transformative capacities.

The principal component analysis of absorptive, adaptive, and transformative capacities showed that the resilience capacities of households were built on risk management, informal safety nets, disaster mitigation and early warning systems, adaptive strategies, wealth, food security, information and training, social networks, and infrastructure use. The simple mean comparison of absorptive, adaptive, transformative, and climate resilience capacities among adopters and non-adopters of CSA innovations revealed that improved variety and crop residue management adoption demonstrated a significant increase in absorptive capacity due to their effect on disaster mitigation and early warning systems as well as informal safety nets, whereas crop rotation adoption demonstrated a significantly lower absorptive capacity due to lower infrastructural capacity. All CSA innovation adoptions showed a significantly increased adaptive capacity because of their higher value for wealth and food security, while lower wealth and food security status correspond to lower adaptive capacity for adopters of SWC. However, of all the CSA innovation adoptions, only row planting showed a significantly increased transformative capacity due to lower information and training, social networks, and infrastructure use. Higher informal safety net support from friends and community during disasters, as well as strong disaster mitigation and early warning systems through strong social communication and access to mobile phone communications resulted in higher absorptive capacity among crop residue management adopters, whereas crop rotation adopters had lower absorptive capacity. Hence, ensuring strong informal safety nets as well as disaster mitigation and early warning systems builds strong climate resilience capacity among smallholder farmers. Similarly, higher wealth, which includes more durable assets, a larger farm size, a larger livestock holding, a greater number of plots, farm income, and improved food security in terms of quantity and quality of food consumption, fosters strong adaptive capacity in all CSA innovation adoptions except SWC, which has lower food security status than non-adopters. In addition, the strong wealth and food security status of farmers may offset lower adaptive strategies among adopters of agroforestry. Thus, strong wealth and food security build a strong climate resilience farming system in the face of climate change. Furthermore, higher information and training indexes through a strong public agricultural extension system and strong infrastructure use led to higher transformative capacity among row-planting adopters. However, a higher information and training index was offset by a lower social network index among adopters of compost. This led to an insignificant difference in transformative capacity between adopters and non-adopters, as compost needs more labor either from higher education or from social networks. Hence, strong information and training through strong public agricultural extension as well as the presence of climate-resilient infrastructure build the climate resilience capacity of smallholder agriculture systems. Strong absorptive, adaptive, and transformative capacities through strong disaster and early warning systems, climate-resilient infrastructure, a strong public agricultural extension system, a strong informal safety net, and social networks build a climate-resilient agriculture system among smallholder farmers. Therefore, improved variety, crop residue management, compost, row planting, and agroforestry adoption showed significant increase in climate resilience capacities.

The true average adoption effects of climate resilience capacity under actual and counterfactual conditions showed that different CSA innovations have different effects on climate resilience capacity of households. Except for SWC adopters, all CSA innovations significantly increased the climate resilience capacity of households. However, improved variety, crop residue management, and SWC have a more profound effect on the non-adopters than adopters, if non-adopters adopted these CSA innovations. Thus, scaling up of CSA innovations may expand the benefit of CSA innovation on building climate resilience capacities of households. Thus, strong risk management, disaster mitigation and early warning systems, adaptive strategies, information and training, informal safety nets, social networks, and infrastructure use may build climate resilience capacity of smallholder farmers by facilitating adoption of CSA innovations. Therefore, policies that strengthen good governance, social cohesion, disaster communication and early warning systems, input supply of drought-resistant varieties, climate smart extension services, and climate-resilient infrastructure are necessary.