Ecological Economics 235 (2025) 108611

Contents lists available at ScienceDirect

Ecological Economics
journal homepage: www.elsevier.com/locate/ecolecon

ANALYSIS

Climate adaptation, perceived resilience, and household wellbeing:

Comparative evidence from Kenya and Zambia
Haseeb Ahmed a,* , Juan Sebastian Correa b , Nicholas J. Sitko b
a
Economics Department, Asian Infrastructure Investment Bank (AIIB), Beijing, China
b
Rural Transformation and Gender Equality Division, Food and Agriculture Organization of the United Nations (FAO), Rome, Italy

A R T I C L E I N F O A B S T R A C T

Keywords: The growing frequency and severity of extreme weather events has spawned a rapid increase in policies and
Climate Adaptation programs designed to enhance the resilience of small-scale producers through the promotion of climate-adaptive
Climate Change agricultural practices. However, gaps exist in the conceptualization and measurement of farm-households’
Farm-Household Production
resilience in face of climatic stress. Furthermore, comparative evidence to understand the relationships between
Resilience
climate-adaptive practices, resilience capacities, and household wellbeing across diverse rural contexts remains
Sub-Saharan Africa
scant. Using a novel approach to measure households’ perceived resilience against climatic events, we empiri­
cally examine the relationship between perceived climate resilience, the adoption of climate-adaptive practices,
and household wellbeing in a pastoralist setting in Kenya and a rain-fed cropping system in Zambia. To enable
comparisons across these diverse settings, we use a typology of climate-adaptive practices based on their relative
factor intensities or diversification decisions. Using the ‘doubly-robust’ inverse-probability-weighted-regression-
adjustment (IPWRA) approach to account for potential selection issues, we find that capital-intensive strategies
are consistently and positively associated with resilience, food security, and income in both contexts. Labor-
intensive and diversification strategies have generally positive but heterogeneous impacts across the two pro­
duction systems, likely governed by contextual differences. Results also highlight the complementarity between
different climate-adaptive practices in improving household welfare in both contexts. The findings suggest that
enhancing resilience and improving overall wellbeing in small-scale producer settings requires multi-
dimensional approaches. These include interventions that reduce the capital constraints that inhibit the adop­
tion of capital-intensive adaptation practices, bundled with approaches that promote the simultaneous adoption
of context specific labor-intensive and diversification practices.

1. Introduction However, the relationships between the adoption of climate-

adaptive practices, household wellbeing, and resilience are not
Climate change is adversely affecting agricultural production and straightforward. First, climate-adaptive agricultural practices are often
food security in sub-Saharan Africa (SSA) (Calzadilla et al., 2013; Stuch highly context specific, what works in one context may not be appro­
et al., 2021). These negative effects are particularly significant for priate in others. Second, the livelihoods of small-scale farmers are
resource-poor farmers of SSA (Azzarri and Signorelli, 2020). To alleviate typically very diverse (Davis et al., 2017), thus adaptive actions through
the burdens that arise from climate volatility, national governments agriculture alone may not be sufficient to build broader livelihood
along with development and humanitarian agencies are developing resilience for many small-scale producers. Third, given that many small-
policies and programs to promote the adoption of climate-adaptive scale producers are not self-sufficient in food production, greater agri­
agricultural practices, with the aim of enhancing the resilience and cultural adaptation may not translate directly into improvements in
wellbeing of smallholder farmers. Climate-adaptive practices are ex­ household wellbeing and food security.
pected to build resilience and food security of farmers by reducing the Another key difficulty in understanding these relationships relates to
sensitivity of their agricultural systems to climate risks (e.g., Di Falco the conceptualization and measurement of resilience. As Barrett et al.
and Veronesi, 2014; Issahaku and Abdulai, 2020c). (2021) note, definitions of resilience are either context-specific or

* Corresponding author.
E-mail addresses: [email protected] (H. Ahmed), [email protected] (J.S. Correa), [email protected] (N.J. Sitko).

https://doi.org/10.1016/j.ecolecon.2025.108611
Received 10 August 2023; Received in revised form 24 October 2024; Accepted 21 March 2025
Available online 9 April 2025
0921-8009/© 2025 Food and Agriculture Organization of the United Nations. Published by Elsevier B.V.
H. Ahmed et al. Ecological Economics 235 (2025) 108611

inconsistently theorized across the empirical literature, which limits community-based organization or self-help group in pastoral production
comparison of resilience outcomes across diverse rural contexts. Addi­ systems of Kenya and conservation agriculture practices in crop pro­
tionally, even when resilience measures are well-grounded in theory and duction system of Zambia are included as system-specific labor-intensive
are validated across diverse contexts (e.g., Food and Agriculture Orga­ practices (more detail in section 4.2.1). The differing factor intensities of
nization’s (FAO) Resilience Indicators for Measurement and Analysis climate-adaptive practices may constrain their adoption or enable a
(RIMA) and TANGO International’s multidimensional resilience index preference for certain types of practices based on households’ relative
(RCI)), they require substantial amount of data for their estimation and factor endowment (Amadu et al., 2020; Sertse et al., 2021).
are focused on observable dimensions of resilience (e.g., see Smith and Given that the adoption of practices is a choice, impact estimates of
Frankenberger (2018); Upton et al. (2022) for RCI, and Jones and climate-adaptive practices are prone to misspecification due to selection
d’Errico (2019); D’Errico et al. (2018)) for RIMA). Perceived resilience bias (i.e., there may be observed and unobserved heterogeneity across
measures can potentially overcome these issues, and more importantly, adopters and non-adopters). To overcome these endogeneity issues, we
offer valuable insights into how individuals experience and evaluate use the “doubly-robust” IPWRA method that relies on ‘selection on ob­
their own capacity to cope with shocks, providing a more comprehen­ servables’ and controls for selection and confoundedness at the inverse-
sive understanding of unobservable dimensions of resilience, including probability weighting (IPW) and regression-adjustment (RA) stage
psychological, social, and behavioral dimensions (Jones and Tanner, (Wooldridge, 2010). The IPW helps by weighting households based on
2017; Clare et al., 2017). how likely they were to adopt climate-adaptive practices, given their
Building on this, our study draws on the literatures related to resil­ characteristics. Households who were unlikely to adopt practices but did
ience measurement (e.g., Jones and Tanner, 2017; Clare et al., 2017; (or vice versa) get a bigger weight, while those whose adoption status
Spiegel et al., 2021) as well as the empirical evidence associated with the was more predictable get a smaller weight. This weighting procedure
effects of climate-adaptive practices (e.g., Di Falco and Veronesi, 2014; creates a suitable counterfactual group against which the group of
Scognamillo et al., 2022) to: (a) conceptualize and measure farm- adopters can be compared, thus reducing bias. The RA stage runs
households’ perceived resilience; and (b) examine the interrelationships weighted regressions controlling for several observable confounders,
between climate-adaptive practices and broader household wellbeing further reducing bias.
including perceived resilience. Specifically, the paper asks the following Another important contribution of our study relates to generating
questions: (1) how can we measure farm-households’ perceived resil­ evidence regarding the impacts of climate-adaptive practices in the
ience to climate shocks? and (2) what are the impacts of climate- context of pastoralists of eastern Africa. We know of only one study,
adaptive practices on household wellbeing including perceived resil­ Ndiritu and Muricho (2021), which estimates the impact of feed storage,
ience to climatic stressors? We answer these questions using data from better watering practices, and destocking on food security of pastoralist
rural communities in Zambia and Kenya. households. We extend their work by not only assessing impacts of on-
The study contributes to the literature by introducing a cost-effective farm adaptive practices, but also examine impacts of off-farm and col­
way (in terms of data needs) of eliciting self-assessed resilience across a lective action strategies on a wider array of farm and household well­
broad spectrum of rural contexts. In addition, the study provides being variables. This allows us to document the trade-offs between on-
comparative evidence regarding the relationships between resilience, farm and off-farm strategies as well as provide insights related to im­
household wellbeing, and climate-adaptive practices from small-scale pacts on economic as well as psychological dimensions of wellbeing.
crop (Zambia) and livestock production systems (Kenya), making it Finally, we provide evidence that the simultaneous adoption of practices
relevant for researchers as well as development practitioners working in offer greater benefits to households. While most papers model impacts of
the space of rural development, resilience, and climate change. While adaptive practices individually (e.g., Asfaw et al., 2019; Mulwa and
several studies document the relationships between household well­ Visser, 2020), we only know of Issahaku and Abdulai (2020b) and Tabe-
being and climate-adaptive practices in country-specific case studies (e. Ojong et al. (2023) that document the joint effects of climate-smart
g., Di Falco and Veronesi, 2014; Abdulai and Huffman, 2014; Khanal practices in crop production systems.
et al., 2018; Issahaku and Abdulai, 2020b; Issahaku and Abdulai, 2020c;
Mulwa and Visser, 2020; Alfani et al., 2021; Maggio et al., 2022; 2. Conceptualizing resilience and the role of climate-adaptive
Scognamillo et al., 2022), comparative evidence regarding the role of practices
climate-adaptive action in shaping household wellbeing, including
resilience, is scarce. Subjective resilience measures offer a unique and valuable perspec­
To conceptualize resilience, we adapt the resilience framework based tive in understanding how households perceive their capacity to cope
on adaptive cycles in social-ecological systems for the case of farm- with shocks. These measures capture crucial social, cultural, and psy­
households under the risk of climate shocks (Folke et al., 2010; chological dimensions—such as risk perception, self-efficacy,1 and
Darnhofer, 2014; Meuwissen et al., 2019). We postulate that farm- aspirations—that are often overlooked by traditional objective resil­
households need to build both on- and off-farm resilience capacities to ience metrics. As Béné et al. (2019) highlight, individuals’ perceptions of
successfully navigate risks posed by climatic change. Therefore, through risk and their ability to manage stressors directly influence critical de­
a novel survey-based scale, we elicit information on households’ cisions, such as whether to adopt adaptive strategies. Likewise, Jones
perceived capacity to bounce-back (robustness), adapt at the farm-level and Tanner (2017) emphasize that subjective resilience reflects in­
(farm adaptability), transform production (transformability), and adapt dividuals’ evaluations of their ability to cope with future shocks, shaped
their off-farm livelihoods (livelihood adaptability) to measure their by factors like social cohesion or feelings of marginalization. Although
resilience to droughts (in Kenya and Zambia) and floods (in Zambia).
We then estimate the impact of climate-adaptive practices on cost of
production, income, food security, and resilience in Kenya and Zambia,
1
respectively. To explore these relationships and draw comparisons Perceived resilience, much like the concept of perceived self-efficacy (PSE),
across these production systems, we develop a simple typology of can significantly influence a household’s adaptive capacity. PSE refers to an
individual’s belief in their ability to manage and respond to challenges
climate-adaptive practices being promoted in each context based on
(Wuepper and Lybbert, 2017). In the context of climate change adaptation,
their relative factor intensity. In our analysis, analogous practices across
higher PSE has been shown to drive the adoption of adaptive measures,
the Kenyan and Zambian contexts include herd and crop diversification enabling households to better respond to climatic risks. Similarly, perceived
strategies, system-specific capital-intensive strategies such as adoption resilience can empower households to take more proactive and informed ac­
of feed storage and use of improved/hybrid seeds, and off-farm diver­ tions, thus enhancing their overall adaptive capacity to climatic events
sification strategies, respectively. Collective production within the (Gebrehiwot and van der Veen, 2015; Burnham and Ma, 2017).

