African Crop Science Journal, Vol. 20, Issue Supplement s2, pp. 349 - 360 ISSN 1021-9730/2012 $4.

00
Printed in Uganda. All rights reserved ©2012, African Crop Science Society

CLIMATE VARIABILITY AND YIELDS OF MAJOR STAPLE FOOD CROPS IN

NORTHERN GHANA

J. AMIKUZUNO and S.A. DONKOH

Department of Agricultural and Resource Economics, P. O. Box TL1882, University for Development
Studies, Tamale, Ghana
Corresponding author: amikj26@yahoo.com

ABSTRACT

Climate variability, the short-term fluctuations in average weather conditions and agriculture affect each other.
Climate variability affects the agroecological and growing conditions of crops and livestock, and is recently
believed to be the greatest impediment to the realisation of the first Millennium Development Goal of reducing
poverty and food insecurity in arid and semi-arid regions of developing countries. Conversely, agriculture is a
major contributor to climate variability and change by emitting greenhouse gases and reducing the agroecology’s
potential for carbon sequestration. What however, is the empirical evidence of this inter-dependence of climate
variability and agriculture in Sub-Sahara Africa? In this paper, we provide some insight into the long run relationship
between inter-annual variations in temperature and rainfall, and annual yields of the most important staple food
crops in Northern Ghana. Applying pooled panel data of rainfall, temperature and yields of the selected crops
from 1976 to 2010 to cointegration and Granger causality models, there is cogent evidence of cointegration
between seasonal, total rainfall and crop yields; and causality from rainfall to crop yields in the Sudano-Guinea
Savannah and Guinea Savannah zones of Northern Ghana. This suggests that inter-annual yields of the crops
have been influenced by the total amounts of rainfall in the planting season. Temperature variability over the
study period is however stationary, and is suspected to have minimal effect, if any, on crop yields. Overall, the
results confirm the appropriateness of our attempt in modelling long-term relationships between the climate and
crop yield variables.

Key Words: Causality, cointegration, limate variability, rainfall, yield

RÉSUMÉ

La variabilité du climat, les fluctuations à court terme des conditions météorologiques moyennes et l’agriculture
s’affectent mutuellement. La variabilité climatique affecte les conditions agroécologiques et les conditions de
croissance des cultures ainsi que l’élevage, et serait le principal obstacle à la réalisation du premier objectif
principal du ‘Millennium Development Goal’ pour la réduction de la pauvreté et l’insécurité alimentaire dans des
régions arides et semi arides des pays en développement. Inversement, l’agriculture contribue de façon importante
à la variabilité climatique et changement par l’émission des gaz à effet de serre et réduction du potentiel agroécologique
pour la séquestration du carbone. Ceci constitue une évidence empirique de l’interdépendance de la variabilité
climatique et l’agriculture en Afrique sub-saharienne. Cet article présente une relation à long terme entre les
variations interannuelles de la température et de la pluviométrie ainsi que les rendements annuels de la plupart des
cultures de base dans le Nord du Ghana. L’application d’un ensemble des données pluviométriques, température
et rendements des cultures sélectionnées depuis 1976 à 2012 pour la co-intégration et simulation par le modèle de
causalité Granger, li ya une forte évidence de la co-intégration entre les saisons, la pluviométrie totale et les
rendements des cultures, ainsi que la causalité de la pluviométrie aux rendements des cultures dans les savanes
soudano-Guinéennes et les savanes guinéennes du nord du Ghana. Ceci suggère les rendements interannuels des
cultures ont été influencé par les quantités totales des pluies au cours de la saison culturale. La variabilité de la
température pendant la période d’étude est par ailleurs stationnaire, et aurait un effet minimal ou non sur les
350 J. AMIKUZUNO and S.A. DONKOH
rendements des cultures. En somme, les résultats confirment la pertinence de cette tentative de modélisation des
relations à long terme entre le climat et les variables du rendement des cultures.

