Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Global Environmental Change 81 (2023) 102697

Contents lists available at ScienceDirect

Global Environmental Change


journal homepage: www.elsevier.com/locate/gloenvcha

“Climate-smart agriculture and food security: Cross-country evidence from


West Africa”
Martin Paul Jr. Tabe-Ojong a, *, Ghislain B.D. Aihounton b, c, Jourdain C. Lokossou d
a
International Food Policy Research Institute (IFPRI), Cairo, Egypt
b
Department of Food and Resource Economics, University of Copenhagen, Rolighedsvej 25, 1958 Frederiksberg C, Denmark
c
Laboratory of Analysis and Research on the Economic and Social Dynamics, University of Parakou, BP 123 Arafat, Parakou, Benin
d
Department of Agri-Food Economics and Consumer Sciences, Faculty of Food Science and Agriculture, Laval University, Quebec, Canada

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

JEL codes: In the face of climate change and extreme weather events which continue to have significant impacts on agri­
O12 cultural production, climate-smart agriculture (CSA) has emerged as one important entry point in reducing the
O13 emission of greenhouse gases and building climate resilience while ensuring increases in agricultural produc­
Q12
tivity with ensuing implications on food and nutrition security. We examine the relationship between CSA, land
Q54
Q56
productivity (yields), and food security using a survey of farm households in Ghana, Mali, and Nigeria. To un­
derstand the correlates of the adoption of these CSA practices as well as the association between CSA, yields, and
Keywords:
Adoption
food security, we use switching regressions that account for multiple endogenous treatments. We find a positive
Yields association between the adoption of CSA practices and yields. This increase in yields translate to food security as
Climate change we observe a positive association between CSA and food consumption scores. Although we show modest asso­
Climate-smart agriculture ciations between the independent use of CSA practices such as adopting climate-smart groundnut varieties,
Climate resilience cereal-groundnut intercropping, and the use of organic fertilizers, we find that bundling these practices may lead
West Africa to greater yield and food security gains. Under the different combinations, the use of climate-smart groundnut
varieties exhibit the strongest association with yields and food security. We also estimate actual-counterfactual
relationships where we show that the adoption of CSA practices is not only beneficial to CSA adopters but could
potentially be beneficial to non-CSA adopters should they adopt. These results have implications for reaching
some of the sustainable development targets, especially the twin goals of increasing agricultural productivity and
maintaining environmental sustainability.

1. Introduction agriculture could be one avenue to drive mitigation efforts for climate
change but also it is important that farmers continually adapt to these
Climate change, expressed through both high and reduced rainfall changing climates (Cohn et al. 2017; Fujimori et al. 2019). Climate-
and rising temperatures, coupled with droughts continue to be envi­ smart agriculture (CSA) has emerged as a viable alternative to conven­
ronmental challenges plaguing smallholder agriculture in many tropical tional agriculture given its ability to simultaneously reduce GHG emis­
regions (Lobell et al. 2011; Ortiz-Bobea et al. 2021; Qiao et al. 2022). sions while tackling agricultural production and food insecurity as well
Smallholder agriculture and climate change exhibit an interconnected as building climate resilience (Lipper et al. 2014; Partey et al. 2018;
relationship with visible effects on each other (Cohn et al. 2017). It has Ntinyari and Gweyi-Onyango 2021). Despite some attempts, empirical
been established that climate change could greatly affect agriculture insights on the relationship between CSA, land productivity, and food
with material impacts on food and nutrition security (Wheeler and von security are not fully understood especially given the significant varia­
Braun, 2013; Hertel 2016; Michler et al. 2019). On the other hand, tions and heterogeneities in various production systems and agro-
agriculture could further exacerbate climate change by releasing ecological zones.
greenhouse gases (GHGs) from unsustainable farm practices and inputs In the paper, we examine the relationship between CSA, land pro­
(Barrios et al. 2008; Bennetzen et al. 2016; IPCC 2019). In this regard, ductivity (yields) and food security. Rainfall anomalies (shortfalls and

* Corresponding author.
E-mail addresses: m.p.tabe-ojong@cgiar.org (M.P.Jr. Tabe-Ojong), aihountong@gmail.com (G.B.D. Aihounton), jourdain.lokossou.1@ulaval.ca (J.C. Lokossou).

https://doi.org/10.1016/j.gloenvcha.2023.102697
Received 13 September 2022; Received in revised form 11 April 2023; Accepted 12 May 2023
Available online 24 May 2023
0959-3780/© 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
M.P.Jr. Tabe-Ojong et al. Global Environmental Change 81 (2023) 102697

surpluses) have been highlighted to reduce crop yields in many small­ food security. Based on these results, we conclude that CSA is food and
holder settings (Di Falco and Chavas 2008; Michler et al. 2019). Given nutrition-sensitive and could be a critical entry and leveraging point for
the extended dry seasons in many arid and semi-arid regions of Africa, climate change adaptation and mitigation. In terms of heterogeneity, we
CSA practices such as improved crop varieties (climate-smart)1 have show that different correlates matter for the adoption of CSA practices in
been developed and disseminated. These improved climate-smart crop the different study countries.
varieties are tolerant to droughts and heat stresses and could survive and We offer several contributions to the growing literature on the
lead to higher crop yields in varying production systems (Michler et al. relationship between climate change adaptation, agriculture, and food
2019; Simtowe et al. 2019). Besides these improved crop varieties, other security. In the first place, we contribute to the evidence on the role of
climate-smart practices such as cereal-groundnut intercropping (inter­ CSA in increasing agricultural productivity and building climate resil­
cropping groundnuts with maize, millet or sorghum) and the use of ience in smallholder farming systems. Our focus is groundnut-based
organic fertilisers (farm and animal manure, and composts) also exists production systems in West Africa. About 34 % of the total world pro­
(Lipper et al. 2014; Tabe-Ojong et al., 2023a). At this point, it is duction of groundnut comes from Africa (FAOSTAT 2022). Here, the
important to already mention that CSA is built on existing knowledge, major groundnut producing countries are Nigeria, Sudan, Senegal,
technologies, and principles of sustainable intensification, making it Congo, Ghana, Chad and Niger, most of which form the West African
necessary to acknowledge the relevance of cultural ecology and local belt (FAOSTAT 2022). Secondly, we provide evidence that the com­
indigenous knowledge systems. Although similar to other agronomic plementary use of these practices offers greater productivity and food
practices, CSA is somewhat different as it seems to emphasize the use of security gains than their independent use. For instance, the use of
these practices but insists on the flexibility, context dependency and the climate-smart varieties and organic fertilisers offer greater yield gains
role of climate financing to close the deficit in investments in climate than using either of the practices independently. The third contribution
change adaptation (Lipper et al. 2014). More importantly, CSA focuses is a more policy oriented one as we establish actual-counterfactual re­
more on climate change and considers the complementarities and trade- lationships and show what the potential yields and food security gains
offs between agricultural productivity, climate change adaptation as would be for non-adopters of CSA should they adopt these CSA practices.
well as climate change mitigation. The last contribution is providing evidence beyond a single case study.
We begin by investigating the relationship between the use of the Most previous studies have looked at the implications of CSA on yields
above three CSA practices and yields. We then proceed to examine and food security using single case studies. Here, we improve on these
whether increases in yields translate to increases in food security, studies and provide evidence from three different countries (Ghana,
proxied using the food consumption score. In establishing the above, we Mali, and Nigeria) enabling us to have a wide external validity.
also indirectly examine the adoption of CSA by highlighting their drivers The rest of the study is structured as follows: Section two provides
and constraints. We use a rich cross-sectional, cross-country dataset the context and data as well as how the variables of interest are
from Ghana, Mali, and Nigeria for which we observe the cultivation of measured. The descriptive summary of these variables is also presented
groundnut, an important and strategic proteinous and oil crop which has in this section. While section three shows the empirical and estimation
been referred to as a “women’s crop (Njuki et al. 2011; Orr et al. 2016). strategy, section four discusses the results. It starts by discussing the
We use switching regressions with multiple treatments to account for the main estimation results and then the cross-country heterogeneity anal­
use of climate-smart varieties, cereal-groundnut intercropping and ysis and some robustness checks. The article concludes with some policy
organic fertilisers. This method is known as the multinomial endogenous implications in section five.
switching regression model (MESR). It is a two-stage model that uses a
multinomial logit model in the first stage to estimate the drivers of 2. Context and data
adoption of these CSA practices. A bias-adjusted ordinary least squares2
(OLS) model is then used in the second stage to model the relationship 2.1. Farm household survey
between the adoption of these practices, yields and food security. We
consider different combinations of these practices and test their impli­ This study relies on a farm household survey conducted as part of the
cation for yields and food security. We perform some robustness checks USAID-funded groundnut upscaling project meant to increase the pro­
on both the estimation strategy and the measurement of food security. ductivity and profitability of groundnut in Ghana, Mali, and Nigeria
For instance, we use the multinomial endogenous treatment effect (Fig. 1). The project follows up on earlier projects such as the Tropical
model (METE) and food security levels based on different thresholds of Legumes projects I, II, and III funded by the Bill and Melinda Gates
the food consumption score. Foundation and led by the International Crops Research Institute for the
We find a positive association between CSA and yields. These results Semi-arid tropics (ICRISAT). The groundnut upscaling project was
highlight the inherent benefits in using these CSA practices. We also implemented in the Northern, Upper East, and Upper West regions in
show that farmers that adopt these CSA practices (climate-smart vari­ Ghana; Koulikoro, and Sikasso regions in Mali; Jigawa, Katsina, Kano,
eties, cereal-groundnut intercropping and organic fertilisers) are likely Kebbi, and Sokoto States in Nigeria. Our study regions are mostly
to consume more diverse and nutritionally valued foods. Thus, adopting located in the Sudano-Savannah zone of Africa, whose rainy season lasts
households have greater caloric intake than their non-adopting coun­ from June to October with an average yearly rainfall ranging between
terparts. Considering that these CSA practices can be adopted either 800 and 1200 mm. The areas have a mean annual temperature of about
partially or bundled, we further show that bundling matters more and is 26.9 ◦ C with monthly temperatures ranging between 24 ◦ C and 38 ◦ C.
associated with higher productivity and food security gains, especially Sampling and data collection followed a multistage sampling pro­
combining the adoption of climate-smart varieties with organic fertil­ cedure where six districts/local government areas (LGAs) were selected
isers. We also find strong evidence on the important role of climate- based on their intensity of groundnut production. Some districts were
smart varieties given their huge associations with productivity and left out due to accessibility and security constraints, especially in Mali.
From these districts, about 4–6 villages were randomly selected. These
villages formed the primary sampling unit from which households were
1 interviewed. From these randomly selected villages, households were
Throughout the manuscript, we interchangeably use climate-smart crop
varieties and improved crop varieties. further randomly selected after a village census, and household listing
2
The model is usually identified by the non-linearity of the first stage model. exercises were undertaken in the selected villages. Thirty households
Additionally, exclusion restrictions are used to improve this identification. An were then randomly selected from these villages. Households were
inverse mills ratio generated from the first stage and used in the second stage interviewed in 2017 and followed in 2018 and 2019. In the first year of
controls for selection bias. data collection (2017), a total of 900 households from Ghana, 1350

