EFFECT OF CLIMATE CHANGE ON MAIZE SUPPLY IN 
GHANA, 1970-2002
BY
AMA ASANTEWAH AHENE
THIS THESIS IS SUBMITTED TO THE UNIVERSITY OF GHANA, LEGON, IN 
PARTIAL FULFILMENT OF THE REQUIREMENT FOR THE AWARD OF M.Phil
DEGREE IN ECONOMICS.
AUGUST 2003
This work is dedicated to my mother, Akosua Otubea (O.T).
DEDICATION
1
-O ' 3 7 4 1 3 8
DECLARATION
I, Ama Asantewah Ahene, author o f this MPhil thesis do hereby declare that the work 
presented in this thesis titled, “Effect o f  Climate Change on Maize Supply in Ghana, 1970- 
2002” was done entirely by me in the Department o f Economics, University o f  Ghana, Legon 
from October 2002 to August 2003.
This work has never been presented either in whole or in part for any other degree in this 
University or elsewhere.
Ama Asantewah Ahene 
(Student)
This thesis has been presented for examination with our appro v?15,5 a™ ™ '™
K. Yerfi Fosu
(Major Supervisor)
Kwame Baah-Nuakoh 
(Co-Supervisor)
ACKNOWLEDGEMENTS
I wish to express my profound thanks to my Major Supervisor, K. Yerfi Fosu, for supervising 
this work. His advice and encouragement made this work possible. I also thank Kwame 
Baah-Nuakoh for his co-supervision.
I also wish to thank Professor Amoah Baah-Nuakoh, Mr. G.K. Tsikata and Dr. Barfour Osei 
for their encouragement and advice during my entire stay at the Department of Economics.
My family has been a real rock o f support. God bless you all. My sincere thanks also go to 
Nana Yaw A. Achiampong -  you were very supportive of me and my work. I am grateful to 
Emmanuel A. Codjoe, Donatus Ayitey and Christian Ahortor for their help.
All Power, Glory and Honour belong to God. Thank you Jesus.
ABSTRACT
Societies which depend upon agriculture for their livelihood may face the perils o f climate 
change (Global Wanning), especially developing countries in the tropics and subtropics 
where some crops are already near their maximum temperature tolerance and where dry land 
and non-irrigated agriculture predominate. Yields will tend to decrease with even nominal 
amounts o f climate change. Agriculture in Ghana is largely rain fed, therefore changes in 
climate in the form of changes in rainfall for example, may affect agricultural supply 
especially crops, maize inclusive. This study examines the effects of climate change, using 
rainfall and temperature as climate variables, on maize supply in Ghana. Annual time series 
data on temperature, rainfall, and output of maize, prices o f  maize, rice, cassava and fertiliser 
covering the period 1970-2002 are used. The error correction model and Granger Causality 
Test were employed in the analysis. The results o f the study show that climate change in the 
form of change in rainfall exerts a statistically significant effect on supply o f maize. The 
study further confirms that rainfall is essential during crop growth, and the critical major 
period o f rainfall is the best time to plant to avoid water stress and pest infestation. The study 
closes with some policy recommendations and suggestions for future research.
IV
TABLE OF CONTENTS
CONTENT Page
DEDICATION i
DECLARATION »
ACKNOWLEDGEMENTS iii
ABSTRACT iv
TABLE OF CONTENTS v
LIST OF TABLES vii
LIST OF FIGURES viii
LIST OF ABBREVIATIONS ix
CHAPTER 1. INTRODUCTION 1
1.1. Background and Problem Statement 1
1.2. Objectives of the Study 5
1.3. Relevance o f the Study 5
1.4. Organisation o f the Study 6
CHAPTER 2. LITERATURE REVIEW 7
2.1. General Effects o f Climate Change 8
2.2. Future Effects of Climate Change 10
2.3. Effects of Climate Change on Agriculture 11
2.4. General Empirical Literature 15
2.5. Literature on Ghana 20
V
3.1. Time Profiles o f Temperature and Rainfall 25
3.2. Time Profile o f Maize Supply 25
3.3. Theoretical Framework 26
3.4. Statement o f Hypotheses 28
3.5. Empirical Model 28
3.6. Method o f Estimation 29
3.7. Description of Variables and Sources of Data 33
CHAPTER 4. EMPIRICAL RESULTS 35
4.1. Time Profile of Annual Temperature 35
4.2. Time Profile o f Annual Rainfall 37
4.3. Time Profile o f Maize Supply 39
4.4. Descriptive Statistics of the Variables in the Econometric Model 42
4.5. Empirical Results o f Stationarity Tests 47
4.6. Empirical Results of Cointegration Analysis 50
4.7. Results o f Error Correction Modelling 58
4.8. Results o f Granger-Causality Tests 66
CHAPTERS. SUMMARY AND RECOMMENDATIONS 67
5.1. Summary and Policy Recommendations 67
5.2. Limitations o f the Study and Suggestions for Future Research 68
APPENDICES 70
LIST OF REFERENCES 80
CHAPTER 3. METHODOLOGY 25
vi
Page
45
48
49
51
53
55
56
58
60
62
63
64
65
66
LIST OF TABLES
Descriptive Statistics of the Relevant Variables: Raw Data, 1970-2002
Results o f Unit-Root Test (Augmented-Dickey Fuller) for 1970-2002
Results o f Unit-Root Test (Phillip-Peron) for 1970-2002
Cointegrating Equation: Model 1
Cointegrating Equation: Model 2
Cointegrating Equation: Model 3
Cointegrating Equation: Model 4
Cointegrating Equation: Model 5
Error Correction Model 1
Error Correction Model 2
Error Correction Model 3
Error Correction Model 4
Error Correction Model 5
Results o f the Granger Causality Tests between the Supply of Maize, 
Rainfall and Temperature
vii
LIST OF FIGURES
Figure Page
Figure 1. Annual Mean Temperature in Maize Growing Areas in Ghana, 1970-2002 36
Figure 2. Annual Mean Deviations in Temperature in Ghana, 1970-2002 37
Figure 3. Annual Rainfall in Maize Growing Areas in Ghana, 1970-2002 38
Figure 4. Annual Rainfall Deviations from the Mean, 1970-2002 39
Figure 5. Comparison o f Maize Supply with Annual Rainfall, 1970-2002 40
Figure 6. Comparison o f Maize Supply with the Price of Rice, 1970-2002 41
Figure 7. Comparison o f Maize Supply with Price o f Cassava, 1970-2002 42
viii
LIST OF ABBREVIATIONS
ADF Augmented Dickey-Fuller
AIC Akaike Information Criterion
CERES Crop Environment Resource Synthesis
CFC Chlorofluorocarbon
ECM Error Correction Model
ECT Error Correction Term
EPA Environmental Protection Agency
FAO United Nations Food and Agriculture Organization
GDP Gross Domestic Product
GHG Greenhouse Gas
GNP Gross National Product
IEA International Energy Agency
IPCC Intergovernmental Panel on Climate Change
OECD Organization for Economic Cooperation and Development
OLS Ordinary Least Square
SBIC Schwartz-Bayes Information Criterion
UNEP United Nations Environment Programme
UNFCCC United Nations Framework Convention on Climate Change
IX
CHAPTER 1
INTRODUCTION
"Continued global warming is in nobody's interest, but the simple facts o f  the matter are that 
developing countries will suffer the most damage, and their poor will be at an even greater 
disadvantage. I see the Bank's role in climate change as providing every opportunity to 
developing countries to benefit from the huge investment OECD must make in reducing 
climate change."
James Wolfensohn, World Bank President 
United Nations General Assembly, June 1997
1.1. Background and Problem Statement
Climate is the average weather that describes the condition of different weather 
variables for a specified area during a specified time interval (van der Geest, 2002). Many 
natural factors influence climate and have the potential to change it. One of these factors is 
the greenhouse effect. Climate has always been changing throughout the history o f  the Earth. 
Between 2500 B.C. and 2300 B.C., for example, the climate in the present Sahara changed 
rapidly from a situation in which wheat, barley, millet and guinea com could be cultivated 
into a situation in which only livestock could be kept (Curtin et al., 1978). Climate change is 
a normal phenomenon (Ribot et al., 1996). Natural changes in climatic conditions have 
resulted in Ice Ages, relatively warm periods in temperate regions and dry, intermitted with 
wet periods in Africa (Kemp, 1994). Hence, climate change is not new.
The average global temperatures will increase by 1.4 to 5.8 degree Celsius over the 
next 100 years if no action is taken globally to control it (World Bank, 1996). The earth's 
temperature has been rising progressively for the past 25 years ([bid). The increase in the 
earth’s temperature recently has a more drastic consequence than all other natural climate 
changes that have been documented over die last 100,000 years due to the intensity o f human 
activities (World Bank, 1996; Lean and Mahlman, 1999).
l
The term "climate change" is sometimes used to refer to all forms of climatic 
inconsistency, but because the Earth’s climate is never static, the term is more properly used 
to imply a significant change from one climatic condition to another. In some cases, climate 
change has been used synonymously with the term, global warming. Scientists, however, 
tend to use the term in the wider sense to also include natural changes in climate. The 
progressive rise o f the earth’s surface temperature and changes in global climate patterns is 
thought to be caused by the greenhouse effect. Global warming has occurred in the distant 
past as a result o f  natural influences, but the term is often used to refer to the warming 
predicted to occur as a result o f increased emissions o f greenhouse gases (van der Geest, 
2002).
According to The Intergovernmental Panel on Climate Change (IPCC), global climate 
change is a rise in temperature caused by emissions of greenhouse gases (IPCC, 1990). The 
IPCC, composed o f hundreds o f scientists worldwide, released a scientific assessment o f 
climate change in the early 1990s. The IPCC was jointly set up by the United Nations 
Environment Programme (UNEP) and the World Meteorological Organisation to provide an 
authoritative international statement of scientific opinion on climate change.
The mechanism o f the greenhouse effect is that, the sun is the driving force behind the 
weather and climate. By heating the Earth’s surface, the sun provides the energy for the cycle 
between oceans, atmosphere, glaciers, surface water, groundwater and vegetation through 
precipitation and evapotranspiration. Solar radiation is absorbed by the Earth’s surface and it 
later leaves the Earth as outgoing radiation. Part o f the outgoing radiation is, however, 
absorbed by greenhouse gases in the atmosphere and re-emitted to the earth’s surface 
(Arrhenius, 1896). This makes the earth warmer. Without the natural greenhouse gases, the 
earth would presently be approximately 33 degrees Celsius colder (Ibid). Current life on 
Earth could not be sustained without the natural greenhouse effect. The earth's temperature is
2
affected by numerous natural factors, such as variations in solar radiation and emissions from 
volcanoes, among others. However, the scientific community generally agrees that human 
activity has intensified the greenhouse effect, and it is to blame for the recent increase in 
global average temperature (Houghton el al. 1990; Ribot et al. 1996; Lean and Mahlman,
1999).
Some human activities increase the concentrations of greenhouse gases in the 
atmosphere. Consequently, more outgoing radiation is re-emitted and this is how humanity is 
capable of increasing the global temperature. This is called the enhanced greenhouse effect 
(Houghton et al., 1990), or the anlhropogenically enhanced greenhouse effect (Ribot el al., 
1996). The speed with which greenhouse gases are emitted through human activities like 
deforestation and industrialisation are predicted to cause an accelerated climate change that 
has no precedent (Ibid). Changes in land use patterns such as the clearing o f forests also 
release carbon dioxide that has been stored in trees into the atmosphere. According to 
Lonergan (1998), there is a strong relationship between atmospheric levels o f carbon dioxide 
and temperature.
Temperature alters precipitation patterns, triggers extreme weather conditions which 
could alter forests, crop yields, and water supplies, as well as disrupt farming, fishing, and 
many other industries reliant on weather and the natural world (IPCC, 1990; Downing, 1992). 
Climate change is likely to have numerous negative effects on human development and 
welfare. The per capita fossil energy use (greenhouse gas emitter) by Sub-Saharan Africa in 
general and Ghana in particular is very low; yet the region is most susceptible, due to poverty 
and inability to adapt to continuing climate change.
Although the issue o f global warming has been debatable, die following could be 
argued in favour of its occurrence. One, anthropogenic releases of greenhouse gases are 
increasing at a constant rate (van der Geest, 2002). Second, these gases, in the absence of
3
other changes, will result in general global warming (Ibid). Third, there will be regional 
variation in the amount o f wanning and changes in other climate parameters, such as 
precipitation (Ibid). Finally, the consequences of such warming could be very great, 
depending on the ability o f systems to adapt to these potential changes. Governments, 
environmentalists and many scientific bodies are also calling for severe reduction in the 
emissions o f  these gases with the interests o f  future generations in mind. The response was 
the signing o f the United Nations Framework Convention on Climate Change (UNFCCC) by 
over 150 countries in June 1992. The signing of the convention indicates widespread 
recognition that climate change is potentially a major threat to the World’s environmental and 
economic development. A key objective of the convention is the stabilisation of Green House 
Gas (GHG) concentration in the atmosphere at a level that would prevent dangerous 
anthropogenic interference with the climate system (UNEP/OECD/IEA/IPCC, 1995).
Ghana covers about 238,539 square kilometres and lies between latitude 4.5°N and
11.5°N and longitude 3.5°W and 1.3°E. The country shares borders on the West with Cote 
d’Ivoire, East with Togo, North with Burkina Faso and to the South lies the Gulf of Guinea. 
Three main agro-ecological zones can be found in Ghana. These are the Coastal Plain, the 
Middle Semi-equatorial Forest and the Northern Savannah. There are two major types of 
crops grown in Ghana, cash crops and food crops. Agriculture contributed about 36 percent 
o f the Gross Domestic Product (GDP) in Ghana in 2001 and 2002 (GSS, 2002a).
The Environmental Protection Agency (EPA) in Ghana has observed a 20 percent 
decrease in annual total rainfall and a 1 degree Celsius increase in temperature over the thirty- 
year period of 1961-1990 (EPA, 2000). There is also a general perception of climate change 
in Ghana by both farmers and non-farmers especially the elderly who, by memory, recall 
higher rainfall in their days (Ofori-Sarpong, 2001; van der Geest, 2002).
4
The agricultural sector in Ghana is likely to suffer, most adversely, from climate 
change because o f  its high dependence on rainfall, temperature and soil type. Maize is a 
staple in Ghana and it is cultivated in virtually all the regions in Ghana. This raises the 
following issues: What has been the nature o f climate change in Ghana? What have been the 
effects o f climate change on maize supply in Ghana during the period 1970-2002? These are 
the issues which the present study addresses.
1.2. Objectives of the Study
The primary objective o f this study is to analyse the effects o f climate change on 
maize supply in Ghana over the period 1970-2002. The specific objectives are the following:
a. To describe the time profiles o f rainfall and temperature in Ghana during the period
1970 - 2002.
b. To describe the time profile of maize supply in Ghana over the period 1970-2002.
c. To quantify the effect o f climate change on maize supply in Ghana during the period
1970-2002.
13. Relevance of the Study
Climate change would indeed affect agricultural supply and hence the economic 
development o f  the country. There is not much empirical analysis done in this area, 
especially in Africa in general and Ghana in particular. The present research therefore 
contributes to narrowing this gap in knowledge.
Ghana is an appropriate country for the research since it is a developing country with 
different agro-ecological regions and climatic zones, which would clearly depict the impact 
o f climate change in the country. Also, the country cultivates diverse crops with varying 
temperature tolerance levels, so that farmers and policy makers could identify which crops to
fall on when temperature increases. This could also be a model for other countries in Africa 
which have similar agro-ecological conditions.
The agricultural sector in Ghana is the focus o f  this research because it contributes 
about 36 percent to GDP and employs about 60 percent o f the labour force (GSS, 2002). 
Furthermore, it is highly dependent on rainfall and temperature and it is likely to be the 
hardest hit when climate changes. Also, agriculture is an appropriate sector for this analysis 
because it could be the most important market impact from warming (Mendelsohn and 
Neumann, 1998, Mendelsohn and Schlesinger, 1999; Mendelsohn et al., 2000). Hence, policy 
makers must be made aware, through rigorous research, o f  the adverse effects o f  climate 
change and suitable measures taken to mitigate these adverse effects.
Maize is the crop studied in this research because it is a crop which can be found in all 
the agro-ecological zones in Ghana (GSS, 2000). Maize accounts for about 52 percent o f the 
total output o f cereals in Ghana and occupies about 44 percent o f the total area devoted to 
cereals. It is grown on more than 500,000 hectares o f land in Ghana and has a national 
average yield o f about one ton/ha (Marfo and Read, 1985). Maize is also the staple food o f a 
majority o f  the population in Ghana, and government has sometimes imported large 
quantities o f  this cereal periodically to supplement local production for both human and 
animal consumption (GOG, 1982).
1.4. Organisation of Study
This study is organised into five chapters. In Chapter 2, a review o f the existing 
relevant literature on the theme o f the study is presented. Chapter 3 outlines the methodology 
employed to accomplish the objectives of the study. Empirical results of the study are 
presented in Chapter 4. Summary and policy recommendations are made in Chapter 5.
6
CHAPTER 2 
LITERATURE REVIEW
This chapter reviews the theoretical and empirical studies done on the subject of 
climate change and its effects on agriculture globally, in sub-Saharan Africa and Ghana. It 
highlights the conclusions drawn by earlier researchers concerning how a change in climate 
would or does affect developing countries under limited capital and technology.
The Intergovernmental Panel on Climate Change (IPCC) reports that the impact of 
climate change on people’s livelihood will be greatest in the tropics and subtropics, 
particularly in Africa due to the high dependency o f many poor smallholders on agriculture 
(IPCC, 2001b). According to Watson (2001), 800 million people are already malnourished 
and food production has to double in the next thirty-five years to meet future needs. These 
statistics coupled with the event o f  climate change pose a great threat to development and 
food security. In an increasingly interdependent world, developing countries cannot avoid 
being strongly influenced by what happens in the rest o f the world, particularly in the 
industrialized (Organization for Economic Cooperation and Development - OECD) countries 
(Killick, 1989). This view can be strongly related to climate vis-a-vis development o f the 
world. In the process o f developing, the Developed World destroys the environment through 
increasing CO2 levels in the atmosphere (global warming) and developing countries bear the 
consequences o f their actions.
In identifying possible causes of the observed warming of the earth's climate over the 
20th century, a series of modd calculations were used to examine the differing effects of 
natural variability, carbon dioxide and other greenhouse gases, sulphate particles, and 
changing solar output on the climate o f the 20th century (Lean and Mahlman, 1999). In 
general, these calculations made it clear that it is scientifically very difficult to construct an 
explanation for the 20th century warming that does not include a major role for the added
7
greenhouse gases resulting from human activities (Ibid). Based on the model calculations and 
the observational records o f climate change for the 20th century, the conclusions are that 
added greenhouse gases provide, by far, the most plausible hypothesis for explaining the 
warming o f the 20th century (Ibid). Solar irradiance variations are large enough to shape, but 
not dominate, the observed wanning. Natural variability plus added greenhouse gases and 
added greenhouse gases plus increased solar irradiance can explain the extended warming 
period between 1910-1940 (Ibid).
2.1. General Effects of Climate Change
Rising global temperatures are expected to raise sea level and change precipitation 
and other regional climate conditions. Changing regional climate could alter forests, crop 
yields and water supplies. It could also affect human health, animals, and many types of 
ecosystems (IPCC, 1996).
A French scientist first described the natural greenhouse effect, drawing the parallel 
between the actions of the atmosphere and the effect o f glass in a greenhouse (Fourier, 1824). 
The sun’s radiation reaching the earth is in the shorter wavelength whereas outbound 
radiation from the earth is in the infrared long-wave bands. Clouds, water vapour and natural 
greenhouse gases such as carbon dioxide (CO2), methane, nitrous oxide and ozone are more 
opaque to long-wave than short-wave, trapping 80 to 90 percent o f  the outbound radiation 
from the earth’s surface. This trapping influence is called the greenhouse effect. Without the 
greenhouse effect, the average surface temperature on earth would be -18°C instead o f the 
15°C observed today, and this would be too low for any sort of life (Ibid).