2
H. Ahmed et al. Ecological Economics 235 (2025) 108611

subjective assessments may be prone to biases, they still offer mean­ adaptive strategies have greater impacts on perceived robustness and
ingful insights into actual capacities, as households’ perceptions are farm adaptability capacities, while off-farm practices are more likely to
often grounded in their lived experiences and resource availability. affect perceived livelihood adaptability and transformative capacities.
These perceptions shape their immediate responses to shocks as well as However, the impacts of on- or off-farm practices may not be confined to
longer-term adaptive strategies, offering a more holistic understanding a single perceived resilience capacity. Instead, there may be correlations
of resilience in various contexts (Clare et al., 2017; Béné et al., 2019). By among perceived resilience capacities, leading to complementarities and
focusing on subjective resilience, this study broadens the scope of spillovers in the impacts of climate-adaptive practices across the various
resilience measurement, integrating the cognitive, social, and psycho­ perceived resilience capacity dimensions. For example, off-farm liveli­
logical processes that influence how farm households navigate and hood diversification strategies may allow for greater income generation,
recover from crises. thus directly impacting perceived livelihood adaptability and trans­
Our subjective resilience measure is based on the conceptual foun­ formability capacities, while also reducing capital constraints to in­
dations of adaptive cycles in social-ecological systems theory and goes vestments in farm production, thus indirectly impacting farm-related
beyond the narrow definitions of resilience that focus on maintaining an perceived resilience capacities. Because of these potential synergies, on-
equilibrium (Folke et al., 2010). This stream of literature posits that a and off-farm climate-adaptive practices may be adopted together by
system’s resilience is based on its ability to (a) bounce-back from a shock households.
(robustness), (b) adapt to the shock (adaptability), and (c) transform While adaptation actions are expected to affect perceived resilience,
production processes under pressure from external shocks (trans­ the direction of the impact may also go in the opposite direction (as
formability) (Darnhofer, 2014; Meuwissen et al., 2019; Roy et al., 2019). illustrated in Fig. 1). A ‘resilient’ household, benefiting from improved
We build on this conceptual framework and operationalize it for the economic conditions, reduced sensitivity and vulnerability to climatic
case of smallholder farm-households (resilience of what) that increasingly events, and accumulated experience, is likely to have the resources and
bear the brunt of climate shocks (resilience to what). We postulate that capacity to invest more in adaptation actions; creating feedback loops
there are four key perceived resilience capacities that determine the between the adaptive actions, perceived resilience capacities, and
overall perceived resilience of farm-households in SSA. Robustness (or wellbeing of the farm-household (Bartelet et al., 2022; Wouterse et al.,
absorptive capacity) is the ability to withstand (un)anticipated climate 2022). Perceived resilience may even moderate or mediate the re­
shocks. (b) Farm Adaptability is the capacity to change the composition lationships between adaptation actions and household welfare (Ansah
of inputs, production, and risk management strategies without changing et al., 2019; Ansah et al., 2023). Furthermore, perceived robustness,
the structure of the farm. (c) Transformability is the capacity to signifi­ adaptability, and transformative capacities may be interrelated, with
cantly change the farming structure in response to consistent stressors; potential correlations and spillover effects between them. This concep­
and (d) Livelihood Adaptability is the capacity to engage in off-farm ac­ tual framework provides a basis for thinking about and developing the
tivities to supplement and diversify household income. empirical strategy to estimate the relationship between climate-adaptive
While subjective resilience measures have been developed in the practices and resilience and identify potential limitations in the esti­
literature (e.g.,Nguyen and James, 2013; Lockwood et al., 2015; Jones mation of this relationship.
and d’Errico, 2019), they do not explicitly measure different on- and off-
farm perceived resilience capacities of farm-households, which are 3. Econometric framework
important given the varied livelihood activities of these entities. This
framework not only allows us to elicit households’ overall perceived Our aim is to estimate the relationships between climate-adaptive
resilience, but also allows us to understand the heterogeneous re­ practices and household wellbeing. Adoption of climate-adaptive prac­
lationships between on-farm (e.g., adoption of inputs and diversification tices is a choice variable and is influenced by various observed and
strategies) and off-farm climate-adaptive practices (e.g., livelihood unobserved characteristics that are likely to differ systematically be­
diversification) and the different perceived resilience capacities outlined tween adopters and non-adopters and correlate with welfare outcomes,
in the above paragraph (Fig. 1).2 characterizing the self-selection problem in the estimation. Therefore, to
For example, we may expect that investments in on-farm climate- assess the relationship between climate-adaptive practices and various
wellbeing outcomes, we need a suitable counterfactual against which to
compare the group of adopters.
In order to generate a suitable counterfactual against which adopters
can be compared, we apply the ‘doubly-robust’ inverse-probability-
weighted-regression-adjustment (IPWRA) method proposed by Wool­
dridge (2010). The IPWRA estimator is obtained by combining inverse
probability weighting (IPW) with regression adjustment (RA) in the
second stage of the empirical model. The IPW stage re-weights the data
such that the group of adopters are compared against a suitable coun­
terfactual (adopters under hypothetical non-adoption), reducing
selection-bias. Then, in the RA stage, weighted regressions are used with
a rich set of covariates to further control for confounding. The identifi­
cation is based on the conditional independence assumption (CIA) or
selection on observables, which assumes that after controlling for ob­
servables that explain selection into treatment, the treatment assign­
ment is independent of potential outcomes. We include a rich set of
Fig. 1. Conceptual framework of adaptation and resilience.
covariates that explain selection into treatments as well as outcomes to
satisfy the CIA. Moreover, this econometric approach does not mandate
strict adherence to the CIA for estimating the coefficients of interest;

2
Figure 1 provides a conceptual schematic to think about the relationships
between resilience capacities, adaptation action, and household welfare.
However, it should not be read as causal pathway diagram given the complex
and sometimes bidirectional relationships between these latent variables.

3
H. Ahmed et al. Ecological Economics 235 (2025) 108611

instead, it requires only a less stringent condition.3 of Zambia and Kenya, unobservable heterogeneity (e.g., farmer ability,
Practically, IPWRA involves a three-step procedure. In the IPW stage knowledge, quality of local institutions, etc.) is likely to play a role.
(step 1), the inverse probability weights are calculated based on the Although conditioning on a rich set of observable covariates may help to
estimated probability of being an adopter, using a logit (for binary reduce selection bias due to unobservables (Imbens and Wooldridge,
treatment variables) or a multinomial-logit (for categorical treatment 2009), the potential threat from unobserved heterogeneity remains.
variables) model. For this purpose, we estimate propensity scores as Therefore, our results should be interpreted as associations with strong
defined by Rosenbaum and Rubin (1983), using a range of observable controls that are suggestive, but are not conclusive of causal effects.4
covariates. Then in the RA stage (step 2), the regression adjustment
method fits separate weighted regression models for adopters and non- 4. Data
adopters, and then predicts the covariate-specific outcomes for every
observation under each adoption status. Finally (step 3), we obtain the 4.1. Context and sampling
average differences between predicted outcomes for adopters under
adoption and hypothetical non-adoption, which is the Average Treat­ The target region for data collection in Isiolo, Kenya was chosen in
ment Effect on Treated (ATTIPWRA A
). Formally, the IPWRA model can be collaboration with officials from Ministry of Agriculture, Livestock, and
expressed as follows (Hörner and Wollni, 2021; Manda et al., 2018; Fisheries and National Disaster Management Authority (NDMA). We
Wooldridge, 2010): collected this data from October – November 2022 in collaboration with
∑n FAO Kenya from four wards of Isiolo, namely: Chari, Cherab, Sericho,
ATTAIPWRA = n−A 1 i=1
Ti [rA (X, δA ) − rN (X, δN ) ] (1) and Oldonyiro. Rural households in these arid and semi-arid areas of
Isiolo are pastoral and agropastoral smallholders who rely substantially
where nA is the number of adopters, while ri (X) describes the regression on their livestock for income and food security. Isiolo is an ethnically
model for adopters (A) and non-adopters (N) with covariates X and diverse county and rural households in these areas face considerable
estimated parameters δi , where δi are obtained from weighted regression risks from droughts and drought-induced conflicts (Quandt, 2021).
procedures. The standard errors are clustered at the village- and Com­ The collected data in Kenya forms the baseline for a clustered ran­
munity Welfare Assistance Committee (CWAC)-level for Kenya and domized impact evaluation where the intervention will go through rural
Zambia, respectively, to account for intergroup correlation in errors. community-based or self-help groups.5 Therefore, as a first step, 75
In order to empirically validate the IPWRA approach, we test the groups were randomly selected from the four wards. In the second step,
overlap assumption and check whether balance is achieved. To ascertain about 11 households were randomly selected within each group for data
whether the balance assumption holds after the IPW procedure, we run collected. We ended up with a total of 788 observations with an average
an overidentification test, and also calculate normalized differences for of 16 observations per village. However, we had to drop 120 observa­
each covariate as Imbens and Wooldridge (2009) propose. Both tions and ended up with 668 observations for this analysis since these
normalized weighted differences (in Tables A1 and A2) as well as households do not own any livestock and the chosen climate-adaptive
overidentification tests (in Tables 5 and 6) show that balance conditions practices in Kenya are primarily associated with adaptation within the
hold. To ensure the overlap assumption holds (i.e., conditional on livestock sector.6 For the households that own livestock, the average
covariates, each observation has a positive probability of being an herd size is 5.85 in TLUs with 61 % owning cattle, 82 % owning goats,
adopter), we run our models with the tolerance level set such that p[T = and 81 % owning sheep, respectively.
1|X] < 1 for being an adopter, where p[T = 1|X] is the propensity score. In Zambia, data were collected from villages belonging to 68
In the literature (e.g., Imbens and Wooldridge, 2009; Uysal, 2015; Heiler CWACs7 from 26 different wards and within three districts (Chipangali,
and Kazak, 2021), it is recommended to graphically plot the propensity Katete, and Sinda) of Eastern Province. The data were collected as part
scores to determine the validity of the overlap assumption. In the graphs of a quasi-experimental evaluation clustered at the CWAC level of the
reported in Figs. A1 – A4, we report the kernel density of the proba­ Food Security Package (FSP), which provides seeds and fertilizer to
bilities of being in any group, after reweighting. No spikes are detected eligible farmers. This area was selected for an expansion of the wetland’s
at the extremes of the distributions, meaning that all observations have a subprogram of the FSP in 2022, which targets households with access to
positive probability of being in each group. This implies that the ob­
servations are well matched and do not require any common support
adjustment. 4
This literature has made use of binary and multinomial endogeneous
We also estimate the combined impacts of these strategies on switching regressions (e.g., Abdulai and Huffman, 2014; Issahaku and Abdulai,
household wellbeing to assess if combinations of these practices are 2020b etc.) to overcome selection issues that arise from observable and unob­
complementary in improving wellbeing. Thus, the treatment variable in servable heterogeneity. However, in the absence of strong and valid exclusion
this case is the number of practices adopted by each household, char­ restrictions, estimates from endogenous switching models can suffer from non-
acterizing the intensity of adoption. To do this, we rely on the IPWRA robustness and collinearity problems (Moffitt, 1999).
5
multivalued treatment effects model to measure the impacts of different Rural community-based or self-help groups are formal organizations
registered with the Department of Social Development in Kenya. These groups
levels of adoption of practices on household wellbeing (Wooldridge,
act as forums for citizens in the region to come together and address common
2010; Cattaneo, 2010; Cattaneo et al., 2013; Esposti, 2017; Smale et al.,
issues facing their communities. Therefore, these organizations can be helpful
2018; Issahaku and Abdulai, 2020a). The multivalued treatment effects in planning, implementing, and monitoring rural development programmes.
approach is simply an extension of the binary approach to the multi­ 6
Table A3 provides the pairwise comparisons across livestock owners and
valued case, where the inverse probability weighting is done using a non-owners across our sample. We find that non-owners of livestock have lower
multinomial logit regression instead of logit or probit regressions. gross incomes. However, this is not a concern for our estimation because our
One limitation of the IPWRA method is that it relies on a version of climate-adaptive practices concern mainly with livestock owners. For the
the CIA (i.e., selection on observables) to reduce selection bias and climate-adaptive practices that are shared between livestock owners and non-
confounding and is, therefore, vulnerable to heterogeneity due to un­ owners, we provide results for these practices for livestock owners as well as
observables. In the context of climate-adaptive practices of smallholders the whole sample to illustrate robustness of our estimates.
7
These committees consist of volunteer members elected from the commu­
nity/communities they cover that help identify potential beneficiaries. The
committees were initially part of the structure of the Public Welfare Assistance
3
Only the conditional mean independence assumption (CMIA) restriction Scheme (PWAS), which aimed at supporting the most disadvantaged members
needs to be satisfied. The CMIA allows the conditional variance to depend on of the population to enable them to fulfill their essential requirements and
the treatment, unlike the CIA (Wooldridge, 2010). fostering community resilience in tackling poverty and vulnerability.