2
M.P.Jr. Tabe-Ojong et al. Global Environmental Change 81 (2023) 102697

Fig. 1. Map of study area.

households from Mali and 2500 households from Nigeria were surveyed. Several indicators have been used to capture one or more of these di­
In 2018, the sample was again re-sampled due to financial constraints mensions. FCS is one of the best proxies to capture food security in most
that led to downscaling of the initial survey. Households in the Mopti contexts (Maxwell et al., 2013). It is a composite score based on
region of Mali were also dropped in 2018 due to a deteriorating security households’ dietary diversity, food frequency, and relative nutritional
situation. There were also some low rates of attrition (8% in Ghana, 7% importance of different food groups. It has been validated against
in Mali, and 4% in Nigeria) throughout the panel years which reduced quantity and caloric intake (Wiesmann et al., 2009). FCS is computed at
the sample to a total of 2,868 households, of which 498 were from the household level by inspecting the consumption frequency of food
Ghana, 840 from Mali and 1,530 from Nigeria as shown in Fig. 1. The items from the different food groups during a 7-day reference period as
sampling procedure makes adopters and non-adopters to be under follows:
similar administrative, environmental, and climatic conditions.

8
Interviews were usually conducted with household heads or their FCS = wi fi
spouse in the native language of the villages. The interviews were made 1
easy with the use of survey-based tablets by trained enumerators. The
Where wi indicates the weight of the various food groups and is equal
survey collected rich information on household characteristics such as
to four for meat and milk, three for pulses, two for staples, one for
age, educational level, gender, household size, access to institutional
vegetables and fruits, 0.5 for oil and sugar. fi varies between 0 and 7,
services like extension (private and public), training, credit as well as
indicating the consumption frequency (number of days) of different
social and physical capital. Information was also obtained on produc­
foods from a specific food group. The FCS can also be used to classify
tivity measures and the use of various farm inputs and farm character­
households into one of the three food security categories: poor (0–21),
istics. Food consumption was only captured in the last panel survey
borderline (21.5–35) and acceptable (>35) (Wiesmann et al. 2009).
(2019). As a result of this, we do not use the panel dataset but rather just
In the study area, groundnut is consumed in various forms almost
the 2019 cross-sectional survey for the analysis.
daily by households. It is part of the pulse group and is weighted three in
the calculation of the FCS, to reflect its high protein and fat concentra­
tion. Given that it is consumed daily by households, a week’s con­
2.2. Measurement of food security and land productivity
sumption of groundnut multiplied by the weight of its group is equal to a
consumption score of 21. Suppose this diet is combined with the con­
To measure food security, we rely on the food consumption score
sumption of other food categories during the week, households will be
(FCS). Food security is defined as when “all people, at all times, have
classified as borderline or acceptable food consumers regardless of the
physical, social, and economic access to sufficient, safe, and nutritious
low diversity. To account for this dietary pattern in the study area, we
food to suit their dietary needs and food preferences for an active and
follow Wiesmann et al. (2009) to adjust the thresholds by adding 21 to
healthy life.” (FAO, 1996). This definition, which is one of the most
each threshold. Households’ food security status is then set as poor
popular, includes the four main dimensions of food security: food
(0–42), borderline (42.5–56), and acceptable (>56).
availability, accessibility, utilization, and stability (Lawlis et al., 2018).

3
M.P.Jr. Tabe-Ojong et al. Global Environmental Change 81 (2023) 102697

Land productivity is also an outcome variable in this study. It is developing countries.


defined and measured as the ratio of farm production to land size. In In Figs. 2 and 3 respectively, we show the relationship between CSA
other words, land productivity is the quantity of crops harvested (kg) per and land productivity as well as CSA and FCS using box plots. We
unit of land (ha). It is expected that the food security impacts of the use observe high land productivity associated with the adoption of the CSA
of CSA would arise from increased land productivity. As highlighted practices (Fig. 2). On average, households obtain 720.60 kg/ha of
earlier, increasing land productivity is one of the triple wins of CSA. groundnut, and all adopters of the different CSA packages obtain higher
Greater food production and increased land productivity is expected to yields than non-adopters. Adopters of the full package obtain greater
boost smallholder incomes and increase food and nutrition security. yields than non-adopters. Non-adopters of CSA packages have the lowest
food consumption score, and adopters of the three CSA practices (Fig. 3)
2.3. Measurement of CSA obtain the highest food consumption score at the household level. This
trend in food consumption score corroborates with different levels of
CSA includes a suite of different sustainable agricultural practices food security, where most adopters of CSA are found to have an
relevant for increasing agricultural productivity and food security while acceptable food consumption score, and non-adopters are on average at
also ensuring environmental sustainability. For this study, we consider the borderline or poor line of the food consumption score.
three CSA practices used by farmers in the study area: climate-smart
varieties, cereal-groundnut intercropping, and organic fertilizers. For 3. Empirical estimation
the use of the climate-smart varieties, a farmer is considered as an
adopter if s(he) cultivates at least one climate-smart groundnut variety We are interested in estimating the relationship between (1) CSA and
during the last agricultural season. Some of the improved varieties used yields and (2) CSA and food security. While one could in principle use
by farmers are Samnut 22, Yenyawoso, and Nkatiesari in Ghana; ICGV linear models like OLS to get at these objectives, the use of OLS may lead
86124 (Niètatiga), ICGV 86015 (Yriwatiga), ICGV 86024 (Bonitiga) and to biased estimates given that adoption of the CSA practices are not
Fleur 11 (Allason) in Mali; Samnut 23, Samnut 24, Samnut 25, and random, and farmers self-select to adopt these CSA practices based on
Samnut 26 in Nigeria. These varieties are the most popular and available different characteristics. These could be both observed and unobserved
improved groundnut varieties in the study area, due to extensive characteristics such as skills, preferences, risks, and managerial abilities
dissemination campaigns conducted by the USAID-funded groundnut that could be correlated with yields and food security. Different methods
upscaling project and the Gates Foundation-funded Tropical Legumes I, such as weighting and matching estimators exist in reducing selection
II, and III projects. For cereal-groundnut intercropping, we captured bias but these estimators only control for observed self-selection.
adoption as a dummy variable that takes the value of one for adopters Switching regression models offer the promise of controlling both
and zero otherwise. Some of the cereals intercropped with groundnut observed and unobserved selection. Given our multiple treatments
are maize, millet, and sorghum. Cereal-groundnut intercropping which (three CSA practices and combinations of these), we follow Kassie et al.
involves different cereals could also be argued to proxy crop diversifi­ (2015) and Hörner and Wollni (2022) in using the multinomial endog­
cation which is relevant for food and nutrition security. Organic fertil­ enous switching regression model (MESR). MESR is a two-stage simul­
izers which refer to the use of different organic soil amendments like taneous estimation approach that combines a multinomial logit model
farmyard manure, animal manure, green manure, and composting are (MNL) in the first stage and an OLS model in the second stage. In the first
also recorded as a binary variable taking the value of one for households stage, factors associated with the use of these CSA practices are esti­
using these farm supplements and zero otherwise. mated with a MNL model. The second stage involves the estimation of
the relationship between the CSA practices and their combinations on
2.4. Descriptive review of data yields and food security. This step involves the addition of the inverse
mills ratio generated from the first stage equation which corrects for
The descriptive statistics of the respondents are summarized in selection bias, leading to the estimation of unbiased effects. Given that
Table 1. About three-quarters of surveyed households have adopted at these models have been extensively discussed in the empirical literature
least one CSA practice. Among the adopter groups, about one-fifth of (Kassie et al. 2015; Khonje et al. 2018; Hörner and Wollni 2022), we do
adopters have adopted the full CSA package3, which is the largest share. not repeat this here but more fully relate it to our objectives.
CSA packages that include organic fertilisers only on the one hand and a
combination of cereal-groundnut intercropping and improved 3.1. Multinomial logit model
groundnut varieties on the other hand, are the least adopted in the lo­
cations. The average household in the sample is about 50 years old with For the first stage of this estimation set-up, we begin by assuming a
slight heterogeneity between adopters of the CSA practices. Male random utility model where farmers adopt these CSA practices, or their
farmers are more likely than female farmers to adopt the CSA package. combination based on their perceived utility. We consider a latent model
While non-adopters of CSA practices are the least educated, adopters of (Q*ki ) to represent this unobserved utility differences between adopting
the full CSA package are more educated. This may indicate that edu­ and not adopting and represent this as:
cation level seems to play a key role in the optimal adoption of CSA
practices. The average household has about 12 household members, Q*ki = θi X ki + εki (1)
where on average 1.83 are dependent members. Adopters tend to have Where Q*ki is a latent variable representing utility differences in
larger plot sizes, participate in off-farm activities, and apply more adopting and not adopting CSA, Xki is a vector of explanatory variables
chemical fertilisers and pesticides on their plots than non-adopters. such as household and farm characteristics, network variables, access to
About 46% of the surveyed households are willing to purchase institutional services and wealth variables. Its parameter estimate is
improved seeds, which can be argued to be a pre-requisite for technol­ represented as θi . εki is the stochastic error term. Although Q*ki is not
ogy adoption. Moreover, 58% of the surveyed households receive advice
observable, the adoption decision of the farmer Qi can be represented as:
on agronomic practices from private extension services. Indeed, these ⎧ ( )
CSA practices are often promoted by NGOs and development projects in ⎪
⎪ 1 if Q*1i > max Q*mi or δ1i < 0