Mots Clés: Causalité, co-intégration, variabilité climatique, pluviométrie, rendement

INTRODUCTION already shows that in semi-arid regions of Africa,

rainfall distribution over the past few years is
Climate variability – the short term changes in low and highly variable spatially and inter-
the average weather patterns and agriculture temporally.
affect each other. On the one hand, climate Few empirical climate science studies in
variability affects agroecological, growing Ghana have revealed distinctive inter-annual and
conditions of crops and livestock. Climate inter-decadal variability in climatic variables like
variability and change are believed to be the precipitation and temperature. For instance, high
greatest impediment to the realisation of the first resolution regional climate simulation using
Millennium Development Goal of reducing explicit dynamic downscaling of global climate
poverty and food insecurity globally, via scenario indicates a slight increase in total annual
increased agricultural production in developing precipitation of 5%, but also a significant
countries. decrease (up to 70%) of precipitation in April, the
Climate variability results from our use of traditional month for land preparation and
energy, but its impact is manifested through commencement of crop productions in Northern
changes in agroecological conditions and climatic Ghana. The total duration of the rainy season is
factors, particularly precipitation and temperature. also revealed to have shortened, while Dry
Precipitation, especially rainfall and atmospheric Season and Rainy Season temperature increased
temperature, the most important weather around 1°C and 2°C respectively (Kunstmann and
variables affected by climate change, play a Jung, 2005).
crucial role in agricultural production in Sub- High climatic variability (e.g. low and variable
Sahara Africa (SSA). Having no glaciers to melt, distribution of rainfall) represents a delicate
it is believed that the most profound and direct balance between agricultural production and
impacts of climate variability over the next few food security. Studies show that the changes in
decades in Africa will be on agriculture the agriculturally-relevant variables of climate
(Kurukulasuriya and Mendelsohn, 2008). This is (e.g. increasing temperatures and declining levels
because, unlike in the industrialised regions of and distribution of rainfall), are likely to reduce
the world, a majority of SSA’s largely smallholder yields of maize, rice, wheat and other food crops
farmers lack yield-improving techniques like in semi-arid regions of the world (Lobell et al.,
greenhouse and irrigation facilities for crop 2009). The amount and distribution of rainfall
production. In addition, the usually self-bred crop affects many other aspects of agricultural
varieties of SSA’s subsistent farmers have a more production among smallholder farmers in SSA,
limited genetic diversity and resilience to namely farm sizes, crop enterprises, cropping
unfavourable weather conditions than those of calendars, incidence and growth of weeds, crop
developed countries where crops are virtually pests and diseases (Yengoh et al., 2010).
bred to withstand changes in climatic factors. In Ghana, the variability of rainfall is a threat
Although SSA, like other developing regions to the livelihood of smallholder farmers. Over the
of the world, contributes the least to climate past few years, rainfall-related crop failure due to
change and variability, it is known that semi-arid episodes of late rains for planting, variability in
countries in SSA will be the most affected by the the pattern and levels of rainfall, and intermittent
phenomenon in terms of agricultural production. droughts and floods in Northern Ghana are
Africa is expected to be the continent where the common phenomena. Thus climate variability
impact of global warming on agriculturally- entails risks in Ghana, especially in the dryer
relevant climatic variables and production will be Northern part of Ghana.
greatest (IWRM, 2001). Prevailing evidence
 Climate variability and yields of major staple food crops 351
The objective of this study is to assess contributes about 70% to livelihood in Northern
quantitatively how climate variability affects Ghana.
yields (ha-1 per annum) of major staple crops, viz.
millet, maize, sorghum and rice (cereal and C4 Study area and data description. The northern,
plants), and groundnuts and yam (C3 legume and Savannah region of Ghana lies between latitudes
tuber crops respectively) in Northern Ghana. The 8°N and 11°N and has a land area of 97702 km².
motivation is to contribute to the knowledge Northern Ghana’s rainfall is characterised by a
dearth on the effects of changes in temperature long dry period of about seven months from
and rainfall in the planting season on the October/November to April/May called the Dry
production of major staple crops in Northern Season with no appreciable rainfall level, and a
Ghana. To the best of our knowledge, there has Rainy Season from May to October with adequate
not been any study on climate variability impacts levels of rainfall for agricultural production.
on crop yields in Northern Ghana. Describing the Northern Ghana, comprising the Northern,
consequences of climate variability on the yield Upper East and Upper West regions is located in
of major food crops in Northern Ghana will help the sub-humid to semi-arid Guinea Savannah and
institute policy measures, for instance under arid Sudan Savannah zones of Ghana, where the
Ghana’s Government Savannah Accelerated annual precipitation ranges between 400 and 1200
Development Authority (SADA), to improve mm, and agriculture, which employs about 70%
upon the resilience of subsistence agriculture to of the population, is around 95% purely rain-fed.
covariate, weather shocks, and in predicting long- Since colonial times, Northern Ghana has been
run economic impacts of climate change on the poorest part of the country because of its
smallholder agriculture. This is particularly vulnerability to climate change and precarious
important because smallholder agriculture climatic conditions like the long dry season of

Source: Yengoh et al., 2010.