4
H. Ahmed et al. Ecological Economics 235 (2025) 108611

wetland areas, known locally as dambos. Data collection was completed their relative factor intensities. In particular, feed storage and adoption
between October – December 2022. The districts belong to the same of improved seeds are considered analogous given the capital-intensive
agroecological zone, characterized by a rainy season that generally oc­ nature of these practices. Feed storage can be an effective adaptive
curs from November to April, with peak rainfall between December and strategy in the pastoralist setting to ensure animal health and nutrition
February. Farmers engage in rainfed agriculture, utilizing the rainfall during lean seasons. However, procuring and storing feed is capital
pattern to determine planting and harvesting seasons, making them intensive, a factor of production that may be constrained in this context
highly vulnerable to climatic shocks (Arslan et al., 2015). We obtained a (Ndiritu and Muricho, 2021; Berhe et al., 2017). In Zambia, adoption of
total of 2996 valid surveys with an average of 44 observations per improved/hybrid seeds can be an effective welfare enhancing strategy,
CWAC. Farmers in the sample have access to 2.23 ha of land for agri­ but its adoption may be constrained due to lack of capital (Emerick et al.,
cultural activities. The principal crops cultivated include maize (90 % of 2016).
households), soybeans (58 % of households), and sunflower (25 % of Herd and crop-diversification are comparable on-farm diversifica­
households). Agriculture accounts for 73 % of the total income, with tion strategies for the livestock and crop production systems studied in
production being entirely reliant on family labor for 75 % of households. this paper, respectively. Off-farm wage or enterprise work are again
In both countries, survey instruments were developed in a way that comparable livelihood diversification strategies in the two contexts.
allows comparison across contexts. In particular, the survey instruments Finally, collective production and conservation agriculture are system-
include comparable modules to capture farming and non-farming in­ specific labor-intensive climate-adaptive strategies. Collective produc­
comes and expenditures, and adoption of climate-adaptive practices. tion within the community-based organization provides an additional
Both surveys also included modules on food consumption score (FCS) avenue of producing food for the household. However, in the presence of
and subjective resilience. Furthermore, both surveys included questions labor constraints faced by the households, not all households may be
on household demographics, access to extension services, access to able to practice it (Fischer and Qaim, 2012). Similarly, conservation
climate information services, and details on farm-level characteristics agriculture may be important in terms of adaptation action, but its
such as land ownership and number of productive assets owned. adoption can be constrained by lack of labor (Pannell et al., 2014).
Comparable strategies across the two contexts are illustrated in Fig. 2.
Moreover, our data only allows us to ascertain whether adaptive
4.2. Variables
practices were employed in the agricultural season prior to the survey.
Given this data’s cross-sectional nature, our analysis is confined to
4.2.1. Adoption variables
examining only the contemporaneous effects of these practices. Addi­
Climate adaptation is a comprehensive concept that does not
tionally, in the context of Kenya, data collection coincided with a pro­
comprise a clearly defined, fixed set of practices. It can include a wide
longed drought in Isiolo and other northern regions, resulting in
range of site-specific, improved technologies depending on local con­
minimal variation in drought status among households. In Zambia, all
ditions, resource levels and the production system in question. In this
observations stemmed from the same agroclimatic region (region IIa).
article, the climate-adaptive practices that are analyzed are based on the
Analysis of drought conditions during the 2021/2022 agricultural sea­
country’s national adaptation plans that prioritize these adaptive ac­
son, including the Standardized Precipitation Index (SPI), number of
tions. In the pastoral setting of Kenya, feed storage, herd-diversification,
rainless days, and dry spells, revealed deviations from long-term aver­
and livelihood diversification are prioritized in the Kenya National
ages, but not severe stressors. Consequently, the limited variability in
Adaptation Plan as important climate-adaptive practices (Govt. of
these metrics prevents our ability to evaluate the relevance of climate-
Kenya, 2016). In addition, collective action and production can facilitate
adaptive practices in helping absorb contemporaneous climatic shocks.
informal risk sharing as well as diversification and is considered an
important climate-adaptive practice in rural settings (Adger, 2003). In
4.2.2. Outcome variables
our sample, 29 %, 57 %, 37 %, and 16 % of the households adopt feed
The Food Consumption Score (FCS) and the subjective resilience
storage, herd-diversification, off-farm work, and collective production,
measure that elicits the self-assessed resilience of farm-households to
respectively (Table 1). Similarly, according to Zambia National Adap­
droughts (in Kenya and Zambia) and floods (in Zambia) are considered
tation Plan (Govt. of Zambia, 2023), adoption of improved/hybrid
as outcome variables that proxy household wellbeing. Furthermore, to
seeds, crop-diversification, off-farm diversification, and conservation
understand the channels through which household wellbeing is affected
agriculture are considered important adaptation practices and 55 %, 53
by adoption of climate-adaptive practices, we examine impacts on
%, 39 %, and 40 % of the households in Zambian sample adopt these
annual farm-related expenditures, farm income, and total income of the
practices, respectively (Table 1).
household in Zambia and Kenya. Table 2 provides details on the defi­
To conduct a comparative study in these diverse production systems,
nitions, means, and standard deviations of these outcome variables.
we develop a simple typology of climate-adaptive practices based on

Table 1
Summary statistics (Adoption Variables).
Country Variable Description Mean

Kenya Feed Storage Indicator variable = 1 if household has access to individual or communal livestock feed storage, 0 otherwise. 0.29
Herd Diversification Indicator variable = 1 if household owns more than one type of livestock, 0 otherwise. 0.57
Off-farm Indicator variable = 1 if any member of the household works or runs an enterprise outside the farm, 0 otherwise. 0.37
Diversification
Collective Production Indicator variable = 1 if the household engages in collective production (of crops, poultry etc.) with the community-based organization 0.16
or self-help group, 0 otherwise.
Zambia Improved Seed Indicator variable = 1 if households have used improved and/or hybrid seeds in the previous year, 0 otherwise. 0.55
Crop Diversification Indicator variable = 1 if household has a summary crop diversification index a la Anderson (2008) larger than the median. The 0.53
summary crop diversification index is a standardized weighted average of the number of cultivated crops and of both a Simpson and
Shannon index of crop diversity.
Off-farm Indicator variable = 1 if any member of the household operates an enterprise outside the farm, 0 otherwise. 0.39
Diversification
Conservation Indicator variable =1 if household has adopted at least 3 out of 4 conservation agriculture practices (crop residue are spread over plot, 0.40
Agriculture zero tillage, crop rotation, intercropping), 0 otherwise.

N = 668 for Kenya and 2996 for Zambia.

5
H. Ahmed et al. Ecological Economics 235 (2025) 108611

Fig. 2. Comparable climate-adaptive practices across Kenya and Zambia.

Table 2
Summary statistics (Outcome Variables).
Variable Description Kenya Zambia

Food Consumption Standardized weighted average of the number of times a certain food group was consumed by the 0 (1) 0 (1)
Index household in the last 7 days.
Subjective Resilience Standardized weighted average of 11 Likert scale items eliciting subjective resilience of farm-household. 0 (1) 0 (1)
Index
Livestock Expenditure Inverse hyperbolic sine of total livestock expenditure in the last 12 months (Kenya). 9447 Ksh. (20,867) –
Agricultural Inverse hyperbolic sine of the total agricultural input expenditure from May 2021 to April 2022. – 1689 ZMW (2579)
Expenditure
Livestock Income Inverse hyperbolic sine of the total livestock income in the last 12 months. 76,127 Ksh. –
(109,531)
Agricultural Income Inverse hyperbolic sine of the total agricultural income from May 2021 to April 2022. – 13,260 ZMW
(43,582)
Gross Income Inverse hyperbolic sine of the total household income in the last 12 months. 130,428 Ksh. 22,302 ZMW
(175,458) (55,010)

N = 668 for Kenya and 2996 for Zambia.