⎨ m∕
=1
Qi = ::: for all m ∕
=k (2)
⎪ ( )
3 ⎪ K if Q*Ki > max Q*mi or δKi < 0


We denote by full CSA package, the adoption of the combination of the m∕
=K
three CSA practices such as: climate-smart seed, cereal-groundnut intercropping
and organic fertilizer. Where δki = max(Q*mi − Q*ki ) < 0. Eq. (2) highlights that a
m∕
=k

4
M.P.Jr. Tabe-Ojong et al. Global Environmental Change 81 (2023) 102697

Table 1
Descriptive statistics of farm households in Ghana, Mali, and Nigeria.
Variables All Not adopt IC OF IS IC + OF IS + IC IS + OF IS + IC + OF P-value

Household characteristics
Age 49.56 50.87 53.05 46.83 50.69 47.04 53.35 48.29 47.38 <0.001
Sex (1 = male) 0.94 0.92 0.85 0.97 0.92 0.98 0.85 0.97 0.98 <0.001
Years of education 2.84 1.20 1.63 3.29 2.20 3.24 2.56 5.19 4.34 <0.001
Household size 11.76 13.85 16.51 8.25 12.33 9.94 13.59 8.24 10.12 <0.001
Dependency ratio 1.83 1.65 1.61 1.91 1.60 1.91 1.90 1.84 2.30 <0.001
Experience in agriculture (years) 19.66 20.40 20.14 18.40 21.51 21.86 18.55 17.98 17.64 <0.001
Contact with extension (1 = Yes) 0.75 0.71 0.78 0.59 0.92 0.68 0.97 0.72 0.79 <0.001
Credit access (1 = Yes) 0.04 0.03 0.02 0.03 0.08 0.04 0.04 0.09 0.04 <0.001
Cooperative membership (1 = Yes) 0.24 0.31 0.16 0.13 0.35 0.22 0.25 0.14 0.27 <0.001
Off-farm income (1 = Yes) 0.12 0.08 0.06 0.09 0.05 0.13 0.09 0.16 0.26 <0.001
Distance to urban market (km) 12.61 12.55 13.46 10.05 12.27 16.56 10.32 7.33 15.91 0.718
Distance to village market (km) 4.28 6.52 4.01 3.19 5.01 3.46 7.37 2.72 2.41 <0.001
Willingness to buy improved seeds 0.46 0.13 0.22 0.04 0.72 0.09 0.83 1.00 1.00 <0.001
Private extension (1 = Yes) 0.58 0.57 0.53 0.53 0.86 0.50 0.85 0.61 0.44 <0.001

Plot characteristics
Groundnut area (ha) 1.65 1.38 1.54 1.87 1.17 1.86 1.55 1.92 1.96 <0.001
Inherited land (1 = Yes) 0.75 0.77 0.78 0.73 0.72 0.76 0.74 0.75 0.76 0.688
Rented land (1 = Yes) 0.04 0.03 0.04 0.04 0.03 0.04 0.04 0.04 0.04 0.987
Applied chemical fertiliser (1 = Yes) 0.46 0.11 0.32 0.79 0.23 0.58 0.61 0.64 0.78 <0.001
Applied pesticide (1 = Yes) 0.29 0.01 0.05 0.32 0.10 0.31 0.18 0.73 0.68 <0.001
Total labor cost (USD/ha) 50.03 46.61 51.77 63.21 50.22 44.74 61.64 43.63 51.82 0.833
Clay soil (1 = Yes) 0.17 0.18 0.17 0.14 0.23 0.17 0.18 0.16 0.14 0.075
Sandy-clay soil (1 = Yes) 0.44 0.41 0.47 0.42 0.37 0.48 0.46 0.43 0.47 0.055
Silty soil (1 = Yes) 0.18 0.20 0.18 0.17 0.17 0.17 0.17 0.17 0.16 0.876

Outcome variables
Land productivity (kg/ha) 720.60 559.08 608.99 574.40 754.79 579.36 803.03 978.40 1012.08 <0.001
Food consumption score 62.95 56.06 57.02 61.61 67.14 62.36 65.01 69.71 71.24 <0.001
Acceptable food security status 0.43 0.23 0.22 0.43 0.55 0.47 0.48 0.63 0.66 <0.001
Borderline food security status 0.41 0.51 0.57 0.40 0.37 0.36 0.42 0.27 0.29 <0.001
Poor food security status 0.16 0.26 0.21 0.17 0.08 0.17 0.10 0.09 0.04 <0.001

Country
Ghana 0.17 0.29 0.36 0.02 0.44 0.01 0.23 0.02 0.00 <0.001
Mali 0.29 0.58 0.55 0.09 0.41 0.05 0.47 0.05 0.02 <0.001
Nigeria 0.53 0.14 0.09 0.89 0.15 0.93 0.30 0.93 0.98 <0.001
Observations 2868 669 386 276 283 339 117 373 425

Notes: P-values of continuous variables are obtained by one-way anova test for equality of mean values and P-values of binary variables are obtained by Pearson’s χ 2-
tests for equal proportions. Country-specific descriptives statistics are reported in Tables A1-A3 in the supplementary material.

Fig. 2. Relationship between yield and CSA practices. Notes: No CSA indicates no adoption of any CSA practices, IC stands for intercropping only, OF for organic
fertiliser only and IS for improved seeds only. Country-specific figures (A1-A3) are shown in the supplementary material.

representative farmer i would adopt a combination of CSAs k to maxi­ 3.2. Multinomial endogenous switching regression
mize his expected utility if these CSAs provide a higher expected utility
than the alternative combination m, m ∕ = k. We assume that εki is inde­ For the second stage, we model the relationship between the adop­
pendent and identically Gumbel distributed (Bourguignon et al. 2007). tion of CSA, yields and food security. Here, we estimate the relationship
As demonstrated by McFadden (1973), the probability that a farmer i between our outcomes of interest (Yki ) and a range of controls (Zki ) for
with observed characteristics Xki adopts combination k can be estimated different combinations of CSAs based on the first stage multinomial
with a multinomial logit model using maximum likelihood estimation. model. The base reference category which signifies non adoption is
represented as k = 1 and for the remaining combinations, at least one
CSA is adopted. Here, we estimate different equations for all the various

5
M.P.Jr. Tabe-Ojong et al. Global Environmental Change 81 (2023) 102697

Fig. 3. Relationship between FCS and CSA practices. Notes: No CSA indicates no adoption of any CSA practices, IC stands for intercropping only, OF for organic
fertiliser only and IS for improved seeds only. Country-specific (figures (A4-A6) are shown in the supplementary material.

