Impact of Climate Change Adaptation on Farm Productivity and

Household Welfare
Prince Maxwell Etwire
etwiremaxwellprince@[Link]

Isaac Koomson
ikoomso2@[Link] / koomsonisaac@[Link]

Edward Martey
eddiemartey@[Link]

Forthcoming: Climatic Change (Jan 2022)

NETWORK FOR SOCIOECONOMIC WORKING PAPER

RESEARCH AND ADVANCEMENT

NESRA nesra/wp/22/002

© 2022 Network for Socioeconomic Research and Advancement

[Link]

Electronic copy available at: [Link]

 Impact of Climate Change Adaptation on Farm Productivity and
Household Welfare

Prince Maxwell Etwire1*

etwiremaxwellprince@[Link]

Isaac Koomson2,3
ikoomso2@[Link] / koomsonisaac@[Link]

Edward Martey1
eddiemartey@[Link]

1
Socioeconomics Section, CSIR-Savanna Agricultural Research Institute, P.O. Box TL 52,
Tamale, Ghana
2
UNE Business School, Faculty of Science, Agriculture, Business and Law, University of New
England, Armidale, NSW, Australia
3
Network for Socioeconomic Research and Advancement (NESRA), Accra, Ghana

Abstract
This paper examines how climate change adaptation impacts on farm productivity and the ability
of households to accumulate assets in the context of a developing economy. We apply an
endogenous switching regression to data obtained from 1,440 farmers in Ghana. Our model,
which accounts for endogeneity and selection bias, allows us to simultaneously determine the
factors that influence maize farmers’ decision to adapt to climate change and the productivity and
household assets that result from both adaptation and otherwise. We estimate an inverse
relationship between rainfall and the decision to adapt to climate change. As expected, we find
that access to information has a positive effect on the decision to adapt. Farms that benefit from
adaptation do not become less productive with increases in temperature or rainfall. Overall, we
find that farmers who adapt to climate change are more productive and have more household
assets than their counterfactual. Farmers who do not adapt obtain less yield and have less
household assets than their counterfactual. These findings have important implications for policy.

Keywords: Adaptation; Climate Change; Endogenous Switching Regression; Ghana; Maize

Productivity.

*
Email for correspondence: etwiremaxwellprince@[Link]

Electronic copy available at: [Link]

Introduction

Analysis of scientific data shows that the globe has been warming especially after the pre-industrial
period resulting in climate change. Adverse impacts of climate change include increased rainfall
variability, drought, desertification, and reduced land and farm productivity (IPCC, 2019).1 There
is sufficient evidence to show that Sub-Saharan Africa, though an insignificant contributor to
global warming, is very vulnerable to the effects of climate change with Ghana being a typical
example (Barrios, Ouattara and Strobl, 2008)2. For the different sectors of the economy, the
agricultural sector is more exposed to global warming due to its direct reliance on the climate. The
vulnerability of the agricultural sector of Ghana to climate change cannot be overemphasized.
First, a significant proportion of the population are poor, live in rural areas and depend directly on
agriculture for their livelihood (Gollin, 2010). Second, about 80% of agricultural production is
done on a smallholder basis (i.e., less than 2 hectares) with limited access to modern inputs,
extension services, infrastructure (Ministry of Food and Agriculture, MoFA 2016) and farm
insurance (Barrios, Ouattara and Strobl, 2008). Third, agricultural production in the country is
driven almost entirely by rain-fed production with irrigation accounting for less than 5% of the
cultivated land area (MoFA, 2016). Therefore, the productivity of the sector varies with changes
in rainfall and other climatic factors. Meanwhile, the contribution of the agriculture to
employment, foreign exchange and the overall GDP of Ghana is not trivial thus any changes in
climate can have very important consequences. A major challenge facing agriculture in Ghana
centres around how to assure a sustained improvement in farm productivity as a basis for
increasing production which can then guarantee wealth generation or asset accumulation and
overall economic wellbeing. Growth in the agricultural sector is critical for overall economic
development.

1
The record for Africa shows a general increase in temperature and decline in rainfall (Barrios, Ouattara and Strobl,
2008). In Ghana, climate change manifestations has been in the form of a 0.21°C rise in temperature every decade
with references to the 1960s, a 13% increase in the number of hot days per annum (i.e., Tmax >35°C), a 20% rise in
the number of hot nights per year (i.e., Tmin >26°C) and a 2.4% decline in rainfall per decade since the 1960s. These
indices are more pronounced for the middle and northern parts of Ghana (World Bank, 2021), our study area. These
areas are characterized by frequent drought, decreased fallow period, deforestation, overgrazing, bush fires (Asante
and Amuakwa-Mensah, 2015) and longer duration of Harmattan (Ministry of Foreign Affairs of the Netherlands,
2018).
2
Ghana, for example, contributes only 0.07% of global greenhouse emissions and is ranked 151 out of 188 countries
in terms of per capita greenhouse emissions (Ministry of Foreign Affairs of the Netherlands, 2018).

Electronic copy available at: [Link]

Maize is one staple food crop that dominates the agricultural sector of Ghana. The crop, an
important food security commodity, is grown in every agroecology of the country and accounts
for more than 50% of total cereal production (Abdulai, Nkegbe and Donkoh, 2018). In the absence
of appropriate technologies and practices, small changes in climate (even when not characterized
by extreme events such as floods and droughts) can have large effects on maize productivity
(Barrios, Ouattara and Strobl, 2008). Therefore, estimating the impact of climate change
adaptation3 on maize productivity, in the context of long-term changes in temperature and rainfall,
have important policy implications for improving or maintaining the welfare of producers given
that maize currently provides 20% of the caloric requirements of Ghanaians aside the income it
generates (Stanturf et al., 2011). Our focus on one crop, as opposed to aggregate analysis of several
crops as is the case with some earlier studies (for example, Barrios, Ouattara and Strobl, 2008; Di
Falco, Veronesi and Yesuf, 2011; Anuga et al., 2019; Issahaku and Abdulai, 2019), allows us to
account for an important fact that different crops respond differently to climate change.

The Government of Ghana has formulated several polices and plans to promote climate change
adaptation including a National Climate Change Adaptation Strategy developed in 2012, a
National Climate Change Policy and a National Environment Policy both formulated in 2013. The
country has also launched a National Climate Change Master Plan (2015-2020), a Biological
Diversity National Strategy and Plan of Action, and a National Plan of Action to Combat
Desertification (Ministry of Foreign Affairs of the Netherlands, 2018). Climate change-related
interventions that have been implemented in Ghana include Climate Change Adaptation of Agro-
ecosystems in Ghana, Integrated Resource and Resilience Planning, Adaptation at Scale in Semi-
Arid Regions, West African Science Service Centre on Climate Change and Adapted Land Use,
Great Green Wall (Ministry of Foreign Affairs of the Netherlands, 2018), Adaptation Learning
Program for Africa, Climate Change Adaptation in Northern Ghana Enhanced, Climate for
Development in Africa Programme, Deltas, Vulnerability and Climate Change: Migration and
Adaptation, Ghana Agricultural Sector Investment Programme, Promoting Value Chain Approach
to Adaptation in Agriculture, Resilient and Sustainable Livelihoods Transformation in Northern

3
We define climate change adaptation as any practice or technology that farmers employ to either take advantage of
changes in the climate or minimize the negative impacts of climate change.

Electronic copy available at: [Link]

Ghana, Resilient Landscapes for Sustainable Livelihoods, and Southern Voices on Adaptation
(Dazé and Echeverría, 2016).

There is evidence that some farmers have already adapted to climate change by altering their
practices or adopting new technologies (Gornall et al., 2010; IPCC, 2019). The major climate
change adaptation practices and technologies that farmers in our study area (savannah and forest-
savannah transition zones of Ghana) employ include expansion of area cultivated, planting of
early-maturing and high-yielding varieties, use of chemical fertilizer (World Bank, 2010; Asante
and Amuakwa-Mensah, 2015); application of organic manure, agroforestry, cultivation of
drought-tolerant varieties (Fagariba, Song and Soule Baoro, 2018; Guodaar, Bardsley and Suh,
2021); mulching, changing of planting dates (Fagariba, Song and Soule Baoro, 2018); use of soil
and water conservation measures (Issahaku and Abdulai, 2019); minimum tillage (Anuga et al.,
2019); cover cropping, and integrated pest and disease management (Limantol et al., 2016).