The Swedish scientist Svante Arrhenius introduced the possibility of an enhanced or 
man-made greenhouse effect one hundred years ago. Arrhenius hypothesized that the 
increased burning of coal would lead to increased concentration of carbon dioxide in the
8
atmosphere and warms the earth (Fleming, 1998; Arrhenius 1896). Since Arrhenius’ time, the 
emissions o f greenhouse gases have increased dramatically. The concentration o f carbon 
dioxide in the atmosphere has increased by 25 percent over pre-industrial levels (Ibid). In 
addition to increased burning o f fossil fuels such as coal, oil and natural gas, man-made 
chemical substances such as Chlorofluorocarbons (CFCs) as well as methane and nitrous 
oxide emissions from agriculture and industry contribute to the greenhouse effect (Cline, 
1992; Frankhauser, 1995).
With respect to the greenhouse effect from human activity, the IPCC noted that “there 
has been a real, but irregular, increase o f global surface temperature since the late nineteenth 
century” amounting to 0.45°C on average (IPCC, 1996). Current emission trends will lead to 
a doubling o f greenhouse gas concentration over pre-industrial levels around the year 2050. 
The IPCC projects a global average temperature increase o f 1 - 3.5 degrees Centigrade, or 2 - 
6 degrees Fahrenheit, which would have significant impact on climate throughout the world.
Average global surface temperature has increased by approximately 0.6°C since the 
late 19th century, with 95 percent confidence that it lies between 0.4 and 0.8°C (IPCC, 1990; 
1996). Most o f this increase has occurred in two periods, from about 1910 to 1945 and since 
1976; the largest warming in recent period has been recorded in the winter extra-tropical 
Northern Hemisphere (Ibid). The warming rate of 0.17°C per decade since 1976 has been 
slightly larger than the rate of warming during the 1910 to 1945 period (specifically, 0.14°C 
per decade), although the total increase in temperature is larger for the 1910 to 1945 period 
(IPCC, 1990). The most recent periods of warming also recorded a faster rate o f warming 
over land compared with the oceans (IPCC, 1990; 1996).
According to the IPCC, Third Assessment Report (Scientist Assessment), warming 
from 1910 to 1945 was initially concentrated in the North Atlantic and nearby regions. 
During 1946 to 1975, the Southern Hemisphere was warm whilst the Northern Hemisphere
9
showed cooling. Temperature trends in the Twentieth Century exhibit a broad pattern of 
tropical warming with extra-tropical trends being more variable. The El-Nino event in 
1997/1998 is associated with the high global temperature in both surface and tropospheric 
temperature. Warming was emphasised in the Northern Hemisphere in winter and spring with 
all year-round cooling in the Southern Hemisphere oceans and the Antarctica between 1976 
and 2000 (IPCC, 2001). The North Atlantic/Arctic Oscillation Westerly phase and the Pacific 
variability are recognised to have caused global temperature change since the 1970s (Ibid).
2.2. Future Effects of Climate Change
It is projected that the equivalent of doubling atmospheric C 02 will occur by 2050 and 
this will lead to a global temperature rise o f between 2.4 and 10.5 degrees Fahrenheit by 2100 
(IPCC, 1990; 1997). The longer it takes humanity to begin reducing greenhouse gas 
emissions, the greater the rise in global temperature will be. Land surfaces and higher 
latitudes will experience larger increases (Ibid). Temperature will continue to rise long after 
emissions are reduced and greenhouse gas concentrations stabilise (Ibid).
Sea level will rise 10-30 inches by 2100, and will continue to rise thereafter (Ibid). 
Wetlands will be flooded. Many low-lying regions and small island states will have to be 
evacuated due to storm surges and saltwater intrusion (IPCC, 1996). Evaporation and 
precipitation will increase about 1 percent for every 1 degree Fahrenheit temperature rise, and 
their distribution will be increasingly uneven and unpredictable (Ibid). There will be more 
frequent and severe heat waves and droughts, and heavier storms and floods. Rapid climate 
change will limit the ability o f many plant and animal species to adapt. However, insects, 
rodents, disease organisms, and other species which reproduce rapidly will increase in 
population (Ibid).
10
2.3. Effects of Climate Change on Agriculture
Recent literature suggests that, in the tropics and subtropics where some crops are 
already near their maximum temperature tolerance and where dry land, non-iirigated 
agriculture predominates, yields will tend to decrease with even nominal amounts o f climate 
change (IPCC, 1998). The literature tends to project further that positive effects on 
agriculture would be concentrated in high latitudes and negative effects in lower latitudes, 
precisely where problems o f hunger already exist (Ibid).
The effects o f climatic variations on agriculture (especially in Africa) have been well 
established through decades o f field experiments, statistical analyses of observed yields, and 
monitoring o f agricultural production (IPCC, 1997; Akong'a et al, 1988). The most important 
climatic element is precipitation, particularly seasonal drought and the length o f the growing 
season. The distribution o f rainfall within the growing season also may affect yields. Low 
temperatures and radiation limit production in some high-elevation regions; frost is a hazard 
in Southern Africa (IPCC, 1997). High temperatures can negatively affect yields and yield 
quality in semi-arid and arid regions, although water is more important. Sea-level rise and 
coastal erosion will affect groundwater through increased salinity, irrigated agriculture, and 
low-lying coastal land in some areas on the basis o f the extent o f land liable to inundation and 
the population being at risk (IPCC, 1997).
Based on experimental research, crop yield responses to climate change vary widely, 
depending upon species and cultivars, soil properties, pests and pathogens, the direct effects 
o f carbon dioxide (CO2 ) on plants, and interactions between CO2 , air temperature, water 
stress, mineral nutrition, air quality, and adaptive responses (IPCC, 1996). Even though 
increased C 02 concentration can stimulate crop growth and yield, this benefit may not always 
outweigh the adverse effects o f excessive heat and drought (IPCC, 1997).
1 1
Yields o f some crops in tropical locations would decrease generally with even 
minimal increases in temperature, because such crops are near their maximum temperature 
tolerance and diyland/rainfed agriculture predominates (Reilly et al, 1996). Where there is 
also a large decrease in rainfall, tropical crop yields would be even more adversely affected. 
Most studies indicate that global mean annual temperature increases o f a few degrees Celsius 
or greater, would prompt food prices to increase due to a slowing in the expansion o f  global 
food supply relative to growth in global food demand (Ibid).
Changes in temperature and in the level and timing o f precipitation may influence the 
yield of crops (Adams et al, 1990; Parry, 1990). It is expected that in some areas, crop 
production will generally suffer from the changes whereas in other areas crop production will 
benefit from the changes (Parry, 1990). In some areas, heat or water stress will reduce the 
yields of certain crops, whereas other crops will benefit from the changes in climate, 
implying that the relative productivity o f different crops will change (Ibid). Some o f the crops 
which may be affected negatively are wheat, maize and rice to mention but a few (Ibid). 
Furthermore, the frequency o f extreme meteorological events may change, causing higher 
yield variability. Increased levels o f CO2 , technological changes, plant breeding, inter alia., 
will also influence the future crop productions (Ibid).
Higher temperatures may increase the need for plant protection (Chen and McCarl,
2001). The conditions are more favourable for the proliferation of insects and pests in warmer 
climates. Temperature as well as increased levels o f C 02 influence the growth o f  weeds, 
implying that control of weeds may be more difficult (Rosenzweig and Liverman, 1992). 
However, the effects o f climate change on the need for plant protection will vary from case to 
case. In addition to temperature and precipitation changes, climate change may also impact 
agriculture through greater competition from weeds, increased plant and animal disease, 
changes in soil nutrients and pests, and increased conflicts for available water. While these
12
damaging effects are probably controllable, no conclusions have been drawn yet as to what 
they may do to the cost o f agricultural production and how they will affect agricultural 
resources and the environment.
Since rain-fed agriculture is the main source o f livelihood for most people in Africa, 
the combination o f decreasing annual amounts o f rainfall, increasing rainfall variability, 
increasing temperature and population growth could cause a serious decline in the 
population’s capacity to secure its food and other needs. Agricultural droughts would occur 
more frequently and without a dramatic shift in agricultural and non-agricultural production 
strategies, the region would become much more prone to famine. That is why climate change 
in Sub-Saharan Africa, albeit being an age-old phenomenon, has to be taken seriously. The 
Sahelian droughts o f the 1970s and the early 1980s may have been a first warning. They were 
much more extreme than the ‘normal’ oscillating pattern (Hulme 1994). Although 
precipitation levels have partly recovered in the 1980s and 1990s, they are still well below the 
average of the first part o f the century.
Where certain varieties of crops are grown near their climatic limits, extreme weather 
events may have dramatic effects on agricultural production (Parry, 2000). This may 
consequently influence the variability of the economic outcome o f farming and increase the 
risk of agricultural investments. The economic and social impacts o f climate change include 
changes in the optimal farming systems, relocation o f farm processing industries, increased 
economic risk and changes in the rural income and heritage (Kane et al, 1992; Molua, 2002). 
Changes in the economic performance o f different farming systems due to, for example, 
changes in crop growth conditions, may lead to substantial changes in the farming systems 
and with agriculture being the main income generator, the economic welfare o f the society 
would be affected.
13
Changes in rainfall and temperature could have direct effects on biological 
productivity, crop yields and economic welfare. These effects depend on the importance of 
the climatic controls o f  production relative to other controls such as provision of plant 
nutrients, availability o f labour and availability o f capital, technology and know-how. 
Climatic control o f productivity is o f crucial importance in certain agricultural zones in 
developing countries. Climatic ‘marginal’ zones, such as the semi-arid savannas are 
particularly susceptible to climate-driven changes in productivity. Even though many 
scientists fear that the most adverse effects are likely to occur in poorer countries, few 
attempts have been made to explore climate change (global warming) in the developing 
world (Schelling, 1992; Nordhaus, 1993). Most of the research has focused on the U.S and 
Europe (Ibid).
Rasmussen (1999) noted that the effects o f  climate change on human lives and 
national economies are likely to be greater in the tropical and subtropical developing 
countries than in the industrialised (and typically temperate) countries. This is not necessanly 
due to differences in the size and severity - in purely physical terms - o f the projected climate 
changes in itself, but rather it is due to the greater importance o f the primaiy sector in these 
countries in terms of contribution to the GNP. This underlines the need to look at climate 
change, supposedly mainly related to greenhouse gas emissions caused by human 
consumption as a problem affecting food security, quality of life and economic development 
in developing countries in particular. Not only the climatic changes, but also the economic 
and political mechanisms applied to limit these changes, have implications for developing 
countries.
14
2.4. Empirical Literature on Effect of Climate Change on Agriculture
Most empirical studies undertaken focus mostly on the impact of climate change on 
agriculture in the United States. Some of the existing studies are Nordhaus (1993), Cline 
(1996), Dinar and Mendelsohn (1999) and Mendelsohn et al (2001). These researchers 
concluded that agriculture systems in the US would readily adapt to climate change, by 
introducing new technologies, new crop varieties, and cultivation practices so that there 
would be minimal changes in yields and net profits. These results would likely extend to 
other Organization for Economic Cooperation and Development (OECD) countries, 
suggesting that agriculture in developed countries is less sensitive to climate change. 
However, less is known o f the effects o f climate change in developing countries, especially 
Africa in general and Ghana in particular.
Molua (2002) observes that an increase in rainfall during crop growth is a positive 
covariate of income in Southwestern Cameroon. This implies that, for a region whose 
agriculture is rain-fed, an increase in rainfall coupled with improved tillage practices could 
enhance farm economic return. Therefore irrigation in the growth period especially during 
dry spells would be valuable for stimulating increases in production. He used an econometric 
function which directly relates farm income and precipitation in order to statistically estimate 
the significance o f  farm-level adaptation methods. Similarly, agroclimatic studies by Akong’a 
et al. (1988) considered the effects o f climatic variability on agriculture, with an emphasis on 
coping with drought. Drought episodes in sub-humid and semi-arid zones were observed to 
have led to the failure of crop production and dependency on other sources of income to buy 
food, or on famine relief.
Schulze et al. (1996) used the ACRU/CERES hybrid model to evaluate the impact of 
climate change on maize in South Africa. The investigators divided the diverse geography of 
South Africa into 712 relatively homogeneous zones, each associated with a specific type of
15
vegetation, soil, and climate. Daily values of temperature (minimum and maximum), rainfall, 
wind speed and solar radiation were used in crop evaluation, based on the CERES-Maize 
model. For three scenarios of climate change, yields tend to decrease in the semi-arid west. 
For most o f the country, C 02 enrichment effect tends to counteract the relatively modest 
changes in temperature and precipitation and potential yields tend to increase. In parts o f the 
eastern highlands, particularly in Lesotho, dramatic increases in yields tend to result from 
higher temperatures.
Endrody-Younga (1968) observed that insects increase during the season o f crop 
growth. However, maize planted early in the major season tended to escape heavy infestation 
and mature before insects could cause severe damage. Therefore, yield reduction due to 
greater plant damage normally occurs when cultivation is done late in the major season or in 
the minor season when insects are on the increase.
Liverman and O'Brien (1991) observed that global warming may present a threat to 
local and national food security in Mexico, especially if farmers are unable to adapt to a drier 
climate and as food imports from other regions become more costly.
Muchena (1994) and Muchena and Iglesias (1995), explained that climate change may 
lower maize yields in Zimbabwe due to shortening o f the favourable growing period and 
increasing water stress. Furthermore, climate change is likely to cause significant spatial 
shifts in agricultural capability and land-use zones, with wet zones diminishing and dry zones 
expanding due primarily to rises in potential evapotranspiration. A large portion of African 
agriculture is rain-fed, but heat-related plant stress may reduce yields in several key crops, 
such as wheat, rice, maize, and potatoes (Burke et al, 1988). These crops most often are 
located in warmer, dryer climates and are quite susceptible to water stress.
The type o f land preparation before planting affects the development and yield of 
maize (Takyi, 1970). Furthermore, planting maize in ridges have greater yields than maize in
16
flat seed beds since there are higher concentrations o f fertiliser in the root zone as a result o f 
ridging (Ibid). He came to this conclusion by examining the planting of maize in ridges, flat 
seed beds and soil not ploughed in central Ghana. Clarke (1962) observed that soil structure 
and climate strongly influenced the nutrient holding capacity o f soils. Also, soil management 
and cultural practices such as fallowing affect the fertility o f soils.
According to Davis and Porter (1936), Doneen and MacGillivray (1943) and 
Delouche (1953), adequate moisture availability is important for good growth and yields of 
plants.
From a review o f various studies on the impact o f climate change on agriculture, 
Dinar and Beach (1998) concluded that the relative lack o f information and insufficient 
economic analysis o f the impact of and adaptation to climate change in developing countries, 
suggests that an important task is to initiate empirical research in developing countries, since 
these countries are likely to be subject to a negative impact o f climate change in the future. 
They were o f  the view that, appropriate measures which might be used to deal with the 
climate change phenomenon are complicated due to a great deal o f controversy surrounding 
the issue; that is, different interpretations o f the evidence, different interests o f policy makers 
and different evaluation o f intervention options.
Using cross-sectional data, Zekri (2002) regressed regional net revenue on climate, 
geographical, soil, economic and demographic variables to determine the intrinsic value of 
climate change. He noted that in the coming fifty years, global temperature will rise by 2 °C 
to 5°C if nothing is done in terms of reducing C 02 and other gases responsible for the Green 
House Effect. Negotiations between the major industrialized countries have taken place to set 
acceptable standards concerning CO2 gas emissions. Nevertheless, many pieces o f the puzzle 
are still missing to come to a reasonable compromise. In effect, decision-makers cannot 
accept any alternative in the lack of quantitative information about the economic impacts of
17
climate change on the world economy, which sectors will benefit, what countries will benefit 
and who will lose and so on. It is obvious that the reduction o f emissions will mean either 
decreases in the industrial activities or adoption of new technologies resulting in increases in 
production costs and shifting part o f the investments in pollution abatement to this sector. On 
the other hand, part emissions reduction should lead to benefits or costs to other sectors such 
as health, agriculture, landscape and water. Thus, what decision-makers should ensure is that 
the stream o f benefits would be greater or equal to the costs o f reducing emissions (Zekri,
2002 ).
Adams et al. (1990) conducted an integrated study for the US, linking models from 
atmospheric science, plant science and agricultural economics. While the outcomes depend 
on the severity o f climate change and the compensating effects of carbon dioxide on crop 
yields, the simulations suggest that irrigated acreage will expand and that regional patterns of 
US agriculture will shift with predicted global warming. With the more severe climate change 
scenario tested, the movement of US production into export markets is substantially reduced.
Rosenberg and Crosson (1990) integrated the Missouri, Iowa, Nebraska, and Kansas 
(MINK) study, both within the agricultural sectors and across other sectors. The study 
incorporated both the physiological effects of CO2 and adaptation by fanners to the climatic 
conditions o f the 1930s. Even with the relatively mild warming (1.1°C) of the 1930s and with 
fanner adaptation and CO2 physiological effects taken into account, regional production 
declined by 3.3 percent. Given the IPCC (1990a) estimate of 2.5°C warming for doubled C 02 
conditions, these results imply agricultural losses of about 10 percent for equilibrium 
warming of doubling o f carbon dioxide-equivalent (Cline, 1991).
According to Hulme (1996), the coefficient of variation for annual maize yields due to 
the effects of the 1984-85 and 1991-92 droughts varies from about 10 percent in Central 
Africa to almost 50 percent in drier countries such as Botswana and Swaziland. A significant
18
component o f the variability is likely to be related to rainfall, although prices and market 
policies are influential. Hulme (1996) presents an integrated view of climate impacts in 
Southern Africa. Prospects for agriculture depend critically upon changes in precipitation. A 
“dry” scenario suggests less-suitable conditions in semi-arid and sub-humid regions. With 
little decreases (or increases) in precipitation, agriculture should be able to cope with the 
average changes. However, shifts in drought risk need to be considered.
Mendelsohn and Williams (2002) concur that tropical nations will be hurt, temperate 
nations will be barely affected, and polar nations will benefit from climate change; since 
these effects are offsetting, the net global impact of climate is relatively small. Compensation 
is clearly called for but only a modest abatement programme appears warranted at this time. 
The large uncertainty surrounding these forecasts further suggests that continued monitoring 
o f  both the climate and impacts is worthwhile.
The empirical results of Sanghi et al (1998) indicated that climate change would have 
an overall negative impact on Indian agriculture, with varying seasonal and regional 
implications. A warming scenario o f 2°C increase in mean temperature and a 7 percent 
increase in mean precipitation levels exerted a 12.3 percent reduction in net revenues for 
India as a whole. When the increases in temperature and precipitation were calculated 
differently, it was realised that a rise in temperature was damaging, whereas an increase in 
precipitation levels was beneficial. However, the positive precipitation effect was dwindled 
by the negative temperature effect. There was a suggestion that, in order to adapt to these 
climate changes, there was the need to develop increased heat tolerance in high-value 
temperature sensitive crops. Also, minimizing run-offs to capture benefits from increased 
rainfall was expected to be a beneficial strategy.
Dinar and Mendelsohn (1999) using a Ricardian Analysis argue that, the damages of 
climate change will be greater in developing countries because their levels of capital and
19
technology are lower. For example, the results cited in Pearce et al. (1996) predict that OECD 
countries will suffer damages o f 1.4 1.6 percent o f GDP but less-developed countries will
have damages between 1.6 percent and 2.7 percent. Economies with less capital and 
technology could be more vulnerable to climate change as they have less control over their 
environments, due to the fact that more o f their economy is in vulnerable sectors, and they 
have warmer climates to begin with.