categories of CSA adoption shown as: households will only adopt if they are not cash-strapped (Suri and Udry
⎧ 2022).
⎨ Regime 1 : Y1i = β1 Z1i + μ1i if Q = 1 Distance to the nearest village market measures the walking distance
::: k = 2, ..., K (3)
⎩ to markets where households buy the climate-smart seeds. Access to
Regime K : YKi = βK ZKi + μKi if Q = K
private extension services is measured as a dummy variable that takes
Where Yi represents the outcomes of interest for the farmer i, and μi the value of one for households that receive agronomic support from
denotes the stochastic error term. If the error terms in Eqs. (1) and (3) private extension services and zero otherwise. The use of this variable is
are uncorrelated, OLS estimations can be directly used to estimate Eq. based on the fact that farmers learn about the use of most of these CSA
(3), but it is hard to maintain this assumption given selection into practices from extension agents (Piñeiro et al. 2020). As explained, all
adoption. Thus, using OLS without correcting for this selection may bias the exclusion restrictions are potentially correlated with the use of the
the estimates. To obtain consistent and unbiased estimates of βk , we CSA practices. However, we are not sure whether and how these ERs
augment Eq. (3) by adding selection terms as shown below: could be associated with the outcomes except through using the CSA
⎧ practices. While the former ER condition, also known as the relevance
⎨ Regime 1 : Y1i = β1 Z1i + γ 1 λ1i + ∊1i if Q = 1
condition, can be empirically tested by checking the correlation of the
::: k = 2, ..., K (4)
⎩ ER with the CSA practices, the latter condition can only be theoretically
Regime K : YKi = βK ZKi + γ K λKi + ∊Ki if Q = K
motivated. However, we follow the procedure of Di Falco et al. (2011)
The error term in Eq. (4), ∊ki has an expected value of zero, γk is the and run a falsification test (Khonje et al. 2018; Tesfaye et al. 2021).
covariance between εki and μki . λk is the inverse mills ratio computed Regressing the ERs on the different combinations of CSA practices, we
from the estimated probabilities in the MNL set up. It is the bias-adjusted show that the ERs may be valid. This insight is further strengthened by
coefficient that corrects for selection bias (Bourguignon et al. 2007). Zi is the falsification tests as we do not find any association between some of
a vector of control variables similar to Xi that may be correlated with the ERs and the outcomes of non-CSA adopters (See Table A7 in sup­
yields and food security. These include both farm and household level plementary material). This makes us conclude that our ERs may be in
controls, access to institutional services, network variables as well as order. That notwithstanding, we refer to the estimated coefficients as
social and physical capital. associations but not causal effects.
The MESR model is identified by the non-linearity of the first stage
MNL model. However, it is standard and recommended to always add
exclusion restrictions. We use three exclusion restrictions (ERs) to 3.3. Actual and counterfactual analysis
identify the relationship between CSA, yields and food security. These
ERs are distance to the nearest village market, access to extension ser­ After estimating the relationship between CSA, yields and food se­
vices and willingness to purchase climate-smart seeds. The use of these curity using the switching regression framework, we compute the
instruments builds largely from the empirical literature that has shown average treatment effect on the treated and the average treatment effect
them to be valid instruments in different contexts (Di Falco et al. 2011; on the untreated. We refer to these as the actual-counterfactual analysis
Kassie et al. 2015; Khonje et al. 2018; Tesfaye et al. 2021). The will­ since it estimates the expected effects for households that adopt CSA as
ingness to purchase climate-smart seeds can be argued to envelop sub­ well as non-adopters. Four different estimations are carried out here: (1)
jective preferences for adopting these climate-smart seeds. This variable the effect of adoption on adopters as observed in the study area; (2) the
is potentially correlated with both observed and unobserved charac­ effect of adoption on non-adopters should they have adopted (counter­
teristics such as skills, preferences, and entrepreneurial skills. Previous factual); (3) the effect of adoption on adopters should they have not
studies have shown that using willingness to pay can help control for adopted (counterfactual); and the effect of adoption on non-adopters as
residual endogeneity (Bellemare and Novak 2017; Ruml and Qaim observed in the sample. These four estimations can be presented as
2021). The willingness to purchase climate-smart seeds is a dummy that conditional expected outcomes. The use of these outcomes makes it
takes the value of one for households that have access to and are willing possible to calculate the average adoption on yields and food security by
to purchase improved groundnut varieties. Accessibility in this case simply subtracting the counterfactual from the actual in their various
implies some aspects of awareness (knowledge exposure) about the categories.
climate-smart groundnut varieties. Of course, households will only
adopt if they know about these varieties, have access to them and are 4. Results and discussion
willing to purchase them. Information exposure usually matters and has
been shown to drive adoption, sometimes coupled with knowledge 4.1. Main baseline results
exposure (Kabunga et al. 2012). Beyond being aware and having access
to these climate-smart varieties, liquidity also matters and has been 4.1.1. Drivers of adoption of CSA
argued to be an important constraint in technology adoption, given that The results of the multinomial selection are summarised in Table 2.
Non-adopters of CSA practices are considered as the base category. We

6
M.P.Jr. Tabe-Ojong et al. Global Environmental Change 81 (2023) 102697

Table 2
Multinomial logit results of the drivers of CSA packages.
VARIABLES IC OF IS IC + OF IS + IC IS + OF IS + IC + OF

Age of household head (years) 0.02** (0.01) − 0.00 (0.01) − 0.01 (0.01) − 0.04** (0.02) 0.03** (0.01) − 0.01 (0.02) − 0.04*** (0.02)
Experience (years) − 0.01 (0.01) − 0.01 (0.01) 0.02 (0.02) 0.04*** (0.01) − 0.03 (0.02) 0.01 (0.02) 0.04* (0.02)
Sex of household head (1 = male) − 0.90*** 0.14 (0.53) − 0.24 (0.44) 0.94 (0.61) − 0.96** (0.47) 0.31 (0.52) 1.20** (0.59)
(0.32)
Education level (years) 0.08*** (0.03) 0.04 (0.04) 0.07** (0.03) 0.07** (0.03) 0.07* (0.03) 0.05 (0.04) 0.03 (0.04)
Household size (number) 0.03** (0.01) − 0.01 (0.03) 0.00 (0.02) 0.04 (0.02) 0.03* (0.02) − 0.04 (0.03) 0.02 (0.02)
Distance to nearest urban market (IHS, km) − 0.03 (0.22) − 0.21 (0.19) 0.07 (0.18) 0.20 (0.18) − 0.21 (0.32) − 0.33** (0.15) 0.32 (0.27)
Extension access (1 = Yes) 0.19 (0.50) − 0.06 (0.53) 0.04 (0.60) 0.91** (0.40) 1.95*** (0.50) 0.64 (0.39) 1.14** (0.45)
Cash credit (1 = Yes) 0.02 (0.51) 2.53*** (0.92) 1.47*** 2.18*** (0.77) 0.26 (0.65) 2.34*** (0.53) 1.50** (0.59)
(0.41)
Chemical fertilizer (1 = Yes) 1.35*** (0.46) 2.48*** (0.48) 0.86** (0.38) 1.50*** (0.36) 2.40*** (0.67) 1.59*** (0.49) 2.07*** (0.43)
Pesticide use (Yes = 1) 2.01*** (0.77) 2.66*** (0.92) 2.53*** 2.45** (1.00) 2.38** (0.95) 3.44*** (0.88) 3.21*** (0.93)
(0.84)
Cooperative membership (1 = Yes) − 0.99*** 0.03 (0.39) 0.59* (0.34) 0.34 (0.44) − 0.16 (0.50) 1.01*** (0.36) 1.07*** (0.40)
(0.38)
log of plot size 0.09 (0.20) 0.16 (0.27) 0.11 (0.26) 0.24 (0.20) 0.10 (0.25) 0.36 (0.26) 0.44* (0.26)
Off-farm income (1 = Yes) 0.13 (0.51) − 1.44*** − 0.08 (0.41) − 1.24** (0.50) − 0.28 (0.48) − 0.62 (0.45) − 0.54 (0.42)
(0.37)
Dependency ratio − 0.08 (0.07) 0.16 (0.10) − 0.01 (0.09) 0.12 (0.11) 0.04 (0.10) 0.03 (0.11) 0.08 (0.11)
Inherited land (1 = Yes) − 0.01 (0.18) − 0.21 (0.21) − 0.30 (0.24) 0.09 (0.19) − 0.23 (0.28) 0.43 (0.26) 0.44 (0.30)
Rented land (1 = Yes) 0.05 (0.48) − 0.27 (0.71) − 0.60 (0.54) 0.35 (0.60) − 0.82 (0.77) − 0.52 (0.78) − 0.31 (0.73)
Clay soil (Yes = 1) 0.19 (0.21) 0.04 (0.32) 0.44** (0.22) 0.51** (0.25) 0.22 (0.26) 0.44 (0.36) 0.46 (0.28)
Sandy-clay soil (1 = Yes) 0.27 (0.25) 0.04 (0.20) − 0.18 (0.20) 0.36** (0.18) 0.18 (0.25) − 0.05 (0.24) 0.11 (0.23)
Silty soil (1 = Yes) 0.03 (0.26) − 0.07 (0.30) − 0.09 (0.20) 0.14 (0.22) 0.03 (0.47) − 0.23 (0.36) − 0.24 (0.41)
Distance to nearest village market (IHS, km) − 0.49*** − 0.35* (0.19) − 0.32* (0.19) − 0.07 (0.24) − 0.16 (0.21) − 0.13 (0.24) − 0.70** (0.31)
(0.11)
Willingness to purchase improved seeds 0.28 (0.60) − 1.99** (0.83) 2.75*** − 0.79 (0.58) 3.18*** (0.45) 7.35*** (1.11) 7.71*** (1.28)
(Yes = 1) (0.46)
Private extension (1 = Yes) − 0.21 (0.39) 1.97*** (0.52) 2.34*** 1.16** (0.54) 1.85*** (0.59) 2.42*** (0.48) 1.72*** (0.62)
(0.42)
Mali − 1.30** (0.58) − 0.65 (1.30) − 1.35** − 0.50 (0.89) − 1.39 (0.93) − 1.21 (1.18) 11.94*** (1.07)
(0.68)
Nigeria − 1.93*** 3.99*** (1.22) 0.21 (0.76) 5.26*** (0.99) 0.86 (0.77) 4.25*** (0.99) 18.56*** (0.85)
(0.74)
Constant 0.65 (0.80) − 3.53** (1.45) − 2.75** − 6.91*** − 6.41*** − 10.32*** − 26.49***
(1.35) (1.74) (1.37) (1.65) (1.90)
Observations 2,804 2,804 2,804 2,804 2,804 2,804 2,804