The impact of any climate change adaptation practice or technology depends on a number of
factors including baseline adaptation capacity or constraints (IPCC, 2019). Given that adaptation
is context specific and its effect can be positive (i.e., enabling farming systems to tolerate, escape
or benefit from climate change) or negative (i.e., maladaptation), this paper makes an empirical
contribution within the context of a developing economy using a rich set of farm, farmer and
climate-related data. In spite of adaptation, the effects of climate change can be further exacerbated
by other productivity-constraining factors such as poor access to agricultural extension services.
Agricultural extension agents provide valuable productivity-enhancing advisory services to
farmers. We examine how access to agricultural extension services, conditioned on climate change
adaptation, influence maize productivity and household asset holdings.

There is already some information on the impact of climate change adaptation. For example,
Diallo, Donkor and Owusu (2020) assessed the factors that influenced Malian farmers to adapt to
climate change and the resulting effects of adaptation on maize productivity and household food
security. They employ propensity score matching for analysis thereby controlling for only
observable factors while failing to account for unobservable characteristics that influence
adaptation. Ojo and Baiyegunhi (2020) examined the impact of climate change adaptation on the
technical efficiency of smallholder rice producers in south-western Nigeria by estimating a
stochastic frontier model that corrects for endogeneity. The main limitation of the stochastic

Electronic copy available at: [Link]

frontier model is its assumption that farmers are either homogenous or apply the same production
technology. Despite the expansive literature on farm-level climate adaptation (see Etwire, 2020
and there references therein), there are only few studies that employ counterfactual analysis to
robustly estimate the impact of climate change adaptation. Even though endogenous switching
regression (ESR) accounts for both observed and unobserved heterogeneity and also allows for
counterfactual analysis, the use of such a robust method of analysis in the climate change
adaptation literature is rare. A few exceptions include Di Falco, Veronesi and Yesuf (2011) who
applied the ESR to household data from Ethiopia in order to identify the factors that influence
farmers’ decisions to adapt to climate change and the resulting impact of adaptation on farm
productivity. Ojo and Baiyegunhi (2019) also relied on the ESR to estimate the influence of
climate change adaptation on net revenue from rice produced in south-western Nigeria. In north-
western Ethiopia, Adego, Simane and Woldie (2019) used the ESR to assess how climate change
adaptation influences the productivity of maize and teff farms. Given the attractive features of the
ESR, this study also applies the technique to examine how the decision to either adapt or not,
impacts on maize productivity and household asset holdings while controlling for temperature,
rainfall and a number of other covariates. We anticipate that maize producers who employ climate
change adaptation measures can maintain or improve their yields and consequently accumulate
household assets with the proceeds from the farm.

Earlier micro-impact assessments have tended to focus on traditional outcomes such as farm
productivity, household income, consumption expenditure and food security (Barrios, Ouattara
and Strobl, 2008; Di Falco, Veronesi and Yesuf, 2011; Adego, Simane and Woldie, 2019; Martey,
Kuwornu and Adjebeng-Danquah, 2019; Ojo and Baiyegunhi, 2019; Diallo, Donkor and Owusu,
2020; Ojo and Baiyegunhi, 2020). There is relatively limited information on how climatic variables
and climate change adaptation impacts on household asset holdings. The sparse literature includes
Ogada et al., (2020) who found that the adoption of stress-tolerant crops has a positive effect on
asset accumulation among smallholder farmers in Kenya. Even though the earlier papers on the
impacts of climate change adaptation are useful, the findings of those studies cannot be generalized
since adaptation capacity is variable and each geographic area is characterized by a unique climate
as it relates to variables such as temperature and rainfall. This study adds to the literature on the
impacts of climate change adaptation by providing evidence from Ghana. Given the national

Electronic copy available at: [Link]

importance of the crop that we examine and the climate vulnerability of the area that we study, the
findings of this research have very important policy implications for Ghana.

We describe our main estimation technique (the ESR model) in the next section within a context
of counterfactual analysis. The section also outlines the alternative methods that we compute to
check the robustness of our main estimates. The method’s section is followed by a data and
variables section where we discuss our sources of primary and climate data, sampling procedure
for the primary data, description of the variables that we measure as well as our model’s
identification strategy. We present and discuss the estimates and diagnostic results of our main
model, as well as those of the alternate models, in the results and discussion section. The final
section of the paper contains the main conclusions of our study and their policy implications.

Model

A straightforward approach to estimating how adapting to climate change impacts on farm

productivity and household asset holding is to estimate a classical least squares regression where
climate change adaptation is included as a covariate. However, such an approach is likely to result
in biased estimates of the impacts of adaptation as it ignores potential endogeneity in farmers’
decision to adapt by assuming that adaptation is exogenous without any selection bias. This
assumption is easily violated given that farm-level adaptation is private, voluntary and has some
expected benefits. Thus, farmers may self-select to adapt or not. Additionally, farmers that choose
to adapt may be systematically different from those that do not. Furthermore, there could be some
unobservable characteristics or innate abilities that affect both the decision to adapt and farm
productivity or household asset holdings. Reverse causality between the decision and outcome
variables could also be an issue of concern. We deal with these potential selection and endogeneity
biases by estimating an endogenous switching regression (ESR). The ESR is a popular method in
the evaluation literature that has been applied to many settings (e.g., Di Falco, Veronesi and Yesuf,
2011; Adego, Simane and Woldie, 2019; Martey, Kuwornu and Adjebeng-Danquah, 2019; Ojo
and Baiyegunhi, 2019). We estimate the decision to adapt to climate change as;

1 𝑖𝑓 𝐶𝑖∗ > 0,
𝐶𝑖∗ = 𝛼𝑍𝑖 + 𝜀𝑖 such that 𝐶𝑖 = { (1)
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

where 𝐶𝑖 is farmer 𝑖’s decision to adapt to climate change when the utility from adaptation exceeds
that from non-adaptation (i.e., 𝐶𝑖∗ > 0), 𝑍 is a vector of covariates that influence the decision to
5

Electronic copy available at: [Link]

adapt including climatic, farm and farmer characteristics, 𝛼 is a vector of parameters to be
estimated while 𝜀 is the error term. In terms of productivity or household asset index, maize
farmers face the following two outcomes;

𝑂𝑢𝑡𝑐𝑜𝑚𝑒 1: 𝑌1𝑖 = 𝛽1 𝑋1𝑖 + 𝜖1𝑖 (2a)

𝑂𝑢𝑡𝑐𝑜𝑚𝑒 2: 𝑌2𝑖 = 𝛽2 𝑋2𝑖 + 𝜖2𝑖 (2b)

where 𝑌1𝑖 and 𝑌2𝑖 is the productivity or household asset index of the 𝑖th farmer in regime (1) and
(2), respectively, 𝑋 is a vector of covariates that determine productivity or household asset index
similar to 𝑍, but with at least one less variable in order to satisfy the exclusion restriction principle
and the 𝜖s are the error terms associated with each regime. We compute a Cobb-Douglas function
(i.e. double-log function) for our maize productivity outcome but estimate the asset index outcome
in levels. Our model allows us to compare the expected productivity or household asset index of
farms that adapted to climate change to farms that did not adapt. The model is also able to estimate
the expected counterfactual productivity or household asset index of farmers that adapted had they
not adapted and of farmers that did not adapt had they adapted. These conditional expectations
for maize productivity or household asset index are presented as Table A1 and summarized as;

𝐸(𝑌1𝑖 |𝐶𝑖 = 1) = 𝑋1𝑖 𝛽1 + 𝜎1𝜀 𝜆1𝑖 (3a)

𝐸(𝑌2𝑖 |𝐶𝑖 = 0) = 𝑋2𝑖 𝛽2 + 𝜎2𝜀 𝜆2𝑖 (3b)

𝐸(𝑌2𝑖 |𝐶𝑖 = 1) = 𝑋1𝑖 𝛽2 + 𝜎2𝜀 𝜆1𝑖 (3c)