The food consumption index is based on the number of times each of Kenya are more food insecure and less resilient to droughts likely due to
the food groups were consumed within the last 7 days (based on WFP the prolonged drought they faced preceding the data collection period.
(2008)), while the subjective resilience index is based on the scale
presented in Fig. 3. Both, the food consumption index and the subjective 4.2.3. Control variables
resilience index are estimated using the standardized inverse-covariance The selection of household- and farm-level control variables is based
weighted average approach (à la Anderson (2008)). This approach uses on the literature on adaptation and resilience in rural Africa (e.g.,
generalized least-square (GLS) weighting, which has two key advan­ Issahaku and Abdulai, 2020b; Maggio and Sitko, 2019; Maja et al.,
tages. First, it increases efficiency by ensuring highly correlated in­ 2023). These variables include age, gender, and education of the
dicators receive less weight than uncorrelated indicators. Intuitively, household-head (in case of Kenya) and main respondent (in case of
uncorrelated indicators, which represent “new” information, receive Zambia), total number of household members, total herd size (in TLUs),
more weight. Second, the procedure uses all available data but ascribes whether the household received any climate information services, and
lower weight to indicators with missing values, which allows for the whether the household was visited by an extension officer. Table 3
calculation of the composite index even for observations with missing provides the summary statistics related to these variables. The average
indicators. The procedure is implemented via Swindex command in household size is 6.59 in Kenya and 5.47 in Zambia. The average age of
Stata® (see Anderson (2008) and Schwab et al. (2020) for details). the household-heads in Kenya is 45, while the average age of a
We find a scale reliability coefficient (Cronbach’s α) of 0.77 for respondent in Zambia is 40. In Kenya, 41 % of the household-heads have
resilience against droughts in Kenya, while in Zambia, this was 0.81 for received some formal schooling and 71 % of them are males. In Zambia,
resilience against droughts and 0.84 for resilience against floods. respondents, on average, have received about 5 years of schooling and
Furthermore, Figs. A5 and A6 show positive correlations between 65 % of them are male. Forty-two and 37 % of the households receive
resilience capacities, suggesting that strengthening particular compo­ information on seasonal forecast on rainfall or drought in Zambia and on
nents or dimensions of resilience may also have positive spillovers on short- and near-term weather forecast (two to three days) or forecasts for
overall resilience. the next two to three months in Kenya, respectively. Finally, 32 % of the
In Kenya, the mean annual livestock expenditure is 9447 Ksh (~ household report extension contact in the last six months in Kenya,
$80), while mean livestock and total income are 76,127 Ksh (~$650) while 45 % of the households in Zambia report receiving extension
and 130,428 Ksh (~$1100), respectively. This total income (~$92 per services.
month) is slightly less than the representative rural incomes in Kenya,
which are reported to be $105 per month likely because Isiolo is far 5. Results & discussion
away from wealthier urban locations, faces ethnic conflict, and faces
higher drought risk than western or southern Kenya (KIHBS, 2018; 5.1. Correlates of adopting climate-adaptive practices
Quandt, 2021). For our sample in Zambia, the mean annual agricultural
expenditure is 1689 ZMW (~$90), while mean agricultural and total Table 4 presents the correlates of the number of climate-adaptive
income are 13,260 ZMW (~$700) and 22,302 ZMW (~$1200), practices adopted by a household based on Poisson regressions.8 While
respectively (Table 2). Agricultural and non-agricultural incomes are
similar across the two samples and, on average, income from agricul­
tural activities makes about 60 % and 75 % of the total income in Kenya 8
The adoption model within the IPWRA approach is a logit model that serves
and Zambia, respectively. While average incomes are similar across the as a method to estimate propensity scores and achieve balance between the
two contexts, cross-country comparisons for food security and subjective observed covariates across adopters and non-adopters (Table A6 and Table A7).
resilience, presented in Table A4, indicate that pastoralist households in However, the Poisson regressions included in the main text allow us to provide
summary results of these adoption regressions in a concise manner (Table 4).

6
H. Ahmed et al. Ecological Economics 235 (2025) 108611

Fig. 3. The subjective resilience measurement scale based on the 5-degree-likert scale (1 = strongly disagree, 2 = disagree, 3 = neither agree nor disagree, 4 = agree,
and 5 = strongly agree). Note: Lines indicate 95 % confidence intervals.

no causal interpretation can be inferred from the results in this table, understanding risk mitigation and promoting climate adaptation in rural
some of the covariates stand out in terms of explaining the adoption of communities. Lastly, herd size, a measure of wealth as well as savings in
climate-adaptive practices. In line with previous literature, access to these contexts, correlates significantly with adoption of adaptive
extension services has a positive relationship with adoption of practices practices.
in Kenya and Zambia, likely through transfer of necessary knowledge In Table A5, we also provide correlates of resilience. We observe that
that make adoption of practices and technologies possible in both con­ some of the covariates associated with climate-adaptive practices, such
texts (e.g., Hasan and Kumar, 2019; Issahaku and Abdulai, 2020b). as labor availability and access to extension services, are also correlated
Labor constraints (which capture high dependency ratio and/or with households’ perceived resilience. These results, in line with pre­
disability in the household) are negatively correlated with adoption of vious literature (Mutabazi et al., 2015; Hasan and Kumar, 2019), are
practices in Kenya (as in Scognamillo et al., 2022), but do not have important from the perspective of designing development interventions
significant correlates in Zambia. These results highlight that some di­ and highlight that strengthening factors, such as rural extension net­
mensions of adaptation in the Kenyan context may be more labor works, can potentially enhance adoption of climate-adaptive practices
intensive and may not be adopted due to labor constraints. Additionally, as well as resilience capacities of rural households in diverse rural
hired labor may be less available in this context due to the prevailing contexts.
drought and associated drought-induced migration.
Among other correlates, being part of a savings group, household
5.2. Relationships between climate-adaptive practices and household
size, formal education and herd sizes (which also indicates with capital
wellbeing
holdings of the household in these contexts) are associated positively
with adoption of adaptive practices in both contexts. Indeed, feed stor­
This section highlights three key findings on the relationships be­
age and hybrid seeds are capital-intensive technologies, and participa­
tween adoption of adaptive practices and household wellbeing, as esti­
tion in savings groups can help farm-households finance the costs of
mated in Eq. 1 (Table 5 and Table 6). First, we find that capital-intensive
purchasing these technologies. Savings may also help households to hire
practices, such as feed storage in Kenya and hybrid seeds in Zambia,
additional labor and make investments in non-farm enterprises to sup­
have large and consistently positive associations with household well­
port livelihood diversification (e.g. Carter et al., 2016). Larger house­
being in both countries. Second, while on-farm diversification strategies
hold size, perhaps due to availability of necessary labor resources for
do not have any significant impacts in these contexts, off-farm diversi­
labor-intensive practices, is positively associated with adoption in
fication strategies show a robust positive association with household
both contexts (Teklewold et al., 2013; Shikuku et al., 2017). For
income and resilience in both contexts. Third, system-specific labor-
example, off-farm diversification in both contexts is positively associ­
intensive strategies, such as conservation agriculture in Zambia and
ated with household size (Table A6 and Table A7), highlighting that a
collective production in Kenya, show heterogeneous associations with
households’ ability to engage in off-farm labor may depend on the
household wellbeing, with differences in relationships being plausibly
availability of labor within the household. Similarly, schooling is posi­
governed by differences in the contexts. These results are unpacked in
tively associated with adoption (see e.g., Walker et al., 2022), illus­
detail in the following paragraphs.
trating that access to formal education may play a significant role in
We start by comparing the effects of adoption of capital-intensive

7
H. Ahmed et al. Ecological Economics 235 (2025) 108611

Table 3
Summary statistics (Control Variables).
Variable Description Kenya Zambia

Herd Size (TLUs) Tropical livestock units (TLUs) owned by the household. 5.85 (8.56) 2.24
(4.40)
Household Size Number of household members. 6.59 (2.33) 5.47
(2.11)
Age Age of the household head in years (Kenya). 45.38 39.8
Age of main respondent in years (Zambia). (13.70) (13.8)
Gender Indicator variable = 1 if household-head is male, 0 otherwise (Kenya). 0.71 0.65
Indicator variable = 1 if main respondent is male, 0 otherwise (Zambia).
Schooling Indicator variable = 1 if household-head received any level of formal schooling, 0 otherwise (Kenya). 0.41 5.33
Number of years of formal schooling attended by the main respondent (Zambia). (3.81)
Market Distance Number of minutes it takes to walk to the nearest market (Kenya). 40.55 –
(46.74)
Climate Information Indicator variable = 1 if the household received information regarding weather, 0 otherwise. 0.42 0.37
Services
Extension Contact Indicator variable = 1 if the household was visited by an extension officer in the last six months, 0 otherwise (Kenya). 0.32 0.45
Indicator variable = 1 if household received information about agricultural practices, soil conservation, and agro-forestry
practices, (Zambia)
Saving Group Indicator variable = 1 if any member of the household is a member of any savings group, 0 otherwise. 0.65 0.17
Labor Constraint Indicator variable = 1 if household does not have an able-bodied adult or the dependency ratio (household members not fit to 0.26 0.23
work / household members fit to work) is greater than 3.
Drought Shock Indicator variable = 1 if household has experienced a drought shock in the past 8 years (self-reported) – 0.78
Flood Shock Indicator variable = 1 if household has experienced a flood shock in the past 8 years (self-reported) – 0.39

N = 668 for Kenya and 2996 for Zambia.

practices, such as feed storage in Kenya (Table 5) and hybrid seeds

Table 4
Zambia (Table 6). Feed storage in Kenya is associated with an increase of
Correlates of climate-adaptive practices in Kenya and Zambia.
food consumption and subjective resilience by 0.350 and 0.256 standard
deviations (SD), over the counterfactual mean, respectively (Table 5). In Kenya Zambia
terms of expenditures and incomes, adoption of feed storage is associ­ # of Climate adaptive # of Climate adaptive
ated with a 20.1 % and 6.4 % increase in livestock expenditures and practices practices
gross income over the counterfactual mean, respectively (estimates Household size 0.052*** 0.010**
statistically significant at 1 % and 5 % level of significance).9 In com­ (0.009) (0.004)
parison, adoption of improved/hybrid seeds in Zambia is associated Gender (Male = 1) 0.007 0.072**
(0.044) (0.029)
with an increase of food consumption and subjective resilience against
Age 0.001 0.001
droughts by 0.27 and 0.14 standard deviations (SD), over the counter­ (0.002) (0.001)
factual mean, respectively (Table 6). Adoption of hybrid seeds is also Schooling 0.188*** 0.015***
associated with an increase of 29.8 % in agricultural expenditures, 7.4 % (0.049) (0.003)
in agricultural income, and 7.0 % in gross income over the counterfac­ Labor constrained − 0.178*** 0.009
(0.048) (0.022)
tual mean (estimates statistically significant at 1 % level of significance). Extension Contact 0.263*** 0.114***
We observe similarities in the relationship between capital-intensive (0.073) (0.022)
practices and outcome variables across the two production systems. Climate Information 0.074 0.079***
While feed storage and hybrid seeds may be effective in improving Services
(0.077) (0.025)
overall household wellbeing, capital requirement can be a major barrier
Savings Group 0.289*** 0.108***
to adoption of these technologies (e.g., Emerick et al., 2016; Amadu (0.069) (0.025)
et al., 2020). Herd Size 0.008*** 0.010***
Herd and crop-diversification strategies are analogous on-farm (0.003) (0.003)
diversification strategies in the two production systems. In Kenya,
herd-diversification has no significant relationships with household Observations 664 2996
wellbeing or farm-related outcomes (Table 5). In comparison, crop- ***, **, * indicate statistical significance at 1, 5 and 10 %, respectively.
diversification has small, positive associations with food consumption Dependent variable is the count of climate-adaptive practices adopted by
as well as subjective resilience outcomes (Table 6). Furthermore, crop- households in Kenya and Zambia.
diversification is associated with increases of 2.6 % and 2.2 % in agri­ Both regressions are Poisson regressions with ward-level fixed effects.
cultural and gross income, respectively. Taken together, the results Standard errors in parenthesis are clustered at the village-level for Kenya and
suggest that crop diversification may be able to generate marginally CWAC-level for Zambia.
better benefits in terms of household wellbeing than livestock diversi­
fication in their respective contexts. This is perhaps due to better market may help with spreading climate or market-related risks and decreasing
integration of Zambian farmers across the two contexts. Furthermore, volatility of production (e.g., Alfani et al., 2021). These outcomes are
while diversification decisions may not always increase incomes, they not fully captured here.
In Kenya, off-farm diversification (typically includes casual labor,
boda-boda taxi service, and small businesses mostly related to charcoal
burning and sales (Quandt, 2021)) is associated with an increase of 0.40
9
The coefficients from the IPWRA regressions can be converted to percent­ SD in food consumption and 0.32 SD in subjective resilience against
ages by dividing the ATT with predicted outcome means of treatment group droughts over the counterfactual mean (Table 5). Furthermore, off-farm
under hypothetical non-adoption. For example, 1.54
7.63*100 = 20.1%. Similarly, diversification is associated with a decrease in livestock income by 15.2
0.668
10.33*100= 6.4%. The rest of the ATT’s are also converted to percentages using %, and an increase in total income by 23.3 % over the counterfactual
this formula.