Figure 1. The savannah agro-ecological zones of Northern Ghana.

352 J. AMIKUZUNO and S.A. DONKOH

about seven months followed by just a five-month the selected crops cultivated within the Rainy
rainy season with recurrent, intermittent droughts Season. This is because there is only one crop
and/or floods in the planting season. production season in Northern Ghana from April/
Apart from land degradation and May to October/November, hence temperature,
intensification of agriculture which lead to rainfall and yield levels measured only within this
declining farm sizes, continuous cropping and period are useful here.
unsustainable farming practices also result in
declining output from crop and livestock systems METHODOLOGY
in this part of Ghana. Analyses of climate patterns
in the last 60 years have shown that precipitation Many models exist in the climate impact
has slightly decreased, while temperatures and assessment literature. Some of these include
evapotranspiration have increased. In addition, computable general or partial equilibrium models,
the onset of the rainy season has often delayed; production functions, and Ricardian-type
staring in June instead of April (Kranjac- statistical or process-based models, which are
Brisavljevic et al., 1999). modified and applied in analysing the impact of
Northern Ghana is chosen as the study area changes in precipitation, temperature and carbon
because it is sensitive to climate variability on agricultural production. A recurring concern
especially erratic rainfall. The region is also the however in using these models to estimate
most agricultural part of Ghana with a high potential impacts of climate change on agriculture
percentage of subsistent, food crop farmers, and is whether or not they are able to capture real
rural poverty incidence of about 52%. Many climate variability scenarios among smallholder
manifestations of climatic instability in the last crop farmers in rain-fed farming systems in dry
two decades are: floods causing extensive gully lands.
erosion and devastating large areas of farmlands To address this concern, we move beyond
and crops; several cases of late rains before; and the usual intergovernmental panel on climate
persistent droughts during the planting season. change (IPCC) modelling techniques of climate
Irrigated agriculture is seen as the most important change impact analysis to using time series
means for coping with the effects of climate models to simulate the long run and causality
change and variability in Ghana, but just about relationships between rainfall and crop yield
3% of Ghana’s farmers practise irrigated variables. We use two time series modelling
agricultural (Kurukulasuriya and Mendelsohn, techniques - cointegration and Granger causality
2008). to simulate the impact of changes in the total
The data used for the study includes levels of seasonal rainfall and average, seasonal
temperature and rainfall time series from 1976 – minimum and maximum temperature on average,
2010 collected from the Ghana Meteorological annual yields of the selected major staple crops
Services (GMS), and crop yield (MT Ha-1) data of Northern Ghana. The models applied are in the
from 1991 – 2010 for maize, millet, rice, sorghum, class of statistical crop modelling approaches
groundnuts and yam, obtained from Ghana’s used in estimating yield sensitivity to variations
Ministry of Food and Agriculture (MoFA). These in rainfall and temperature.
datasets were collected for six climatically-unique Cointegration analysis help determine the
geographical locations viz. Bawku, Bolgatanga long run, dynamic relationship between two or
and Navrongo in the arid Sudano-Guinean more variables. Johansen’s (1991) Variance
Savannah (SGS) zone with rainfall ranging from Autoregressive (VAR) approach is used to
400-1000 mm in the Upper East Region (UER); simulate the dynamic relationship between crop
and Bole, Tamale, and Yendi in the sub-humid/ yields and average annual total rainfall in the
semi-arid Guinea Savannah (GS) zone with an selected locations. Temperature was found to be
annual rainfall of about 1200mm in the Northern fairly stable over the period of the study and thus
Region (NR). The rainfall and temperature figures dropped from the impact analysis
c
are those collected within the Rainy Season. Let the yield per Ha of a given crop be Qt and
Likewise, the yields are average yields per Ha of the seasonal average level of the climate variable
 Climate variability and yields of major staple food crops 353
(e.g. seasonal total rainfall or average in the selected locations. In addition, we estimate
t
temperature) be Wi (i = precipitation or and test for significance of the long run
cointegration coefficient β which indicates the
c
temperature) with t denoting time. If Qt and
t
Wi are time series variables containing sensitivity of yield to variability in rainfall.
stochastic trends, then they are said to be The existence of cointegration between
cointegrated if there is an equilibrium, long run Qtc and Wi t implies Granger causality (Granger,
relationship between them (Prakash, 1997). 1987). Furthermore, since cointegration between
c t
Alternatively, Qt and Wi are cointegrated if β variables does not automatically imply causality
c t
is such that Qt + β Wi β†I (0). between them, the evidence of causality between
Johansen’s approach can be applied to test the variables must be provided by Granger
for the existence of a cointegration relationship causality analysis. The Granger causality models
c t
between Qt and Wi as follows: applied in this study are specified in equations
(4) and (5).
yt = µ + A1yt-1 + A2yt-2 +... + Apyt-p +εt (1)
c t
Where yt = ( Qt and Wi ) is an n x 1 vector of .............. (4)
c t
variables ( Qt and Wi ) and is assumed to follow
a VAR process, µ is a vector of intercept terms,
yt-i are the lagged values of yt, Ai are vector ............. (5)
coefficient matrices and εi is an n x 1 vector of
error terms. The VAR process in (1) is rewritten
c
as: The equation (4) postulates that, Qt is
dependent on Qtc−1 and Wt −c 1 ; conversely,
........... (2)
equation (5) postulates the same (theoretically)
for Wi t . The ε1t and ε 2t are uncorrelated error
is the coefficient matrix
terms. A proof of causality between rainfall and
denoted as is a yield implies that the variability of yield overtime
can be explained by varying episodes of rainfall
matrix of short run coefficients and temperature over the same period of time.
The variant of the Granger causality tests run in
The test for the existence of a cointegration this paper is based on the vector error correction
c t
relationship between Qt and Wi is based on model (VECM).
two likelihood ratio tests – the trace and maximum The data for the analysis is a pooled panel
eigen value tests specified as follows: data of temperature and rainfall figures of the
GMS from 1976 to 2010, and crop yields data of
MoFA from 1991 to 2010. Since the data used is
time series, we performed unit root tests using
the augmented Dickey-Fuller (ADF) and
respectively Kwiatkowski–Phillips–Schmidt–Shin (KPSS)
approaches to determine the stationarity
............................................................................ (3) properties of the series. Testing the null
hypothesis ( H 0 ) that the individual series
Where T is the sample size and λi is the ith largest contain unit root with a drift and trend but without
canonical correlation. In this study we run the structural breaks, the results revealed that all the
trace test which evaluates the null hypothesis series are non-stationary and integrated of the
( H 0 ) that there are at least cointegrating order 1. The analysis is done in the natural
relationships between the rainfall/yield variables logarithm values of the variables.
354 J. AMIKUZUNO and S.A. DONKOH