Note: Robust clustered standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. Country-specific multinomial logit results are reported in Tables A3-A6 in
the supplementary material.

performed three validation analyses. First, the result of the likelihood likely to adopt the other CSA practices to boost their farm productivity.
ratio test (χ2 = 5486.83 and p > χ2 = 0.000) indicates that the co­ We also find that producer organisations such as cooperatives matter in
efficients are different from zero. Second, smhsiao’s test results of the the adoption of different practices such as improved seeds and the
independence of irrelevant alternatives indicate that the CSA practices combination between improved seeds and organic fertilisers. Previous
are significantly different from each other at the 1% significance level. analysis from Cameroon also supports this claim that producer organi­
Finally, we use likelihood ratio tests to analyze whether some of the CSA sations are associated with the use of CSA practices such as organic
can be combined. The results reject the null hypothesis that all co­ fertilisers (Tabe-Ojong 2022).
efficients, except interactions, associated with a given pair of alterna­ We observe that distance to the nearest village market is negatively
tives are zero. Altogether, we conclude that the multinomial logit model associated with the adoption of some CSA packages, since rural house­
is appropriate for the data. holds usually buy improved inputs from village markets. It may thus be
The results also show some significant heterogeneity. Beginning with difficult for households to travel to distant markets to purchase
age, we show the existence of heterogeneity in the age of the adopters, improved inputs. Village market accessibility could also denote neigh­
adopters of the full CSA package are more likely to be younger and more bourhood effects in technology adoption, since larger villages with
experienced in groundnut production than non-adopters. Adopters of numerous producers usually have markets closer to inhabitant dwellings
the single or combined CSA practices are more likely to have higher at the center of the village, where farmers usually sell their products.
years of education than non-adopters. Contact with extension service is Applying chemical fertilizers and pesticides is positively associated with
positively associated with the adoption of most of the combined CSA the adoption of all CSA packages. This could be because of the seeming
packages, which implies knowledge sharing about improved agronomic complementarity between the use of these farm inputs (Yang et al.
practices by extension agents. More importantly, receiving extension 2015). Except in two treatments, households that have access to credit
advice from private extension services largely increases the probability are more likely to adopt most packages than non-adopters.
of adopting most packages. Indeed, in developing countries, numerous
projects and NGOs support farmers towards the diffusion and adoption 4.1.2. Average treatment effects of CSA adoption
of improved agronomic practices, and it goes without saying that access Results of the average treatment effect of the adoption of the CSA
to private extension services could increase the likelihood of adopting packages and their combinations are summarized in Table 3. The esti­
improved farming practices (Piñeiro et al. 2020). mates are compared to the base scenario of non-adoption. Results
Households that are willing to purchase improved seeds are more indicate that the adoption of both individual and combined CSA prac­
likely to adopt most of the CSA practices and their combinations. tices is significantly associated with groundnut yields. Adopting cereal-
Consistently, households that have access to improved seeds are more groundnut intercropping and organic fertilizers in isolation appears to

7
M.P.Jr. Tabe-Ojong et al. Global Environmental Change 81 (2023) 102697

Table 3 improved seeds, adopted in isolation are associated with a 1.36 and 9.83
Average treatment effects of the adoption of CSA practices and their increase in food consumption scores respectively. Once again, this
combination. substantial positive association may be due to the adoption of improved
Outcomes Technology choice Adoption status seeds. Interestingly, combining improved seed with intercropping (5.98)
variables
Non- Adoption Average
and organic fertilisers (3.55) is also positively associated with food
adoption treatment consumption scores. Finally, combining all practices, that is improved
effects seed, intercropping and organic fertilisers, we again observe a positive
Land Intercropping (IC) 461.47 525.46 63.99*** correlation with food consumption score. Our results suggest that food
productivity (3.86) (4.83) (6.18) security can be substantially improved in developing countries by
(kg/ha) Organic fertilisers 431.04 492.80 61.76*** adopting yield increasing CSA practices in combination.
(OF) (3.08) (5.45) (6.26) Our results here are in line with previous studies such as Lopez-
Improved seeds (IS) 455.85 632.31 176.46***
(5.60) (6.97) (8.95)
Ridaura et al. (2018) and Bazzana et al. (2022) who used simulation and
Intercropping + 443.43 486.59 43.16*** agent based models to find such food security implications. It also cor­
Organic fertilisers (3.42) (4.81) (5.90) roborates studies that used regression models to evaluate this relation­
Improved seeds + 427.51 725.75 298.24*** ship between CSA, yields and food security (Di Falco et al. 2011; Wekesa
Intercropping (9.68) (22.09) (24.12)
et al. 2018; Hasan et al. 2018; Mishra et al. 2021; Tabe-Ojong et al.,
Improved Seeds + 359.96 827.35 467.39***
Organic fertilisers (5.42) (8.90) (9.66) 2023a; Tabe-Ojong et al., 2023b). It has also been established that some
Improved seeds + 479.32 872.15 392.84*** of these practices build resilience to climate change and help mitigate
intercropping + (7.05) (6.57) (9.64) yield losses arising from extreme weather events (Di Falco and Chavas
Organic fertilisers 2008; Michler et al. 2019). Overall, we conclude that CSA is food and
nutrition-sensitive as it is associated with increases in agricultural pro­
Food Intercropping (IC) 55.18 56.54 1.36*** duction and food security (Lipper et al. 2014; Partey et al. 2018; Ntinyari
consumption (0.14) (0.20) (0.24)
and Gweyi-Onyango 2021).
score Organic fertiliser 61.83 60.86 − 0.97**
(OF) (0.32) (0.24) (0.40
Improved seeds (IS) 56.21 66.04 9.83*** 4.2. Heterogeneity analysis of CSA adoption
(0.25) (0.42) (0.49)
Intercropping + 59.44 61.56 2.11***
We assess whether the effects of the CSA practices are driven by the
Organic fertilisers (0.21) (0.17) (0.27)
Improved seeds + 57.79 63.78 5.98*** number of observations in Nigeria in Tables 4 and 5. Results show
Intercropping (0.46) (0.70) (0.83) heterogeneous associations of CSA practices, land productivity and food
Improved seeds + 64.81 68.35 3.55*** consumption score across the various countries. Despite the existing
Organic fertilisers (0.22) (0.22) (0.31) differences in the magnitude of the estimates, we only obtain significant
Improved seeds + 62.25 69.79 7.54***
intercropping + (0.19) (0.19) (0.27)
positive associations of intercropping and land productivity in Ghana
Organic fertilisers (48 kg/ha) and Mali (80 kg/ha), but not in Nigeria. The adoption of
organic fertilisers is not significantly associated with land productivity
Clustered robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p <
in Ghana. Among the CSA practices used in isolation, the highest land
0.1.
productivity is observed under the use of improved seeds, but the
magnitude of association is substantially larger in Nigeria, compared to
have smaller effects on land productivity, with yield gains of around 64
Ghana and Mali. Going further, it is intuitive to assume that the positive
kg/ha and 62 kg/ha of groundnuts respectively. When combining the
association of the joint use of intercropping and organic fertilizers
two practices, the effects decline to 43 kg/ha. This could be because both
mainly emerges from Nigeria, as these associations are statistically
intercropping and organic fertilisers offer similar benefits to yield gains.
insignificant in Mali and Ghana. When improved seeds are used with
While intercropping improves soil fertility through the synthesis of at­
either organic fertilisers or intercropping or with both, the magnitude of
mospheric nitrogen, organic fertilisers also improve soil fertility. Thus,
association of the combined practices with land productivity in Nigeria
the joint use of both may indicate some form of overuse which may
becomes more than two times larger than the associations obtained in
rather reduce yield gains. A direct implication of this result would be
each of the two other countries. Similarly, most of the estimates related
that all combinations of CSA are not necessarily optimal. We obtain
to food consumption are systematically different between the countries,
higher yields associated with the adoption of improved seed, with about
with the most being driven by Nigeria. For instance, negative associa­
176 kg/ha of groundnut. Combining intercropping or organic fertilisers
tions between the use of organic fertilisers and food consumption are
with improved seeds, we obtain larger positive associations with yields.
solely obtained in Nigeria. Except for intercropping and the full CSA
Specifically, yields increase largely by 298 and 467 kg/ha when
package, the adoption of the other practices is associated with smaller
improved seed is combined with intercropping and organic fertilisers. A
estimates in Nigeria. Despite this decrease in the food consumption score
combination of the three individual CSA practices also leads to high
in Nigeria, large significant estimates of about 7.60 are obtained when
yields, that is, it increases groundnut yield by 393 kg/ha. However, one
the three practices are combined, the association being insignificant in
might expect to achieve the highest productivity with the adoption of
Mali.
the three CSA practices at once, which is not the case compared to the
use of improved seeds and organic fertilisers. This slight difference could
4.3. Further analysis and robustness check
be due to the complementary benefits of using both intercropping and
organic fertilisers which offer similar benefits.
We also perform some robustness checks to give credence to the
As expected, the adoption of CSA practices exhibits a positive asso­
estimates and findings obtained so far. In the first place, we decompose
ciation with food consumption scores. This is specifically the case for the
the food consumption score into acceptable, borderline, and poor levels
adoption of improved climate-smart varieties, cereal-groundnut inter­
cropping and other CSA combinations. These results suggest that
adopting CSA practices may be important in building the resilience of
poor agricultural households and thus sustaining food security.
Whereas, adopting solely organic fertilisers is associated with decreasing
food consumption scores, other practices such as intercropping and