𝐸(𝑌1𝑖 |𝐶𝑖 = 0) = 𝑋2𝑖 𝛽1 + 𝜎1𝜀 𝜆2𝑖 (3d)

where Equations (3a) and (3b) are the actual expectations while Equations (3c) and (3d) are the
counterfactual expected outcomes. Note that 𝜎 is a coefficient that captures the correlation between
the decision and outcome equations and 𝜆 is the inverse mills ratio generated from the decision
equation as regressors for the outcome equations. Consistent with the literature, we estimate the
average adaptation effect on those that adapted (𝐴𝑇𝑇) as the difference between Equation (3a) and
(3c) such that;

𝐴𝑇𝑇 = 𝐸(𝑌1𝑖 |𝐶𝑖 = 1) − 𝐸(𝑌2𝑖 |𝐶𝑖 = 1) = 𝑋1𝑖 (𝛽1 − 𝛽2 ) + (𝜎1𝜀 − 𝜎2𝜀 ) 𝜆1𝑖 (4)

Similarly, we estimate the average adaptation effect on those that did not adapt (ATU) as the
difference between Equations (3d) and (3b) such that;

Electronic copy available at: [Link]

 𝐴𝑇𝑈 = 𝐸(𝑌1𝑖 |𝐶𝑖 = 0) − 𝐸(𝑌2𝑖 |𝐶𝑖 = 0) = 𝑋2𝑖 (𝛽1 − 𝛽2 ) + (𝜎1𝜀 − 𝜎2𝜀 ) 𝜆2𝑖 (5)

We also use the expected outcomes defined in Equations (3a) – (3d) to estimate the heterogeneity
effects following the literature. There is a possibility that farmers that adapt may be intrinsically
more productive than those that did not adapt because of unobservable characteristics such as
innate ability and skills. The base heterogeneity effect of farms that adapted is the difference
between Equations (3a) and (3d);

𝐵𝐻1 = 𝐸(𝑌1𝑖 |𝐶𝑖 = 1) − 𝐸(𝑌1𝑖 |𝐶𝑖 = 0) = (𝑋1𝑖 − 𝑋2𝑖 )𝛽1𝑖 + 𝜎1𝜀 ( 𝜆1𝑖 − 𝜆2𝑖 ) (6)

Similarly, the base heterogeneity for farmers that did not adapt is the difference between Equations
(3c) and (3b);

𝐵𝐻2 = 𝐸(𝑌2𝑖 |𝐶𝑖 = 1) − 𝐸(𝑌2𝑖 |𝐶𝑖 = 0) = (𝑋1𝑖 − 𝑋2𝑖 )𝛽2𝑖 + 𝜎2𝜀 ( 𝜆1𝑖 − 𝜆2𝑖 ) (7)

Transitional heterogeneity, which shows if the climate change adaptation effects are larger or
smaller for adapting or non-adapting households relative to their respective counterfactuals, is
computed as the difference between Equations 4 (ATT) and 5 (ATU).

We check the consistency of our ESR results by estimating an Inverse-Probability-Weighted

Regression Adjustment model. Just like our main model, the alternate model is also popular in the
literature (see, for example, Wooldridge, 2002) so we do not derive them in this paper in the
interest of brevity.

Data, Variables and Identification

We rely on secondary climate data and primary farm and farmer data for our analyses. The area
that we studied shares border with the republics of Côte d'Ivoire to the West, Burkina Faso to the
North and Togo to the East. Five of the eight regions that we sampled (Savannah, Northern, North
East, Upper East and Upper West regions) fall within the Savannah agroecology characterized by
the presence of drought-tolerant trees such as acacia (Acacia longifolia), baobab (Adansonia
digitata L), shea nut (Vitellaria paradoxa), dawadawa (Parkia biglobosa) and neem (Azadirachta
indica). The Savannah agroecology experiences a single rainy season which normally starts in
May and ends in October. The rest of the year is dry with November to January being the harmattan
season where the mornings are usually cold, dusty and hazy. February to April often experience
intense hot weather prior to the onset of the rains. The Savannah agroecology has a relatively flat

Electronic copy available at: [Link]

topography with most lands being below 300 metres above sea level. The population of the area,
the majority of whom engage in agriculture, rely heavily on rainfall to cultivate crops such as
maize, sorghum, millet, rice, soybean, groundnuts, cotton, yam and cowpea. Livestock commonly
produced include poultry, sheep, goats, pigs and cattle. (Ghana Statistical Service [GSS] 2013a, b,
c). The rest of the three regions that we study, namely Ahafo, Bono and Bono East, experience
two rainy seasons (major and minor). Ahafo and parts of the other two regions have semi-
deciduous forest with the rest being Guinea savannah woodland. The main crops cultivated are
maize, cassava, plantain, yam, cocoyam and rice (GSS, 2013d).

We adopted a multistage sampling technique to identify maize producing households for

enumeration. We randomly sampled three districts from each of the eight study regions given that
maize is produced in all farming communities. We further randomly sampled four communities in
each of the selected districts for the survey. In each selected community, we sampled 15 farmers
for the interviews making a total of 1440 primary observations. The interviews were conducted by
experienced enumerators who have all completed tertiary education and speak the first languages4
of the farmers. The survey questionnaire, which was administered electronically on a handheld
device after three days of training and pretesting, captured information on farmer, farm and
household characteristics such as farm size, maize output, farming experience, household
demography, assets, economic activities, access to agricultural services and GPS record. We
measure maize productivity, one of our outcomes of interest, as the kilograms of maize harvested
over a hectare of land. We compute the other outcome variable, household asset index, using
principal component analysis such that;

𝑎𝑗𝑖 −𝑥𝑖
𝐼𝑗 = ∑𝑘𝑖=1 𝑏𝑖 [ ] (8)
𝑠𝑖

where 𝐼 is the household asset index, 𝑏 is loading from the first principal component, 𝑎 is indicator
value, 𝑥 is the mean indicator value and 𝑠 is the standard deviation of the indicators. A higher
index value suggests a higher accumulation of household assets and vice versa. Note that negative
values do not imply that those households do not own any asset at all, such households rather own

4
The major ethnic groups in the study area are Mole-Dagbon, Akan, Guan, Gurune, Dagaaba, Sissala, Wala, Kassena-
Nankani and Kusasi.

Electronic copy available at: [Link]

comparatively less assets. The assets captured in the data include car/tractor, draught animals,
motor bicycle, bicycle, television, radio, mobile/telephone, refrigerator, fan and jewelry.

We define farm size as the total hectares of land under production. Whereas we capture farming
experience as the total number of years that a head of house has been engaged in production, we
measure household size as the total number of individuals that make up a household. We capture
multiple economic activities, sex, formal education, access to agricultural extension and location
as dummy variables. We assign one if the head of household engages in multiple economic activity
(i.e., has another economic activity aside farming), is a male, had formal education (i.e., attended
at least a primary school), or has had contact with agricultural extension agents. We also collected
information on region of residence and enabled the handheld data collection devices to record the
GPS coordinates after each successful enumeration.

We relied on the collected GPS coordinates to generate 30-year mean temperature and
precipitation data from [Link] We measure
temperature and rainfall in degree Celsius and millimeters, respectively.

In order to identify our model, we assume, as a validity condition, that the mechanism by which
farmers with access to agricultural extension services can improve on their productivity and
household asset holdings is by implementing climate change adaptation practices and
technologies. Therefore, we exclude those two variables from our productivity and asset index
equations in order to satisfy the exclusion restriction principle. As a robustness check, we also
explore intensity of access (or number of neighbours’ with access) to agricultural extension
services as an additional instrument.6 In addition to the validity condition, an instrument must also
satisfy the relevance condition (Wooldridge, 2015) so we argue that a farmer’s adaptation to
climate change is significantly associated with access to extension services7 (Di Falco, Veronesi
and Yesuf, 2011). Given that farmers can only apply adaptation measures that they are aware of,

5
These climate data were generated by the US National Centre for Atmospheric Research Community Climate System
Model for the 4th and 5th Assessment Reports of the Intergovernmental Panel on Climate Change (IPCC). The
Community Climate System Model is a fully-coupled, global climate model that provides historical, present and
projected computer climate simulations.
6
In rural Ethiopia, Di Falco, Doku and Mahajan (2019) observed that peer influence increases the likelihood of
adoption of climate change adaptation strategies.
7
In Ghana, agricultural extension agents are a major source of information on climate change adaptation practices and
technologies, principal facilitators of adult learning and a major source of support for farmers to develop technical and
managerial skills (Danso Abbeam, Ehiakpor and Aidoo, 2018).