RESULTS AND DISCUSSION effects on crop yields in the arid and semi-arid
zones of Ghana. In Figure 2, we illustrate the
This section reports the results of the analysis. pattern of rainfall in four locations within the study
First graphical plots of the variability and patterns area, namely Navrongo and Bawku in the arid
in total rainfall and seasonal average minimum SGS zone, and Tamale and Bole in the semi-arid
and maximum temperature within the farming GS zone.
season, and yields (mt ha-1) of the selected major The seasonal variability in the pattern of
staple crops cultivated in the Rainy Season at rainfall over the study period is quite stark,
the six locations within both agroecological zones exhibiting a near-cyclical pattern with the rainfall
of Northern Ghana are presented. levels alternating quite regularly between peaks
and troughs above the mean rainfall level of
Rainfall and temperature variability in Northern 958.84mm almost triennially. Within the period of
Ghana. Variability in the level and distribution of the analysis, several episodes of rainfall levels
rainfall is the most important determinant of crop as low as about 600 mm or up to 1800 mm can be
yields in smallholder, resource-poor farming seen. These represent periods of droughts and
systems in arid and semi-arid areas, where farmers floods, hence risk for crop yields. Furthermore,
often lack sufficient yield improving inputs or the overall trend in rainfall, as demonstrated by
technology. Thus, analysing the rainfall regime the trend line appears to have slightly increased
is a prerequisite for examining climate variability over the whole period of study.