Although it is commonly believed that vulnerable sectors in developing countries are 
more climate sensitive (Nordhaus, 1991; Schelling, 1992), it has never been rigorously tested 
empirically particularly in Africa. Mendelsohn et al (2001) in their study o f sensitivity of 
climate and development arrived at the conclusion that, development does have an important 
effect on climate sensitivity. Furthermore, farmers in developing countries are currently more 
climate sensitive than farmers in the more-developed countries. Combining this climate 
sensitivity with the higher temperatures at low latitudes suggests that lowrlatitude agriculture 
would be hurt if  warming occurred today (IPCC, 2001). Whether this is the case or not is 
unknown in Africa in general and Ghana in particular. The present study will contribute to 
narrowing this gap in knowledge.
2.5. Literature on Climate Change and Agriculture in Ghana
In Ghana, global warming has precipitated climate change (EPA 2000). For example, 
over the thirty year period covering 1961-90, annual total rainfall declined by 20 percent and 
stream flow or runoff in all the river basin systems declined by 30 percent, as temperature 
increased by 1°C (EPA 2000). Furthermore, it is estimated that mean daily temperature will 
increase by 2.5-3.2°C by the year 2100, whereas annual total rainfall will decline by 9-27 
percent by the same year. Stream flows in all the river basins are expected to decline by 15- 
20 percent by 2020 and 30-40 percent by 2050. Notably, Ghana has been experiencing long
20
periods of drought and erratic rainfall, particularly in the arid north (savannah) and along the 
coast. Irregular climatic conditions like untimely onset o f rainfall and untimely cessation of 
rainfall during the crop-growing season have also been experienced. Erratic rainfall has also 
been observed during the minor crop season (namely, during September and November) in 
the forest-savannah transition zone. Sometimes excessive rainfall causes floods. The current 
levels of temperature and evaporation rates are high in Ghana, particularly in the Guinea 
savannah, Sudan savannah and coastal savannah zones.
Climate change could exert adverse effects on agriculture in Ghana. For example, it 
has been argued that the current high temperatures and high evaporation rates have tended to 
increase water stress and ultimately precipitated declines in crop yields (EPA, 2000). 
Furthermore, high temperatures and declines in surface water resource availability due to 
temperature effects on river discharges or runoff values, as well as declines in groundwater 
resource availability due to climate-induced declining rates o f recharge o f aquifers have also 
occurred in Ghana. It has been estimated, for instance, that climate change is likely to cause 
22 percent, 17 percent and 5 percent declines in groundwater (aquifer) recharge by the year 
2020 in the Volta basin system, the Pra system and the Ayensu system respectively (EPA,
2000).
Adverse climate change could lead to total crop failure. For example, farmers in the 
Upper East Region have lamented the failure o f the millet crop, whereas farmers in the 
Northern Region have recorded rice crop failure due to change in climate (Ofori-Sarpong, 
2001). Farmer perceptions concerning climate change indicate that farmers are aware of 
climate changes like the delayed onset of rains, increase in day length, and intermittent 
droughts during the relevant crop growing season and crop failure due to adverse climate 
change (Ibid).
21
Evidence o f climate change (specifically, reduction in rainfall amount) in the Sekyere- 
West District o f Ghana has been provided by Owusu (2002). Similarly, the impact o f climate 
change on agriculture and fanners’ coping strategies in Navrongo and Bawku (the arid 
regions of Ghana) have been analysed by Ofori-Sarpong (2001). The mean annual rainfall for 
the 30-year period 1931-1960 was 1087.6 mm while the mean for the second 30-year period 
1961-1990 was 986.1 mm, which is an indication that the first 30-years were wetter than the 
latter 30 years in the arid Upper East Region (Ofori-Sarpong, 2001). O f the sixty years, 
thirty-six are drought years (drought is here defined as a period during which rainfall falls 
below the mean). This confirms the fact that in the Upper East Region, drought is endemic 
(Ofori-Sarpong, 2001). Furthermore, over the last 30-40 years, a decrease in annual and 
monthly rainfall figures in the Upper East Region occurred (Ofori-Sarpong, 2001). Also, the 
annual mean and monthly temperatures have increased, which could also be an indication of 
climate change (Ofori-Sarpong, 2001). The study, however, was performed only in the Upper 
East Region, with a particular climatic condition (hot and dry weather). The present study 
contributes to narrowing this gap in knowledge by extending the analysis across the country, 
which would capture the effects o f different climatic conditions; hence, contributing to 
analysing the effects o f climate change on agriculture in Ghana.
Owusu (2002) observed that rainfall totals in the Sekyere-West District decreased 
from 1493.1 mm in 1960 to 1164.0 mm in 1999. He further noted evidence o f  drought 
conditions in the monthly rainfall distribution in some wet years (the growing seasons) and 
that in 1992, which was a typical drought year, all the months except May recorded rainfall 
below the long-term mean, while water balance recorded an eight-month instead o f a five- 
month drought period in the long-term mean water balance. Moreover, in 1998, the humid 
periods were reduced to only five months instead of the mean seven months (Owusu, 2002).
22
In Owusu’s study, rainfall was the only climatic factor considered. The present study includes 
an additional climatic factor, specifically, temperature.
Hamid (1984) concluded in his study in the Upper West area of Ghana that rainfall 
seasonality is the major factor which controls the calendar o f agricultural activities in the 
area. He said the struggle for sufficient food supply by the people o f this area is controlled by 
the start, duration and end o f the rainy season. Therefore, agriculture in the Upper West tends 
to be mainly affected by rainfall variability, and lower rainfall has serious repercussion on 
crop yields.
Leyenaar (1976) in conducting three field experiments on maize at the Legon Farm 
came to the conclusion that good cultural practices increase grain yields and that higher 
yields are obtained during the major rainy season; that is, planting right after the dry season 
as soon as the rains start. He argued that little capital is required in this process but more than 
doubled grain output may be achieved. However, planting in the minor season yields low 
output o f maize unless irrigation and improved cultivars are utilised and stem borers 
controlled. He further argued that maize production is not practical without the addition of 
water in the minor rainfall period, but the capital cost of irrigation would be great for the 
average farmer and might not be economically practical.
A review o f optimum planting dates and maize cultivation in Ghana confirms that 
maize cultivation should only be in the major season since rainfall is fairly stable and reliable 
during this period (Dom-Anin, 1985). The long-term average rainfall distribution indicates 
that maximum yield is recorded when tasseling occurs just before the peak o f rainfall, since 
planting early will minimize moisture deficiency during tasseling and grain-filling period.
Fosu et al (1997) in a study o f agricultural supply response and structural adjustment 
in Ghana and Burkina Faso had a result indicating that the supply response to the own-price 
of a crop is positive only for crops that are grown with non-traditional technologies, example
23
is maize in Ghana. Also, the supply response o f traditional food crops is either insignificant 
or negative particularly in Ghana.
In Heerink et al (1997) farmer’s transport time showed a significant negative effect on 
production and marketed quantity o f cassava and maize. The study concluded that, 25 percent 
reduction o f transport time was needed to increase the output of maize by one percent. Maize 
had a small positive supply response to price changes: where the price elasticity o f maize 
supply equals 0.04.
Jones and Thornton (2003), in their study on the potential impacts of climate change 
on maize production in Africa and Latin America in 2055, observed that a 10 percent 
decrease in maize output by 2055 due to climate change is certainly serious unless the level 
of decrease as stated by Pardey and Beintema (2001) will be compensated by plant breeding 
and technological interventions in the intervening period given the history o f cereal output 
increases since 1950. In this study, the output o f a global circulation model (GCM) was 
generated on surfaces that are characteristic o f the predicted climate to 2055, which gave an 
output decrease in maize for Ghana to be 100,800 tons in smallholder rain-fed production 
systems.
In conclusion, it is evident from the present review of literature that there is a dearth 
of knowledge on the theme o f the present study. The present study therefore contributes to 
narrowing this gap in knowledge by using cointegration and error correction modelling in 
analysing the effect o f climate change on maize supply in Ghana.
24
CHAPTER 3 
METHODOLOGY
This chapter outlines the present study’s methodology. First, the procedures for 
constructing the time profiles o f annual mean temperature and its deviation from the mean, 
annual mean rainfall and its deviation from the mean and maize supply are described. 
Second, the theoretical framework for the econometric analysis is also presented in this 
chapter. Finally, cointegration analysis and error correction modelling are undertaken.
3.1. Time Profiles of Temperature and Rainfall
Climate data are collected from meteorological stations in maize production areas, 
since they contain more information about the weather characteristics of the maize areas. The 
mean monthly rainfall and temperatures at these stations are averaged to get the annual mean 
rainfall and temperatures in the maize belt in Ghana. The major production areas and the 
selected meteorological stations are representative o f Ghana’s maize belt. These data, 
covering 1970-2002, are plotted against time and the time profiles described. The mean o f 
temperature and rainfall are also calculated to derive the deviations from the mean. The time 
profile o f the deviations o f temperature and rainfall from the mean are then plotted and 
described.
3.2. Time Profile of Maize Supply
Data on maize supply, the price o f rice and the price of cassava are also plotted 
against time. Section 3.7 details the sources of these data. The time profile of maize supply is 
compared with the time profiles of the annual rainfall, the price of rice and the price of 
cassava. The various comparisons are described.
25
3.3. Theoretical Framework
The theoretical framework for the analysis o f this study is the conventional theory of 
agricultural supply. This theoiy is based on the profit-maximising behaviour o f a rational 
farmer (Heerink et al, 1997), Based on this theory, increasing prices o f farm inputs would 
lead to decreases in the demand for these inputs, whereas increases in price of fann products 
are expected to lead to increases in market supply. For the purposes o f this study, climate 
variables are added to the set o f arguments o f  the well-behaved production function. 
Economic theory suggests that the supply function o f an economic agent, the farmer in this 
study, starts with a well-behaved production function. In theory, the determinants of 
agricultural production comprise local conditions like soil type, climate, agricultural 
resources and the techniques used. Inputs and output are rates of resources used and the rate 
o f production per unit o f time respectively. The production function could be written as:
q0= /(  Zi, Fi, W ,) (1)
where q0 is the output in a given time period, Zi are the variable inputs, F, are the fixed inputs 
and Wi are the climate variables. Variable inputs are those which change with the volume of 
output over the period of time to which the production function applies; for example, 
fertiliser, seeds and pesticides. The fixed inputs are those which cannot be changed during the 
period for which the production function is defined. The climate variables for this study are 
rainfall and temperature. The profit-maximising function consists of output and input prices 
and production factors, describing the farm’s production technology at profit-maximising 
points in the set o f  production possibilities. Thus, it is assumed in the present study that maize 
farmers attempt to maximize profit defined as the return to the variable factors. The restricted 
profit function (Lau, 1976) is defined as follows:
26
7t(p,c)=  Maxq(p.q |q s V(c),c e C) (2 )
where q is a vector o f outputs (qi > 0) and inputs (q, < 0), p is a vector of output and input 
prices, c is a set o f fixed or environmental factors (such as fixed land endowment, weather 
indicators and government policies, captured by C) and V is the state o f the technological 
constraint.
Applying Hotelling’s Lemma yields the supply functions for output and demand 
functions for input variables:
» , ~ > 0  (3 )
Qk =  0 (4 )
dPk
where i and k represent output and input variables respectively.
From equation (3), the maize supply function is derived with temperature and rainfall, inter 
alia, as the relevant arguments. Notably, within the range o f temperature and rainfall said to 
be ambient with respect to maize supply, increases in rainfall and temperature are likely to 
stimulate increases in maize supply and vice versa. From this theoretical framework, a 
number o f hypotheses arise with prices o f variable inputs and the product price (of maize) 
following the respective conventional negative and positive signs. The main focus o f  the 
present study is climate variables and the relevant hypotheses validated witii respect to the 
climate variables are stated as follows.
27
3.4. Statement of Hypotheses
The hypotheses validated with respect to climate variables in this study are as follows 
(the null and alternative hypotheses are denoted as H0 and Hi, respectively):
I. Ho: Rainfall exerts no effect on die supply of maize, versus
Hi: Rainfall exerts a positive effect on the supply o f maize.
II. H0: Temperature exerts no effect on the supply o f maize, versus
H i: Temperature exerts a negative effect on the supply o f maize.
Here, climate change is captured by changes in temperature and rainfall. When rainfall is yet 
to reach its maximum, an increase in rainfall is likely to increase the supply o f maize 
especially in Ghana where agriculture is largely rainfed. A higher maize supply is obtained 
during major rainy seasons; that is, the period of high rainfall (Leyenaar, 1976). According to 
die EPA (2000), the general decline in crop yields could be attributed to current high 
temperatures causing water stress through evaporation. Moisture stress during flowering, 
pollination and grain-filling is harmful to most crops; hence, a decrease in the supply o f crops 
occurs in drought years when temperatures increase (Ofori-Sarpong, 1985; 2001). Also in the 
tropics where temperatures are already high (near their maximum), increases in temperature 
are likely to result in decreases in cereal (maize) supply (IPCC, 2001).
3.5. Empirical Model
The empirical model employed in this study is specified as in equation 5:
Ln Q, = ao + ai In Ptm + a2 lnP,c + a3 lnPtr + a4 lnP,npk+ a5 InT, + a^  InR, + e, (5)
where lnQt denotes the natural logarithm of the output o f maize, ao is an intercept term, P/" is 
the real producer price of maize, P^ is the real producer price of cassava (specifically, the
28
price o f a competitive crop). Cassava competes with maize for land and other agricultural 
resources in the ecological zone in which maize is cultivated. Ptr is the real producer price of 
rice: this is another crop which competes with maize for agricultural resources. Ptnpk is the 
real price o f the fertiliser used in the production o f maize, T is the mean annual temperature 
in Ghana’s maize belt and R is the mean annual rainfall in Ghana’s maize belt. T and R 
represent climate variables in this study, and et is the error term which satisfies the classical 
normal regression assumptions. The a priori signs o f the respective arguments o f the function 
in equation (5) are as follows: aj > 0, a2 < 0, a3 < 0, a4  < 0, a5< 0, a6> 0.
Notably, the prices o f herbicides, maize seed, fungicides and agricultural wage rate 
relevant to maize production could be important additional arguments o f the maize supply 
function, but lack o f reliable complete time series data on these variables precludes their 
consideration in the present study.
3.6. Method of Estimation
Descriptive statistics of the variables employed in this study are analysed briefly, viz., 
mean, variance, skewness, kurtosis, median, inter alia. The coefficient o f  variation is 
computed in order to ascertain the level o f variation o f each variable. Prior to the estimation 
of the model using annual time series data, the degree of integration o f each of the variables 
is ascertained. Thus, unit root tests are carried out.
The Augmented Dickey -  Fuller (ADF) unit root test and the Phillip-Perron test are 
used in this study to identify the order o f integration o f each variable (that is, the number of 
times a variable needs to be differenced to make it stationary). The ADF procedure involves 
the estimation of the following regression:
29
AXt = a  + pt + ( p -1  ) Xt-i + ZiX; A X(_i + et (6)
where Xt is the variable under consideration and t* is the trend term. The ADF approach 
tests the null hypothesis that a series does contain a unit root (that is, it is non-stationary) 
against the alternative o f stationarity. The order of integration is determined by the stage at 
which the ADF test confirms the stationarity o f the variables. If the ADF test shows that the 
variable is stationary in level, the variable is said to be integrated of order zero [/(0)]. If 
stationarity is confirmed when the variable is in first difference, then the variable is said to be 
integrated o f order one [/(l)]; in second difference, the variable is integrated of order two 
[/(2)]; and so on. The Philip-Perron test is used in the event o f  the existence of structural 
breaks in the time series data.
Cointegration is said to exist when a linear combination o f a set of time series is 
stationary, given that the individual series are non-stationary (Engle and Granger, 1987). 
Cointegration o f two or more time series suggests that there is a long-run or equilibrium 
relationship between them. Two conditions must be satisfied for variables to be co-integrated. 
First, the series for the individual variables must be non-stationary. Second, a linear 
combination o f the non-stationary variables from a static regression involving levels o f the 
variables must be stationary.
It is possible to have a mixture o f different order series when there are three or more 
time series variables in the model. For example, it is possible that cointegration is present 
when there is a mix of /(0), 7(1) and 1(2) variables in a model (Harris, 1995). Stationary, viz., 
1(0) variables, might play a key role in establishing a sensible long-run relationship between 
non-stationary variables, especially if theory suggests a priori that such variables should be 
included in a model (Ibid). According to Pagan and Wickens (1989), in the case where 
variables are integrated of different orders, a subset of the higher-order series must
30
cointegrate to the order o f the lower-order series. So, if yt~ 7(1), xt~/(2) and zt~ /(2) then as 
long as we can find a cointegration relationship between xt and z( such that v, (vt = xt - Xz,) ~ 
/(l), then vt can potentially cointegrate with yt to obtain wt (wt = yt - £vt ) ~ 1(0). Also, if there 
are n > 2 variables in a model, there can be more than one cointegration vector. It is possible 
for up to n -  1 linearly independent cointegration vectors to exist, and this has implications 
for testing and estimating cointegration relationships (Harris, 1995). Johanssen (1985) also 
found some mathematically exact and attractive results for the general case which do not rely 
on the assumption that all components o f xt are integrated o f the same order. He points out 
that, if  xu is 7(1) and x2l is 7(0), then x ]t and the mean o f x2t could be cointegrated; thus, 
expanding the class o f variables that might be tested.
Based on the above literature, the long-run equation of the model in this study can be 
estimated as in equation (5), since as will be evident in Chapter 4, the climate variables are 
integrated o f order 7(0), prices of own and competing commodities integrated at 7(1), maize 
supply integrated at 7(1), as well as price of fertiliser integrated as 7(1). The parameters are 
estimated using the Ordinary Least Square (OLS) method. These parameters describe the 
nature o f the long-run relationship between the supply o f maize in Ghana and the relevant 
determinants including the climate variables.
The Error Correction Model (ECM) developed by Engle and Granger is a means of 
reconciling the short run behaviour of an economic variable with its long-run behaviour 
(Gujarati, 1995). Granger and Engle (1985) proved that cointegrated series have an ECM 
representation and conversely, that ECMs generate cointegrated series, thus reconciling the 
two approaches as well as clarifying when levels information could be legitimately retained 
in econometric equations. The ECM is closely bound up with the concept o f cointegration. 
Engle and Granger (1987) show that if y( and x, are cointegrated Cl (1, 1), then there must 
exist an ECM.
31
A specification o f the ECM involves the first difference of the dependent variable as a 
function o f the distributed lags o f the first differences o f the independent variables as well as 
the one period lagged equilibrium error, referred to as the error correction term. In effect, it 
reinstates the levels, and hence the long-run considerations into the differences specification 
that describes the short-run relationships between the variables (Gujarati, 1995).
Hallam and Zanoli (1993) have demonstrated the relevance of the ECM to agricultural 
response, and how it avoids the partial adjustment mechanism’s unrealistic assumption o f a 
fixed target supply based on stationary expectations. In the present study, the ECM is formed 
for maize supply in the form:
AQt = ao + aiAPmt + a2AP°t + ajAPrt + a4APnpkt +asATt + agARt + a7en + Ut (7)
where Ut is the error term and the other variables which are in natural logarithms are as 
already specified.
The Granger-Causality Test (Granger, 1969) is conducted to find out the causal 
relationships which exist between the relevant variables (the supply of maize, rainfall and 
temperature) in the study, since cointegration says nothing about the direction o f the causal 
relationship that exist between the variables. The concept of Granger-causality is that there 
must be causality in at least one direction if  two variables are found to be cointegrated. 
Granger-causality involves regressing a variable say Z on lagged values o f itself and another 
variable say W. The model is:
Z t  =  Z i = i n O iW ,_ i  +  Z i = i n p i Z i - i + U t  ( 8 )
32
where n represents the number o f lags and Ut is the error term. Causality is determined by 
performing the relevant F-test.
In the present study, a series o f experiments are conducted using the cointegration 
equation as specified in equation 5. Thus, different versions of this equation are experimented 
with. First, the dependent variable, supply of maize is regressed on the independent variables, 
namely, prices o f  maize, rice, cassava, fertiliser, the annual mean temperature and rainfall. 