8
M.P.Jr. Tabe-Ojong et al. Global Environmental Change 81 (2023) 102697

Table 4
Heterogeneous effects across countries.
CSA combination Land productivity (kg/ha) Food consumption score

Ghana Mali Nigeria Ghana Mali Nigeria

IC 47.62*** (10.34) 79.97*** (6.55) 29.74 (22.43) 0.44* (0.22) 1.18*** (0.21) 6.33*** (0.64)
OF 94.07 (64.04) 131.33*** (35.51) 54.26*** (5.69) 2.51* (1.30) 1.89 (1.26) − 1.33*** (0.38)
IS 114.38*** (9.93) 171.09*** (3.26) 367.66*** (24.78) 7.35*** (0.31) 15.30***(0.85) 3.03*** (1.10)
IC + OF 25.24 (104.77) − 7.36 (40.59) 46.40*** (5.64) − 0.72 (1.43) 2.67*** (0.92) 2.13**(0.24)
IS + IC 181.24*** (33.50) 229.46*** (26.92) 489.30*** (41.32) 6.02*** (1.30) 9.27*** (1.24) 0.98 (1.17)
IS + OF 183.50*** (43.66) 107.16** (39.78) 493.13*** (9.00) 10.03*** (1.27) 5.40*** (1.38) 3.31*** (0.28)
IS + IC + OF 305.94** (120.56) 394.12*** (9.48) 3.43 (3.85) 7.60*** (0.26)

Clustered robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. Blank case indicates that no observation has adopted the combined package that
involves the three practices in Ghana.

practices, strengthening the insight that the bundled use of these CSA
Table 5 practices matters for food security.
Heterogeneous effects excluding observations from Nigeria.
CSA Land productivity (kg/ha) Food consumption score 4.4. Discussion
combination
Excluding Only Nigeria Excluding Only Nigeria
Nigeria Nigeria Our study on the relationship between CSA, land productivity and
IC 67.25*** 29.74 (22.43) 0.89*** 6.33*** food security offer a couple of new insights that support different strands
(6.40) (0.17) (0.64) in the empirical literature on CSA, climate change adaptation, climate
OF 124.91*** 54.26*** 2.00* (1.07) − 1.33*** resilience and food security in Africa. In the first place, insights from our
(31.13) (5.69) (0.38)
analysis corroborates and bolsters early evidence on the various ways of
IS 141.03*** 367.66*** 11.09*** 3.03***
(8.63) (24.78) (0.48) (1.10) adapting to climate change in smallholder production settings (Lipper
IC + OF − 0.28 46.40*** 1.93** (0.84) 2.13** et al. 2014; Partey et al. 2018; Ntinyari and Gweyi-Onyango 2021).
(38.07) (5.64) (0.24) Here, we add to empirical studies that have examined the implications of
IS + IC 213.59*** 489.30*** 8.20*** 0.98 (1.17)
CSA on yields and food security both from an ex-ante and ex-post
(22.70) (41.32) (0.94)
IS + OF 127.72*** 493.13*** 6.65*** 3.31***
perspective. While some studies such as Lopez-Ridaura et al. (2018)
(33.41) (9.00) (1.16) (0.28) and Bazzana et al. (2022) have used simulation and agent based models
IS + IC + OF 305.94** 394.12*** 3.43 (3.85) 7.60*** to examine these relationships, others have used regression models to
(120.57) (9.48) (0.26) show the positive relationship between CSA, yields and food security (Di
Clustered robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < Falco et al. 2011; Wekesa et al. 2018; Hasan et al. 2018; Mishra et al.
0.1. 2021). It has also been established that some of these practices build
resilience to climate change and help mitigate yield losses arising from
of food security and estimate a Multinomial treatment effect4 model extreme weather events (Di Falco and Chavas 2008; Michler et al. 2019).
proposed by Deb & Trivedi (2006) as a robustness check to the MESR We provide additional insights from unexplored productions systems
model. in West Africa. More importantly, we consider legume-based
Fig. 4 shows the distribution of the effects of the CSA packages on the (groundnut) production systems. Legumes such as groundnuts offer a
food consumption score and the binary variable that denotes acceptable couple of social, environmental and food implications for rural house­
food consumption level. Results confirm that CSA practices such as holds in many smallholder settings. From a social perspective, legumes
intercropping, and the use of climate-smart seeds have the potential to offer immense opportunities for women who have usually been sidelined
sustain food security. The effects are more pronounced when improved in the control of farm resources and income. Legumes have been high­
seed is used together with either intercropping or both intercropping lighted to be a “women’s crop” with significant potentials of boosting
and organic fertiliser. We note some sensitivity regarding the magnitude smallholder commercialization by improving women’s access to re­
of the estimates, but more importantly, the prior negative association sources (Njuki et al. 2011; Orr et al. 2016). It could also influence intra-
between the adoption of organic fertiliser and food consumption score household dynamics with some implications for female autonomy and
becomes statistically insignificant. Fig. 5 further supports the notion that empowerment. From an environmental perspective, groundnuts have
the use of improved seeds offers the biggest gains to productivity and been described as a climate-resilient crop with immense potentials to
food security. alleviate food insecurity in Africa (Soumare et al. 2022). Being a legume,
The robustness checks are largely in line with the main findings that it could also help in the synthesis of atmospheric nitrogen which may be
the use of CSA practices matters in increasing yields with ensuing im­ relevant for soil fertility when intercropped with other cereals as is
pacts on food security. The partial use as well as the combined use of currently the case in the study area. This inherent characteristic of le­
these practices matter in increasing the food consumption levels of gumes could lead to the reduction of the use of chemical fertilisers with
households. The use of improved climate-smart varieties offers the some reductions in GHG emissions. Finally, groundnut is a staple for
biggest productivity and food security gains as an individual practice, many households in rural production systems throughout sub-Saharan
but it is also immensely relevant when in combination with other Africa (Tabe-Ojong et al. 2021). It is one crop whose food group clas­
sification overlaps as it is categorized both as proteins as well as fats and
oils (Tabe-Ojong et al., 2022). Given all these positive characteristics
4 and attributes, improved groundnut varieties have been described as
The multinomial treatment effect model METE, in contrast to multinomial
pro-poor and environmentally friendly (Tabe-Ojong et al., 2022).
endogenous treatment effects, that model continuous variables, accommodate
both quantitative and qualitative outcomes variables, and used a mixed Our analysis also provides evidence that the bundled use of these
multinomial logit in the first stage, and OLS in the second stage in a simulta­ CSA practices may offer more land productivity and food security gains
neous procedure. We have applied METE here as our variable denoting the than the piecemeal (independent) use of the practices. This is not to
acceptable food consumption level is binary, which we were unable to estimate place a low eye on the independent use of these practices as some of
with our prior MESR. them like the use of climate-smart varieties are high yielding with strong

9
M.P.Jr. Tabe-Ojong et al. Global Environmental Change 81 (2023) 102697

Fig. 4. METE analysis of the effect of CSA practices on food consumption score.