Source: Own Plots from GMS Data.

Figure 2. Seasonal patterns of rainfall in Northern Ghana (1976-2010).
 Climate variability and yields of major staple food crops 355
The total mean of rainfall per Rainy Season fluctuations above mean minimum or maximum
from 1976 to 2010 is 958.84mm, and the cumulative temperature. The strong 1997/1998 El Nino effect
change in the seasonal total rainfall over the caused just a slight increase in average minimum
period is 11.41mm.This supports Boko (2007) and and maximum temperatures, but this increase was
Nicholson (2005) who recorded slight increases not significant enough to affect the long term
in rainfall over recent years following declining trend in the observed temperatures in the area.
trends and frequent dry spells between the 1960s As expected, the average minimum and maximum
and 1980s in Ghana. It is also evident that climate temperatures of the SGS zone (UER) are about
change does cause changes in the rainfall pattern 0.9oC and 1.54oC respectively higher than the
of Northern Ghana. Even though the study did corresponding values in the GS zone (NR).
not record the number of rainy days, these, This finding suggests that Northern Ghana
together with lengths of planting periods, are has not experienced notable temperature
believed to be declining over time (Yengoh et al., increases over the study period. The absence of
2010). The question here is whether or not these significant varying temperature trends or
variations in the levels and distribution of rainfall fluctuations above the mean minimum or maximum
affect the yield performance of crops in the study temperatures implies that no fundamental causal
area. relationship should be expected between
Threshold levels of environmental temperature variability and yields of the selected
temperature also affect crop production activities staple crops in Northern Ghana. For this reason,
and yields in semi-arid and arid areas (Kunstmann we dropped the temperature variable from the
and Jung, 2005; Kurukulasuriya et al., 2006). cointegration and causality analysis.
There is a suitable temperature range for optimal
yields in every crop. Therefore, examining the Crop yield variability in Northern Ghana. In
variability of temperature in climate-related Figures 4 and 5, we plot the yield levels and
studies in crop production is necessary. The variability of four (4) major staple crops viz. maize,
study examined the pattern of seasonal, average rice, sorghum and groundnut for each zone. Like
minimum and maximum temperatures from 1991 the rainfall series, the graphical analysis of the
to 2010 in the study area. The temperature curve yield data shows high and non-stationary
is presented in Figure 3. variability in the yields of the selected crops over
The figure shows a stationary variation in the period of study.
temperature over the study period. Unlike the First, the average yields of the crops in the
rainfall series, there are no explosive temperature arid, SGS zone (Fig. 4) show an expected inter-

Figure 3. Average minimum and maximum production season temperatures in the two agro ecological zones (OC, 1987 - 2000).
356 J. AMIKUZUNO and S.A. DONKOH
Yield (mt ha-1)

Figure 4. Yield variability of selelcted staple crops in the dry Sudano-Guinea agro ecological zone (Mt ha-1, 1991 - 2010).

Source: Own Plots from MoFA Data.

Figure 5. Yield variability of selected staple crops in the semi-arid Guinea savannah agroecological zone (mt ha-1, 1991 - 2010).

annual variability. It can be seen that whereas relationship, the yield of rice, partly rain-fed
the yields of the indigenous crops - maize, (upland rice) and partly irrigated (lowland rice),
sorghum and groundnut - which are purely rain- tends to largely lead the yields of the indigenous
fed appear much related (turning upward and crops over the entire period of the analysis. This
downward together) over time and giving an is expected because irrigation is available for some
insight of the existence of a long run equilibrium rice farms in the SGS zone to safeguard yield
 Climate variability and yields of major staple food crops 357
losses during dry spells in the planting season. semi-arid zones of Northern Ghana, where farmers
In addition, high yielding varieties of rice have lack insufficient inputs and improved technology.
been consistently released by different rice It appears that as rainfall increased slightly over
improvement research programmes within the the last few years, yield levels and variability rose
period of study. as well. Further analysis is however required to
The yield patterns of the crops in the semi- reveal the nature of the empirical linkage between
arid GS zone are similar to that in the arid zone rainfall and the yield performance of the crops.
(Fig. 5). This is the tasks of the next section.
Even though the output per Ha of groundnut
in the first decade of the analysis lags behind Testing cointegration and causality between crop
that of the other two indigenous crops, the yield yields and rainfall in Northern Ghana. Tables
levels of the three purely indigenuous crops 1and 2 contain the three different categories of
(maize, sorghum and groundnuts) appear again estimation results of the cointegration and
to co-move over the study period, while that of causality analyses. These are: 1) results of the
rice, also partly rain-fed and irrigated in this zone, test for cointegration relationships between
again leads the other yield curves over the entire rainfall and crop yields in the chosen locations,
period. 2) estimated elasticity coefficients of
In both figures, yield levels for all the crops cointegration obtained from a LS estimation of
appear to be rising in the last five years. This the long-run cointegration equation
suggests that agro-climatic conditions, especially Q tc + β W t i I ( 0 ) , and 3) uni-directional
rainfall, in the last few years are yield-improving. Granger causality test statistics of the causal
As noted above, favourable weather, especially relationship between rainfall and crop yields. The
rainfall is the predominant underlying factor Akaike Information Criteria (AIC) recommended
affecting the output of most crops in the arid and lag values of the cointegartion test are also