8
H. Ahmed et al. Ecological Economics 235 (2025) 108611

Table 5
Relationship between climate-adaptive practices and household welfare outcomes in Kenya: ATT from IPWRA model.
Food Consumption Index Subjective Resilience Livestock Expenditure Livestock Income Gross Income

Feed Storage
Predicted outcome under non-adoption − 0.059 − 0.094 7.63 9.49 10.98
ATT 0.350*** (0.088) 0.256*** (0.095) 1.536*** (0.254) 0.591* (0.337) 0.680** (0.281)
Over identification test for covariate balance (p – value) 0.578

Herd-Diversification
Predicted outcome under non-adoption 0.119 0.052 9.36 11.16 11.18
ATT − 0.100 (0.145) − 0.062 (0.236) − 0.823** (0.357) − 1.185 (0.725) − 0.190 (0.669)
Over identification test for covariate balance (p – value) 0.566

Off-farm Diversification
Predicted outcome under non-adoption − 0.118 − 0.144 7.97 10.06 10.14
ATT 0.401*** (0.091) 0.318*** (0.103) 0.566** (0.227) − 1.532*** (0.307) 2.366*** (0.308)
Over identification test for covariate balance (p – value) 0.481

Collective Production
Predicted outcome under non-adoption − 0.018 − 0.191 8.10 9.05 10.63
ATT 0.568*** (0.121) 0.565*** (0.112) 0.107 (0.429) − 0.014 (0.615) 0.736* (0.411)
Over identification test for covariate balance (p – value) 0.968

N = 664.
***, **, * indicate statistical significance at 1, 5 and 10 %, respectively.
Standard errors in parenthesis are clustered at the village-level.

Table 6
Relationship between climate-adaptive practices and household welfare outcomes in Zambia: ATT from IPWRA model.
Food Consumption Subjective Resilience Subjective Resilience Agricultural Gross Agricultural
Index Drought Flood Expenditure Income Income

Hybrid or improved seed

Predicted outcome under non-adoption − 0.120 − 0.065 − 0.060 5.801 9.640 9.186
ATT 0.273*** 0.138** 0.116* 1.731*** 0.683*** 0.677***
(0.043) (0.049) (0.049) (0.093) (0.043) (0.043)
Over identification test for covariate 0.4453
balance (p – value)

Crop Diversification
Predicted outcome under non-adoption − 0.047 − 0.051 − 0.042 6.526 9.846 9.380
ATT 0.114* 0.096** 0.076* 0.188 0.214*** 0.243***
(0.044) (0.037) (0.034) (0.104) (0.041) (0.037)
Over identification test for covariate 0.9679
balance (p – value)

Off-Farm Diversification
Predicted outcome under non-adoption 0.041 − 0.059 − 0.015 6.294 9.638 9.416
ATT − 0.027 0.220*** 0.085 1.123*** 0.905*** 0.318***
(0.048) (0.043) (0.046) (0.114) (0.047) (0.041)
Over identification test for covariate 0.3590
balance (p – value)

Conservation Agriculture
Predicted outcome under non-adoption 0.014 0.131 0.092 6.842 9.962 9.485
ATT 0.011 − 0.271*** − 0.210*** − 0.390** 0.085 0.152**
(0.044) (0.053) (0.060) (0.130) (0.047) (0.048)
Over identification test for covariate 0.5528
balance (p – value)
# of Observations 2996 2996 2996 2996 2996 2996

***, **, * indicate statistical significance at 1, 5 and 10 %, respectively.

Standard errors in parenthesis are clustered at the CWAC-level.

9
H. Ahmed et al. Ecological Economics 235 (2025) 108611

mean. Generating off-farm income has an overall positive relationship

with total income and household wellbeing. However, with greater off-
farm diversification, livestock income goes down likely because house­
holds have less time to dedicate to livestock management. In compari­
son, off-farm diversification does not have statistically significant
associations with food consumption in Zambia. Still, it is associated with
an increase of 0.22 SD in subjective resilience against drought.
Furthermore, off-farm diversification in Zambia is associated with an
increase in agricultural income (by 3.4 %) and total income (by 9.4 %).
In contrast to Kenya, off-farm diversification in Zambia, which typically
involves work in agricultural retail and trading businesses, complements
agricultural income. The heterogeneity in relationships in the two con­
texts arises because in Kenya, non-farm work is often outside of the
agrifood system, while in Zambia, non-farm work often includes mar­
keting and trading of own farm produce.
Lastly, participation in collective production with community orga­
nizations and conservation agriculture practices are system specific
labor-intensive practices in Kenya and Zambia, respectively. Collective
production is associated with an increase in food consumption and Fig. 4. Relationships between climate-adaptive practices and resilience ca­
subjective resilience of 0.57 SD and 0.56 SD over the counterfactual pacities in Kenya. Note: Lines indicate 95 % confidence intervals.
mean, respectively (Table 5). As expected, we find no associations be­
tween collective production and farm-level outcomes since collective
production generally occurs on communal lands. However, it is associ­
ated with an increase of 7 % in total income over the counterfactual
mean. Collective production may provide an additional avenue of live­
lihood diversification by supplying households with supplementary
agricultural goods (mostly vegetables from collective kitchen gardens)
that can be sold in the nearby market for extra income. Indeed, the
ability to act collectively has been considered an important aspect in
building adaptive capacity in varied contexts (Adger, 2003; Villamayor-
Tomas and García-López, 2017).10
Conservation agriculture practices in Zambia are not significantly
associated with food security outcomes and the impacts on farm income
are not economically significant. Furthermore, these practices are
negatively associated with subjective resilience to droughts and floods.
This is likely because conservation agriculture benefits, if any, take time
to be observed (Thierfelder et al., 2015; Thierfelder et al., 2017). Both
the improvement of soil quality and the acquisition of proper knowledge
by farmers regarding the implementation of these practices require
multiple agricultural seasons. While our survey does not provide infor­
mation on the duration of farmers’ adoption of conservation agriculture
practices, it is evident from the data that their implementation is not yet
prevalent in the study area. Despite a nearly eight-fold increase in the
number of hectares dedicated to conservation agriculture between 2009
and 2016 in Zambia, the set of practices were utilized on less than 1 % of
the country’s arable land in 2016 (Kassam et al., 2019), suggesting
adoption is probably recent. Additionally, the potential for increased
yields frequently relies on a range of factors and the integrated adoption
of diverse practices and use of inputs (Giller et al., 2009). The lack of
short-term welfare effects of adoption of conservation practices may
break the feedback loop between increased resilience and climate-
adaptive practices.11

5.3. Relationships between climate-adaptive practices and resilience

capacities

In this section, we estimate the impacts of adopting different on- and Fig. 5. Relationships between climate-adaptive practices and resilience ca­
off-farm practices on the different dimensions (robustness, farm pacities in Zambia.

adaptability, transformability, and livelihood adaptability) of resilience

10
Collective production as well as off-farm diversification can be effective for capacities based on Eq. 1. We expect these strategies to have heteroge­
livestock owners as well as non-owners of livestock. Table A8 shows that the neous relationships with different resilience capacities. Additionally, the
treatment effects for full sample are similar to that of the livestock owners in the impacts of on- or off-farm strategies may not necessarily be concentrated
sample, illustrating robustness of our impact estimates. among particular dimensions of resilience. Rather, there may be com­
11
Tables A1 and A2, and Figs. A1 and A2 illustrate that balance and overlap plementarities in the relationships of these practices across resilience
assumptions hold for treatment effect models.