Fig. 5. METE analysis of the effect of CSA practices on food consumption score.

implications for food security (Simtowe et al. 2019; Michler et al. 2019). selection bias. We also examine some of the household, farm and insti­
The adoption of independent CSA practices such as organic fertilisers is tutional characteristics associated with the use of CSA. We show that the
associated with farm performance and food security (Tabe-Ojong et al., adoption of CSA is associated with yield increases which translates to
2023a; Tabe-Ojong et al., 2023b). Our analysis supports these different increased household consumption. Specifically, farmers who adopt
studies but also recommends bundling. The use of practices such as climate-smart varieties, cereal-groundnut intercropping, and organic
cereal-groundnut intercropping may also be additionally suggestive of fertilisers report higher yields and consumed more nutritionally
efforts towards crop diversification which is relevant for food and adequate foods than their non-adopting counterparts. Despite the posi­
nutrition security (Madsen et al. 2021: Molua, 2022). tive association between the use of these individual practices and yield
Furthermore, this analysis provides some insights on policy relevant and food security, we observe that greater yield and food security gains
questions regarding the use of CSA. Development policy is always are observed under the bundled use of these practices. We also observed
interested in understanding the implications of many novel in­ that most of the yield and food security gains arise from the use of
terventions that are being rolled out for different development and improved groundnut varieties. This is not surprising given that the use of
environmental outcomes. Most of these interventions are always esti­ improved seeds has been argued to be crucial in increasing productivity
mated such that we have restricted evidence on just those who are and maintaining environmental sustainability. This is even the case for
treated, in our case those who are using CSA practices (actual). What is legumes that are increasingly referred to as pro-poor and environmen­
the case for those who are not treated (counterfactual)? Do we expect tally friendly.
similar outcomes as in the case of the treated? The use of the MESR Our findings have important implications for some of the sustainable
enables us to estimate slope coefficients and highlight actual- development targets like reducing hunger and ensuring healthy diets for
counterfactual relationships. This is an important contribution in the all. It also has implications for meeting the twin goals of increasing
face of many policy interventions on various ways of pushing adoption agricultural productivity while maintaining environmental sustainabil­
rates and availing these CSA practices to farmers. ity. As shown in this analysis, CSA practices have the potential to in­
crease productivity with ensuing implications on food security. But
5. Conclusion and policy implications given that these practices are climate and environmentally friendly, they
may also have the potential to reduce the emissions of GHGs. This is
This study examines the relationship between climate-smart agri­ necessary to not jeopardise the needs of future generations.
cultural practices, yield and food security in Ghana, Mali, and Nigeria. Given these findings, our analysis supports the upscaling of CSA
We use a farm household survey in these countries and estimate practices as they have the potential to trigger sustainable agricultural
regression models that account for multiple treatments as well as production with immense concern for the environment. Some of these

10
M.P.Jr. Tabe-Ojong et al. Global Environmental Change 81 (2023) 102697

practices like the use of climate-smart varieties are capital intensive, project “Increasing Groundnut Productivity of Smallholder farmers in
which may explain their limited adoption for most households that are Ghana, Mali, and Nigeria” which was implemented by ICRISAT. This
usually cash-strapped. Subsidizing these seeds would go a long way to paper has benefitted from comments after presentations at the Norwe­
increase productivity which may have greater welfare implications on gian Council for Africa in Norway and the 2022 edition of the African
smallholder livelihoods. Beyond being capital intensive, some of these Economic Conference in Mauritius. We are grateful to Marilou Goussard
practices are also knowledge and sometimes labour sensitive. As our Vincent for her research assistance. The usual disclaimer applies.
results highlight, access to information especially on the use of these
sustainable practices is important for adoption. In this regard, Appendix A. Supplementary data
strengthening agricultural extension systems as well as the provision of
technical and knowledge-based information would greatly enhance the Supplementary data to this article can be found online at https://doi.
sustainability of smallholder production systems. Strengthening and org/10.1016/j.gloenvcha.2023.102697.
empowering extension systems with the right framework and effective
follow-up are steps in the right direction of maintaining environmental References
sustainability while increasing agricultural productivity. This could take
the form of solidifying public-private partnerships in extension service Barrios, S., Ouattara, B., Strobl, E., 2008. The impact of climatic change on agricultural
production: Is it different for Africa? Food Policy 33 (4), 287–298. https://doi.org/
delivery as well as increasing geographical reach and coverage. 10.1016/j.foodpol.2008.01.003.
We end by mentioning some of the limitations of the study that may Bazzana, D., Foltz, J., Zhang, Y., 2022. Impact of climate smart agriculture on food
be interesting in better understanding the analysis but more importantly security: An agent-based analysis. Food Policy 111, 102304. https://doi.org/
10.1016/j.foodpol.2022.102304.
for future studies that may want to build on some of these limitations. Bellemare, M.F., Novak, L., 2017. Contract Farming and Food Security. American Journal
We refer to all the estimates as associations as we have used only a single of Agricultural Economics 99 (2), 357–378. https://doi.org/10.1093/ajae/aaw053.
period data set. Adoption is dynamic and it would have been important Bennetzen, E.H., Smith, P., Porter, J.R., 2016. Agricultural production and greenhouse
gas emissions from world regions—The major trends over 40 years. Global
to model these dynamics using panel data. Panel data would have Environmental Change 37, 43–55. https://doi.org/10.1016/j.gloenvcha.2015.12.004.
additionally enabled us to control for unobserved heterogeneity more Bourguignon, F., Fournier, M., Gurgand, M., 2007. Selection bais corrections based on
rigorously and move towards causality. Future studies may want to use the multinomial logit model: Monte Carlo comparisons. J Economic Surveys 21 (1),
174–205. https://doi.org/10.1111/j.1467-6419.2007.00503.x.
longitudinal data to answer these sustainability questions and provide
Cohn, A.S., Newton, P., Gil, J.D.B., Kuhl, L., Samberg, L., Ricciardi, V., Manly, J.R.,
more policy relevant causal insights. Secondly, CSA as an approach in­ Northrop, S., 2017. Smallholder Agriculture and Climate Change. Annu. Rev. Environ.
volves many more practices than we have covered in this analysis. While Resour. 42 (1), 347–375.
we have shown that the use of climate-smart varieties, intercropping Di Falco, S., Chavas, J.-P., 2008. Rainfall Shocks, Resilience, and the Effects of Crop
Biodiversity on Agroecosystem Productivity. Land Economics 84 (1), 83–96. https://
groundnuts with cereals and the use of organic fertilisers matter, other doi.org/10.3368/le.84.1.83.
practices do exist, but we do not consider them in the analysis due to Di Falco, S., Veronesi, M., Yesuf, M., 2011. Does Adaptation to Climate Change Provide
data limitations. It may be in the best interest of future studies to take Food Security? A Micro-Perspective from Ethiopia. American Journal of Agricultural
Economics 93 (3), 829–846.
this direction and further build on the external validity of this analysis. FAO (1996): Food and Agriculture Organization. Rome Declaration on World Food
Finally, we caution against an absolute generalisation of our analysis. Security and World Food Summit Plan of Action. Rome: FAO. 1996.
We provide evidence from three West African countries which supports Faostat, 2022. Crops and livestock products. Food and Agriculture Organization of the
United Nations, Rome, Italy. Available online at https://www.fao.org/faostat/en/
the external validity of our analysis, but it may not be ideal to generalise #data/QCL/visualize.
since context matters and production systems are different. Moreover, Fujimori, S., Hasegawa, T., Krey, V., Riahi, K., Bertram, C., Bodirsky, B.L., Bosetti, V.,
CSA is context specific and varies from place to place as well as with the Callen, J., Després, J., Doelman, J., Drouet, L., Emmerling, J., Frank, S., Fricko, O.,
Havlik, P., Humpenöder, F., Koopman, J.F.L., van Meijl, H., Ochi, Y., Popp, A.,
local environment. In this regard, it would be relevant to conduct other Schmitz, A., Takahashi, K., van Vuuren, D., 2019. A multi-model assessment of food
studies focusing on different production systems and environments in security implications of climate change mitigation. In Nat Sustain 2 (5), 386–396.
order to provide more robust insights. Hasan, M.K., Desiere, S., D’Haese, M., Kumar, L., 2018. Impact of climate-smart
agriculture adoption on the food security of coastal farmers in Bangladesh. In Food
Sec. 10 (4), 1073–1088.
CRediT authorship contribution statement Hertel, T.W., 2016. Food security under climate change. In Nature Clim Change 6 (1),
10–13. https://doi.org/10.1038/nclimate2834.
Martin Paul Jr. Tabe-Ojong: Conceptualization, Methodology, Hörner, D., Wollni, M., 2022. Does integrated soil fertility management increase returns
to land and labor? In Agricultural Economics 53 (3), 337–355. https://doi.org/
Software, Formal analysis, Investigation, Writing – original draft, 10.1111/agec.12699.
Writing – review & editing. Ghislain B.D. Aihounton: Methodology, IPCC (2019): Climate Change and Land, an IPCC Special Report on Climate Change,
Software, Formal analysis, Investigation, Writing – original draft, Desertification, Land Degradation, Sustainable Land Management, Food Security,
and Greenhouse Gas Fluxes in Terrestrial Ecosystems.
Writing – review & editing. Jourdain C. Lokossou: Data curation, Kabunga, N., Ssentamu; Dubois, Thomas; Qaim, Matin, 2012. Heterogeneous
Writing – review & editing. information exposure and technology adoption: the case of tissue culture bananas in
Kenya. In Agricultural Economics 43 (5), 473–486. https://doi.org/10.1111/j.1574-
0862.2012.00597.x.
Declaration of Competing Interest Kassie, M., Teklewold, H., Marenya, P., Jaleta, M., Erenstein, O., 2015. Production Risks
and Food Security under Alternative Technology Choices in Malawi: Application of a
The authors declare that they have no known competing financial Multinomial Endogenous Switching Regression. In J Agricultural Economics 66 (3),
640–659. https://doi.org/10.1111/1477-9552.12099.
interests or personal relationships that could have appeared to influence Khonje, M.G., Manda, J., Mkandawire, P., Tufa, A., Hirpa; Alene, Arega D, 2018.
the work reported in this paper. Adoption and welfare impacts of multiple agricultural technologies: evidence from
eastern Zambia. In Agricultural Economics 49 (5), 599–609. https://doi.org/10.1111/
agec.12445.
Data availability Lawlis, T., Islam, W., Upton, P. (2017). Achieving the four dimensions of food security
for resettled refugees in Australia: A systematic review. https://doi.org/10.1111/
Data will be made available on request. 1747-0080.12402.
Lipper, L., Thornton, P., Campbell, B.M., Baedeker, T., Braimoh, A., Bwalya, M.,
Caron, P., Cattaneo, A., Garrity, D., Henry, K., Hottle, R., Jackson, L., Jarvis, A.,
Funding Kossam, F., Mann, W., McCarthy, N., Meybeck, A., Neufeldt, H., Remington, T.,
Sen, P.T., Sessa, R., Shula, R., Tibu, A., Torquebiau, E.F., 2014. Climate-smart
We thank the United States Agency for International Development agriculture for food security. In Nature Clim Change 4 (12), 1068–1072.
Lobell, David B.; Schlenker, Wolfram; Costa-Roberts, Justin (2011): Climate trends and
(USAID) and the International Crops Research Institute for the Semi- global crop production since 1980. In Science (New York, N.Y.) 333 (6042),
Arid Tropics (ICRISAT) for survey funding and implementation of the pp. 616–620. DOI: 10.1126/science.1204531.
data respectively. The data was collected as part of the USAID-funded