Electronic copy available at: [Link]

several studies have used information-related variables to show how access to information
significantly influence climate change adaptation ((Di Falco, Veronesi and Yesuf, 2011; Adego,
Simane and Woldie, 2019; Issahaku and Abdulai, 2019; Martey, Kuwornu and Adjebeng-
Danquah, 2019).

We rely on a falsification test to determine the statistical validity of our instruments. An instrument
is deemed to be valid if it has a significant influence on the adaptation decision but an insignificant
effect on either maize productivity or the asset index of non-adapting households. Table A2, in the
appendix, shows that access to agricultural extension service can be considered as a valid selection
instrument as it significantly determines the decision to adapt to climate change but has an
insignificant effect on both maize productivity and the asset index of non-adapting households.

Table 1 presents a description of the variables that we model for the total sample and a
disaggregation of the data by climate change adaptation status. Farmers who apply any climate-
smart-agricultural practice or technology (such as use of stress-tolerant varieties, soil water
conservation, and soil amendments) are deemed to have adapted to climate change.

For the combined sample, the data shows that an average household, regardless of climate change
adaptation status, produces maize on 2 hectares of land under mean annual temperature of 31oC
and average annual precipitation of 1250mm. Maize farmers, on the average, obtain a yield of
about 700Kg/Ha. A typical household has 11 members and is headed by a male with 21 years of
farming experience. The data shows that the majority of farmers did not have multiple economic
activities, formal education or access to agricultural extension. In addition to the aggregate
descriptive analysis, we apply student’s t and Pearson’s chi-square tests to our quantitative and
qualitative variables, respectively, to test if farmers significantly differ by adaptation status.

Results of the unconditional tests (Table 1) show that farmers that adapted to climate change
obtained significantly higher maize yield, had a higher asset index, produce on bigger plots and
had slightly lower household members. A higher proportion of households that adapted to climate
change were headed by males, had formal education and also had access to agricultural extension
services. Compared to middle Ghana (Bono and Bono East regions), there were significantly lower
proportion of households that adapted to climate change in northern Ghana (Northern and
Savannah regions). These findings do not suggest a causal relationship between climate change
adaptation and the aforementioned variables. Farmers who adapted to climate change may also,

10

Electronic copy available at: [Link]

for example, had access to other productivity-enhancing farm attributes and services that impacted
substantially on yield relative to adaptation. We did not find significant bivariate unconditional
differences between farmers with respect to variables such as temperature, precipitation, farming
experience and engagement in multiple economic activities.

Figure A1 provides additional information on the distributional relationship between maize

productivity and climate change adaptation status as well as other binary variables such as
participation in multiple economic activities, sex, formal education, and access to agricultural
extension services. Regardless of adaptation status, Figure A1 shows that the distribution of maize
productivity is slightly skewed to the right implying that only few farms are highly productive
thereby providing some support for log-transformation of the data. Similarly, the data is skewed
to the right when we plot maize productivity by participation in multiple economic activities,
formal education, access to extension services or sex. In all instances, the mode of the maize
productivity distribution is higher than the superimposed normal curve. Figure A2 suggests an
approximately normal distributional relationship between household asset index and climate
change adaptation status. A similar approximately normal relationship is established for the other
binary variables that we model.

11

Electronic copy available at: [Link]

Table 1: Description of variables by climate change adaptation status
Adapted Did not adapt Combined t P-value
Quantitative variables Mean values with standard
deviations in italics
Maize yield (Kg/Ha) 819.8 603.2 696.1 -7.213 0.000
608.7 511.1 565.1
Asset index 0.158 -0.137 -0.009 -3.929 0.000
1.392 1.341 1.371
0
Temperature ( C) 31.2 31.1 31.1 -0.630 0.529
1.2 1.0 1.1
Precipitation (mm) 1249.8 1250.9 1250.4 0.147 0.883
153.9 149.7 151.5
Farm size (Ha) 2.6 1.7 2.1 -7.209 0.000
3.0 1.6 2.4
Farming experience (years) 21.2 20.7 21.0 -0.778 0.437
13.0 11.7 12.3
Household size (number) 10.1 11.0 10.6 1.915 0.056
8.4 9.5 9.0
Adapted Did not adapt Combined Chi2 P-value
Qualitative variables Frequencies with percentages in
italics
Multiple economic activities = yes 289 376 665 0.019 0.890
45.7 46.0 45.9
Male = yes 545 637 1,182 15.646 0.000
86.1 78.0 81.5
Formal education = yes 330 365 695 7.947 0.005
52.1 44.7 47.9
Access to extension = yes 315 317 632 17.432 0.000
49.8 38.8 43.6
Location = Northern 34 146 180 51.247 0.000
5.4 17.9 12.4
Location = North East 81 101 182 0.061 0.805
12.8 12.4 12.6
Location = Savannah 67 113 180 3.458 0.063
10.6 13.8 12.4
Location = Upper East 70 110 180 1.898 0.168
11.1 13.5 12.4
Location = Bono 98 85 183 8.339 0.004
15.5 10.4 12.6
Location = Bono East 143 42 185 97.572 0.000
22.6 5.1 12.8
Total 633 817 1,450
Notes: We apply student t and Pearson chi square tests to our quantitative and qualitative variables, respectively, to
examine if maize productivity differ significantly by climate change adaptation status. Upper West Region serves as
the base category for location while otherwise/=no serves as the base category for the other qualitative variables.

12

Electronic copy available at: [Link]

Climate Change Adaptation and Household Welfare

The estimates of the ESR are presented in Table 2. The results show that whereas some variables
have crosscutting effects on the decision to adapt to climate change and its consequential
productivity or asset index, there are other variables that only have specific effects. We find a
negative relationship between the decision to adapt and rainfall. Farmers are more likely to adapt
when they have access to additional land, are males and have access to agricultural extension
services, our instrument.

The inverse relationship between rainfall and the decision to adapt appears to be intuitive for the
northern and middle belts of Ghana. Rainfall is one of the most important agro-climatic variables
especially in developing economies such as Ghana where farmers rely heavily on rain-fed
production for over 95% of farms. Therefore, farmers are less likely to have need for adaptation if
rainfall is good and vice versa if rainfall is poor. In addition, long-term increases in rainfall in the
savannah (northern Ghana) and forest-savannah transition (middle Ghana) zones of Ghana will
likely convert the agroecology of the former to the latter and then convert the latter to a forest
agroecology (as is the present case in southern Ghana). Hence, farmers may not need to employ a
new adaptation measure with an increase in rainfall. Conversely, long-term decline in rainfall will
likely convert the forest-savannah agroecology to a savannah agroecology and then convert the
savannah agroecology to a Sahel agroecology. A Sahel or arid agroecology will be new to farmers
(as Ghana currently does not have such an agroecology) which implies that farmers may need to
adapt to this new type of agroecology resulting from a decline in rainfall. Furthermore, the mean
rainfall observed in the data (1250mm) appears to be just about adequate for the production of
cereals and legumes. This implies that whereas farmers may not need to adapt to an increase in
rainfall as agricultural production will likely benefit from a decline in water stress, farmers may
have to adapt to water stresses, such as drought, with a decline in rainfall.

Farmers who have access to an additional plot are more likely to employ a climate change
adaptation measure. It is well known that farmers generally experiment with technologies and
practices on a piece of land prior to large scale adoption. This is usually done to minimize risks
and to validate the technologies and practices before full-scale adoption. Consequently, farmers
with bigger plots have more space to try out climate change adaptation measures and so, more
likely to adapt. Also, some adaptation measures (such as agroforestry or tree crop integration)

13

Electronic copy available at: [Link]

require large farm sizes implying that farmers with smaller plots may be constrained from adopting
those practices and technologies. Unlike farmers with smaller fields, farmers with bigger plots are
more likely to benefit from economies of scale and thus more likely to implement climate change
adaptation measures.