10
H. Ahmed et al. Ecological Economics 235 (2025) 108611

capacities. Table 7
Figures 4 and 5 show that climate-adaptive practices indeed have Combined effects of climate-adaptive practices on household wellbeing in
heterogeneous relationships with different resilience capacities. In Kenya: ATT from IPWRA multivalued treatment effects model.
Kenya, feed storage has a positive association with on-farm adaptability Food Consumption Subjective Gross
as well as transformability, while off-farm diversification has positive Index Resilience Income
associations with robustness and adaptability capacities, with its effect Predicted outcome under − 0.234 − 0.164 9.42
on transformability narrowly missing statistical significance at the 5 % non-adoption
level (Fig. 4). Furthermore, collective production shows positive re­ ATT1vs0 0.121 (0.108) 0.069 (0.165) 0.985**
(0.503)
lationships with all resilience capacities, illustrating the importance of
ATT2vs0 0.269** (0.118) 0.227 (0.171) 2.090***
formal or informal risk-sharing and production networks in rural set­ (0.503)
tings. On the other hand, in Zambia, the associations related to on-farm ATT3vs0 0.745*** (0.181) 0.554* (0.298) 2.587***
strategies are concentrated among on-farm resilience capacities and off- (0.601)
farm diversification has statistically significant association only with ATT4vs0 1.551*** (0.205) 1.982*** 3.112***
(0.240) (0.469)
livelihood adaptability and no correlations with on-farm resilience ca­ # of Observations 664 659 662
pacities (Fig. 5). These results highlight that, depending on the context,
climate-adaptive practices may strengthen all or some resilience ***, **, * indicate statistical significance at 1, 5 and 10 %, respectively.
Standard errors in parenthesis are clustered at the village-level.
capacities.
The treatment variable ranges from 0 to 4, with 0 = no adoption, 1 = any one of
The heterogenous relationships between climate-adaptive practices
the four practices is adopted, 2 = any two of the four practices are adopted, 3 =
and resilience capacities raise further questions regarding the causality
any three of the four practices are adopted, and 4 = all four practices are
of these relationships and policy implications. Development programs adopted.
need to carefully consider several aspects of absorptive, adaptive, and ATT1vs0 is the average treatment effect on treated for households that adopt any
transformative capacities to meaningfully enhance the overall resilience one practice versus households that do not adopt any practice, and so on.
of farm-households against climatic extremes. Policies, such as
strengthening agricultural extension, alleviating capital constraints
through credit markets that work for the rural poor, or improving the
overall educational outcomes in rural areas, can simultaneously affect
adaptation action as well as multiple perceived resilience capacities,
effect models is that, while they allow us to estimate the differential
improving the overall resilience of farm-households in SSA (as in Table 4
impacts of additional adoption, they do not indicate what combination
and Table A5). Furthermore, evidence from programs like the “gradu­
of strategies is the most effective for improving household wellbeing.12
ation approach,” which address both external constraints and internal
factors such as agency, locus of control, and perceived self-efficacy
6. Conclusions
(closely related to perceived resilience), suggests that enhancing both
external resources and internal capacities can significantly improve
In this paper, we used comparable farm-level data from pastoralists
beneficiaries’ well-being. These interventions may help households
of Kenya and smallholder rain-fed crop producers of Zambia to (i)
break free from poverty traps and foster sustained resilience to climatic
measure farm-households’ resilience to climatic stressors using the
and economic shocks (Wuepper and Lybbert, 2017).
conceptual framework of adaptive cycles in socio-ecological systems, (ii)
assess the correlates of adoption of climate-adaptive practices, and (iii)
5.4. Multivalued treatment effects of climate-adaptive practices
examine the relationship between climate-adaptive practices and
household wellbeing in diverse production systems.
Given that adaptive practices may be adopted in combination as part
We show that our resilience measure is conceptually and statistically
of a broader farm or livelihood adaptation strategy, it is important to
robust. While the scale offers a snapshot view of households’ perceived
understand how the impact estimates change at different intensities of
resilience, it has the potential to measure household resilience without
adoption. Tables 7 and 8 provide the ATT for different adoption in­
requiring copious amounts of data, making it suitable for use as a
tensities in Kenya and Zambia, respectively. In both countries, adoption
monitoring and evaluation tool in rural development projects. Further­
of one strategy has no statistically significant impact on food security
more, because of its ability to be replicated across diverse rural contexts,
and resilience, although income increases in both contexts relative to
it can be used to conduct comparative analysis that can help in the
non-adoption. Most of the complementarities in effects of adopting
design of broad-based development interventions that aim to enhance
bundles of practices are observed in terms of food consumption and
the resilience of rural households.
income in both countries. However, the magnitude of these effects dif­
Our findings demonstrate that capital-intensive climate-adaptive
fers across Kenya and Zambia. In Kenya, we also observe this comple­
practices, such as feed storage in Kenya and hybrid seeds in Zambia,
mentarity in effects for subjective resilience. However, we do not
consistently correlate with improved household wellbeing in both
observe this for Zambia, likely due to the negative association between
countries. Similarly, off-farm diversification strategies and system-
one of the climate-adaptive practices (conservation agriculture) and
specific labor-intensive strategies, such as conservation agriculture in
subjective resilience measures, which may cancel out the positive im­
Zambia and collective production in Kenya, also show heterogeneous
pacts of other strategies.
but generally positive relationships with household wellbeing. Addi­
Results from multivalued treatment effect models in Tables 7 and 8
tionally, households may benefit the most when they adopt a
show that combining several adaptive practices simultaneously is
associated with increases in household wellbeing and income for most
outcomes. These results, in line with Tabe-Ojong et al. (2023), highlight
the increasing returns to additional adoption of adaptive practices. Thus, 12
In terms of structure, there are 16 possible combinations per country,
programs that promote the adoption of multiple strategies together are
making detailed reporting challenging. However, given that capital intensive
most likely to have the highest returns for rural households in SSA,
strategies yield the highest impacts (Table 5 and Table 6), and that greater
though they must also consider potential trade-offs between farm and intensity of adoption yields complementary welfare benefits (Table 7 and
off-farm activities. However, a key limitation of multivalued treatment Table 8), we can deduce that combinations that join capital-intensive strategies
with other strategies may yield the largest benefits in the two contexts under
investigation.

11
H. Ahmed et al. Ecological Economics 235 (2025) 108611

Table 8
Combined effects of climate-adaptive practices on household wellbeing in Zambia: ATT from IPWRA multivalued treatment effects model.
Food Consumption Subjective resilience index, Subjective resilience index, Gross Agricultural
Index drought flood Income Income

Predicted Outcome under Non- − 0.137 − 0.120 − 0.111 9.055 8.781

adoption
ATT1vs0 0.003 0.040 0.083 0.383*** 0.318***
(0.077) (0.093) (0.116) (0.070) (0.095)
ATT2vs0 0.102 0.113 0.094 0.942*** 0.776***
(0.089) (0.083) (0.104) (0.076) (0.099)
ATT3vs0 0.177 0.142 0.167 1.209*** 0.930***
(0.094) (0.086) (0.120) (0.081) (0.101)
ATT4vs0 0.446*** 0.084 0.020 1.795*** 1.419***
(0.100) (0.102) (0.122) (0.129) (0.144)
# of Observations 2996 2996 2996 2996 2996

***, **, * indicate statistical significance at 1, 5 and 10 %, respectively.

Standard errors in parenthesis are clustered at the CWAC-level.
The treatment variable ranges from 0 to 4, with 0 = no adoption, 1 = any one of the four practices is adopted, 2 = any two of the four practices are adopted, 3 = any
three of the four practices are adopted, and 4 = all four practices are adopted.
ATT1vs0 is the average treatment effect on treated for households that adopt any one practice versus households that do not adopt any practice, and so on.

combination of these practices, given the complementarity of on- and Understanding these relationships is critical to designing and imple­
off-farm climate-adaptive practices. menting effective programs that address the multiple dimensions of risk
However, in the presence of capital, labor, and informational con­ to wellbeing of farmers posed by a rapidly changing climate.
straints, which are ubiquitous in these settings, the adoption of climate-
adaptive practices may be hampered. Addressing these constraints CRediT authorship contribution statement
individually is often insufficient, as they are interconnected and can
compound one another, creating further barriers to adoption. This im­ Haseeb Ahmed: Writing – review & editing, Writing – original draft,
plies that integrated rural development programs designed to address Visualization, Project administration, Methodology, Formal analysis,
multiple economic constraints simultaneously are critical for improving Data curation, Conceptualization. Juan Sebastian Correa: Writing –
the adoption rates of climate-adaptive practices. Programs that provide review & editing, Writing – original draft, Visualization, Project
a combination of financial and psychological support, training, access to administration, Methodology, Formal analysis, Data curation, Concep­
agricultural inputs, and market information are better equipped to tualization. Nicholas J. Sitko: Writing – review & editing, Writing –
overcome these challenges and generate more significant and sustained original draft, Supervision, Project administration, Methodology,
benefits for rural communities. Such holistic approaches, which tackle Investigation, Conceptualization.
several barriers at once, have the potential to create synergies, allowing
households to leverage resources more effectively and increasing their Declaration of competing interest
resilience in the face of climate-related shocks. Evidence from integrated
interventions, such as social protection programs in Bangladesh, high­ We confirm that the contents of this manuscript have not been
lights the effectiveness of combining different forms of support to ach­ published, nor are they under review elsewhere. All co-authors have
ieve long-term resilience outcomes (Ahmed et al., 2024). Therefore, consented to the submission of this paper. We declare no conflicts of
future policy interventions should aim to implement comprehensive, interest.
multi-dimensional support systems that simultaneously address internal
capacities (such as perceptions about wellbeing) and external con­ Acknowledgements
straints (e.g., capital, labor, and informational) to foster broader adop­
tion of climate-adaptive practices and build more resilient rural This research was funded by the Food and Agriculture Organization
communities. of the United Nations through the Flexible Multi-Partner Mechanism
This analysis also highlights areas of future research, such as un­ (FMM), under the project “Building Back Better and Greener: Integrated
derstanding the underlying mechanisms and causal relationships be­ approaches for an inclusive and green COVID-19 recovery in rural
tween adaptive action, resilience, and household wellbeing. spaces” (project symbol: FMM/GLO/203/MUL).

12
H. Ahmed et al. Ecological Economics 235 (2025) 108611

Appendix A. Appendix

Fig. A1: Overlap graphs for climate adaptive practices (Kenya).

13
H. Ahmed et al. Ecological Economics 235 (2025) 108611

Fig. A2: Overlap graphs for climate adaptive practices (Zambia).

14
H. Ahmed et al. Ecological Economics 235 (2025) 108611

Fig. A3: Overlap graphs for multivalued treatment effects model (Kenya).

15
H. Ahmed et al. Ecological Economics 235 (2025) 108611

Fig. A4: Overlap graphs for multivalued treatment effects model (Zambia).

16
H. Ahmed et al. Ecological Economics 235 (2025) 108611

Fig. A5: Quadratic fit plots for subjective resilience capacities (Kenya).

17
H. Ahmed et al. Ecological Economics 235 (2025) 108611

Fig. A6: Quadratic fit plots for subjective resilience capacities (Zambia).

Table A1: Normalized differences of covariates between adopters and non-adopters after IPW in Kenya.

Feed Storage Herd Diversification Off-farm Work Collective Production

Herd Size 0.070 − 0.004 0.070 0.015

Age of Household Head − 0.002 − 0.064 0.019 − 0.008
Household Size − 0.007 0.049 − 0.004 − 0.001
Male-Headed Household 0.008 0.045 − 0.033 0.011
Market Distance − 0.002 0.053 − 0.021 0.004
Saving Group − 0.022 0.008 0.035 − 0.002
Formal Schooling 0.042 0.010 0.008 0.006
Climate Information Services − 0.017 − 0.012 − 0.062 − 0.025
Extension Contact − 0.030 − 0.014 0.044 0.002

Table A2: Normalized differences of covariates between adopters and non-adopters after IPW in Zambia.

Crop Diversification Conservation Agriculture Hybrid or improved seed Off-Farm Diversification

Age 0.00 0.01 0.03 − 0.01

Household Size 0.02 0.01 − 0.01 − 0.01
Gender 0.00 − 0.01 0.03 0.01
Saving Group − 0.00 − 0.01 0.01 − 0.02
Schooling − 0.01 − 0.00 0.02 0.02
Climate Information Services 0.00 − 0.01 0.02 − 0.02
Extension Contact − 0.01 − 0.02 0.04 − 0.01

18
H. Ahmed et al. Ecological Economics 235 (2025) 108611

Table A3: Differences in means of outcome variables for livestock owners and non-owners in Kenya Data.

Owners Non-owners P-value

Food Consumption Index − 0.016 0.094 0.269

(0.038) (0.100)
Food Security Index − 0.004 0.022 0.787
(0.049) (0.052)
Gross Income 130,429 59,149 0.000
(175,450) (12,190)
P-values for two-tailed pairwise t-tests are provided in the third column.
N for owners = 668.
N for non-owners = 120.

Table A4: Comparisons for food security and subjective resilience against droughts in Kenya and Zambia.

Kenya Zambia P-values

Food Security Index − 0.288 0.076 0.000

(0.732) (1.046)
Subjective Resilience Index, Droughts − 0.686 0.177 0.000
(0.867) (0.954)
# of Obs 788 3002

Table A5: Correlates of perceived resilience in Kenya and Zambia.