TABLE 1. Cointegration and causality tests statistics and cointegration elasticity coefficients in the Sudano-Guinean savannah zone

Est. Equation Coint. Test Statistics No. of Lags Normalised Coint. Granger causality
(rainfall -) Coefficients (β)
r0 = 0 r1 = 1

Bawku
Maize 27.53** 8.18 4 -0.954*** 7.033***

Millet 31.79 8.05 4 -

Sorghum 27.39** 10.65 4 1.271*** 5.539**
Groundnuts 26.52** 10.3 4 1.476 3.193*

Bolgatanga
Maize 26.04** 8.36 4 0.115** 1.592
Millet 13.21 3.51 4 - -
Sorghum 23.9 4.91 4 -
Groundnuts 25.94** 10.13 4 0.519*** 3.959**

Navrongo
Maize 24.74* 9.85 3 -1.710*** 12.472***
Millet 15.76 3.29 4 - -
Sorghum 26.99** 8.26 4 -0.701 0.85
Groundnuts 26.62** 9.35 3 -3.077*** 4.901**

The asterisk *, ** and *** indicate significance at the at the 10%, 5% and 1% levels, respectively

Source: Own Estimation

358 J. AMIKUZUNO and S.A. DONKOH
TABLE 2. Cointegration and causality tests statistics and cointegration elasticity coefficients in the Guinea savannah zone

Est. Equation Coint. Test Statistics No. of Lags Normalised Coint. Causality
(rainfall -) Coefficients (β) (Test Stat.)
r0 = 0 r1 = 1

Bole
Maize 35.52*** 8.98 4 0.541*** 19.212***
Rice 18.03* 5.07 4 0.318*** 3.758*

Groundnuts 28.09** 6.99 4 -0.135 1.8626

Yam 27.87** 8.16 2 0.06 0.944

Tamale
Maize 26.64** 6.28 4 -0.213* 1.121
Rice 19 4.37 4 - -
Groundnuts 38.46*** 5.66 4 -0.212*** 0.048
Yam 18.33 2.36 4 - -

Yendi
Maize 35.54*** 13.46** 4 -0.942*** 37.865***
Rice 34.99** 11.66* -0.569*** 21.260***
Groundnuts 25.04* 7.87 -0.048 2.21
Yam 40.7 3.3 2 -0.299*** 3.450**

The asterisk *, ** and *** indicate significance at the at the 10%, 5% and 1% levels respectively
Source: Own Estimation