The same regression procedure is performed with the squares o f  annual mean temperature 
and rainfall. The process is repeated with the annual mean temperature being replaced by 
annual mean maximum and minimum temperature. The regression is further repeated with 
the rainfall in the critical (major and minor) periods and non-critical period in place o f the 
annual mean rainfall. The linear and square o f annual mean temperature and rainfall could 
not be regressed together because the results showed a near-singular matrix.
3.7. Description of Variables and Sources of Data
Annual time series data on maize supply in Ghana, the producer prices o f maize, rice 
and cassava, and the price o f  fertiliser (NPK.15-15-15) are obtained from the Ministry of 
Food and Agriculture, Ghana. The climate data for the major maize growing areas are 
obtained from the Ghana Meteorological Services Department. The real producer prices of 
maize, rice, cassava and fertiliser are obtained by deflating the available wholesale prices, 
since the relevant farm gate prices are not readily available. The deflator is the Consumer 
Price Index (CPI) and it is obtained from various issues o f the International Financial 
Statistics (IFS), published by the International Monetary Fund (IMF). The base year is 1994. 
The various annual time series data used in this study cover the period 1970 to 2002.
Eleven weather stations are selected across the country. The selection is based on the 
weather stations located within the major maize producing areas in each region. Mean
33
monthly temperature and rainfall data are obtained from the Ghana Meteorological Services 
Department for these weather stations. Mean monthly maximum and minimum temperatures 
are also similarly acquired. The mean monthly temperature in each year is computed for each 
station and then averaged again using the eleven stations to get the annual mean monthly 
temperature for the country. A similar method is used for rainfall and the maximum and 
minimum temperatures.
To acquire the relevant data on rainfall during the critical periods (major and minor) 
and the non-critical period for maize, the rainfall during the months of major and minor rains, 
as well as the months with no or very few rains are computed for each station and then 
averaged for the whole country. The months o f the major rains are March, April, May and 
June, which also represent the major planting season for maize. The minor rains are 
experienced in the months o f July, August and September. Cultivation during the minor 
season is only recommended for the forest and transition zones and not for the coastal 
savannah. The non-critical periods are therefore the months o f October, November, 
December, January and February.
34
CHAPTER 4 
EMPIRICAL RESULTS
In this chapter, the time profiles o f  annual mean temperature, the annual mean 
deviation o f  temperature, the annual mean rainfall, the annual mean deviation o f rainfall, and 
the output o f maize in Ghana over the period o f the study are described. The empirical 
estimates o f  the effect o f  climate change on maize supply in Ghana are also presented.
4.1. Time Profile of Annual Temperature
Figure 1 depicts the annual mean temperature in the maize growing areas in Ghana 
during the period 1970 -  2002. Temperature has been increasing gradually from the 
beginning o f the stipulated time period. It however decreased between 1973 and 1976, but the 
general rise continued from that period. The highest temperature over the period o f the study 
occurred in 1999 when it reached 28°C. Peak temperatures were also recorded in 1983,1987, 
1990,1995 and 1998. Global warming due to the greenhouse effect may be part o f the causes 
o f the observed general increase in temperature.
Temperature increase may accelerate the rate at which plants release carbon dioxide 
(CO2 ) in the process o f respiration, resulting in less than optimal conditions for net growth 
(Ofori-Sarpong, 2001). Crops including maize often respond negatively with a steep drop in 
net growth and yield when temperatures exceed the optimal for biological processes (Ofori- 
Sarpong, 2001). The optimum range of temperature for the growth of maize is 16-32 degrees 
Celsius (Rouanet, 1992), but the most favourable temperature during germination is 18.3 
degrees Celsius (Berger, 1962). Temperature below 10 degrees Celsius would result in lower 
crop yields (Berger, 1962), and fertilisation is disrupted when temperature exceeds 35 
degrees Celsius (Rouanet, 1992).
35
Mean Temperature in Degrees Celsius
Figure 1. Annual Mean Temperature in Maize Growing Areas in Ghana, 1970-2002
Years
Source: Based on data obtained from the Ghana Meteorological Services Department Accra, Ghana.
The deviations o f annual temperatures from the mean temperature are depicted in 
Figure 2. A continuous increase in annual temperatures around the mean temperature is 
observed. The deviations o f temperature from 1993 to 2002 have tended to move above the 
mean. This confirms Ontoyin’s study on temperature at a number of stations in Ghana 
(Ontoyin, 1993). Increasing temperature due to global warming would further increase annual 
temperatures and their deviations from the mean.
36
Figure 2. Annual Mean Deviations in Temperature in Ghana, 1970-2002
Deviations in Degrees Celsius
Source: Author’s computations based on data obtained from the Ghana Meteorological Services Department,
Accra,Ghana.
4.2. Time Profile of Annual Rainfall
In Ghana, annual mean rainfall in the maize growing areas has tended to fluctuate 
around a trend with a zero slope (Figure 3). Peak rainfall amounts were recorded in 1974, 
1979, 1985, 1989, 1991, 1995, 1997 and 1999. The lower periods of decline in rainfall 
amounts occurred during 1977 and 1983.
The lowest amount of rainfall was recorded in 1983, which was a drought year 
(Figure 3). During times of drought, the length of the growing period for crops (maize 
inclusive) reduces due to lack or inadequate rains. Farmers are tempted to delay planting with 
the late onset o f rains, since water availability strongly influences agriculture. With the abrupt 
ending of the rainy period, hence inadequate water, the growing period too becomes short for 
crops to give good harvest (Ofori-Sarpong, 1985). Moisture stress during flowering, 
pollination and grain, filling is harmful to most crops (Leyenaar, 1976). Increasing 
evaporation from the soil and accelerated transpiration in the plants themselves cause
37
moisture stress. The intensified evaporation increases the hazard of salt accumulation in the 
soil, thus rendering it infertile. Here, inadequate rainfall precipitates lower levels of crop 
output.
Figure 3. Annual Rainfall in Maize Growing Areas in Ghana, 1970-2002 
Annual Rainfall (mm)
Years
Source: Based on data obtained from the Ghana Meteorological Services Department, Accra, Ghana.
The deviations o f rainfall around the mean show fluctuations around the mean (Figure 
4). Rainfall amounts fell below the mean in some years. The lowest deviation was recorded in 
the drought year (specifically, 1983) and the highest was recorded in 1979. In some years, the 
amount o f rainfall hovered around the mean, and did not deviate much from the mean.
38
Figure 4. Annual Rainfall Deviations from the Mean. 1970-2002 
Rainfall Deviations (mm)
Source: Author’s computations based on data obtained from the Ghana Meteorological Services Department, 
Accra,Ghana.
4.3. Time Profile of Maize Supply
In 1983 when there was drought, maize supply decreased precipitously (Figure 5). 
This indeed is the lowest supply o f maize recorded during the period o f the study. In years of 
decreased rainfall, supply of maize falls whereas increases in rainfall result in increases in the 
supply o f maize. It is observed that from 1985 to the end of the period o f the study, though 
the amount of rainfall recorded was not far from the mean, the supply of maize increased 
gradually. This could be explained by the use of fertilisers by farmers, improved cultivars of 
maize and early planting of seeds to mention but a few. Generally, increases in the annual 
mean temperature tend to be associated with increases in the supply o f maize. Maize needs 
light and solar energy for photosynthesis, which is the food-producing process for plants. 
This increases plant growth and output of crops. However, it should be noted that other
39
factors could influence this relationship; for example, the prices of competitive crops like the 
price o f rice and cassava.
Figure 5. Comparison of Maize Supply with Annual Rainfall, 1970-2002
Source: Data on maize supply are obtained from the Ministry o f Food and Agriculture, whereas data on rainfall 
are obtained from the Ghana Meteorological Services Department, Accra, Ghana.
The output o f maize and the price o f rice are generally indirectly related as can be 
observed in Figure 6. This suggests that maize and rice are competing crops; they compete 
for the land and other factor inputs of farmers. Farmers in anticipation of prices, carry out 
their cultivation accordingly. Hence, when they have a favourable anticipation of future
40
prices o f a crop like maize, they tend to plant more at the expense of say rice and vice versa. 
Hence, a prediction o f a high price of rice could precipitate a decrease the cultivation of 
maize and the supply of maize grain. An inverse relationship is observed between the price of 
rice and the output of maize (Figure 6). A notable observation in the time profile of the price 
o f rice is the fluctuation witnessed between 1980 and 1994. A steady rise in the price of rice 
is observed after 1995.
Figure 6. Comparison of Maize Supply with the Prices of Rice, 1970-2002
Source: Data on maize supply and price of rice are obtained from the Ministry of Food and Agriculture, Accra, 
Ghana.
The time profile o f the price of cassava is not very different from that o f the price of 
rice. The shock in prices that was witnessed in 1983 is also revealed in the price of cassava
41
(Figure 7). As with the price of rice, there seem to exist an inverse relationship between the 
price o f cassava and the output of maize. Here again, an indication of cassava being a 
competing crop in relation to maize, specifically in land use.
Figure 7. Comparison of Maize Supply with Price of Cassava, 1970-2002
Source: Data on maize supply and price of cassava are obtained from the Ministry of Food and Agriculture, 
Accra, Ghana.
4.4. Descriptive Statistics of the Variables in the Econometric Model
The definitions of variables used are presented in Appendix 1. The descriptive 
statistics of the raw data concerning the relevant variables are presented in Table 1. The
42
Jarque-Bera statistic is significant for rain, maximum temperature, prices o f cassava, maize, 
fertiliser (NPK: 15-15-15), the square o f  maximum temperature and the square of the non- 
critical rainfall period. This indicates that the null hypothesis o f a normal distribution for each 
o f these variables is rejected in favour o f the alternative hypothesis that these variables are 
not normally distributed. This corroborates the fact that the skewness coefficient (which is 
equal to zero for normal distribution), is not equal to zero for each o f these variables. In 
addition, the estimated kurtosis coefficients for these variables are generally greater than 3.
Supply o f  maize, mean rainfall, mean rainfall during the non-critical period, price o f 
cassava, the square o f  the critical major and minor rainfall period, the square o f rain and the 
square o f  maximum temperature have a kurtosis o f less than 3. The data therefore seem not to 
follow a normal distribution. The estimated coefficient o f  variation shows low variance in 
temperature and its square and maximum temperature and its square (Table 1).
The descriptive statistics for the transformed data, viz., the natural logarithm o f the 
price o f  rice, the natural logarithm o f the square o f temperature, the natural logarithm o f the 
square o f the non-critical rainfall period, the logarithm o f the critical major rainfall period, 
the natural logarithm o f temperature, the natural logarithm o f the maximum and minimnm 
temperature and the natural logarithm o f the price o f fertiliser (npk) are statistically 
significant according to the Jarque-Bera statistic and therefore not normally distributed 
(Appendix 2).
The correlation matrix shows that the climate variables, namely, temperature and 
rainfall, and their squares are characterised by high correlation (Appendix 3). There is a high 
positive correlation (0.776) between mean rainfall and mean critical minor rainfall as well as 
the square o f  mean rainfall and the square o f mean critical minor rainfall. Similarly, mean 
temperature is highly correlated with mean minimum temperature (0.934). A high correlation 
is observed between the price fertiliser and the supply of maize (0.871). The price o f  maize
43
and the prices o f  cassava are also highly correlated (0.966) and the correlation between the 
prices o f  rice and cassava is 0.909.
The price o f  maize and the price o f  rice have a correlation of 0.939. Similarly, high 
correlation is observed between the square o f  minimum temperature and the square mean 
temperature (0.934). Minimum temperature and mean temperature are also highly correlated 
(0.934) as well as the square o f  the mean minimum temperature and mean temperature.
44
able 1. Descriptive Statistics of the Relevant Variables: Raw Data, 1970-2002
Maize Rain Temp Maxt Mint NonC Pxcass Pxmaize Pxnpk Pxrice
MEAN 631.8 93.9 27.03 31.96 20.43 38.11 87.83 204.09 1.219 483.574
MEDIAN 559.1 94.9 27.05 31.93 20.43 39.22 75.38 176.74 0.668 470.074
MAXIMUM 1400.0 115.5 28.03 33.16 21.08 63.44 374.25 964.50 5.148 1503.00
MINIMUM 140.8 55.07 26.31 31.34 19.77 17.09 -108.43 -195.82 0.035 -309.62
STD. DEV. 315.72 11.654 0.350 0.438 0.284 9.240 85.999 209.438 1.450 410.773
SKEWNESS 0.472 -0.898 0.294 0.993 -0.184 0.157 1.123 1.305 1.420 0.746
KURTOSIS 2.184 5.141 3.88 3.99 3.421 3.906 5.773 6.512 4.190 3.364
C. V. 0.500 0.124 0.013 0.014 0.014 2.242 0.979 1.026 1.190 0.849
J. B. 2.142 10.74 1.54 6.772 0.430 1.266 17.510 26.323 13.035 3.242
PROB 0.343 0.005 0.463 0.034 0.806 0.531 0.000 0.000 0.001 0.198
OBS 33 33 33 33 33 33 33 33 33 33
Maize denotes output of maize, Pxmaize denotes real price of maize, Pxcass denotes real price o f cassava, Pxnpk denotes real price of fertiliser, Pxrice denotes real price of 
rice, Rain denotes annual mean rainfall, Temp denotes annual mean temperature, Maxt denotes annual maximum mean temperature, M int denotes annual minimum mean 
temperature, NonC denotes annual non-critical rains, C.V denotes coefficient of variation, J.B denotes Jarque-Bera statistic, PROB denotes probability, OBS denotes 
observation.
Source: Author’s computations.
45
Continuation of Table 1.
Sqcrmaj Sqcrmin Sqrain Sqtemp Sqmaxt Sqmint SqnonC Crmaj Crm in
MEAN 133725.8 245838.6 8951.03 730.64 1021.56 417.288 1535.47 354.91 457.96
MEDIAN 116262.4 207941.5 8999.97 731.68 1019.74 417.188 1538.35 340.97 456.01
MAXIMUM 256582.0 597230.5 13353.81 785.59 1099.68 44.142 4024.92 506.54 772.81
MINIMUM 48866.1 19849.3 3033.50 692.31 982.015 390.794 292.099 221.06 140.89
STD. DEV. 65432.7 178592.6 2083.11 18.93 28.192 11.603 731.648 89.49 192.99
SKEWNESS 0.497 0.477 -0.291 0.348 1.034 -0.135 1.163 0.191 -0.037
KURTOSIS 2.074 2.081 3.770 3.959 4.090 3.415 5.490 1.895 1.841
C. V. 0.489 0.726 0.233 0.026 0.028 0.028 0.476 0.252 0.421
J. B. 2.540 2.414 1.280 1.929 7.512 0.337 15.957 1.879 1.854
PROB. 0.281 0.299 0.527 0.381 0.023 0.845 0.000 0.391 0.396
OBS 33 33 33 33 33 33 33 33 33
Sqmaxt denotes square annual maximum temperature, Sqmint denotes square annual minimum temperature, SqnonC denotes square non-critical annual mean rains, C.V denotes coefficient of 
variation, J.B denotes Jarque-Bera PROB denotes probability, OBS denotes observation.
Source: Author’s computations.
46
Before undertaking the econometric estimations, unit root tests using the Augmented 
Dickey-Fuller (ADF) and Phillip-Perron (PP) approaches are conducted to find out whether 
each variable is characterised by stationarity (or non-stationarity) and the order of integration 
o f the variable. The Philip-Perron test is carried out to deal with some o f the weaknesses of 
the Augmented Dickey-Fuller test, like dealing with the existence o f structural breaks.
In conducting the ADF test on the variables under consideration, an attempt is made 
to determine the appropriate number o f lagged difference terms to include in the ADF model. 
In this study, using the general ADF model with trend and intercept, unit root tests were 
conducted using lags from 1 to 5. The Akaike Information Criterion (AIC) and Schwartz- 
Bayes Information Criterion (SBIC) values for each model o f a given lag were examined and 
compared. The number o f  lags in the model which had the smallest SBIC and AIC values 
with the DW statistic close to 2 was chosen as the appropriate number o f lags to be used for 
the test (Johnston and DiNardo, 1997). For all the variables, both information criteria suggest 
that the ADF model should contain only 1 lag of the differenced term. Then the trend and 
intercept terms are closely examined. Models with insignificant trend and intercept terms 
were dropped for models with only the intercept In cases where the intercepts were also not 
significant, the unit root tests were conducted using models with no intercept and trend terms. 
Column 2 o f Table 2 shows the type o f ADF model used for each variable tested for 
stationarity. It is clear from Column 5 of Table 2 that all the price variables and maize 
quantity are integrated of order 1 viz., 1(1), while the climate variables are integrated o f order
To corroborate the results of the ADF unit root tests, the study further conducted 
stationarity tests using the Phillip-Perron approach (Table 3). The results o f the Phillip-Perron 
test are consistent with the results o f the Augmented Dickey-Fuller test.
4.5. Empirical Results o f Stationarity Tests
47
Table 2. Results of Unit-Root Test (Augmented-Dickey Fuller) for 1970 to 2002
Variables Model t-statistic in level t-statistic in Is1 
difference
Order of Integration
Maize output T & I -2.466458 -5.221123*** I d )
Real price o f maize None 0.835426 -835426*** KD
Real price o f rice None 0.751117 -3.921808*** KD
Real price o f cassava None 0.966162 -5.091725*** KD
Real price o f fertiliser T & I -2.433526 -4.838637*** KD
Temperature T & I -4.826504*** 1(0)
Temperature square T & I -4.826504*** 1(0)
Rain I -4.097348*** 1(0)
Rain square I -4.097348*** 1(0)
Max. temp. T & I -5.916444*** 1(0)
Min. temp T & I -4.772944*** 1(0)
Max.temp.square T & I -5.916444*** 1(0)
Min.temp.square T & I -4.772944*** 1(0)
Critical major rains I -4.164124*** 1(0)
Critical major rains square I -4.164124*** 1(0)
Critical minor rains I -4.176165*** 1(0)
Critical minor rains square I -4.176165*** 1(0)
Non-critical rains I -3.221440*** 1(0)
Non-critical rains square I -3.221440*** 1(0)
McKinnon Tables are used for the rejection Critical Values 
T & I represent Trend and Intercept 
I represent Intercept
“None” represents no Trend and Intercept
*** denotes significance at the 1 percent level, **denotes significance at the 5 percent level 
All variables are in natural logarithms and climate variables are in annual means as well. 
E-views econometric software is used in the performance of the tests.
Source: Author’s computations.
48
Table 3. Results of Unit-Root Test (Phillip-Perron) for 1970 to 2002
Variables Model t-statistic in 
level
t-statistic in I” 
difference
Order o f Integration
Maize output T & I -3.369688 -10.47194*** 1(1)
Real price o f maize None 0.885123 -6.688509*** KD
Real price of rice None 0.909598 -4.615568*** KD
Real price of cassava I -1.832079 -7.409832*** UD
Real price of fertiliser T & I -2.412061 -5.758878*** 1(1)
Temperature T & I -6.217564*** 1(0)
Temperature square T & I -6.217564*** 1(0)
Rain I -6.037438*** 1(0)
Rain square I -6.037438*** 1(0)
Max. temp. T & I -7.159166*** 1(0)
Min. temp T & I -5.209286*** KO)
Max.temp.square T & I -7.159166*** 1(0)
Min.temp.square T & I -5.209286*** KO)
Critical major rains I -5.886277*** 1(0)
Critical major rains square I -5.886277*** 1(0)
Critical minor rains I -6.099075*** 1(0)
Critical minor rains square I -6.099075*** KO)
Non-critical rains I -4.694971*** 1(0)
Non-critical rains square I -4.694971*** 1(0)
McKinnon Tables are used for the rejection Critical Values 
T & I represent Trend and Intercept 
I represents Intercept
“None” represents no Trend and Intercept
*** denotes significance at the 1 percent level, ** denotes significance at the 5 percent level 
All variables are in natural logarithm and climate variables are in annual means as well 
E-views econometric software is used in the performance of the tests.
Source: Author’s computations.