Kenya Zambia

Subjective resilience index, Subjective resilience index, Subjective resilience Subjective resilience index, Subjective resilience
drought drought index, flood drought index, flood

Household size − 0.005 0.008 0.015* 0.009 0.017*

(0.017) (0.008) (0.009) (0.008) (0.009)
Gender (Male = 1) − 0.076 0.102** 0.115*** 0.099** 0.109***
(0.083) (0.042) (0.041) (0.042) (0.041)
Age − 0.006** − 0.005*** − 0.008*** − 0.005*** − 0.008***
(0.002) (0.001) (0.001) (0.001) (0.001)
Schooling − 0.013 0.004 − 0.003 0.004 − 0.004
(0.071) (0.005) (0.005) (0.005) (0.005)
Labor constrained − 0.173** − 0.027 − 0.035 − 0.031 − 0.035
(0.084) (0.050) (0.046) (0.049) (0.045)
Extension Contact 0.170** 0.147*** 0.064 0.147*** 0.069
(0.082) (0.049) (0.044) (0.048) (0.044)
Climate Information 0.152 − 0.060 0.008 − 0.045 0.029
Services
(0.122) (0.055) (0.051) (0.055) (0.050)
Savings Group − 0.089 0.011 0.115** 0.013 0.116**
(0.095) (0.053) (0.051) (0.054) (0.051)
Herd Size 0.0006 0.013** 0.007 0.013** 0.007
(0.003) (0.005) (0.005) (0.005) (0.005)
Drought shock – – – − 0.124***
(0.046)
Flood shock – – – − 0.142***
(0.050)
Observations 668 2996 2996 2996 2996
***, **, * indicate statistical significance at 1, 5 and 10 %, respectively.
Standard errors in parenthesis are clustered at the CWAC-level for Zambia and village-level for Kenya.
All regressions are OLS regressions with ward-level fixed effects to account for unobserved heterogeneity at the ward-level.

Table A6: Adoption of climate adaptive practices in kenya: marginal effects from treatment model (Logit Regressions).

Feed Storage Herd Diversification Off-farm Work Collective Production

Herd Size − 0.017 0.071*** 0.002 − 0.007*

(0.012) (0.009) (0.002) (0.004)
Age of Household Head 0.001 − 0.005 − 0.003* 0.002**
(0.008) (0.001) (0.001) (0.0009)
Household Size 0.016*** 0.007 0.029*** 0.003
(0.005) (0.006) (0.007) (0.005)
Male Headed Household 0.084* 0.030 − 0.115*** 0.032
(0.044) (0.029) (0.038) (0.030)
Market Distance 0.002 0.007*** − 0.009 − 0.005
(0.019) (0.001) (0.018) (0.003)
Saving Group 0.171*** 0.027 0.136*** 0.073*
(0.047) (0.036) (0.041) (0.042)
(continued on next page)

19
H. Ahmed et al. Ecological Economics 235 (2025) 108611

(continued )
Feed Storage Herd Diversification Off-farm Work Collective Production

Formal Schooling 0.006 − 0.018 0.230*** 0.041

(0.003) (0.039) (0.034) (0.031)
Climate Information Services 0.248*** − 0.063 0.099* 0.039
(0.041) (0.053) (0.057) (0.044)
Extension Contact 0.070 0.052* 0.068** 0.014
(0.054) (0.031) (0.031) (0.048)
Chari − 0.053 − 0.159* 0.146** 0.081
(0.065) (0.088) (0.059) (0.124)
Oldonyiro 0.053 − 0.094 0.188*** 0.022
(0.054) (0.070) (0.058) (0.091)
Sericho − 0.091 0.021 − 0.035 − 0.152**
(0.061) (0.063) (0.054) (0.060)
N = 664
***, **, * indicate statistical significance at 1, 5 and 10 %, respectively.
Standard errors in parenthesis are clustered at the village-level.

Table A7: Adoption of climate adaptive practices in Zambia: marginal effects from treatment model (Logit Regressions).

Crop Conservation agriculture Hybrid or improved seed Off-farm Diversification

diversification

Household size − 0.002 − 0.008* 0.021*** 0.012***

(0.005) (0.004) (0.005) (0.004)
Gender 0.026 0.069*** 0.055** 0.005
(0.024) (0.019) (0.025) (0.024)
Age 0.002*** 0.001 0.001 − 0.001*
(0.001) (0.001) (0.001) (0.001)
Years of Education 0.004 − 0.000 0.018*** 0.005**
(0.003) (0.002) (0.003) (0.003)
Labor Constraint − 0.018 − 0.015 − 0.004 0.023
(0.025) (0.022) (0.028) (0.024)
Extension Contact 0.008 0.018 0.059*** 0.143***
(0.019) (0.021) (0.021) (0.021)
Climate Information Services − 0.005 0.083*** 0.021 0.048*
(0.025) (0.024) (0.021) (0.026)
Saving Group 0.005 0.004 0.108*** 0.085***
(0.025) (0.026) (0.029) (0.030)
Observations 2996 2996 2996 2996
***, **, * indicate statistical significance at 1, 5 and 10 %, respectively.
Standard errors in parenthesis are clustered at the CWAC-level.

Table A8: Relationship between collective production, off-farm diversification and household welfare outcomes for full sample in Kenya: ATT from IPWRA model.

Food Consumption Index Subjective Resilience Livestock Expenditure Livestock Income Gross Income

Off-farm Diversification
Predicted outcome under non-adoption − 0.162 − 0.128 7.179 8.260 8.411
ATT 0.471*** 0.239*** 0.297 − 1.236*** 3.358***
(0.090) (0.060) (0.305) (0.328) (0.363)

Herd-Diversification
Predicted outcome under non-adoption − 0.044 − 0.290 7.369 7.596 9.572
ATT 0.575*** 0.522*** 0.018 0.058 0.715*
(0.114) (0.100) (0.365) (0.529) (0.394)
N = 771
***, **, * indicate statistical significance at 1, 5 and 10 %, respectively.
Standard errors in parenthesis are clustered at the village-level.

Data availability DEEP Working paper 25, Data and Evidence to End Extreme Poverty Research
Programme, Oxford. https://doi.org/10.55158/DEEPWP25.
Alfani, F., Arslan, A., McCarthy, N., Cavatassi, R., Sitko, N., 2021. Climate resilience in
Data will be made available on request. rural Zambia: evaluating farmers’ response to El Niño-induced drought. Environ.
Dev. Econ. 26 (5–6), 582–604.
Amadu, F.O., McNamara, P.E., Miller, D.C., 2020. Understanding the adoption of
References
climate-smart agriculture: a farm-level typology with empirical evidence from
southern Malawi. World Dev. 126, 104692.
Abdulai, A., Huffman, W., 2014. The adoption and impact of soil and water conservation Anderson, M., 2008. Multiple inference and gender differences in the effects of early
technology: an endogenous switching regression application. Land Econ. 90 (1), intervention: a reevaluation of the abecedarian, perry preschool, and early training
26–43. projects. J. Am. Stat. Assoc. 103 (484), 1481–1495.
Adger, W.N., 2003. Social capital, collective action, and adaptation to climate change. Ansah, I.G.K., Gardebroek, C., Ihle, R., 2019. Resilience and household food security: a
Econ. Geogr. 79 (4), 387–404. review of concepts, methodological approaches and empirical evidence. Food Secur.
Ahmed A., Bakhtiar M., Hoddinott J., Roy S. (2024), Does Nutrition-Sensitive Social 11 (6), 1187–1203.
Protection Build Long-Term Resilience? Experimental Evidence from Bangladesh,