reported. It is recognised that this lag selection period and the corresponding yields of maize, β = 1.7
1.27an
technique is suited for the nonlinear models we rice and sorghum over the period of the analysis.
estimated and the nonlinear Data Generation Millet yield in the long-run however appears to
Process (DGP) under study. be insensitive to rainfall. This segmentation
Table 1 presents the etimated results for the between rainfall and millet yield may be because
staple crops of the SGS zone - millet, maize, rice millet is drought-tolerant and as such, low levels
and groundnuts. The test statistics of of precipitation are not necessarily yield-
cointegration, signifying the existence and nature reducing.
of the long run relationships between rainfall and The cointegration elasticity coefficients,
crop yields (presented in the column 2), provide estimated for only the cases with significant
evidence in favour of cointegration between cointegration relationships, are the long run
rainfall and crop yields for all cases under the responsiveness of the crop yields to rainfall
three locations except for late millet in all the variability. Note that the rainfall and yield
locations and Sorghum in Bolgatanga. variables are estimated in their logarithm values.
The H 0 of r = 0 , tested at the 1%, 5% and 10% These are significant in all the included cases
levels, and implying an absence of a cointegration except for groundnut under Bawku and Sorghum
relationship between rainfall and yields of the under Navrongo. There is especially a high
given crops is rejected for all the cases except sensitivity of the yields of sorghum in Bawku
those mentioned above. We cannot however ( ), and of maize and groundnuts
reject the null hypothesis of one cointegrating ( respectively) in
relation, i.e. r = 1 in all cases. Navrongo, with estimated elasticity coefficients
The findings suggest that, there exists at least above 1. The estimated elasticity of 1.27 for
one equilibrium cointegration relationship ((r = sorghum yield in Bawku implies that a 1% increase
1) between the amount of rainfall in the planting in rainfall (within the suitable range) will cause a
 Climate variability and yields of major staple food crops 359
1.27% upward shock in yield. Positive and responsive to rainfall in the GS zone (all elasticity
negative signs of the elasticity coefficients are coefficients < 1) with more negative sign than is
expected since only rainfall within the suitable the case under the GSG zone. It is expected that
range is yield-improving, while both flood and yield sensitivity to rainfall in the semi-arid zone
drought events, ceteris paribus, exert downward where latent soil moisture content is higher should
shocks on crop yield. be less than that of crops in arid areas with lower
Having determined the cointegration soil moisture content. The negative signs of the
relationships and elasticity coefficients, we elasticity coefficients is an insight to the effect
performed the Granger causality test to check of drought, but particularly floods on crop yields.
whether significant causal, equilibrium This is because, having a higher level of soil
relationships from rainfall to the crop yields in moisture content and subsoil hard pan, the effects
the chosen locations exist. The test is of extremely high rainfall is more yield-reducing
unidirectional; it only examines shocks of rainfall here than will be the case in the SGS zone with
variability on crop yields and not vice versa; arid conditions, low hard pan and well drained
though doing the reverse is theoretically possible, farmlands.
it is practically only artificial. The is that The causality test statistics are significant in
“there is no causality from rainfall to crop yields five out of the ten cases analysed. There is
in the SGS zone”. The results are presented in particularly a strong causation of yields of maize
the last column of Table 1. and rice in Bole and Yendi, while no empirical
For the vast majority of the cases tested, we evidence of causality exists from rainfall to the
establish the existence of causality by rejecting yields of the all crops in Tamale. This suggests
H 0 . Thus, for most of the staple crops cultivated that rainfall does not have a long-term causal
in the zone, there is a significant unidirectional impact on the yield of staple crops in Tamale
causal relationship from rainfall to their yields .Tamale is a big meteropolitan city and the yields
(ha-1 per annum). Therefore, recorded planting of even smallholders here unlike in the rural
season rainfall levels can be used to predict the locations may not be solely climate-forcing i.e.
H0 yields of the selected crops. Rainfall especially determined solely by rainfall, but also by the use
has a strong predictive power for maize and of yield-improving external inputs, seed and
sorghum in Bawku, groundnuts in Bolgatanga, technology.
and maize and groundnuts in Navrongo where
the estimated test statistics are high and highly CONCLUSION AND RECOMMENDATION
significant. The independencies of the yields of
maize in Bolgatanga and sorghum in Navrongo The objective of this paper was to examine long-
cannot be empirically explained. term cointegration and causality between rainfall
The results of the SG zone are presented in and yields of the most important staple crops in
Table 2. Similar to those of the SGS zone, we find two distinct agroecological zones in Northern
strong evidence of cointegration between rainfall Ghana. The analysis was conducted using pooled
and yields of the four major staple crops – maize, panel data of total rainfall and annual yields of
rice, yam and groundnuts. The cases in which the selected crops over the period 1991 to 2010.
we cannot establish cointegration are rice and The data was applied to the Johansen VAR and
yam in Tamale and Yendi respectively. The Granger causality techniques.
indigenous crops – maize, yam and groundnuts High seasonal variability in rainfall with
particularly show strong evidence of cyclical fluctuations above the mean of about
cointegration with rainfall. This confirms the 959 mm, coupled with several episodes of floods
results of the graphical analysis in figure 5 where and droughts over the study period, were
the yields of these crops are shown to drift revealed. The cumulative average increase in
together over the period of the analysis. rainfall from 1976 to 2010 was 11 mm. In contrast,
The estimated elasticity coefficients are no notable variability in minimum and maximum
significant in all cases except three. In all temperature was seen. In the core analysis, we
significant cases, crop yields appear to be less found evidence of cointegration and causality
360 J. AMIKUZUNO and S.A. DONKOH