49
4.6. Empirical Results of Cointegration Analysis
Unit root tests are performed on the residuals o f the present study’s econometric
models as specified in equation 5 on page 28 to find out whether the variables in each o f the 
models are cointegrated. Testing for the stationarity o f the error term in order to establish 
whether the relevant variables are cointegrated provides a way to determine whether there is a 
long run relationship between these variables (Johnston and DiNardo, 1997). The error terms 
are significant and therefore integrated o f order zero [namely, I  (0)].
The variables in Model 1 are thus cointegrated; hence, Model 1 shows the relevant 
long-run equilibrium relationship between maize supply, mean annual rainfall and 
temperature in the maize belt, the prices of maize, rice, fertiliser (npk: 15-15-15), and cassava 
(Table 4). The relevant ADF statistic of -3.529888 is significant at the 1 percent level, 
implying a rejection o f  the null hypothesis o f non-stationary in favour of the alternative 
hypothesis o f stationarity.
The R2 is approximately 95 percent, meaning that the model explains 95 percent of 
the variation in maize supply. The F statistic with a probability value o f (0.000) shows that 
the explanatory variables jointly explain maize supply. The estimated Durbin-Watson statistic 
of 2.23 shows the absence of an autocorrelation problem.
The econometric results for Model 1 further show that annual mean rainfall exerts a 
significant effect on the output o f maize at the 1 percent level in the long-run. The positive 
coefficient of rainfall indicates that an increase in rainfall during the growth period o f maize 
tends to stimulate an increase in maize supply, whereas a reduction in rainfall decreases 
maize supply. The significance of rainfall indicates how critical it is for maize to receive 
adequate rain at the developmental stage and its dependence on rainfall. Maize production in 
Ghana is largely rainfed. A 1 percent increase (decrease) in mean annual rainfall stimulates a 
0.799 percent increase (decrease) in maize supply. A similar result was found by Muchena
50
(1994) as well as Muchena and Iglesias (1995) who observe that lower rainfall leads to lower 
maize output.
The significant negative coefficient o f the price of rice at the 1 percent level shows 
that rice competes with maize for agricultural resources. A 1 percentage increase (decrease) 
in the price o f  rice tends to precipitate a 0.3058 percent decrease (increase) in the supply of 
maize.
Mean annual temperature, the prices o f maize and cassava are not significant even at 
the 10 percent level. The coefficient o f the price o f fertiliser (NPK: 15-15-15), though 
statistically significant, does not conform to the a priori expectation. It is worth noting that 
this result is rather intriguing, since farmers are likely to purchase less fertiliser in the event 
of a rise in the price o f  fertiliser. Future studies could re-examine this issue.
Table 4. Cointegrating Equation: Model 1
Dependent Variable: Imaize 
Sample: 1970 - 2002 
Included observations: 33
Variable Coefficient Std. Error t-Statistic Prob.
C -5.089971 9.887331 -0.514797 0.6110
Lrain 0.799244 0.211853 3.772631 0.0008
Ltemp 2.334991 3.010569 0.775598 0.4450
Lrpxmaize -0.026529 0.161078 -0.164694 0.8705
Lrpxrice -0.305780 0.111928 -2.731938 0.0112
Lrpxcassava 0.021508 0.125097 0.171933 0.8648
Lrpxnpk 0.403369 0.028034 14.38841 0.0000
R-squared 0.947318 Mean dependent var 6.315912
Adjusted R-squared 0.935161 S.D. dependent var 0.542588
S.E. of regression 0.138162 Akaike info criterion -0.934943
Sum squared resid 0.496310 Schwarz criterion -0.617502
Log likelihood 22.42657 F-statistic 77.92113
Durbin-Watson stat 2.231349 Prob(F-statistic) 0.000000
Jarque-Bera 3.530558 Prob(Jarque-Bera) 0.171139
Breusch-Godfrey 1.053805 Prob(Breusch-Godfrey) 0.314463
White Heteroscedasticity Test 0.353929 Prob(White Heteroscedasticity Test) 0.965717
Ramsey RESET Test 0.825026 Prob(Ramsey RESET Test) 0.372388
ADF Test Statistic of the Error Term -3.529888 1% Critical Value* -2.6395
5% Critical Value -1.9521
10% Critical Value -1.6214
. r  f ricc. Lrpxcassava denotes
the log of the real price of cassava. Lrpxnpk denotes the log of the real price of fertiliser (npk). Ltemp denotes the log of 
annual mean temperature. Lrain denotes the log of annual mean rainfall. I maize denotes output of maize.
Source: Author’s computation.
51
To analyse the effects o f  minimum temperature and maximum temperature on the 
supply o f maize, Model 1 is revised by substituting minimum and maximum temperatures for 
mean temperature to give Model 2. The error term o f Model 2 is found to be stationary at the 
1 percent level (Table 5). This implies the existence o f a long-run relationship between 
supply o f maize and mean annual rainfall, maximum and minimum temperature, prices of 
maize, rice, cassava and fertiliser. The high value o f the R2 demonstrates a high explanatory 
power of the independent variables in the model. The F-statistic with a low probability value 
o f  (0.000) shows that the explanatory variables jointly influence maize supply. The Durbin- 
Watson statistic with a value o f  2.488 lies in the inconclusive region o f the DW statistic table 
o f range 2.224 and 3.184 in this model. The Breusch-Godfrey Test for first order 
autocorrelation shows no autocorrelation problem. Similarly, the Ramsey RESET Test does 
not reject the null hypothesis o f  no model misspecification.
In this model, the coefficient o f rainfall is again positive and significant at the 1 
percent level (Table 5). A 1 percent increase (decrease) in rainfall tends to stimulate 
(precipitate) a 0.857 percent increase (decrease) in maize supply in the long-run. An increase 
in the elasticity o f maize with respect to rainfall is observed here. The increase is from 0.799 
in the previous model to 0.857 in this model. Thus, maize supply seems to be more sensitive 
to rainfall when temperature is analysed using its maximum and minimum components. The 
price of rice still exhibits the significant competitive relationship between rice and maize 
supply.
Maximum temperature exerts a positive effect whereas minimum temperature exerts a 
negative effect, but both are not significant even at the 10 percent level. The insignificance of 
maximum and minimum temperatures cannot be attributed to multicollinearity because their 
correlation coefficient o f 0.6 and those between minimum temperature and the other variables 
as well as those between maximum temperature and the other variables in the model are not
52
high. The relevant correlation coefficients hardly exceed 0.75 (Appendix 3) with the 
exception o f mean square temperature and mean temperature (0.934). The coefficient o f the 
price o f  maize in model 2 is seen to be positive but still not significant (Table 5). The price of 
cassava is not significant even at the 10 percent level. The price o f fertiliser, though 
significant at the 1 percent level, still shows a positive coefficient, which is contrary to 
expectation.
Table 5. Cointegrating Equation: Model 2
Dependent Variable: Imaize 
Sample: 1970 - 2002 
Included observations: 33
Variable Coefficient Std Error t-Statistic Prob.
C -8.167712 8.754656 -0.932956 0.3598
Lrain 0.857075 0.218731 3.918390 0.0006
Lmaxtemp 3.529747 2.378796 1.483838 0.1504
Lmintemp -0.607273 3.244894 -0.187147 0.8531
Lrpxmaize 0.024804 0.166946 0.148576 0.8831
Lrpxcassava 0.009843 0.123777 0.079524 0.9372
Lrpxrice -0.334321 0.116685 -2.865149 0.0083
Lrpxnpk 0.405554 0.025282 16.04131 0.0000
R-squared 0.951321 Mean dependent var 6.315912
Adjusted R-squared 0.937691 S.D. dependent var 0.542588
S.E. of regression 0.135439 Akaike info criterion -0.953368
Sum squared resid 0.458596 Schwarz criterion -0.590578
Log likelihood 23.73057 F-statistic 69.79576
Durbin-Watson stat 2.487669 Prob(F-statistic) 0.000000
Jarque-Bera 3.093605 Prob( Jarque-B era) 0.212928
Breusch-Godfrey 3.146688 Prob(Breusch-Godfrey) 0.088771
White Heteroscedasticity Test 0.328683 Prob(White Heteroscedasticity Test) 0.980139
Ramsey RESET Test 0.459465 Prob(Ramsey RESET Test) 0.504357
ADF Test Statistic of the Error Term -4.078141 1% Critical Value* -2.6395
5% Critical Value -1.9521
10% Critical Value -1.6214
Lrpxmaize denotes log of the real price of maize. Lrpxrice denotes the log of the real price of rice. Lrpxcassava denotes 
the log of the real price of cassava. Lrpxnpk denotes the log of the real price of fertiliser (npk). Lmaxtemp denotes the log 
of annual mean maximum temperature. Lmintemp denotes the log of annual mean minimum temperature Lrain denotes the 
log of annual mean rainfall. Source: Authors compulation.
In order to analyse the effects o f three other types of rainfall indicators, Model 1 is 
revised to include the amount o f rainfall during the critical major rainfall periods, amount of 
rainfall during the critical minor rainfall periods and the amount o f rainfall during the non-
53
critical rainfall periods instead o f the mean annual rainfall, giving Model 3 (Table 6). Thus, 
climate variables included in the model are mean temperature, amount of rain in the critical 
major rainfall period (March, April, May and June) and minor rainfall period (My, August 
and September) and the non-critical rainfall period (January, February, October, November 
and December).
The ADF test statistic on the error term in Model 3 is significant, implying a 
significant long-run relationship between the variables (Table 6). The adjusted R2 value is 
about 0.9366, showing an improvement in explanatory power over those of the previous 
models. The F statistic is significant at the 1 percent level, indicating that the explanatory 
variables in the model jointly explain maize supply. The estimated Durbin-Watson statistic 
with a value o f 2.3687 lies in the inconclusive zone in the DW table. The coefficient of 
critical major rainfall period is positive as expected and it is significant at the I percent level.
Higher maize yields are obtained during major rainy seasons (Layenaar, 1976). The 
maize supply elasticity with respect to the mean critical major season rainfall is 0.614 (Table 
6). It is lower than the elasticity with respect to annual mean rainfall. The critical major 
rainfall is that recorded during the period when plants receive the maximum amount o f water 
which it needs for development. This is in consonance with Endrody-Younga (1968), who 
observed that maize planted in the major season have higher output since crop growth 
escapes heavy pest infestation and mature before insects could cause damage. Water 
availability also makes plant use o f fertilisers more effective. The minor critical rainfall 
period is not significant even at the 10 percent level. This could be due to the fact that it is not 
always economically viable to plant in the minor critical period since planting during this 
period has to be done with irrigation, fertilisers, insecticides and improved cultivars of maize. 
The capital cost o f irrigation is too much for the average farmer and the market price of
54
maize does not warrant such investment (Leyenaar, 1976). The coefficient o f the non-critical 
period o f rainfall is not significant even at the 10 percent level.
The price o f  rice is significant at the 5 percent level and has a negative coefficient. 
The price o f  fertiliser is significant but has an unexpected positive coefficient. The prices of 
cassava and maize are not significant even at the 10 percent level.
Table 6. Cointegrating Model 3
Dependent Variable: lmaize 
Sample: 1970 - 2002 
Included observations: 33
Variable Coefficient Std. Error t-Statistic Prob.
C -10.19682 10.30724 -0.989286 0.3324
Ltemp 3.660503 3.106074 1.178498 0.2502
Lcnnj 0.614095 0.207304 2.962296 0.0068
Lcrminor 0.159845 0.128204 1.246798 0.2245
Lnoncrit 0.175061 0.107150 1.633795 0.1154
Lrpxmai2e -0.150204 0.181168 -0.829086 0.4152
Lrpxrice -0.266200 0.114138 -2.332260 0.0284
Lrpxcassava 0.121986 0.140520 0.868101 0.3939
Lrpxnpk 0.388703 0.029115 13.35051 0.0000
R-squared 0.952477 Mean dependent var 6.315912
Adjusted R-squared 0.936636 S.D. dependent var 0.542588
S.E. of regression 0.136581 Akaike info criterion -0.916797
Sum squared resid 0.447705 Schwarz criterion -0.508659
Log likelihood 24.12715 F-statistic 60.12763
Duibin-Watson stat 2.368713 Prob(F-statistic) 0.000000
Jarque-Bera 3.407320 Prob(Jarque-Bera) 0.182016
Breusch-Godfrey 1.401034 Prob(Breusch-Godfrey) 0.248640
White Heteroscedasticity Test 0.291865 Prob(White Heteroscedasticity Test) 0.990744
Ramsey RESET Test 0.259101 Prob(Ramsey RESET Test) 0.615587
ADF Test Statistic of the Error Term -4.568020 1% Critical Value* -2.6395
5% Critical Value -1.9521
10% Critical Value -1.6214
Lrpxmaize denotes log of the real price of maize. Lrpxrice denotes the log of the real price of rice. Lrpxcassava denotes 
the log of the real price of cassava Lrpxnpk denotes the log of the real price of fertiliser (npk). Ltemp denotes the log of 
annual mean temperature. Lcrmj denotes log of the amount of rainfall in the critical major rainfall periods. Lcrminor 
denotes log of the amount of rainfall in the critical minor rainfall periods. Lnoncrit denotes log of the amount of rainfall 
during the non-critical rainfall periods.
Source: Author’s computation
Results o f another variant o f Model 1 are presented in Table 7. Here, a different set o f 
climate indicators are included. The ADF statistic on the error term in Model 4 is significant 
and hence the variables in the model are cointegrated, implying a long-run relationship 
between the variables (Table 7). The value of the R2 is 0.95, meaning that the model explains
55
95 percent o f the variation in the supply o f  maize. The Durbin-Watson statistic o f 2.61 lies in 
the inconclusive region using the DW table. The Breusch-Godfrey test shows the absence of 
first order autocorrelation. The F statistic is significantly different from zero, implying a 
significant joint effect o f  the explanatory variables.
Notably, the new climate variables in Model 4 are the amount o f rainfall during the 
critical major rainfall period (March, April, May and June), minor rainfall period (July, 
August and September), and non-critical rainfall (January, February, October, November and 
December) periods, the maximum and minimum temperatures. In this model, both the critical 
major rainfall and the critical minor rainfall are positive and significant at the 1 percent level 
and the 10 percent level respectively.
Table 7. Cointegrating Equation: Model 4
Dependent Variable: Imaize 
Sample: 1970 - 2002 
Included observations: 33
Variable Coefficient Std. Error t-Statistic Prob.
C -10.47062 9.115326 -1.148683 0.2625
Lcrmj 0.572933 0.196929 2.909344 0.0079
Lcrminor 0.246431 0.135083 ' 1.824288 0.0811
Lnoncrit 0.151779 0.111460 1.361743 0.1865
Lmaxtemp 3.788078 2.460339 1.539657 0.1373
Lmintemp -0.350872 3.373874 -0.103997 0.9181
Lipxmaize -0.056122 0.189732 -0.295799 0.7700
Lrpxcassava 0.087818 0.140334 0.625776 0.5376
Lipxrice -0.319306 0.118631 -2.691583 0.0130
Lrpxnpk 0.397301 0.025947 15.31205 0.0000
R-squared 0.955589 Mean dependent var 6.315912
Adjusted R-squared 0.938210 S.D. dependent var 0.542588
S.E. of regression 0.134874 Akaike info criterion -0.923904
Sum squared resid 0.418393 Schwarz criterion -0.470417
Log likelihood 25.24442 F-statistic 54.98727
Durbin-Watson stat 2.610053 Prob(F-statistic) 0.000000
Jarque-Bera 2.940236 Prob(Jarque-Bera) 0.229898
Breusch-Godfrey 3.617221 Prob(Breusch-Godfrey) 0.070359
White Heteroscedasticity Test 0.252774 Prob(White Heteroscedasticity Test) 0.996353
Ramsey RESET Test 0.166679 Prob(Ramsey RESET Test) 0.687026
ADF Test Statistic of the Error -4.942032 1% Critical Value* -2.6395
5% Critical Value -1.9521
10% Critical Value -1.6214
Lrpxinaize denotes log of the real price of maize. Lrpxrice denotes the log of the real price of rice. Lrpxcassava denotes the log of
the real price of cassava. Lrpxnpk denotes the log of the real price of fertiliser (npk)- Lcrmj denotes log of the critical major
rainfall periods. Lcnninor denotes log of the critical minor rainfall periods. Lnoncrit denotes log of non-critical rainfall periods. 
Lmaxtemp denotes the log of annual mean maximum temperature. Lmintcmp denotes the log of annual mean minimum 
temperature
Source: Author’s computations.
56
The non-critical period is not significant even at the 10 percent level. The price of 
rice, which is significant at the 5 percent level, has the expected negative coefficient. The 
maximum and m in imum  temperatures are not significant. The price o f fertiliser is significant 
but unexpectedly positive. The price o f cassava and maize are not significant even at the 10 
percent level.
Table 8 reports the results o f a model with the square o f mean temperature and the 
square o f  rainfall as the climate variables. The error term in Table 8 is significant at the 1 
percent level and shows a long-run relationship between the variables in the model. The value 
o f the R2 shows that the explanatory variables in the Model explain 95 percent of the 
variation in maize supply. The F statistic is significantly different from zero. The Durbin- 
Watson statistic is 2.23 indicates that there is no autocorrelation problem in this model. The 
table further shows a similar outcome as with mean temperature and rainfall except that the 
magnitudes o f  the coefficients o f  the square o f temperatures and rainfall changed. The 
coefficient o f  temperature increased from 2.335 in the case o f  mean temperature to 1.167 in 
the case o f  the square o f  mean temperature. The coefficient o f rainfall however decreased 
from 0.799 in the mean rainfall to 0.399 in the case o f the square o f mean rainfall. Similar 
outcomes are observed for other cointegrating equations estimated, viz., Model 6, Model 7 
and Model 8 (Appendix 4 to 6).
57
Table 8. Cointegrating Equation: Model 5
Dependent Variable: Imaize 
Sample: 1970 2002 
Included observations: 33
Variable Coefficient Std. Error t-Statistic Prob.
C -5.089971 9.887331 -0.514797 0.6110
Lsqrain 0.399622 0.105927 3.772631 0.0008
Lsqtemp 1.167495 1.505284 0.775598 0.4450
Lrpxmaize -0.026529 0.161078 -0.164694 0.8705
Lipxrice -0.305780 0.111928 -2.731938 0.0112
Lipxcassava 0.021508 0.125097 0.171933 0.8648
Lrpxnpk 0.403369 0.028034 14.38841 0.0000
R-squared 0.947318 Mean dependent var 6.315912
Adjusted R-squared 0.935161 S.D. dependent var 0.542588
S.E. of regression 0.138162 Akaike info criterion -0.934943
Sum squared resid 0.496310 Schwarz criterion -0.617502
Log likelihood 22.42657 F-statistic 77.92113
Durbin-Watson stat 2.231349 Prob(F-statistic) 0.000000
Jarque-Bera 3.530558 Prob(Jarque-Bera) 0.171139
Breusch-Godfrey 1.053805 Prob(Breusch-Godfrey) 0.314463
White Heteroscedasticity Test 0.353929 Prob(White Heteroscedasticity Test) 0.965717
Ramsey RESET Test 0.825026 Prob(Ramsey RESET Test) 0.372388
ADF Test Statistic of the Error Term -3.529888 1% Critical Value* -2.6395
5% Critical Value -1.9521
10% Critical Value -1.6214
Lrpxmaize denotes log of the real price of maize. Lrpxrice denotes the log of the real price of rice. Lrpxcassava denotes 
the log of the real price of cassava. Lrpxnpk denotes the log of the real price of fertiliser (npk). Lsqtemp denotes the log of 
the square of annual mean temperature. Lsqrain denotes the log of the square of annual mean rainfall.
Source:Author’s computations.
4.7. Results of Error Correction Modelling
The error correction model (ECM) provides a useful link between long-run 
equilibrium relationships and short-run (disequilibria) dynamics (Gujarati, 1995). The 
specification o f  the ECM involves expressing the first difference o f the supply o f maize as a 
function o f the first difference o f the independent variables in the cointegrating equation 
(specifically, prices o f maize, rice, cassava and fertiliser, and the climate variables), as well 
as the one period lagged equilibrium error term. This error term is called the error correction 
term (ECT). It re-instates the levels, and hence the long-run considerations into the 
differences specification that describes the short-run relationships between the variables 
(Abdulai and Rieder, 1995).