20
H. Ahmed et al. Ecological Economics 235 (2025) 108611

Ansah, I.G.K., Kotu, B.H., Manda, J., Muthoni, F., Azzarri, C., 2023. Mediation and Jones, L., Tanner, T., 2017. ‘Subjective resilience’: using perceptions to quantify
moderation roles of resilience capacity in the shock–food-security nexus in northern household resilience to climate extremes and disasters. Reg. Environ. Chang. 17,
Ghana. Ecol. Econ. 211, 107894. 229–243.
Arslan, A., McCarthy, N., Lipper, L., Asfaw, S., Cattaneo, A., Kokwe, M., 2015. Climate Kassam, A., Friedrich, T., Derpsch, R., 2019. Global spread of conservation agriculture.
smart agriculture? Assessing the adaptation implications in Zambia. J. Agric. Econ. Int. J. Environ. Stud. 76 (1), 29–51.
66 (3), 753–780. Khanal, U., Wilson, C., Hoang, V.N., Lee, B., 2018. Farmers’ adaptation to climate
Asfaw, S., Scognamillo, A., Di Caprera, G., Sitko, N., Ignaciuk, A., 2019. Heterogeneous change, its determinants and impacts on rice yield in Nepal. Ecol. Econ. 144,
impact of livelihood diversification on household welfare: cross-country evidence 139–147.
from sub-Saharan Africa. World Dev. 117, 278–295. KIHBS, 2018. Basic Report. Kenya Integrated Household Budget Survey 2015–2016.
Azzarri, C., Signorelli, S., 2020. Climate and poverty in Africa south of the Sahara. World Kenya National Bureau of Statistics.
Dev. 125, 104691. Lockwood, M., Raymond, C.M., Oczkowski, E., Morrison, M., 2015. Measuring the
Barrett, C.B., Ghezzi-Kopel, K., Hoddinott, J., Homami, N., Tennant, E., Upton, J., Wu, T., dimensions of adaptive capacity: a psychometric approach. Ecol. Soc. 20 (1).
2021. A scoping review of the development resilience literature: theory, methods Maggio, G., Sitko, N., 2019. Knowing is half the battle: seasonal forecasts, adaptive
and evidence. World Dev. 146, 105612. cropping systems, and the mediating role of private markets in Zambia. Food Policy
Bartelet, H.A., Barnes, M.L., Cumming, G.S., 2022. Determinants, outcomes, and 89, 101781.
feedbacks associated with microeconomic adaptation to climate change. Reg. Maggio, G., Mastrorillo, M., Sitko, N.J., 2022. Adapting to high temperatures: effect of
Environ. Chang. 22 (2), 59. farm practices and their adoption duration on total value of crop production in
Béné, C., Frankenberger, T., Griffin, T., Langworthy, M., Mueller, M., Martin, S., 2019. Uganda. Am. J. Agric. Econ. 104 (1), 385–403.
‘Perception matters’: new insights into the subjective dimension of resilience in the Maja, M.M., Idiris, A.A., Terefe, A.T., Fashe, M.M., 2023. Gendered vulnerability,
context of humanitarian and food security crises. Prog. Dev. Stud. 19 (3), 186–210. perception and adaptation options of smallholder farmers to climate change in
Berhe, M., Hoag, D., Tesfay, G., Tadesse, T., Oniki, S., Kagatsume, M., Keske, C.M., 2017. eastern Ethiopia. Earth Syst. Environ. 7 (1), 189–209.
The effects of adaptation to climate change on income of households in rural Manda, J., Gardebroek, C., Kuntashula, E., Alene, A.D., 2018. Impact of improved maize
Ethiopia. Pastoralism 7 (1), 1–15. varieties on food security in eastern Zambia: a doubly robust analysis. Rev. Dev.
Burnham, M., Ma, Z., 2017. Climate change adaptation: factors influencing Chinese Econ. 22 (4), 1709–1728.
smallholder farmers’ perceived self-efficacy and adaptation intent. Reg. Environ. Meuwissen, M.P., Feindt, P.H., Spiegel, A., Termeer, C.J., Mathijs, E., De Mey, Y.,
Chang. 17, 171–186. Finger, R., Balmann, A., Wauters, E., Urquhart, J., Vigani, M., 2019. A framework to
Calzadilla, A., Zhu, T., Rehdanz, K., Tol, R.S., Ringler, C., 2013. Economywide impacts of assess the resilience of farming systems. Agric. Syst. 176, 102656.
climate change on agriculture in sub-Saharan Africa. Ecol. Econ. 93, 150–165. Moffitt, R.A., 1999. New developments in econometric methods for labor market
Carter, M.R., Laajaj, R., Yang, D., 2016. Savings, Subsidies, and Technology Adoption: analysis. In: Handbook of Labor Economics, vol. 3. Elsevier, pp. 1367–1397.
Field Experimental Evidence from Mozambique. NBER Working Paper, 20465. Mulwa, C.K., Visser, M., 2020. Farm diversification as an adaptation strategy to climatic
Cattaneo, M.D., 2010. Efficient semiparametric estimation of multi-valued treatment shocks and implications for food security in northern Namibia. World Dev. 129,
effects under ignorability. J. Econ. 155 (2), 138–154. 104906.
Cattaneo, M.D., Drukker, D.M., Holland, A.D., 2013. Estimation of multivalued treatment Mutabazi, K.D., Sieber, S., Maeda, C., Tscherning, K., 2015. Assessing the determinants of
effects under conditional independence. Stata J. 13 (3), 407–450. poverty and vulnerability of smallholder farmers in a changing climate: the case of
Clare, A., Graber, R., Jones, L., Conway, D., 2017. Subjective measures of climate Morogoro region, Tanzania. Reg. Environ. Chang. 15, 1243–1258.
resilience: what is the added value for policy and programming? Glob. Environ. Ndiritu, S.W., Muricho, G., 2021. Impact of climate change adaptation on food security:
Chang. 46, 17–22. evidence from semi-arid lands, Kenya. Clim. Chang. 167 (1), 1–20.
Darnhofer, I., 2014. Resilience and why it matters for farm management. Eur. Rev. Agric. Nguyen, K.V., James, H., 2013. Measuring household resilience to floods: a case study in
Econ. 41 (3), 461–484. the Vietnamese Mekong River Delta. Ecol. Soc. 18 (3).
Davis, B., Di Giuseppe, S., Zezza, A., 2017. Are African households (not) leaving Pannell, D.J., Llewellyn, R.S., Corbeels, M., 2014. The farm-level economics of
agriculture? Patterns of households’ income sources in rural sub-Saharan Africa. conservation agriculture for resource-poor farmers. Agric. Ecosyst. Environ. 187,
Food Policy 67, 153–174. 52–64.
D’Errico, M., Romano, D., Pietrelli, R., 2018. Household resilience to food insecurity: Quandt, A., 2021. Coping with drought: narratives from smallholder farmers in semi-arid
evidence from Tanzania and Uganda. Food Secur. 10 (4), 1033–1054. Kenya. Intern. J. Disaster Risk Reduct. 57, 102168.
Di Falco, S., Veronesi, M., 2014. Managing environmental risk in presence of climate Rosenbaum, P.R., Rubin, D.B., 1983. The central role of the propensity score in
change: the role of adaptation in the Nile Basin of Ethiopia. Environ. Resour. Econ. observational studies for causal effects. Biometrika 70 (1), 41–55.
57, 553–577. Roy, R., Gain, A.K., Samat, N., Hurlbert, M., Tan, M.L., Chan, N.W., 2019. Resilience of
Emerick, K., De Janvry, A., Sadoulet, E., Dar, M.H., 2016. Technological innovations, coastal agricultural systems in Bangladesh: assessment for agroecosystem
downside risk, and the modernization of agriculture. Am. Econ. Rev. 106 (6), stewardship strategies. Ecol. Indic. 106, 105525.
1537–1561. Schwab, B., Janzen, S., Magnan, N.P., Thompson, W.M., 2020. Constructing a summary
Esposti, R., 2017. The heterogeneous farm-level impact of the 2005 CAP-first pillar index using the standardized inverse-covariance weighted average of indicators.
reform: a multivalued treatment effect estimation. Agric. Econ. 48 (3), 373–386. Stata J. 20 (4), 952–964.
Fischer, E., Qaim, M., 2012. Linking smallholders to markets: determinants and impacts Scognamillo, A., Sitko, N., Bandara, S., Hewage, S., Munaweera, T., Kwon, J., 2022. The
of farmer collective action in Kenya. World Dev. 40 (6), 1255–1268. challenge of making climate adaptation profitable for farmers: evidence from Sri
Folke, C., Carpenter, S.R., Walker, B., Scheffer, M., Chapin, T., Rockström, J., 2010. Lanka’s rice sector. Environ. Dev. Econ. 1–19.
Resilience thinking: integrating resilience, adaptability and transformability. Ecol. Sertse, S.F., Khan, N.A., Shah, A.A., Liu, Y., Naqvi, S.A.A., 2021. Farm households’
Soc. 15 (4). perceptions and adaptation strategies to climate change risks and their determinants:
Giller, K.E., Witter, E., Corbeels, M., Tittonell, P., 2009. Conservation agriculture and evidence from Raya Azebo district, Ethiopia. Intern. J. Disaster Risk Reduct. 60,
smallholder farming in Africa: the heretics’ view. Field Crop Res. 114 (1), 23–34. 102255.
Government of Kenya, 2016. Kenya National Adaptation Plan 2015–2030, Government Shikuku, K.M., Winowiecki, L., Twyman, J., Eitzinger, A., Perez, J.G., Mwongera, C.,
of Kenya, July 2016. NAP_Final-Signed_22022017.pdf (unfccc.int). Läderach, P., 2017. Smallholder farmers’ attitudes and determinants of adaptation to
Government of Zambia, 2023. National Adaptation Plan for Zambia. October 2023. NAP- climate risks in East Africa. Clim. Risk Manag. 16, 234–245.
Zambia-2023.pdf (unfccc.int). Smale, M., Assima, A., Kergna, A., Thériault, V., Weltzien, E., 2018. Farm family effects
Hasan, M.K., Kumar, L., 2019. Comparison between meteorological data and farmer of adopting improved and hybrid sorghum seed in the Sudan savanna of West Africa.
perceptions of climate change and vulnerability in relation to adaptation. J. Environ. Food Policy 74, 162–171.
Manag. 237, 54–62. Smith, L.C., Frankenberger, T.R., 2018. Does resilience capacity reduce the negative
Heiler, P., & Kazak, E. (2021). Valid inference for treatment effect parameters under impact of shocks on household food security? Evidence from the 2014 floods in
irregular identification and many extreme propensity scores. J. Econ., 222(2), northern Bangladesh. World Dev. 102, 358–376.
1083–1108. Spiegel, A., Slijper, T., de Mey, Y., Meuwissen, M.P., Poortvliet, P.M., Rommel, J.,
Hörner, D., Wollni, M., 2021. Integrated soil fertility management and household welfare Hansson, H., Vigani, M., Soriano, B., Wauters, E., Appel, F., 2021. Resilience
in Ethiopia. Food Policy 100, 102022. capacities as perceived by European farmers. Agric. Syst. 193, 103224.
Imbens, G.W., Wooldridge, J.M., 2009. Recent developments in the econometrics of Stuch, B., Alcamo, J., Schaldach, R., 2021. Projected climate change impacts on mean
program evaluation. J. Econ. Lit. 47 (1), 5–86. and year-to-year variability of yield of key smallholder crops in sub-Saharan Africa.
Issahaku, G., Abdulai, A., 2020a. Household welfare implications of sustainable land Clim. Dev. 13 (3), 268–282.
management practices among smallholder farmers in Ghana. Land Use Policy 94, Tabe-Ojong, M.P., Aihounton, Ghislain B.D., Lokossou, Jourdain C., 2023. Climate-smart
104502. agriculture and food security: cross-country evidence from West Africa. Glob.
Issahaku, G., Abdulai, A., 2020b. Adoption of climate-smart practices and its impact on Environ. Chang. 81, 102697.
farm performance and risk exposure among smallholder farmers in Ghana. Aust. J. Teklewold, H., Kassie, M., Shiferaw, B., Köhlin, G., 2013. Cropping system
Agric. Resour. Econ. 64 (2), 396–420. diversification, conservation tillage and modern seed adoption in Ethiopia: impacts
Issahaku, G., Abdulai, A., 2020c. Can farm households improve food and nutrition on household income, agrochemical use and demand for labor. Ecol. Econ. 93,
security through adoption of climate-smart practices? Empirical evidence from 85–93.
northern Ghana. Appl. Econ. Perspect. Policy 42 (3), 559–579. Thierfelder, C., Matemba-Mutasa, R., Rusinamhodzi, L., 2015. Yield response of maize
Jones, L., d’Errico, M., 2019. Whose resilience matters? Like-for-like comparison of (Zea mays L.) to conservation agriculture cropping system in southern Africa. Soil
objective and subjective evaluations of resilience. World Dev. 124, 104632. Tillage Res. 146, 230–242.
Thierfelder, C., Chivenge, P., Mupangwa, W., Rosenstock, T.S., Lamanna, C., Eyre, J.X.,
2017. How climate-smart is conservation agriculture (CA)?–its potential to deliver

21
H. Ahmed et al. Ecological Economics 235 (2025) 108611

on adaptation, mitigation and productivity on smallholder farms in southern Africa. Walker, S.E., Bruyere, B.L., Zarestky, J., Yasin, A., Lenaiyasa, E., Lolemu, A.,
Food Secur. 9, 537–560. Pickering, T., 2022. Education and adaptive capacity: the influence of formal
Upton, J., Constenla-Villoslada, S., Barrett, C.B., 2022. Caveat utilitor: a comparative education on climate change adaptation of pastoral women. Clim. Dev. 14 (5),
assessment of resilience measurement approaches. J. Dev. Econ. 157, 102873. 409–418.
Uysal, S.D., 2015. Doubly robust estimation of causal effects with multivalued WFP, 2008. Food Consumption Analysis: Calculation and Use of the Food Consumption
treatments: an application to the returns to schooling. J. Appl. Econ. 30 (5), Score in Food Security Analysis. WFP, Rome, Italy.
763–786. Wooldridge, J.M., 2010. Econometric Analysis of Cross Section and Panel Data. MIT
Villamayor-Tomas, S., García-López, G., 2017. The influence of community-based press.
resource management institutions on adaptation capacity: a large-n study of farmer Wouterse, F., Andrijevic, M., Schaeffer, M., 2022. The microeconomics of adaptation:
responses to climate and global market disturbances. Glob. Environ. Chang. 47, evidence from smallholders in Ethiopia and Niger. World Dev. 154, 105884.
153–166. Wuepper, D., Lybbert, T.J., 2017. Perceived self-efficacy, poverty, and economic
development. Ann. Rev. Resour. Econ. 9 (1), 383–404.

22