between total rainfall and crop yields in the spectral methods. Econometrica 37 (3):424–
selected locations, meaning yields of the selected 438.
staple crops have been influenced by the total Johansen, S. 1991. Estimation and hypothesis
mounts of rainfall in the planting season. Overall, testing of cointegration vectors in Gaussian
the results confirm the appropriateness of the Vector Autoregressive Models.
attempt in modelling long-term relationships Econometrica 59(6,):1551–1580.
between the climate and crop yield variables. Kasei, G.N. and Afuakwa, J.J. 1991. Determination
The paper’s strengths lie in being able to use of optimum planting date and growing season
the available data to provide evidence of the long of aize in the northern savanna zone of Ghana.
run and causal link of crop yields to variability in Soil Water Balance in the Sudano-Sdhelian
rainfall by running Granger causality tests based Zone. Proceedings of the Niamey Workshop,
on a VECM, having established long run February 1991.
equilibrium relationships and estimated elasticity Kranjac-Berisavljevic’, G., Blench, R.M.,
coefficients between the variables. In this way, Bayorbor, T. B.,Turton, C. N., Abdulai, A. S.,
the paper links long-term causality and Boyd, C., Obeng, F. and Drake, E. 1999.
cointegration more logically than previous Rethinking natural resource degradation in
studies in Africa. The core limitation of this paper semi-arid sub-Saharan Africa: The case of
is the unavailability of long time series data. semi-arid Ghana. London, Overseas
Ideally much longer time series is required in Development Institute”, http://www. odi. org.
capturing seasonal trends and ensuring both the uk/ rpeg/ soil_ degradation/ghlit.pdf.
consistency and efficiency of the estimated Kunstmann, H. and Jung, G. 2005. Impact of
parameters. However, a longer complete series regional climate change on water availability
for both the weather and agricultural data do not in the Volta Basin of West Africa. pp 75-85.
exist at the moment in Northern Ghana. IAHS Publication No. 295. Regional
An interesting extension of this paper in future Hydrological Impact of Climate variability and
research would be to broaden the scope of the Change.
study to encompass other possible climate Kurukulasuriya, P. and Mendelsohn, R. 2008. A
variables such as CO2, and their interactions with Ricardian analysis of the impact of climate
other environmental variables (e.g. nutrient change on African cropland. African Journal
availability, farmer resource endowments etc). of Agricultural and Resource Economics 2(1).
Policy measures to counteract yield variability Lobell, D.B., Cassman, K.G. and Field, C.B. 2009.
such as encouraging flexible land use and crop Crop yield gaps: Their importance,
insurance should be implemented to protect magnitudes, and causes. Annual Review
farmers from exposure to increasingly climate- Environmental Resource 34:179–204.
related risks. Nicholson, S. 2005. On the Question of the
“Recovery” of the rains in the West African
REFERENCES Sahel. Journal of Arid Environments, Special
Issue on the “Greening” of the Sahel 63(3):
Boko, M., Niang, I., Nyong, A., Githeko, C.A., 615-641.
Medany, M., Osman-Elasha, B., Tabo, R. and Prakash, G. 1996. Pace of market integration. PhD
Yanda, P. 2007. Climate change 2007: Impacts, Thesis, Northwestern University, September
adaptation and vulnerability, contribution of 1996.
Working Group II. The Fourth Assessment Yengoh, G. T., Armah, F.A., Onumah, E.E. and
Report of the Intergovernmental Panel on Odoi, J.O. 2010. Trends in agriculturally-
Climate Change. relevant rainfall characteristics for small-scale
Granger, C.W.J. 1969. Investigating causal agriculture in Northern Ghana. Journal of
relations by econometric models and cross- Agricultural Science 2(3).