58
The results o f the ECM for the supply o f maize with mean annual temperature and 
mean annual rainfall as climate variables, the prices of maize, rice, cassava and fertiliser 
(NPK: 15-15-15) which capture the short-run relationships are reported in Table 9. The 
general performance o f  the Error Correction Model 1 can be seen, inter alia, from the 
coefficient o f  determination (R2) and the F statistic o f  32.2264 which show the joint 
significance o f  the explanatory variables in explaining the variation in the change in maize 
supply in Ghana in the model (Table 9). The R2 value o f 0.9038 means that the model 
explains about 90 percent o f  the total variation in the change in maize supply. With respect to 
the diagnostic tests, the Jarque Bera statistic is not significant at the 5 percent level, implying 
that the null hypothesis o f normally distributed residuals is not rejected. The F-statistic o f the 
Breush-Godfrey serial correlation test is insignificant, suggesting that the residuals o f  the 
Error Correction Model 1 are not serially correlated. The White’s heteroscedasticity test also 
shows that the null hypothesis o f  homoscedastic residuals is not rejected, since the F statistic 
reported is not significant at the 5 percent level. Furthermore, the null hypothesis o f no model 
misspecification is not rejected, as indicated by the insignificant F statistic o f the Ramsey 
reset test. On the whole the diagnostic tests indicate that there is no misspecification o f the 
Error Correction Model 1.
The change in rainfall is significant with a positive coefficient (Table 9). Thus, a 1 
percent change in rainfall tends to stimulate a 0.753 percent change in maize supply. This is 
similar to the results o f Molua (2002), Muchena, (1994) and Muchena and Iglesias (1995), 
who conclude that a change in climate (for example a decrease in rainfall) may lower maize 
yield (and supply) due to water stress and a shortening o f the favourable growing period. 
Change in the price o f maize is significant at the 5 percent level. Change in the mean annual 
temperature, the prices o f rice and cassava are not significant even at the 10 percent level.
59
The change in the price o f  fertiliser is significant but the coefficient is positive, which is not 
as expected.
Table 9. Error Correction Model 1
Dependent Variable; D(lmaize)
Sample(adjusted): 1971 - 2002
Included observations: 32 after adjusting endpoints
Variable Coefficient Std. Error t-Statistic Prob.
C -0.010310 0.022207 -0.464266 0.6466
D(lrain) 0.753548 0.140303 5.370873 0.0000
D(Itemp) 1.314038 2.037076 0.645061 0.5250
D(lrpxmaize) -0.234494 0.102988 -2.276909 0.0320
D(lrpxrice) -0.067061 0.091716 -0.731175 0.4718
D(lrpxcassava) 0.114554 0.089077 1.286015 0.2107
D(lrpxnpk) 0.407435 0.057363 7.102745 0.0000
RESIDOl(-l) -1.227594 0.192034 -6.392569 0.0000
R-squared 0.903840 Mean dependent var 0.033347
Adjusted R-squared 0.875794 S.D. dependent var 0.340841
S.E. of regression 0.120122 Akaike info criterion -1.188295
Sum squared resid 0.346305 Schwarz criterion -0.821861
Log likelihood 27.01272 F-statistic 32.22638
Prob(F-statistic) 0.000000
Jarque-Bera 0.204760 Prob(Jarque-Bera) 0.902686
Breusch-Godfrey 0.531854 Prob(Breusch-Godfrey) 0.473190
White Heteroscedasticity Test 0.594201 Prob(White Heteroscedasticity Test) 0.834805
Ramsey RESET Test 0.586863 Prob(Ramsey RESET Test) 0.451426
DLrpxmaize denotes difference of the log of the teal price of maize. DLrpxrice denotes the difference of the log of the real 
price of rice. DLrpxcassava denotes the difference of the log of the real price of cassava. DLrpxnpk denotes the difference 
of the log of the real price of fertiliser (npk). DLtemp denotes the difference of the log of annual mean temperature. DLrain 
denotes difference of the log of annual mean rainfall RESJD01 denotes the error correcting term.
Source: Author’s computations.
The adjustment coefficient has a negative sign, suggesting that any previous 
disequilibrium in the long-run maize supply relationship will be corrected in the current year. 
However, the magnitude o f  the adjustment coefficient is greater than 1 suggesting that any 
previous disequilibrium in the long-run maize supply is corrected more than proportionately 
in the short-run. In other words, maize supply in Ghana returns to its long-run equilibrium 
path in an oscillating manner following previous year’s deviation from equilibrium. This 
oscillating equilibrium convergence in the maize supply relationship can be explained by the 
fact that in Ghana, maize supply experiences a glut in a particular period and a shortage in 
another period immediately following. Thus, it is not uncommon to experience “excess” 
supply o f maize which dampens its price in one period and in another period shortage which
60
puts an upward pressure on the price. This glut-shortage movement in maize supply can be 
attributed partly to the unreliability and unpredictability o f the weather variables, notably rain 
since generally agriculture in Ghana is rain-fed. Also, the glut-shortage movement could 
partly be due to the high sensitivity o f fanners’ behaviour to price incentives that makes them 
shift resources to the production o f  those farm produce that commanded relatively high price 
in the previous crop season.
The ECM for Model 2 (Table 10) is estimated by replacing the climate variables with 
mean annual rainfall and mean annual maximum and minimum temperatures. The Breusch- 
Godfrey o f the Error Correction Model 2 in Table 10 implies the absence o f autocorrelation. 
The value o f the R2 is 0.93 meaning that the model explains 93 percent o f variation in the 
supply o f  maize. The F statistic is significantly different from zero. The White 
heteroscedasticity test also shows that the null hypothesis o f  homoscedastic residuals is not 
rejected, since the relevant White F statistic reported is not significant at the 5 percent level. 
The Jarque Beta statistic is not significant at the 5 percent level, implying that the residuals 
are normally distributed. The null hypothesis o f  no model misspecification is also not 
rejected, as indicated by the insignificant F statistic o f  the Ramsey RESET test.
Change in mean annual rainfall is significant at the 1 percent level with a positive 
coefficient. A 1 percent increase (decrease) in the first difference in rainfall tends to result in 
a 0.767 percent increase (decrease) in the change in the supply o f maize. This confirms the 
view that, better rainfall conditions tend to enhance the supply o f maize in Ghana (Ofori- 
Sarpong, 2001). Change in the price o f  fertiliser is significant but its effect does not have the 
expected sign. The change in the mean annual maximum and minimum temperatures are not 
significant even at the 10 percent level. The change in the prices o f  maize, rice and cassava 
are not significant even at the 10 percent level.
61
Table 10. Error Correction Model 2
Dependent Variable: D(lmaize)
Sample(adjusted): 1971 - 2002
Included observations: 32 after adjusting endpoints
Variable Coefficient Std. Error t-Statistic Prob.
C -0.009631 0.019433 -0.495589 0.6249
D(lrain) 0.766674 0.123003 6.232976 0.0000
D(lmaxtemp) 1.487902 1.382166 1.076500 0.2929
D(lmintemp) 0.315382 1.965728 0.160440 0.8739
D(lrpxmaize) -0.221234 0.090242 -2.451574 0.0222
D(lrpxcassava) 0.093772 0.074003 1.267149 0.2178
D(lrpxrice) -0.061219 0.085275 -0.717898 0.4800
D(lrpxnpk) 0.396277 0.051910 7.633900 0.0000
RESID02(-1) -1.362613 0.175042 -7.784504 0.0000
R-squared 0.929310 Mean dependent var 0.033347
Adjusted R-squared 0.904723 S.D. dependent var 0.340841
S.E. of regression 0.105207 Akaike info criterion -1.433506
Sum squared resid 0.254578 Schwarz criterion -1.021268
Log likelihood 31.93610 F-statistic 37.79571
Prob(F-statistic) 0.000000
Jarque-Bera 0.296467 Prob(Jarque-Bera) 0.862230
Breusch-Godftey 0.837136 Prob(Breusch-Godfrey) 0.370137
White Heteroscedasticity Test 0.610326 Prob(White Heteroscedasticity Test) 0.831274
Ramsey RESET Test 1.053579 Prob(Ramsey RESET Test) 0.315838
DLrpxmaize denotes difference of the log of the real price of maize. DLrpirice denotes die difference of the log of the real 
price of rice. DLrpxcassava denotes the difference of the log of the real price of cassava. DLrpxnpk denotes the difference 
of the log of the real price of fertiliser (npk). DLmaxtemp denotes the difference of the log of annual mean maximum 
temperature. DLmintemp denotes the difference of the log of annual mean minimum temperature DLrain denotes 
difference of the log of annual mean rainfalL Resid02 denotes the error correcting term.
Source: Author’s computations.
Table 11 depicts the ECM results o f  a different combination o f climate variables. 
These are the mean annual temperature, the mean annual critical major, critical minor and 
non-critical periods o f rainfall. The Breusch-Godfrey LM test suggests an absence o f serial 
correlation o f the residuals. The value o f the R2 shows that 93 percent o f the variation in the 
first difference o f  supply o f maize is explained by the model. The F statistic shows a 
significant joint effect o f  the explanatory variables. The White’s heteroscedasticity test shows 
that the null hypothesis o f  homoscedastic residuals is not rejected since the relevant F statistic 
reported is not significant at the 5 percent level. The Jarque-Bera statistic is not significant at 
the 5 percent level, implying that the residuals are normally distributed. The null hypothesis 
o f  no model misspecification is not rejected as indicated by the insignificant F statistic o f the 
Ramsey RESET test.
62
Change in the mean annual critical major rainfall and change in non-critical rainfall 
are significant at the 1 percent level, with the expected positive sign for the mean annual 
critical major rainfall (Table 11). Change in the mean annual critical minor rainfall is not 
significant even at the 10 percent level. The change in the mean annual temperature, the 
prices o f maize, rice and cassava are not significant even at the 10 percent level.
Table 11. Error Correction Model 3
Dependent Variable: D(lmaize)
Sample(adjusted): 1971 2002
Included observations: 32 after adjusting endpoints
Variable Coefficient Std. Error t-Statistic Prob.
C -0.020612 0.020710 -0.995267 0.3304
D(ltemp) 2.186738 2.080076 1.051278 0.3045
D(lcrmj) 0.540790 0.145349 3.720632 0.0012
D(lcnninor) 0.148198 0.097870 1.514230 0.1442
D (lnoncrit) 0.293725 0.078300 3.751281 0.0011
D (lrpxmaize) -0.311273 0.112702 -2.761905 0.0114
D(lipxrice) -0.027551 0.085252 -0.323164 0.7496
DQipxcassava) 0.235756 0.099459 2.370391 0.0270
D(lipxnpk) 0.458841 0.059686 7.687619 0.0000
RESID03(-1) -1.182686 0.181667 -6.510192 0.0000
R-squaied 0.928942 Mean dependent var 0.033347
Adjusted R-squared 0.899873 S.D. dependent var 0.340841
S.E. of regression 0.107852 Akaike info criterion -1.365814
Sum squared resid 0.255903 Schwarz criterion -0.907772
Log likelihood 31.85303 F-statistic 31.95642
Prob(F-statistic) 0.000000
Jarque-Bera 0.238257 Prob(Jarque-Bera) 0.887694
Breusch-Godfrey 1.614087 Prob(Breusch-God£rey) 0.217814
White Heteroscedasticity Test 0.291865 Prob(White Heteroscedasticity Test) 0.990744
Ramsey RESET Test 0.259101 Prob(Ramsey RESET Test) 0.615587
DLrpxmaize denotes difference of the log of the real price of maize. DLrpxrice denotes the difference of the log of the real 
price of rice. DLrpxcassava denotes the difference of the log of the real price of cassava. DLrpxnpk denotes the difference 
of the log of the real price of fertiliser (npk). DLtemp denotes the difference of the log of annual mean temperature. DLrain 
denotes difference of the log of annual mean rainfall. DLcrmj denotes the difference of the log of critical major rainfall 
periods. DLcrminor denotes the difference of the log of critical minor rainfall periods DLnoncrit denotes the difference of 
the log of non-critical periods RESID03 denotes the error correcting term.
Source: Author’s computations
In Table 12, the ECM is estimated with climate variables, mean annual critical major, 
minor and non-critical rainfall periods, and mean annual maximum and minimum 
temperatures for the short-run equilibrium analysis. The Breusch-Godfrey LM test suggests 
the absence of serial correlation o f the residuals. The model explains 95 percent o f the 
variation in the supply o f  maize. The F statistic is significantly different from zero. The
63
White’s heteroscedasticity test shows that the null hypothesis o f homoscedastic residuals is 
not rejected since the F statistic reported is not significant at the 5 percent level. The Jarque- 
Bera statistic is not significant at the 5 percent, implying that the residuals are normally 
distributed. The null hypothesis o f  no model misspecification is not rejected, as indicated by 
the insignificant F statistic o f the Ramsey reset test even at the 10 percent level.
Table 12. Error Correction Model 4
Dependent Variable: D(Imaize)
Sample(adjusted): 1971 2002
Included observations: 32 after adjusting endpoints
Variable Coefficient Std. Error t-Statistic Prob.
C -0.017894 0.018005 -0.993831 0.3316
D(lcrmj) 0.482369 0.116187 4.151675 0.0005
D(lcnninor) 0.210012 0.085336 2.461005 0.0226
D(lnoncrit) 0.270991 0.070165 3.862188 0.0009
D(lmaxtemp) 1.510380 1.266399 1.192657 0.2463
D(lmintemp) -0.090760 1.882117 -0.048222 0.9620
D(lrpxmaize) -0.259115 0.099585 -2.601953 0.0166
D(lipxcassava) 0.195317 0.086782 2.250651 0.0352
D (lrpxrice) -0.037815 0.078418 -0.482224 0.6346
D(lrpxnpk) 0.435717 0.053141 8.199305 0.0000
RESID04(-1) -1.322840 0.163193 -8.105963 0.0000
R-squared 0.948897 Mean dependent var 0.033347
Adjusted R-squared 0.924563 S.D. dependent var 0.340841
S.E. of regression 0.093615 Akaike info criterion -1.632970
Sum squared resid 0.184038 Schwarz criterion -1.129123
Log likelihood 37.12752 F-statistic 38.99379
Prob(F-statistic) 0.000000
Jarque-Bera 0.109725 Prob(Jaique-Bera) 0.946615
Breusch-Godfrey 2.887945 Prob(Breusch-Godfrey) 0.104747
White Heteroscedasticity Test 0.651497 Prob(White Heteroscedasticity Test) 0.804923
Ramsey RESET Test 0.030338 Prob(Ramsey RESET Test) 0.863475
D Lrpxmaize denotes difference of the log of the real price of maize. DLrpxrice denotes the difference of the log of the real price 
of rice. DLrpxcassava denotes the difference of the log of the real price of cassava. DLrpxnpk denotes the difference of the log of 
the real price of fertiliser (npk). DLmaxtemp denotes the difference of the log of annual mean maximum temperature. DLmmtemp 
denotes the difference of the log of annual mean minimum temperature. DLcrmj denotes the difference of the log of critical major 
rainfall periods. DLcnninor denotes the difference of the log of critical minor rainfall periods DLnoncrit denotes the difference of 
the log of non-critical periods RESID04 denotes the error correcting term.
Source: Author’s computations.
The change in the mean annual critical major rainfall is significant at the 1 percent 
level with a coefficient o f  0.482. This implies that a 1 percent increase in the first difference 
o f rainfall tends to increase the supply o f maize by about 0.482 percent. The change in the 
prices of maize, rice, cassava and the mean annual maximum and minimum temperatures are
64
not significant even at the 10 percent level. The insignificance o f the mean annual maximum 
and minimum temperatures could be due to the small variation in these two variables.
The ECM with the squares o f the mean annual rainfall and temperature as climate 
variables is presented in Table 13. With the exception o f the square annual mean rainfall, 
which is significant at the 1 percent level, all the independent variables are insignificant even 
at the 10 percent level. The value of the R2 is 0.90. The F statistic is significantly different 
from zero. There are no heteroscedastic and autocorrelation problems. Other Error 
Correction Models are presented in Appendix 7 through Appendix 9; these show similar 
results.
Table 13. Error Correction Model 5
Dependent Variable: D(lmaize)
Sample(adjiisted): 1971 - 2002
Included observations: 32 after adjusting endpoints
Variable Coefficient Std. Error t-Statistic Prob.
C -0.010310 0.022207 -0.464266 0.6466
D(Isqrain) 0.376774 0.070151 5.370873 0.0000
D(lsqtemp) 0.657019 1.018538 0.645061 0.5250
D(lrpxmaize) -0.234494 0.102988 -2.276909 0.0320
D(lrpxrice) -0.067061 0.091716 -0.731175 0.4718
D(lrpxcassava) 0.114554 0.089077 1.286015 0.2107
D(lrpxnpk) 0.407435 0.057363 7.102745 0.0000
RESID05(-1) -1.227594 0.192034 -6.392569 0.0000
R-squared 0.903840 Mean dependent var 0.033347
Adjusted R-squared 0.875794 S.D.dependent var 0.340841
S.E. of regression 0.120122 Akaike info criterion -1.188295
Sum squared resid 0.346305 Schwarz criterion -0.821861
Log likelihood 27.01272 F-statistic 32.22638
Prob(F-statistic) 0.000000
Jarque-Bera 0.204760 Prob(Jarque-Bera) 0.902686
Breusch-Godfrey 0.531854 Prob(Breusch-Godfrey) 0.473190
White Heteroscedasticity Test 0.594201 Prob(White Heteroscedasticity Test) 0.834805
Ramsey RESET Test 0.586863_F * ------------------- Prob(Ramsey RESET Test) 0.451426
DLrpxmaize denotes difference of the log of the real price of maize. DLrpxrice denotes the difference of the log of the real 
price of rice. DLrpxcassava denotes the difference of the log of the real price of cassava. DLrpxnpk denotes the difference 
of the log of the real price of fertiliser (npk). DSqLtemp denotes the difference of the log of the square of annual mean 
temperature. DSqLrain denotes difference of the log of the square of annual mean rainfall. RESID05 denotes the error 
correcting term.
Source: Author’s computations
65
4.8. Results o f Granger-Causality Tests
In this study, Granger Causality Tests are performed to determine whether the level o f 
maize supply is caused by temperature and whether rainfall causes maize supply. The results 
in Table 14 show there is a uni-directional causal relationship between maize supply and 
temperature and between maize supply and rainfall at lag two. This is in consonance with 
theory and the relevant existing empirical literature (see for instance, Muchena, 1994, 
Muchena and Iglesias, 1995; Molua, 2002).
Table 14. Results o f the Granger Causality Tests between the Supply of Maize, 
Rainfall and Temperature.
Null Hypothesis: Obs F-Statistic Probability
LRAIN does not Granger Cause LQMAIZE 31 5.37978*** 0.01109
LQMAIZE does not Granger Cause LRAIN 31 0.45611 0.63872
LTEMP does not Granger Cause LQMAIZE 31 4.45421*** 0.02171
LQMAIZE does not Granger Cause LTEMP 31 1.26923 0.29789
*** represents significance at the 1 percent level. LRAIN denotes annual mean rainfall; LQMAIZE denotes 
maize supply.
Source: Author’s computations.
66
CHAPTERS 
SUMMARY AND RECOMMENDATIONS
5.1. Summary and Policy Recommendations
This study has analysed the effects o f  climate change on maize supply in Ghana 
during 1970-2002. The climate variables used are temperature and rainfall. The study 
employs econometric modelling to quantify the effects o f  climate change (specifically, 
changes in rainfall and temperature) on the supply o f maize and provides a theoretical 
foundation for the modelling based on the theory o f the firm. It uses the techniques o f 
cointegration and error correction modelling. Cointegration is applied to validate the long- 
run equilibrium relationships between the time series data employed.