11
M.P.Jr. Tabe-Ojong et al. Global Environmental Change 81 (2023) 102697

Lopez-Ridaura, S., Frelat, R., van Wijk, M.T., Valbuena, D., Krupnik, T.J., Jat, M.L., 2018. Yang, H., Hao, Y., Li, W., Fan, M., 2022. Soil quality both increases crop production
Climate smart agriculture, farm household typologies and food security: An ex-ante and improves resilience to climate change. In Nat. Clim. Chang. 12 (6), 574–580.
assessment from Eastern India. In Agricultural Systems 159, 57–68. https://doi.org/ Ruml, A., Qaim, M., 2021. New evidence regarding the effects of contract farming on
10.1016/j.agsy.2017.09.007. agricultural labor use. In Agricultural Economics 52 (1), 51–66. https://doi.org/
Madsen, S., Bezner Kerr, R., Shumba, L., Dakishoni, L., 2021. Agroecological practices of 10.1111/agec.12606.
legume residue management and crop diversification for improved smallholder food Simtowe, F., Amondo, E., Marenya, P., Rahut, D., Sonder, K., Erenstein, O., 2019.
security, dietary diversity and sustainable land use in Malawi. In Agroecology and Impacts of drought-tolerant maize varieties on productivity, risk, and resource use:
Sustainable Food Systems 45 (2), 197–224. https://doi.org/10.1080/ Evidence from Uganda. In Land use policy 88, 104091. https://doi.org/10.1016/j.
21683565.2020.1811828. landusepol.2019.104091.
Maxwell, D., Coates, J., Vaitla, B., 2013. How Do Different Indicators of Household Food Soumare, A., Diedhiou, A.G., Kane, A., 2022. Bambara groundnut: a neglected and
Security Compare? Empirical Evidence from Tigray. Feinstein International Center, underutilized climate-resilient crop with great potential to alleviate food insecurity
Tufts University, Medford, USA https://fic.tufts.edu/assets/Different-Indicators-of- in sub-Saharan Africa. In Journal of Crop Improvement 36 (5), 747–767. https://doi.
HFS.pdf. org/10.1080/15427528.2021.2000908.
McFadden, Daniel (Ed.) (1973): Conditional logit analysis of qualitative choice behavior. Suri, T., Udry, C., 2022. Agricultural Technology in Africa. In Journal of Economic
Zarembka, P. (ed.), Frontiers in Econometrics. New York, USA: Academic Press. Perspectives 36 (1), 33–56. https://doi.org/10.1257/jep.36.1.33.
Michler, J.D., Baylis, K., Arends-Kuenning, M., Mazvimavi, K., 2019. Conservation Tabe-Ojong, M.P., 2022. Do producer organisations promote environmental
agriculture and climate resilience. In Journal of environmental economics and sustainability through organic soil investments? Evidence from Cameroon. In.
management 93, 148–169. https://doi.org/10.1016/j.jeem.2018.11.008. Journal of Development Effectiveness 1–23. https://doi.org/10.1080/
Mishra, A., Ketelaar, J.W., Uphoff, N., Whitten, M., 2021. Food security and climate- 19439342.2022.2130961.
smart agriculture in the lower Mekong basin of Southeast Asia: evaluating impacts of Tabe-Ojong, M.P.J., Fabinin, A.N., Minkoua Nzié, J.R., Molua, E.L., Fonkeng, E.E.,
system of rice intensification with special reference to rainfed agriculture. In 2023a. Organic soil amendments and food security: Evidence from Cameroon. Land
International Journal of Agricultural Sustainability 19 (2), 152–174. Degrad Dev 34 (4), 1159–1170.
Molua, E.L., 2022. Private farmland autonomous adaptation to climate variability and Tabe-Ojong, M.P.J., Ibarra, L.M., Andrade, R.S., Labarta, R., 2023b. Soil conservation
change in Cameroon. Rural Society 31 (2), 115–135. https://doi.org/10.1080/ and smallholder welfare under cassava-based systems in Thailand. Land Degrad Dev
10371656.2022.2086223. 34 (6), 1795–1805.
Njuki, J., Kaaria, S., Chamunorwa, A., Chiuri, W., 2011. Linking Smallholder Farmers to Tabe-Ojong, M.P.J., Molua, E.L., Ngoh, S.B., Beteck, S.E., 2021. Production, consumption
Markets, Gender and Intra-Household Dynamics: Does the Choice of Commodity and market diversification of grain legumes in the humid forest agroecology of
Matter? In Eur J Dev Res 23 (3), 426–443. https://doi.org/10.1057/ejdr.2011.8. cameroon. In Sustainable Production and Consumption 27, 193–202.
Ntinyari, W., Gweyi-Onyango, J.P., 2021. Greenhouse Gases Emissions in Agricultural Tabe-Ojong, M.P., Smale, M., Jamora, N., Azevedo, V., 2022. Genebanks and market
Systems and Climate Change Effects in Sub- Saharan Africa. In: Oguge, N., Ayal, D., participation: evidence from groundnut farmers in Malawi. CABI Agric Biosci 3 (1).
Adeleke, L., Da Silva, I. (Eds.), African Handbook of Climate Change Adaptation. Tesfaye, W., Blalock, G., Tirivayi, N., 2021. Climate-Smart Innovations and Rural Poverty
Springer International Publishing, Cham, pp. 1081–1105. in Ethiopia: Exploring Impacts and Pathways. In American Journal of Agricultural
Orr, A., Tsusaka, T., Kee-Tui, S., Homann; Msere, Harry, 2016. What Do We Mean by Economics 103 (3), 878–899. https://doi.org/10.1111/ajae.12161.
‘Women’s Crops’? Commercialisation, Gender and the Power to Name. In J. Int. Dev. Wekesa, B.M., Ayuya, O.I., Lagat, J.K., 2018. Effect of climate-smart agricultural
28 (6), 919–937. https://doi.org/10.1002/jid.3224. practices on household food security in smallholder production systems: micro-level
Ortiz-Bobea, A., Ault, T.R., Carrillo, C.M., Chambers, R.G., Lobell, D.B., 2021. evidence from Kenya. Agric & Food Secur 7 (1).
Anthropogenic climate change has slowed global agricultural productivity growth. Wheeler, Tim; Braun, Joachim von (2013): Climate change impacts on global food
In Nature Clim Change 11 (4), 306–312. https://doi.org/10.1038/s41558-021- security. In Science (New York, N.Y.) 341 (6145), pp. 508–513. DOI: 10.1126/
01000-1. science.1239402.
Partey, S.T., Zougmoré, R.B., Ouédraogo, M., Campbell, B.M., 2018. Developing climate- Wiesmann, Doris; Bassett, Lucy; Benson, Todd; Hoddinott, John (2009): Validation of the
smart agriculture to face climate variability in West Africa: Challenges and lessons world food programme’s food consumption score and alternative indicators of
learnt. In Journal of Cleaner Production 187, 285–295. https://doi.org/10.1016/j. household food security. International Food Policy Research Institute. Washington,
jclepro.2018.03.199. DC (IFPRI Discussion Paper, 00870).
Piñeiro, V., Arias, J., Dürr, J., Elverdin, P., Ibáñez, A.M., Kinengyere, A., Opazo, C.M., Yang, J., Gao, W., Ren, S., 2015. Long-term effects of combined application of chemical
Owoo, N., Page, J.R., Prager, S.D., Torero, M., 2020. A scoping review on incentives nitrogen with organic materials on crop yields, soil organic carbon and total nitrogen
for adoption of sustainable agricultural practices and their outcomes. In Nat Sustain 3 in fluvo-aquic soil. In Soil and Tillage Research 151, 67–74. https://doi.org/10.1016/
(10), 809–820. j.still.2015.03.008.
Qiao, L., Wang, X., Smith, P., Fan, J., Lu, Y., Emmett, B., Li, R., Dorling, S., Chen, H.,
Liu, S., Benton, T.G., Wang, Y., Ma, Y., Jiang, R., Zhang, F., Piao, S., Mϋller, C.,

12

You might also like