As can be anticipated for the middle and northern parts of Ghana, households headed by males are
more likely to adapt to climate change. Males in these areas often assume the leadership roles of
their communities and tend to have better access to productivity-enhancing information and
resources. Unlike females, males are allowed to practice polygamy to increase the family labour
required for the implementation of climate change adaptation measures. Also, the normal practice
after marriage is for the woman to join her husband in his community, thus females are often
unable to own and make long-term farm investments.

Our estimates further show that farmers who have access to agricultural extension services are also
more likely to adapt to climate change. This result is expected as awareness is a necessary
condition for adoption or adaptation. As earlier indicated, agricultural extension agents play
important roles in the production process as they often provide several services to farmers
including the provision of climate information, demonstration of climate-smart agricultural
practices and technologies, linking of farmers to other value chain actors and service providers,
among others.

Table 2 further provides information on factors that determine the productivity of maize farmers
that adapted as well as those that did not adapt to climate change. Whereas farmers that employ
climate change adaptation measures benefit from increases in temperature and rainfall, the
productivity and asset holdings of non-adapting farmers are driven by formal education and sex
(being a male). Regardless of adaptation status, we find an inverse relationship between maize
productivity and farm size but estimate a direct relationship for asset index. As can be anticipated,
we observe that participation in multiple economic activities has a positive effect on household
asset holdings irrespective of adaptation status.

Even though climate change is often projected to have an overall negative effect on agriculture,
our findings reaffirms the important roles that adaptation can play to enable farmers to cope with
or even take advantage of long-term increases in temperature and rainfall. Use of stress-tolerant
varieties, soil and water conservation techniques and integrated soil fertility management practices

14

Electronic copy available at: [Link]

are some adaptation measures that farmers can employ to benefit from rising temperature and
rainfall. In Ethiopia, Di Falco, Veronesi and Yesuf (2011) reported that whereas increases in
temperature and rainfall does not harm the productivity of farmers who adapt to climate change,
farmers who do not adapt suffer productivity losses with initial increases in temperature and
rainfall.

Our estimates also show that maize farmers who do not adapt can still rely on attributes linked
with formal education as well as resources associated with being a male to improve their farm
productivity and asset holdings. We find that farm size has a general negative effect on maize
productivity thereby suggesting that smallholder farmers may be better-off concentrating their
meagre resources on effectively managing smaller plots than spreading their resources thinly on
large farms. Abdulai, Nkegbe and Donkoh (2018) observed that relative to farmers with bigger
plots, smallholder maize farmers in northern Ghana have a better chance of securing valuable
services such as rental tractors and hired labour since those service providers are able to complete
their tasks within a shorter time frame for payment. In a worst case scenario, smallholder farmers
can replace external inputs and resources with internal inputs and resources. For example,
smallholder farmers could switch from tractor ploughing to animal ploughing which is prohibitive
for large scale producers. However, Ojo and Baiyegunhi (2019) found a positive relationship
between farm size and rice income in south-western Nigeria. With regards to asset holdings, our
data shows an increase in farm size results in an increase in asset holdings. The mechanism by
which an increase in farm size results in an increase in household assets is not explicitly clear and
therefore requires additional research to identify. However, the effect of engaging in multiple
economic activities is unambiguous as households that have multiple streams of income, regardless
of adaptation status, may still be able to procure assets in poor cropping seasons or in periods of
good alternate income.

In terms of location effects, we find that farms in the Savannah and Upper East regions are less
productive relative to those in the Upper West Region. This implies that maize interventions in
Ghana must specifically target farmers in those regions in order to improve their resilience and
then increase their productivity. Our results also show that households in the Savannah Region
have less assets than those in the Upper West Region. Bridging the yield gap between farmers in

15

Electronic copy available at: [Link]

the Upper West Region and those in the Savannah and Upper East regions will boost maize
production in Ghana.

16

Electronic copy available at: [Link]

Table 2: Determinants of climate change adaptation and resultant maize yields and household asset index

Log(maize Kg/Ha) Asset index

Adaptation Adaptation
Variable Did not Variable Did not
decision (1/0) Adapted decision (1/0) Adapted
adapt adapt
Log(Temperature) 2.275 3.562* 0.226 Temperature 0.063 -0.095 0.122
1.858 1.863 1.325 0.061 0.105 0.075
Log(Precipitation) -3.531*** 4.546*** -0.688 Precipitation -0.003*** 0.001 0.001
0.732 0.788 0.598 0.001 0.001 0.001
Log(Farm size) 0.130*** -0.147*** -0.181*** Farm size 0.041** 0.049* 0.133***
0.050 0.046 0.045 0.019 0.025 0.033
Log(Farming
-0.056 -0.010 0.039 Farming experience 0.002 -0.002 0.006
experience)
0.053 0.049 0.043 0.003 0.005 0.004
Log(Household size) 0.063 -0.016 0.081* Household size 0.002 0.003 0.007
0.064 0.061 0.049 0.004 0.008 0.005
Multiple economic Multiple economic
0.000 -0.067 0.079 -0.012 0.432*** 0.464***
activities = yes activities = yes
0.072 0.070 0.055 0.075 0.122 0.092
Male = yes 0.316*** -0.031 0.256*** Male = yes 0.324*** 0.242 0.256*
0.098 0.099 0.075 0.100 0.177 0.146
Formal education = Formal education =
-0.045 0.068 0.161** 0.038 0.071 0.480***
yes yes
0.082 0.078 0.063 0.085 0.135 0.107
Located in Northern Located in Northern
-0.680*** -0.007 -0.049 -0.631*** 0.646** 0.619**
= yes = yes
0.152 0.162 0.121 0.150 0.285 0.243
Located in North Located in North
-0.227 -0.059 -0.023 -0.338** -0.548** -0.175
East = yes East = yes
0.139 0.133 0.124 0.149 0.237 0.234
Located in Savannah Located in Savannah -
-0.126 -0.987*** -0.378*** -0.107 -0.333*
= yes = yes 1.341***

17

Electronic copy available at: [Link]

 Log(maize Kg/Ha) Asset index
Adaptation Adaptation
Variable Did not Variable Did not
decision (1/0) Adapted decision (1/0) Adapted
adapt adapt
0.129 0.131 0.101 0.132 0.229 0.174
Located in Upper Located in Upper
-0.508*** -0.758*** -0.770*** -0.545*** -0.360 0.383
East = yes East = yes
0.158 0.154 0.142 0.163 0.274 0.280
Located in Bono = Located in Bono =
0.941*** -1.242*** 0.273** 0.935*** -0.338 0.640**
yes yes
0.148 0.162 0.138 0.151 0.312 0.320
Located in Bono Located in Bono -
1.449*** -1.594*** -0.116 1.473*** 0.251
East = yes East = yes 0.983***
0.151 0.158 0.211 0.154 0.327 0.532
Access to extension Access to extension
0.145*** 0.141**
= yes = yes
0.054 0.068
Notes: *, ** and *** denote statistical significance at 1, 5 and 10%, respectively. Upper West Region serves as the base category for location while otherwise/=no
serves as the base category for the other qualitative variables. Values in italics are standard errors.

18

Electronic copy available at: [Link]

Table 3 presents the treatment effects from the ESR model. The table shows that farmers that
adapted to climate change obtained higher maize yields and had larger asset index than they would
have had they no adapted. Maize farms that benefitted from climate change adaptation would have
been 43% less productive without adaptation. Farmers who did not adapt recorded lower yields
and smaller asset index than they would have had they adapted. These farms would have been
190% more productive had they benefitted from adaptation. With a mean yield of 603.2 Kg/Ha
(Table 1), our finding suggests that the average yields of farmers who did not adapt can rise to
1146.1 Kg/Ha if they employ climate change adaptation measures. Note that this projected yield
is still lower than the current national average of 1920 Kg/Ha (MoFA, 2016). Maize productivity
in Ghana is generally low. There is a 65% gap between potential yields and what farmers’ record
(MoFA, 2016). This huge yield gap is caused by biotic and abiotic factors such as poor soil fertility
(owing to both high nutrient depletion and low soil amendment), drought, weed infestation, and
pests and diseases (Lobulu et al., 2019). These constraints are severer in our study area because of
the marginal soils in the zone, harsher climatic conditions and higher levels of poverty (Stanturf et
al., 2011). Martey, Etwire and Kuwornu (2020) found that the adoption of drought-tolerant maize
varieties in the Northern Region of Ghana leads to a 150% increase in yield.