The present study observes that rainfall and change in rainfall have consistently been 
found to be significant in all the models with coefficients having the expected positive signs. 
This is consistent with the relevant existing literature (see for instance, Hamid, 1984; Schulze 
et al, 1996; Muchena, 1994 and Ofori-Sarpong 2001). These clearly suggest a large 
dependence o f  maize supply on rainfall in Ghana. In the present study, it is observed that 
inadequate rainfall in Ghana adversely affects the supply o f  maize, and income o f  maize 
formers and food security in the country, ceteris paribus. The economic welfare o f inhabitants 
in general would decrease, ceteris paribus. Some o f the strategies adopted by farmers in the 
case o f decreased rainfall as a result o f  climate change are the use o f drought resistant crops, 
shifting the planting season, relying on more legumes than cereals and selling their labour for 
cash and food (Ofori-Sarpong, 2001).
Critical major periods o f rains were significant in the present study’s models because 
these are the periods when crop growth escapes heavy infestation and mature before insects 
could cause severe damage. Maize planted during this period are said to be planted early and
67
this activity minimizes moisture stress during flowering and grain-filling period and insect 
population are also at low levels at the beginning o f this period. Temperature, however, has 
not been significant in determining the supply o f maize. This could be due to the small 
variation in the annnnl mean temperature. This in turn, may be because the period covered in 
the present study, viz., 1970 -  2002, may be too short for an observation o f  significant 
variation in temperature.
The empirical results o f  the present study suggest that policy makers should take the 
issue o f  climate change (Global Wanning) seriously. The effect o f  climate change on maize 
farmers who have the ability to adapt and with the availability o f  technology is likely to be 
less serious. Improvement in technology such as establishment o f  irrigation facilities and the 
rehabilitation o f  old ones, as well as teaching adaptive measures and making it part o f  
national policies even before the onset o f  adverse effect o f  climate change would help maize 
farmers in particular and the nation in general.
The present study’s observation that rainfall is significant in all the estimated models 
shows how important water is to the production o f maize. Hence, maize farmers should be 
encouraged to plant their maize early during the major rainy season during which time the 
rains are fairly stable and reliable. Due to the significant influence o f  rainfall on maize 
supply, the Meteorological Stations and the Ministry o f Food and Agriculture should have 
regular consultations and effectively monitor current climate and environmental conditions in 
the country to establish an early warning system.
5.2. Limitations of the Study and Suggestions for Future Research
Due to lack o f data, the study could not incorporate information like the effect of 
actual agricultural wage rate on maize supply. The fees charged by the Ghana Meteorological 
Service Department on the delivery o f meteorological data was a constraint to the acquisition
68
o f a longer-term timf> series data set for rainfall and temperature from the various weather 
stations selected. A larger time series data on climate variables as well as the inclusion of 
data on relative humidity, evaporation rates may improve the understanding o f  the effect o f 
climate change on maize supply. The inclusion o f annual time series data on agricultural 
wages, land rent, and soils may improve the outcome o f the study; this was not possible due 
to lack o f  data. Most farmers in Ghana are small-scale and do not normally keep records of 
their assets. In addition, annual time series data on the quantities and prices o f insecticides 
used in maize production are not available.
Increases in rainfall normally increase the output o f crops (maize inclusive), but this 
increase in output is only to a point, which is referred to as its maximum. Continuous increase 
in rainfall from this maximum rather leads to decreases in output. This theory may be 
captured by climate variables (rainfall and temperature) and their squares in the relevant 
models. The estimation o f this model was however impossible due to the existence o f  a near 
linear correlation relationship in the climate variables and their squares, hence a singular 
matrix.
69
APPENDICES
Appendix 1. Definition o f Variables
Variable Definition
Lmaize Output o f  maize
Lpxmaize Real price o f  maize
Lpxrice Real price o f  rice
Lpxcassava Real price o f  cassava
Lpxnpk Real price o f  fertiliser (npk)
Lrain Rainfall
Ltemp Temperature
Lcrmaj Critical major rainfall
Lcrmin Critical minor rainfall
LnonC Non-critical rainfall
Lmaxt. Maximum temperature
lmint. Minimum temperature
Lsqrain Square rainfall
Lsqtemp Square temperature
Lsqmaxt Square maximum temperature
Lsqmint Square minimum temperature
Lsqcnnaj Square o f  critical major rainfall
Lsqcnnin Square o f critical minor rainfall
LsqnonC Square o f non-critical rainfall
ecm I The error correction term
*Climate variables are in annual means and all real’ prices are in natural logarithms.
Appendix 2 Descriptive Statistics o f the Relevant Variables: Transformed data (The natural log of data)
Maize Rain Temp Maxt M int NonC Pxcass Pxmaize Pxnpk Pxrice
MEAN 6.316 5.840 1.505 3.609 4.534 3.297 6.021 1.310 4.198 5.029
MEDIAN 6.326 5.832 1.504 3.669 4.552 3.298 6.123 1.310 4.352 5.232
MAXIMUM 7.244 6.228 1.521 4.150 4.750 3.333 6.650 1.324 5.925 6.872
MINIMUM 4.947 5.398 1.496 2.839 4.009 3.270 4.948 1.296 0.950 1.822
STD. DEV. 0.543 0.258 0.006 0.264 0.136 0.013 0.497 0.006 0.964 0.994
SKEWNESS -0.330 -0.141 0.953 -0.927 -1.655 0.242 -0.661 -0.233 -1.063 -0.949
KURTOSIS 2.437 1.909 3.891 4.679 7.831 3.806 2.364 3.432 5.161 4.548
C. V. 0.385 0.136 0.044 0.004 0.083 0.005 0.073 0.030 0.004 0.086
J.B . 1.035 1.746 6.090 8.600 47.147 1.215 2.958 0.556 12.638 8.255
PROB. 0.596 0.148 0.048 0.014 0.000 0.545 0.228 0.757 0.002 0.016
OBS 33 33 33 33 33 33 33 33 33 33
Maize denotes output of maize, Pxmaize denotes real price of maize, Pxcass denotes real price o f cassava, Pxnpk denotes real price of fertiliser, Psrice denotes real price of 
rice, Rain denotes annual mean rainfall, Temp denotes annual mean temperature, Maxt denotes annual maximum mean temperature, M int denotes annual minimum mean 
temperature, NonC denotes annual non-critical rains, C.V denotes coefficient of variation, J.B  denotes Jarque-Bera, PROB denotes probability, OBS denotes observation. 
Source: Author’s computations
71
pendix 2. Continued
Sqcrmaj Sqcrmin Sqrain Sqtemp Sqmaxt Sqmint SqnonC Crmaj Crm in
EAN 3.768 5.940 11.680 6.929 12.042 6.033 7.218 9.068 6.594
EDIAN 4.269 6.153 11.664 6.927 12.245 6.034 7.338 9:105 6.595
AXIMUM 5.856 7.315 12.455 7.003 13.300 6.096 8.300 9.500 6.666
[NIMUM 1.386 4.363 10.797 6.890 9.896 5.968 5.677 8.017 6.540
STD. DEV. 1.450 0.807 0.516 0.027 0.993 0.028 0.527 0.273 0.026
SKEWNESS -0.201 -0.167 -0.141 0.953 -0.661 -0.233 -0.927 -1.655 0.242
KURTOSIS 1.425 1.997 1.909 3.891 2.364 3.432 4.679 7.831 3.806
C. V. 0.044 0.004 0.073 0.030 0.004 0.083 0.005 0.230 0.198
J.B . 3.634 1.536 1.746 6.090 2.958 0.556 8.600 47.147 1.215
PROB. 0.162 0.464 0.418 0.048 0.008 0.757 0.014 0.000 0.545
OBS 33 33 33 33 33 33 33 33 33
Sqcrmaj denotes square of annual critical major rains, Sqcrmin denotes annual critical minor rains, Sqrain denotes square annual mean rainfall, Sqtemp denotes square annual temperature, 
Sqmaxt denotes square annual maximum temperature, Sqmint denotes square annual minimum temperature, SqnonC denotes square non-critical annual mean rains, C.V denotes coefficient of 
variation, J.B denotes Jarque-Bera PROB denotes probability, OBS denotes observation.
Source: Author’s computations
72
Appendix 3. Correlation Matrix of Variables
lermin lermj Imaxt Imint InonC Imaize lrain lpxcas lpxmaize lpxnpk
Icrmin 1.000 0.304 -0.23 -0.03 0.066 0.279 0.776 -0.12 -0.14 0.130
lcrmj 0.304 1.000 -0.20 -0.16 0.169 0.351 0.739 -0.34 -0.23 0.106
Imaxt -0.232 -0.21 1.000 0.608 -0.05 0.341 -0.22 0.377 0.342 0.406
lmint -0.030 -0.16 0.608 1.000 -0.13 0.439 -0.13 0.669 0.660 0.603
InonC 0.067 0.169 -0.05 -0.13 1.000 0.067 0.460 -0.48 -0.50 -0.21
Imaize 0.279 0.351 0.341 0.439 0.067 1.000 0.361 0.162 0.165 0.871
lrain 0.776 0.739 -0.22 -0.13 0.460 0.361 1.000 -0.38 -0.35 0.061
lpxcas -0.117 -0.34 0.377 0.669 -0.48 0.163 -0.38 1.000 0.966 0.543
lpxmaize -0.142 -0.23 0.342 0.660 -0.51 0.165 -0.35 0.966 1.000 0.554
lpxnpk 0.131 0.106 0.407 0.603 -0.21 0.871 0.061 0.543 0.554 1.000
lpxrice -0.025 -0.16 0.335 0.575 -0.48 0.254 -0.24 0.909 0.939 0.640
lsqcrmj 0.304 1.000 -0.20 -0.16 0.169 0.351 0.739 -0.34 -0.23 0.106
lsqcrmin 1.000 0.304 -0.23 -0.03 0.067 0.279 0.775 -0.11 -0.14 0.130
lsqmaxt -0.232 -0.20 1.000 0.608 -0.05 0.341 -0.22 0.377 0.342 0.406
lsqmint -0.030 -0.16 0.608 1.000 -0.13 0.439 -0.13 0.669 0.660 0.603
IsqnonC 0.067 0.169 -0.05 -0.13 1.000 0.067 0.460 -0.48 -0.51 -0.210
lsqrain 0.776 0.739 -0.22 -0.13 0.460 0.361 1.000 -0.38 -0.35 0.061
lsqtemp 0.008 -0.23 0.617 0.934 -0.28 0.511 -0.02 0.62 0.611 0.672
ltemp 0.008 -0.23 0.618 0.934 -0.28 0.511 -0.20 0.62 0.611 0.672
lermin denotes critical minor rainfall; Icrmj denotes critical major, Imaxt denotes maximum temperature; lmint denotes 
minimum temperature; InonC denotes non-critical rainfall; Imaize denotes maize output; Irain denotes rainfall; Ipxcas 
denotes prices of cassava; Ipxmaizc denotes prices of maize; Ipxnpk denotes prices of fertilizer.
Source: Author;s computations
Appendix 3. Continued
Ipxrice Isqcrmj Isqcrmin Isqmaxt lsqmint IsqnonC lsqrain lsqtemp ltemp
Icimin -0.024 0.304 1.000 -0.232 -0.030 0.066 0.775 0.008 0.008
lcrmj -0.160 1.000 0.304 -0.205 -0.165 0.169 0.739 -0.230 -0.23
Imaxt 0.335 -0.205 -0.232 1.000 0.608 -0.055 -0.220 0.617 0.617
lmint 0.575 -0.165 -0.030 0.608 1.000 -0.129 -0.128 0.934 0.934
InonC -0.482 0.169 0.066 -0.055 -0.129 1.000 0.460 -0.284 -0.28
Imaize 0.253 0.351 0.279 0.341 0.439 0.067 0.361 0.511 0.511
lrain -0.243 0.739 0.775 -0.220 -0.128 0.460 1.000 -0.202 -0.20
Ipxcas 0.909 -0.340 -0.116 0.377 0.669 -0.486 -0.386 0.625 0.625
Ipxmaize 0.939 -0.234 -0.141 0.342 0.660 -0.506 -0.356 0.611 0.611
lpxnpk 0.640 0.106 0.130 0.406 0.603 -0.210 0.061 0.672 0.672
lpxrice 1.000 -0.160 -0.024 0.335 0.575 -0.482 -0.243 0.561 0.561
lsqcrmj -0.160 1.000 0.304 -0.205 -0.165 0.169 0.739 -0.230 -0.23
lsqcrmin -0.024 0.304 1.000 -0.232 -0.030 0.066 0.775 0.008 0.008
Isqmaxt 0.335 -0.205 -0.232 1.000 0.608 -0.055 -0.220 0.617 0.617
lsqmint 0.575 -0.165 -0.030 0.608 1.000 -0.129 -0.128 0.934 0.934
lsqnonC -0.482 0.169 0.066 -0.055 -0.129 1.000 0.460 -0.284 -0.28
lsqrain -0.243 0.739 0.775 -0.220 -0.128 0.460 1.000 -0.202 -0.20
Isqtemp 0.561 -0.230 0.008 0.617 0.934 -0.284 -0.202 1.000 1.000
ltemp 0.561 -0.230 0.008 0.617 0.934 -0.284 -0.202 1.000 1.000
Ipxrice denotes prices of rice; Isqcrmj denotes square of critical major rainfall; Isqcrmin denotes square of the critical 
minor rainfall; Isqmaxt denotes square of maximum temperature; Isqmint denotes square minimum temperature; IsqnonC 
denotes square of non-critical period of rainfall; Isqrain denotes square of rainfall; Isqtemp denotes square of temperature; 
ltemp denotes temperature.
Source: Author’s computations
73
Appendix 4. Cointegrating Equation: Model 6
Dependent Variable: lmaize 
Sample: 1970 - 2002 
Included observations: 33
Variable Coefficient Std. Error t-Statistic Prob.
c -8.167712 8.754656 -0.932956 0.3598
lsqrain 0.428537 0.109366 3.918390 0.0006
Isqmaxtemp 1.764873 1.189398 1.483838 0.1504
lsqmintemp -0.303636 1.622447 -0.187147 0.8531
lrpxmaize 0.024804 0.166946 0.148576 0.8831
lrpxrice -0.334321 0.116685 -2.865149 0.0083
lrpxcassava 0.009843 0.123777 0.079524 0.9372
lrpxnpk 0.405554 0.025282 16.04131 0.0000
R-squared 0.951321 Mean dependent var
Adjusted R-squared 0.937691 S.D. dependent var
S.E. of regression 0.135439 Akaike info criterion
Sum squared resid 0.458596 Schwarz criterion
Log likelihood 23.73057 F-statistic
Durbin-Watson stat 2.487669 Prob(F-stalistic)
ADF Test Statistic -4.078141 1% Critical Value* -2.6395
5% Critical Value -1.9521
10% Critical Value -1.6214
Lrpxmaize denotes log of the real prioe of maize. Lrpxrice denotes the log of the real price of rice. Lrpicassava denotes 
the log of the real price of cassava. Lrpxnpk denotes the log of the real price of fertiliser (npk). Lsqmaxtemp denotes the 
log of animal mean maximum temperature. Lsqmintemp denotes the log of annual mean minimum temperature Lsqrain 
denotes the log of annual mean rainML 
Source: Author’s computation.
74
Appendix 5. Cointegrating Equation: Model 7
Dependent Variable: Imaize 
Sample: 1970 - 2002 
Included observations: 33
Variable Coefficient Std. Error t-Statistic Prob.
c -10.19682 10.30724 -0.989286 0.3324
Isqcrmj 0.307048 0.103652 2.962296 0.0068
Isqcrminor 0.079923 0.064102 1.246798 0.2245
lsqnoncrit 0.087531 0.053575 1.633795 0.1154
lsqtemp 1.830251 1.553037 1.178498 0.2502
lrpxmaize -0.150204 0.181168 -0.829086 0.4152
Irpxcassava 0.121986 0.140520 0.868101 0.3939
lrpxrice -0.266200 0.114138 -2.332260 0.0284
Irpxnpk 0.388703 0.029115 13.35051 0.0000
R-squared 0.952477 Mean dependent var 6.315912
Adjusted R-squared 0.936636 S.D. dependent var 0.542588
S.E. of regression 0.136581 Akaike info criterion -0.916797
Sum squared resid 0.447705 Schwarz criterion -0.508659
Log likelihood 24.12715 F-statistic 60.12763
Durbin-Watson stat 2.368713 Prob(F-statistic) 0.000000
ADF Test Statistic -4.568020 1% Critical Value* 
5% Critical Value 
10% Critical Value
-2.6395
-1.9521
-1.6214
DLrpxmaize denotes difference of the log of the real price of maize. DLrpxrice denotes the difference of the log of the real 
price of rice. DLrpxcassava denotes the difference of the log of the real price of cassava. DLrpxnpk denotes the difference 
of the log of the real price of fertiliser (npk). DsqLtemp denotes the difference of the log of annual mean temperature. 
DsqLrain denotes difference of the log of annual mean rainfall. DsqLcrmj denotes the difference of the log of critical 
major rainfall periods. DsqLcrminor denotes the difference of the log of critical minor rainfall periods DsqLnoncrit 
denotes the difference of the log of non-critical periods RESID03 denotes the error correcting term.
Source: Author’s computation
75
Appendix 6. Cointegrating Equation: Model 8
Dependent Variable: Imaize 
Sample: 1970 - 2002 
Included observations: 33
Variable Coefficient Std. Error t-Statistic Prob.
c -10.47062 9.115326 -1.148683 0.2625
lsqcrmj 0.286467 0.098464 2.909344 0.0079
lsqcrminor 0.123215 0.067542 1.824288 0.0811
lsqnoncrit 0.075890 0.055730 1.361743 0.1865
lsqmaxtemp 1.894039 1.230170 1.539657 0.1373
lsqmintemp -0.175436 1.686937 -0.103997 0.9181
lrpxmaize -0.056122 0.189732 -0.295799 0.7700
lrpxrice -0.319306 0.118631 -2.691583 0.0130
Irpxcassava 0.087818 0.140334 0.625776 0.5376
lrpxnpk 0.397301 0.025947 15.31205 0.0000
R-squared 0.955589 Mean dependent var 6.315912
Adjusted R-squared 0.938210 S.D. dependent var 0.542588
S.E. o f regression 0.134874 Akaike info criterion -0.923904
Sum squared resid 0.418393 Schwarz criterion -0.470417
Log likelihood 25.24442 F-statistic 54.98727
Durbin-Watson stat 2.610053 Prob(F-statistic) 0.000000
ADF Test Statistic o f -4.942032 1% Critical Value* -2.6395
the Error Term
5% Critical Value -1.9521
10% Critical Value -1.6214
Lrpxmaize denotes log of the real price of maize. Lrpxrice denotes the log of the real price of rice. Lrpxcassava denotes 
the log of the real price of cassava. Lrpxnpk denotes the log of the real price of fertiliser (npk). Lsqcrmj denotes log of the 
critical major rainfall periods. Lsqcrminor denotes log of the critical minor rainfall periods. Lsqnoncrit denotes log of non- 
critical rainfall periods. Lsqmaxtemp denotes the log of annual mean maximum temperature. Lsqmintemp denotes the log 
of annual mean minimum temperature 
Source: Author’s computations.
76
Appendix 7. Error Correction Model 6
Dependent Variable: D(lmaize)
Sample(adjusted): 1971 - 2002
Included observations: 32 after adjusting endpoints
Variable Coefficient Std. Error t-Statistic Prob.