A disaggregation of the treatment effects by age group (i.e., youth versus elderly), as presented in
Table A4 in the appendix, shows that the productivity gains for younger farmers that adapted to
climate change was higher than that of the elderly farmers that adapted when each age group is
compared to their counterfactual implying that younger farmers are better-off adapting to climate
change. Similarly, the productivity of elderly farmers who did not adapt to climate change was
much lower than younger farmers who did not adapt when both categories are compared to their
counterfactual.

Our results confirm that climate change adaptation provides clear benefits as it impacts positively
on farm productivity and household asset holdings. The benefits of adaptation are much more
pronounced among younger farmers suggesting that this gendered difference in adaptation must
be considered in the design of climate change adaptation projects and programs. Adaptation
measures can enable farmers to escape, cope or deal with the negative effects of climate change
while amplifying the positive effects. Adaptation strategies that, among others, maintain or
enhance soil fertility and moisture, improve the performance of crop varieties or increase stress

19

Electronic copy available at: [Link]

tolerance, have positive effects on yields. Note that the heterogeneity effects with respect to our
productivity model are negative implying that the productivity gains from climate change
adaptation is smaller for farmers that adapted relative to those that did not thereby suggesting that
non-adapting farmers will likely record large productivity gains if they adapt. Therefore, there is
a need to encourage and assist non-adapting farmers to adopt climate change adaptation measures.
It is important to expose non-adopting farmers to the performance of various adaptation practices
and technologies through, for example, practical field sessions and then either assist such exposed
farmers with resources to adopt or address the constraints associated with the use of such climate
change adaptation measures.

Our findings are consistent with the literature. Maize farmers who adapted to climate change in
southern Mali recorded higher yields than those that did not adapt. Compared to maize producers
that did not adapt, farmers who applied organic fertilizers had additional yield of 150 Kg/Ha while
those that planted early-maturing varieties had an additional yield of about 250 Kg/Ha (Diallo,
Donkor and Owusu, 2020). There is evidence that rice farmers in south-western Nigeria who
employed climate change adaptation measures were more productive than those that did not (Ojo
and Baiyegunhi, 2019; Ojo and Baiyegunhi, 2020). There is similar evidence of the positive effects
of climate change adaptations on farm-level productivity in Ethiopia (Di Falco, Veronesi and
Yesuf, 2011; Adego, Simane and Woldie, 2019). With reference to assets, Ogada et al., (2020)
observed that the uptake of climate change adaptation measures resulted in a buildup of household
assets in Kenya. Similarly, Awotide et al., (2015) observed that adoption of improved cassava
varieties in south-western Nigeria impacts positively on ownership of assets.

As a robustness check to the ESR estimates, we applied the Inverse-Probability-Weighted

Regression Adjustment (IPWRA)8 to the data and report the results in Table 3. The IPWRA
estimates are consistent with that of our main model, ESR, except that the mean ATTs are smaller
with larger standard errors. In addition to the different model, we also check for robustness by re-
estimating our main model based on the use of an alternate instrument that captures direct and all
indirect (up to 14) access to agricultural extension services. A farmer is deemed to have one
indirect access to agricultural extension services (through peer learning) for every neigbhour that

8
The IPWRA is doubly robust as it allows one of the equations (either treatment status or outcome equation) to be
mis-specified.

20

Electronic copy available at: [Link]

reported having access to agricultural extension services. The maximum number of neighours that
a farmer can have in our sample is 14. The results of our re-estimated model is presented in the
appendix as Table A3. We find that the estimates for the maize productivity and asset index
equations, adaptation selection equation, mean expected maize productivity and asset index, as
well as treatment and heterogeneity effects are similar to that of our main model. Overall, the
robustness checks suggest that our estimates are quite insensitive to different model and instrument
specifications.

Table 3: Mean expected maize productivity, treatment and heterogeneity effects

Productivity model Asset index model
Adaptation
Adaptation decision
Variable decision Treatment Treatment
Not to effect Not to effect
To adapt To adapt
adapt adapt
ATT = ATT =
Adapted 6.416 5.989 0.187 -0.014
0.427*** 0.202***
0.018 0.013 0.022 0.027 0.027 0.038
ATU = ATU =
Did not adapt 8.016 6.114 2.072 -0.134
1.903*** 2.207***
0.020 0.011 0.023 0.023 0.020 0.031
Heterogeneity effects -1.600 -0.125 -1.885 0.120
0.028 0.017 0.036 0.033

Inverse-Probability-
ATT = ATT =
Weighted Regression
0.267*** 0.134*
Adjustment
0.046 0.080
Notes: We estimate a double-log model and rely on the Welch's t-test to examine if the treatment effects are
statistically significant. *, ** and *** denote statistical significance at 1, 5 and 10%, respectively. Values in italics are
standard errors.

Conclusions and Policy Implications

Even though there are several studies that estimate the impact of climate change adaptation in the
literature, there is limited information on expected and counterfactual differences in productivity
and asset holdings of farmers that adapt relative to those that do not. This paper examined how the
decision to adapt to climate change impacts on farm productivity and household assets. We applied
Endogenous Switching Regression (ESR) to farm-level data from Ghana to estimate the factors
that influence the decision to adapt to climate change and then the resulting maize productivity

21

Electronic copy available at: [Link]

and household assets that arise from both adaptation and non-adaptation. The ESR addresses
unobservable factors that influence both climate change adaptation and maize productivity or asset
accumulation within the context of counterfactual analysis.

The results of our first-stage choice equation show that farm size, sex (being a male) and access to
agricultural extension services increase the probability of a farmer adapting to climate change.
Lack of climate change adaptation information is often been cited as an important constraint that
limits or prevents farmers from employing climate change adaptation measures. Our results
confirm that farmers who have access to productive resources or have features that make it easier
to access these resources are more likely to employ climate change adaptation measures on their
farm. We also find that the probability of farmers adapting to climate change declines with rainfall.
This result suggests that rainfall is, perhaps, the most important climatic variable that influences
farmers’ adaptation decision. Therefore, farmers may not have any need to adapt to climate change
as rainfall improves. However, adaptation becomes very critical as rainfall declines and water
stress becomes real.

In terms of maize productivity, our estimates show that farmers who employ climate change
adaptation measures benefit from increases in rainfall and temperature. This result is not surprising
as the essence of farm adaptation is to enable farmers to escape, cope or deal with the effects of
changes in climatic conditions mainly temperature and rainfall. For non-adapting farmers, we find
a positive relationship between maize productivity or asset holdings and farmer characteristics
such as sex (being a male) and formal education. Regardless of adaptation status, we estimate an
inverse relationship between maize productivity and farm size. The inverse relationship suggests
that farmers, considering their current levels of resources and technical efficiencies, cannot
improve on their productivity by engaging in large scale production of maize.

Our counterfactual analysis shows clearly that climate change adaptation improves the
productivity of maize farms and household asset holdings. The average treatment effects show that
farmers who employed climate change adaptation measures recorded yields that were higher than
what would have been the case had they not adapted. We found that farmers who did not adapt
would have obtained yields that are higher than what they recorded had they adapted. These effects
are even higher for younger farmers. Therefore, non-adapting farmers stand to benefit the most

22

Electronic copy available at: [Link]

from climate change adaptation. We get similar findings when we use a different instrument or
estimate alternate models.

Our findings have several policy implications as it relates to climate change adaptation and farm
productivity. Research and funding institutions (such as Ghana’s Council for Scientific and
Industrial Research, the various agricultural faculties of universities in Ghana, the CGIAR, the
African Development Bank and the World Bank, among others) have to prioritize any research or
intervention that seeks to improve the resilience and tolerance of agricultural systems to declining
rainfall and climate-induced water stresses. In addition to improved technologies and practices,
some investments should be channeled to the Ghana Meteorological Agency to develop and make
reliable rainfall forecasts readily available to farmers. Nonetheless, any investment that would
make farmers less reliant on rainfall would likely be very beneficial. Therefore, the Government
of Ghana must mobilize resources and dedicate a fair proportion of its gross domestic product to
finance the country’s national adaptation plan and also consider establishing a national climate
research fund.

There is a need to build capacity and resource agricultural development agents (such as staff of the
Government of Ghana’s Department of Agriculture, Ministry of Environment, Science,
Technology and Innovation, Environmental Protection Agency, Ministry of Local Government
and Rural Development, field workers of non-governmental organizations, staff of climate change
adaptation programs and projects) to implement a number of measures to ensure widespread use
of climate change adaptation practices and technologies. There is still the need to expose farmers,
especially the youth, to the clear benefits of climate change adaptation especially as it relates to
changes in rainfall and temperature. Community-level awareness campaigns, practical on-farm
demonstrations, video shows and radio programs are some tools that can be used to create
awareness on climate change adaptation measures. Such dissemination efforts must, however,
have gender dimensions that aim to target and empower women as it appears males in the middle
and northern parts of Ghana still have better access to information and productive resources. Even
though creating awareness on climate change adaptation measures using novel approaches is
useful, sustainable adoption of those practices will likely occur if farmers have access to external
resources and support such as irrigation schemes, production credit, farm insurance, reliable
weather forecasts, access to agro-inputs and output markets. In the absence of such external

23

Electronic copy available at: [Link]

support, farmers can be encouraged to cultivate plot sizes that they can easily manage within their
resource constraints.

References

Abdulai, S., Nkegbe, P. K., & Donkoh, S. A. (2018). Assessing the technical efficiency of maize
production in northern Ghana: The data envelopment analysis approach, Cogent Food &
Agriculture, 4:1, 1512390, doi: 10.1080/23311932.2018.1512390
Adego, T., Simane, B., & Woldie, G. A. (2019). The impact of adaptation practices on crop
productivity in northwest Ethiopia: an endogenous switching estimation. Development
Studies Research, 6(1), 129–141. doi:10.1080/21665095.2019.1678186
Anuga, S. W., Gordon, C., Boon, E., & Surugu, J. M-I. (2019). Determinants of climate smart
agriculture adoption among smallholder food crop farmers in the Techiman Municipality,
Ghana. Ghana Journal of Geography, 11(1), 124 – 139.
Asante, F., & Amuakwa-Mensah, F. (2015). Climate change and variability in Ghana: Stocktaking.
Climate, 3, 78–99. doi:10.3390/cli3010078
Awotide, B. A., Alene, A. D., Abdoulaye, T., & Manyong, V. M. (2015). Impact of agricultural
technology adoption on asset ownership: the case of improved cassava varieties in Nigeria.
Food Security, 7(6), 1239–1258. doi:10.1007/s12571-015-0500-7
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. doi:10.1016/[Link].2008.01.003
Danso-Abbeam, G., Ehiakpor, D. S., & Aidoo, R. (2018). Agricultural extension and its effects on
farm productivity and income: insight from Northern Ghana. Agric & Food Secur, 7.
[Link]
Dazé, A., & Echeverría, D. (2016). Review of current and planned adaptation action in Ghana.
CARIAA Working Paper no. 9. International Development Research Centre, Ottawa,
Canada and UK Aid, London, United Kingdom.

24

Electronic copy available at: [Link]

Di Falco, S., Veronesi, M., & Yesuf, M. (2011). Does adaptation to climate change provide food
security? A micro-perspective from Ethiopia. American Journal of Agricultural
Economics, 93(3), 829–846. doi:10.1093/ajae/aar006
Di Falco, S., Doku, A., & Mahajan, A. (2019). Peer effects and the choice of adaptation strategies.
Agricultural Economics, 51(1), 17–30. doi:10.1111/agec.12538
Diallo, A., Donkor, E., & Owusu, V. (2020). Climate change adaptation strategies, productivity
and sustainable food security in southern Mali. Climatic Change. doi:10.1007/s10584-020-
02684-8
Etwire, P.M. (2020). The impact of climate change on farming system selection in Ghana.
Agricultural Systems. 179, 102773. [Link]
Fagariba, C., Song, S., & Soule Baoro, S. (2018). Climate Change Adaptation Strategies and
Constraints in Northern Ghana: Evidence of Farmers in Sissala West District.
Sustainability, 10(5). doi:10.3390/su10051484
Ghana Statistical Service, GSS, (2013a). 2010 Population and housing census. Upper West
Regional Analytical Report. Accra, Ghana.
Ghana Statistical Service, GSS, (2013b). 2010 Population and housing census. Upper East
Regional Analytical Report. Accra, Ghana.
Ghana Statistical Service, GSS, (2013c). 2010 Population and housing census. Northern Regional
Analytical Report. Accra, Ghana.
Ghana Statistical Service, GSS, (2013d). 2010 Population and housing census. Brong Ahafo
Regional Analytical Report. Accra, Ghana.
Gollin, D. (2010). Chapter 73 Agricultural productivity and economic growth. Handbook of
agricultural economics, 3825–3866. doi:10.1016/s1574-0072(09)04073-0
Gornall, J., Betts, R., Burke, E., Clark, R., Camp, J., Willett, K., & Wiltshire, A.
(2010). Implications of climate change for agricultural productivity in the early twenty-
first century. Philosophical Transactions of the Royal Society B: Biological Sciences,
365(1554), 2973–2989. doi:10.1098/rstb.2010.0158
Guodaar, L., Bardsley, D. K., & Suh, J. (2021). Indigenous adaptation to climate change risks in
northern Ghana. Climatic Change, 166. doi:10.1007/s10584-021-03128-7

25

Electronic copy available at: [Link]

IPCC (2019). Summary for policymakers. In: Climate change and land: an IPCC special report
on climate change, desertification, land degradation, sustainable land management, food
security, and greenhouse gas fluxes in terrestrial ecosystems. Shukla et al., (eds.). In press.
Issahaku, G., & Abdulai, A. (2019). Adoption of climate-smart practices and its impact on farm
performance and risk exposure among smallholder farmers in Ghana. Australian Journal
of Agricultural and Resource Economics. doi:10.1111/1467-8489.12357
Limantol, A. M., Keith, B. E., Azabre, B. A., & Lennartz, B. (2016). Farmers’ perception and
adaptation practice to climate variability and change: a case study of the Vea catchment in
Ghana. SpringerPlus, 5(1). doi:10.1186/s40064-016-2433-9
Lobulu, J., Shimelis, H., Laing, M., & Mushongi, A. A. (2019). Maize production constraints,
traits preference and current Striga control options in western Tanzania: farmers’
consultation and implications for breeding. Acta Agriculturae Scandinavica, Section B —
Soil & Plant Science, 1–13. doi:10.1080/09064710.2019.1652680
Martey, E., Kuwornu, J. K. M., & Adjebeng-Danquah, J. (2019). Estimating the effect of mineral
fertilizer use on Land productivity and income: Evidence from Ghana. Land Use Policy,
85, 463–475. doi:10.1016/[Link].2019.04.027
Martey, E., Etwire, P. M., & Kuwornu, J. K. (2020). Economic impacts of smallholder farmers’
adoption of drought-tolerant maize varieties. Land Use Policy, 94, 104524.
Ministry of Food and Agriculture (2016). Agriculture in Ghana: Facts and figures (for 2015).
Accra, Ghana.
Ministry of Foreign Affairs of the Netherlands (2018). Climate change profile: Ghana. The Hague,
the Netherlands.
Ogada, M. J., Rao, E. J. O., Radeny, M., Recha, J. W., & Solomon, D. (2020). Climate-smart
agriculture, household income and asset accumulation among smallholder farmers in the
Nyando basin of Kenya. World Development Perspectives, 100203.
doi:10.1016/[Link].2020.100203
Ojo, T. O., & Baiyegunhi, L. J. S. (2019). Determinants of climate change adaptation strategies
and its impact on the net farm income of rice farmers in south-west Nigeria. Land Use
Policy, 103946. doi:10.1016/[Link].2019.04.007

26

Electronic copy available at: [Link]

Ojo, T. O., & Baiyegunhi, L. J. S. (2020). Impact of climate change adaptation strategies on rice
productivity in South-west, Nigeria: An endogeneity corrected stochastic frontier model.
Science of The Total Environment, 745, 141151. doi:10.1016/[Link].2020.141151
Stanturf, J. A., Warren, M. L., Charnley, S., Polasky, S. C., Goodrick, S. L., Armah, F., & Nyako,
Y. A. (2011). Ghana climate change vulnerability and adaptation assessment. United
States Agency for International Development. Ghana.
Wooldridge, J. M. (2002). Inverse probability weighted m-estimators for sample selection,
attrition, and stratification. The Institute for Fiscal Studies Department of Economics,
UCL Cemmap Working Paper CWP11/02
Wooldridge, J. M. (2015). Introductory econometrics: A modern approach. Nelson Education,
Toronto, Canada.
World Bank (2010). Economics of adaptation to climate change: Ghana. Washington DC, USA.
World Bank. (2021). Climate risk profile: Ghana. Washington DC, USA.

27

Electronic copy available at: [Link]

Appendix
Table A1: Conditional expectations, treatment and heterogeneity effects
Adaptation decision
Status Treatment effect
To adapt Not to adapt
Adapted 3a: 𝐸(𝑌1𝑖 |𝐶𝑖 = 1) 3c: 𝐸(𝑌2𝑖 |𝐶𝑖 = 1) 𝐴𝑇𝑇
Did not adapt 3d: 𝐸(𝑌1𝑖 |𝐶𝑖 = 0) 3b: 𝐸(𝑌2𝑖 |𝐶𝑖 = 0) 𝐴𝑇𝑈
Heterogeneity effects 𝐵𝐻1 𝐵𝐻2 𝑇𝐻
Notes: Whereas 3a and 3b are observed expected maize productivities, 3c and 3d are the counterfactual expected
maize productivities. 𝐶𝑖 = 1 if a farmer adapted to climate change, 𝐶𝑖 = 0 if a farmer did not adapt, 𝑌1𝑖 is the
productivity of farms that adapted, 𝑌2𝑖 is the productivity of farms that did not adapt, 𝐴𝑇𝑇 is the average adaptation
effect of farms that adapted, 𝐴𝑇𝑈 is the average adaptation effect of farms that did not adapt, 𝐵𝐻 is the base
heterogeneity effect while 𝑇𝐻 is the transitional heterogeneity.

28

Electronic copy available at: [Link]

Table A2: Falsification test of validity of instrument
Did not adapt
Variable
Log(maize Kg/Ha) Log(maize Kg/Ha) Asset index
Access to extension 0.078 0.006
0.056 0.094
Neighbour has access to extension 0.015
0.008
F statistic 1.94 3.74* 0.01
Sample size 794
Notes: * denote statistical significance at 10%. Values in italics are standard errors. We control for all the other
covariates.

29

Electronic copy available at: [Link]

Table A3: Estimates of the ESR model with different instrument
Main Alternate Main Alternate Main Alternate
Log(maize Kg/Ha)
Variable Adaptation decision
(1/0) Did not Did not
Adapted Adapted
adapt adapt
Log(Temperature) 2.28 1.28 3.562* 3.10* 0.23 0.22
1.86 1.88 1.86 1.85 1.33 1.31
Log(Precipitation) -3.53*** -3.66*** 4.55*** 4.25*** -0.69 -0.63
0.73 0.74 0.79 0.79 0.60 0.57
Log(Farm size) 0.13*** 0.13*** -0.15*** -0.15*** -0.18*** -0.18***
0.05 0.05 0.05 0.05 0.05 0.04
Log(Farming
-0.06 -0.05 -0.01 -0.02 0.04 0.04
experience)
0.05 0.05 0.05 0.05 0.04 0.04
Log(Household
0.06 0.05 -0.02 -0.02 0.08* 0.08
size)
0.06 0.06 0.06 0.06 0.05 0.05
Multiple economic
0.00 0.01 -0.07 -0.07 0.08 0.08
activities = yes
0.07 0.07 0.07 0.07 0.06 0.05
Male = yes 0.32*** 0.32*** -0.03 -0.02 0.26*** 0.25***
0.10 0.10 0.10 0.10 0.08 0.07
Formal education =
-0.05 -0.04 0.07 0.06 0.16** 0.16***
yes
0.08 0.08 0.08 0.08 0.06 0.06
Located in
-0.68*** -0.71*** -0.01 -0.01 -0.05 -0.04
Northern = yes
0.15 0.15 0.16 0.16 0.12 0.11
Located in North
-0.23 -0.16 -0.06 -0.08 -0.02 -0.02
East = yes
0.14 0.14 0.13 0.13 0.12 0.12
Located in
-0.13 -0.10 -0.99*** -0.98*** -0.38*** -0.38***
Savannah = yes
0.13 0.13 0.13 0.13 0.10 0.10
Located in Upper
-0.51*** -0.41*** -0.76*** -0.79*** -0.77*** -0.76***
East = yes
0.16 0.16 0.15 0.15 0.14 0.14
Located in Bono =
0.94*** 0.88*** -1.24*** -1.20*** 0.27** 0.26**
yes
0.15 0.15 0.16 0.16 0.14 0.13
Located in Bono
1.45*** 1.40*** -1.59*** -1.56*** -0.12 -0.14
East = yes
0.15 0.15 0.16 0.16 0.21 0.19

30

Electronic copy available at: [Link]

 Access to extension
0.15***
= yes
0.05
Neighbour has
0.04***
access to extension
0.01
Main Alternate Main Alternate Main Alternate
Status
To adapt Not to adapt Treatment effect
ATT = ATT =
Adapted 6.42 6.42 5.99 5.95
0.43*** 0.47***
0.02 0.02 0.01 0.01 0.02 0.02
ATU = ATU =
Did not adapt 8.02 7.99 6.11 6.11
1.90*** 1.88***
0.02 0.02 0.01 0.01 0.02 0.02
Heterogeneity
-1.60 -1.57 -0.13 -0.17
effects
0.03 0.03 0.02 0.02
Notes: *, ** and *** denote statistical significance at 1, 5 and 10%, respectively. Values in italics are standard errors.

31

Electronic copy available at: [Link]

Table A4: Mean expected maize productivity, treatment and heterogeneity effects by age group
Youth (≤35 years) Elderly (>35 years)
Adaptation decision Adaptation decision
Treatment Treatment
Status Not to Not to
To adapt effect To adapt effect
adapt adapt
ATT = ATT =
Adapted 6.568 5.856 6.356 6.010
0.712*** 0.347***
0.032 0.045 0.055 0.022 0.018 0.028
ATU = ATU =
Did not adapt 6.449 6.109 8.053 6.114
0.340*** 1.938***
0.034 0.027 0.043 0.024 0.015 0.028
Heterogeneity
0.119 -0.253 -1.697 -0.105
effects
0.047 0.050 0.034 0.023
Observations 411 980
Notes: We estimate a double-log model and rely on the Welch's t-test to examine if the treatment effects are statistically
significant. *, ** and *** denote statistical significance at 1, 5 and 10%, respectively. Values in italics are standard
errors.

32

Electronic copy available at: [Link]

Figure A1: Kernel density plot of maize productivity by pairs of covariates

33

Electronic copy available at: [Link]

Notes: The broken vertical lines represent mean value of the individual distributions. We use the same epanechnikov
bandwith value and axis rule for all the graphs in order to allow for easy cross comparison of the various kernel density
plots.

34

Electronic copy available at: [Link]

Figure A2: Kernel density plot of household asset index by pairs of covariates

35

Electronic copy available at: [Link]

Notes: The broken vertical lines represent mean value of the individual distributions. We use the same epanechnikov
bandwith value and axis rule for all the graphs in order to allow for easy cross comparison of the various kernel density
plots.

36

Electronic copy available at: [Link]