C -0.009631 0.019433 -0.495589 0.6249
D(Isqrain) 0.383337 0.061501 6.232976 0.0000
D(lsqmaxtemp) 0.743951 0.691083 1.076500 0.2929
D(lsqmintemp) 0.157691 0.982864 0.160440 0.8739
D(lrpxmaize) -0.221234 0.090242 -2.451574 0.0222
D(lrpxrice) -0.061219 0.085275 -0.717898 0.4800
D(lrpxcassava) 0.093772 0.074003 1.267149 0.2178
D(lrpxnpk) 0.396277 0.051910 7.633900 0.0000
RESID10(-1) -1.362613 0.175042 -7.784504 0.0000
R-squared 0.929310 Mean dependent var 0.033347
Adjusted R-squared 0.904723 S.D. dependent var 0.340841
S.E. o f regression 0.105207 Akaike info criterion -1.433506
Sum squared resid 0.254578 Schwarz criterion -1.021268
Log likelihood 31.93610 F-statistic 37.79571
Jarque-Bera 
Breusch-Godfrey 
White Heteroscedasticity Test 
Ramsey RESET Test
0.296467
0.837136
0.610326
1.053579
Prob(F-statistic)
Prob(Jarque-Bera)
Prob(Breusch-Godfrey)
Prob(White Heteroscedasticity Test) 
Prob(Ramsey RESET Test)
0.000000
0.862230
0.370137
0.831274
0.315838
DLrpxmaize denotes difference of the log of the real price of maize. DLrpxrice denotes the difference of the log of the real 
price of rice. DLrpxcassava denotes the difference of the log of the real price of cassava. DLrpxnpk denotes the difference 
of the log of the real price of fertiliser (npk). DsqLmaxtemp denotes the difference of the log of annual mean maximum 
temperature. DsqLmintemp denotes the difference of the log of annual mean minimum temperature DsqLrain denotes 
difference of the log of annual mean raihfall. Resid02 denotes the error correcting term.
Source: Author’s computation
77
Appendix 8 Error Correction Model 7
Dependent Variable: D(taiaize)
Sample(adjusted): 1971 - 2002
Included observations: 32 after adjusting endpoints
Variable Coefficient Std. Error t-Statistic Prob.
C -0.020612 0.020710 -0.995267 0.3304
D(lsqcnnj) 0.270395 0.072674 3.720632 0.0012
D(lsqcnninor) 0.074099 0.048935 1.514230 0.1442
D(lsqnoncrit) 0.146863 0.039150 3.751281 0.0011
D(lsqtemp) 1.093369 1.040038 1.051278 0.3045
D(lrpxmaize) -0.311273 0.112702 -2.761905 0.0114
D(lrpxcassava) 0.235756 0.099459 2.370391 0.0270
D(lrpxrice) -0.027551 0.085252 -0.323164 0.7496
D(lrpxnpk) 0.458841 0.059686 7.687619 0.0000
RESID07(-1) -1.182686 0.181667 -6.510192 0.0000
R-squared 0.928942 Mean dependent var 0.033347
Adjusted R-squared 0.899873 S.D. dependent var 0.340841
S.E. o f  regression 0.107852 Akaike info criterion -1.365814
Sum squared resid 0.255903 Schwarz criterion -0.907772
Log likelihood 31.85303 F-statistic 31.95642
Jarque-Bera 
Breusch-Godfrey 
White Heteroscedasticity Test 
Ramsey RESET Test
0.238257
1.614087
0.291865
0.259101
Prob(F-statistic)
Prob(Jarque-Bera)
Prob(Breusch-Godfrey)
Prob(White Heteroscedasticity Test) 
Prob(Ramsey RESET Test)
0.000000
0.887694
0.217814
0.990744
0.615587
SLrpxmaize denotes difference of the log of the real price of maize. DLrpxrice denotes the difference of the log of the real 
price of rice. DLrpxcassava denotes the difference of the log of the real price of cassava. DLrpxnpk denotes the difference 
of the log of the real price of fertiliser (npk). DsqLtemp denotes the difference of the log of annual mean temperature. 
DsqLcrmj denotes the difference of the log of critical major rainfall periods. DsqLcrminor denotes the difference of the 
log of critical minor rainfall periods DsqLnoncrit denotes the difference of the log of non-critical periods RESID03 denotes 
the error correcting temL Source; Author’s computation 
Source: Author’s computation
78
Appendix 9 Error Correction Model 8
Dependent Variable: D(lmaize)
Sample(adjusted): 1971 - 2002
Included observations: 32 after adjusting endpoints
Variable Coefficient Std. Error t-Statistic Prob.
C -0.017894 0.018005 -0.993831 0.3316
D(lsqcrmj) 0.241185 0.058093 4.151675 0.0005
D(lsqcrminor) 0.105006 0.042668 2.461005 0.0226
D(lsqnoncrit) 0.135496 0.035083 3.862188 0.0009
D(Isqmaxtemp) 0.755190 0.633200 1.192657 0.2463
D(lsqmintemp) -0.045380 0.941058 -0.048222 0.9620
D(lrpxmaize) -0.259115 0.099585 -2.601953 0.0166
D(lrpxrice) -0.037815 0.078418 -0.482224 0.6346
D(lrpxcassava) 0.195317 0.086782 2.250651 0.0352
D(Irpxnpk) 0.435717 0.053141 8.199305 0.0000
RESID08(-1) -1.322840 0.163193 -8.105963 0.0000
R-squared 0.948897 Mean dependent var 0.033347
Adjusted R-squared 0.924563 S.D. dependent var 0.340841
S.E. o f regression 0.093615 Akaike info criterion -1.632970
Sum squared resid 0.184038 Schwarz criterion -1.129123
Log likelihood 37.12752 F-statistic 38.99379
Jarque-Bera 
Breusch-Godfrey 
White Heteroscedasticity Test 
Ramsey RESET Test
0.109725
2.887945
0.651497
0.030338
Prob(F-statistic)
Prob(J arque-Bera)
Prob(Breusch-Godfrey)
Prob(White Heteroscedasticity Test) 
Prob(Ramsey RESET Test)
—' — . —■ . . .  ---- ------------------ -----------------------------i=L_
0.000000
0.946615
0.104747
0.804923
0.863475
DLrpxmaize denotes log of the real price of maize. DLrpxrice denotes the log of the real price of rice. DLrpxcassava 
denotes the log of the real price of cassava. DLrpxnpk denotes the log of the real price of fertiliser (npk). DsqLcrmj 
denotes log of the critical major rainfall periods. DsqLcrminor denotes log of the critical minor rainfall periods. 
DsqLnoncrit denotes log of non-critical rainfall periods. DsqLmaxtemp denotes the log of annual mean maximum 
temperature. DsqLmintemp denotes the log of annual mean minimum temperature 
Source: Authors computation
79
REFERENCES
Abdulai, A. and P. Rieder (1995). “The Impacts of Agricultural Price Policy on Cocoa 
Supply in Ghana: An Error Correction Estimation”, Journal o f  African Economies, 
4(3): 315-35.
Adams, R., C. Rosenzweig, R. Pearl, J. Ritchie, B. McCarl, D. Glyer, B. Cuny, J. Jones, K. 
Boote and H. Allen (1990). “Global Climate Change and U.S Agriculture”, Nature, 
17:219-24.
Akong'a, J., T.E. Downing, N.T. Konijn, D.N. Mungai, H.R. Muturi and H.L. Potter. 
(1988).“The Effects o f  Climatic Variations on Agriculture in Central and Eastern 
Kenya”. In, M L. Parry, T.R. Carter, and N.T. Konijn (eds.), “The Impact o f  Climatic 
Variations on Agriculture, Volume 2: Assessments in Semi-Arid Region.s”, Kluwer, 
Dordrecht, The Netherlands.
Arrhenius, S. (1896). “On the Influence o f  Carbonic Acid in the Air Upon the Temperature of 
the Ground”, Philosophical Magazine, 41: 237-76.
Berger, J. (1962). “Maize Production and the Manuring o f  Maize”. Centre D’etude L’Azote, 
Geneva, Switzerland.
Burke, J.J., J.R- Mahan and J.L. Hatfield (1988). “Crop-specific Thermal Kinetic Windows 
in Relation to Wheat and Cotton Biomass Production”, Agronomy Journal, 80: 553- 
556.
Chen, C-C. and B. A. McCarl (2001). “An Investigation o f the Relationship Between 
Pesticide Usage and Climate Change”, Climatic Change, 50: 475-487.
Clarke, R.T. (1962). “The Effect of Some Resting Treatments on a Tropical Soil”, Empir. J. 
Exp. Agric. 30: 57-62.
Cline, W. (1996). ‘The Impact o f Global Warming on Agriculture: A Ricardian Analysis: 
Comment”, American Economic Review, 86: 1308-1311.
80
Cline, W.R. (1991). “The Economics o f the Greenhouse Effect”, Economic Journal, July : 
920-37.
Cline, William R. (1992). “The Economics o f Global Warming”. Institute for International 
Economics, Washington D.C., USA.
Curtin, P., S. Feierman, L. Thompson and J. Vansina (1978). African History. Longman, 
London, U.K.
Davis, G.N. and R.H. Porter (1936). “Comparative Absorption o f  Water by Endosperm and 
Embryo o f Com Kernels”, Assoc. Offic. Seed Anal. Proc., 28: 62-67.
Delouche, J. C. (1953). “Influence o f Moisture and Temperature Levels on the Germination 
of Com, Soybeans and Watermelons”, Assoc. Offic. Seed Anal. Proc., 43: 117-126.
Dinar, A and H. Beach (1998). “Measuring the Impact o f and the Adaptation to Climate 
Change in Agriculture and Other Sectors -  Literature Review” In A. Dinar, R. 
Mendelsohn, R. Evenson, J. Parikh, A. Sanghi, K. Kumar, J. McKinsey and S. 
Lonergan (eds.), ‘Measuring the Impact o f  Climate Change on Indian Agriculture", 
World Bank Technical Paper 402, Washington DC, USA.
Dinar, A and R. Mendelsohn (1999). “Climate Change, Agriculture and Developing 
Countries: Does Adaptation Matter?”, The World Bank Research Observer, 14(2): 
277-93.
Dom-Anin, K. (1985). “Optimum Planting Dates and Maize Cultivation in Ghana: A 
Review”. B.A Thesis Presented to the Department o f Geography, University of 
Ghana, Legon, Accra, Ghana.
Doneen, L.D. and MacGillivray, J.H. (1943). “Germination (Emergence) o f Vegetable Seed 
as Affected by Different Soil Moisture Conditions”, Plant Physiol., 18: 524-529.
81
Downing, T.E. (1992). “Climate Change and Vulnerable Places: Global Food Security and 
Country Studies in Zimbabwe, Kenya, Senegal and Chile” Research Report I, 
Environmental Change Unit, University of Oxford, Oxford, U.K.
Endrody-Younga, S. (1968). “The Stem Borer Sesamia botanephaga Tams and Bowden 
(Lep., Noctuidae) and the Maize Crop in Central Ashanti”, Ghana J. Agric. Sci., 1: 
103-131.
Engle, R.F. and C.W.J. Granger (1987). “Co-integration and Error Correction: 
Representation, Estimation and Testing”, Econometrica, 55 (2): 251-278.
EPA (2000). “Ghana’s  Initial National Communication Under the United Nations 
Framework Convention on Climate Change” Environmental Protection Agency 
(EPA), Accra, Ghana.
Fankhauser, S. (1995). Valuing Climate Change: the Economics o f  the Greenhouse. 
Earthscan Publications, London, U.K.
Fleming, J. R. (1998). Historical Perspectives on Climate Change. Oxford University Press, 
Oxford, U .K
Fosu, K.Y., N. Heerink, K.E. Ilboudo, M. Kuiper and A  Kuyvenhoven (1997). 
“Agricultural Supply Response and Structural Adjustment in Ghana and Burkina 
Faso: Estimates from Macro-level Time Series Data”. In W.K. Asenso-Okyere, G. 
Benneh and W. Tims (eds.) Sustainable Food Security in Central West Africa, 
Kluwer, Boston, USA
Fourier, J.B.J. (1824). “General Remarks on the Temperature o f the Terrestrial Globe and 
Planetary Spaces”, Annales de Chimie et de Physique. 11-12.
GOG. (1982). Economic Survey. Government o f Ghana (GOG), Central Bureau o f Statistics, 
Accra, Ghana.
82
GSS (2000). Ghana Living Standards Survey Report: Round 4, Ghana Statistical Service 
(GSS), Accra, Ghana.
GSS (2002a). Quarterly Digest o f  Statistics. Ghana Statistical Service (GSS), Accra, Ghana.
GSS (2002b). 2000 Population and Housing Census: Summary Report o f  Final Results 
Ghana Statistical Service (GSS), Accra, Ghana.
Granger, C.W.J. (1969). “Investigating Causal Relationships by Econometric Models and 
Cross-Spectral Methods”, Journal o f  Econometrics, 2:111-20.
Granger, C.W.J. and R.F. Engle (1985). “Dynamic Model Specification with Equilibrium 
Constraints: Cointegration and Error-Correction” Mimeo, University o f California, 
San Diego, USA.
Gujarati, D.N. (1995). Basic Econometrics. Third edition. McGraw-Hill, Inc., Singapore.
Hallam, D. and R. Zanoli (1993). “Error Correction Models and Agricultural Supply 
Response”, European Review o f  Agricultural Economics, April: 151-66.
Hamid, O.B. (1984). Rainfall Variability and its Effects on Agriculture in the Upper West 
Area. B. A Thesis Presented to the Department o f Geography, University o f Ghana, 
Legon, Accra, Ghana.
Harris R.I.D. (1995). Using Cointegration Analysis in Econometric Modelling. Prentice 
Hall/Harvester Wheatsheaf, Hertfordshire, Great Britain.
Heerink, N., P. Atsma and K.Y. Fosu (1997). “Farmers’ Transport Costs and Agricultural 
Production in Ghana” . In, W.K. Asenso-Okyere, G. Benneh and W. Tims (eds.), 
Sustainable Food Security in Central West Africa. Kluwer, Boston, USA.
Houghton, J.T., G.J. Jenkins and J.J. Ephraums (1990). “Executive Summary and 
Introduction”. In, J.T. Houghton, G.J. Jenkins and J.J. Ephraums (eds.), Climate 
Change: The Intergovernmental Panel on Climate Change: Scientific Assessment. 
(pp.xi-xxxix), Cambridge University Press, Cambridge, U.K.
83
Hulme, Mike (1994). “Using Climate Information in Africa: Some Examples Related to 
Drought, Rainfall Forecasting and Global Warming”. IDS Bulletin. 25(2): 59-68.
Hulme, M. (ed.) (1996). “Climate Change in Southern Africa: An Exploration o f  Some 
Potential Impacts and Implications in the SADC Region”. Climatic Research Unit, 
University o f East Anglia, Norwich, U.K.
Intergovernmental Panel on Climate Change (1990). Climate Change: The IPCC Scientific 
Assessement. Cambridge University Press, Cambridge, U.K.
Intergovernmental Panel on Climate Change (IPCC) (1996). J.T. Houghton, L.G. Meira 
Filho, B.A  Callander, N. Harris, A. Kattenberg and K. Maskell (eds.). Climate 
Change 1995: The Science o f  Climate Change. Cambridge University Press, 
Cambridge, U.K.
Intergovernmental Panel on Climate Change (IPCC) (1997). “Guidelines for National 
Greenhouse Gas Inventories”, Intergovernmental Panel on Climate Change /  
Organization for Economic Cooperation and Development, Paris, France.
Intergovernmental Panel on Climate Change (IPCC) (1998). Intergovernmental Panel on 
Climate Change (IPCC): An Assessment o f  Vulnerability. In, R.T. Watson, M.C. 
Zinyowera and R.H. Moss (eds.), A Special Report o f  IPCC Working Group II. 
Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.
Intergovernmental Panel on Climate Change (IPCC) (2001a). Climate Change 2000: The 
Science o f  Climate Change. Contribution o f  Working Group I  to the Third Assessment 
Report o f  the Intergovernmental Panel on Climate Change. Cambridge University 
Press, Cambridge, U.K.
Intergovernmental Panel on Climate Change (IPCC), (2001b). “The Regional Impacts of 
Climate Change: An Assessment o f  Vulnerability”. Cambridge University Press, 
Cambridge, U.K.
Johanssen, S. (1985). “The Mathematical Structure o f Error-Correction Models” . Discussion 
Paper, Mathematics Department, University o f Copenhagan, Copenhagan, Denmark.
Johnston, J. and J. DiNardo (1997). Econometric Methods. McGraw-Hill, Inc., New York, 
USA.
Jones, P.G. and P.K. Thornton (2003). “The Potential Impacts o f  Climate Change on Maize 
Production in Africa and Latin America in 2055”, Global Environmental Change, 13: 
5 1 -5 9 .
Kane, S., J. Reilly and J. Tobey (1992). “An Empirical Study o f the Economic Effects of 
Climate Change on World Agriculture”, Climatic Change, 21: 17-35.
Kemp, David D. (1994). “Global Environmental Issues: A Climatological Approach”, 
(Chapter 3), Drought, Famine and Desertification. London and Routledge, New 
York, USA.
Killick, T. (1989). Economic Development and the Adaptive Economy. Working Paper 31, 
November. ODI, London, U.K.
Lau, L.J. (1976). “A Characterisation o f the Normalised Restricted Profit Function”, Journal 
o f  Economic Theory, 12: 131-63.
Lean J. and J.D. Mahlman (1999). “What's Driving Climate Change in the 20th Century - 
Changes in Solar Radiation or the Buildup o f Greenhouse Gases?” US Global Change 
Research Program (USGCRP) Seminar, 23 November.
Leyenaar, P. (1976). “Maize Production in the Coastal Savanna Zone o f  Ghana” , A M.Sc 
Thesis Presented to the Faculty of Graduate Studies o f the University o f  Guelph, 
Guelph, Canada.
85
Liverman, D.M and K.L. O ’B rien  (1991). “Global Warming and Climate Change in Mexico” , 
Global Environmental Change 1: 351-64.
Lonergan, S. (1998). ■'Climate Warming and India” In, A. Dinar, R. Mendelsohn, R. 
Evenson, J. Parikh, A. Sanghi, K. Kumar, J. McKinsey and S. Lonergan (eds.), 
“Measuring the Impact o f  Climate Change on Indian Agriculture", World Bank 
Technical Paper 402, Washington DC, USA.
Marfo, K.A and M.D. Read (1985). “Adoption o f  Improved Maize Production Technology in 
the Ashanti Region o f  Ghana'. A Follow-up o f Verification/Demonstration Trials. 
Paper Presented at the 5th National Maize and Cowpea Workshop, Kumasi, Ghana.
Matarira, C. H., F. C. Mwamuka and J. M. Makadho (1996). “Adaptive Measures for 
Zimbabwe's Agricultural Sector” In, J.B. Smith el al. (eds.), Adapting to Climate 
Change: An International Perspective (pp. 129-147), Springer-Verlag, New York, 
USA.
Mendelsohn, R. and L. Williams (2002). “ Comparing Forecasts o f the Global Impacts o f 
Climate Change” Paper Presented to the Workshop on Regional Climate, Water and 
Agriculture: Impacts on and Adaptation o f Agro-Ecological Systems in Africa. Cape 
Town, South Africa, 4-7 December.
Mendelsohn, R. A. Dinar and A. Sanghi (2001). “ The Effect o f Development on the Climate 
Sensitivity o f Agriculture” , Environment and Development Economics. 6: 85-10 1.
Mendelsohn, R., W. Morrison, N. Andronova and Schlesinger, M. (2000). "Country-Specific 
Market Impacts from Climate Change” , Climatic Change, 45: 553-569.
Mendelsohn, R. and M. Schlesinger (1999). “Climate Response Functions” , Ambio, 28: 362 
-36 6 .
Mendelsohn, R. and J. Neumann (eds.) (1998). The Impact o f  Climate Change on the 
Economy o f  the United Stales. Cambridge University Press, Cambridge, U.K.
86
Molua, E.L. (2002). “ Climate Variability, Vulnerability and Effectiveness o f Farm-Level 
Adaptation Options: the Challenges and Implications for Food Security in 
Southwestern Cameroon” , Environment and Development Economics. 7: 529-545.
Muchena, P. (1994). “ Implications o f Climate Change for Maize Yields in Zimbabwe” . In, C. 
Rozenweig, A. Iglesias (eds.). Implications o f Climate Change fo r  Internationa! 
Agriculture. US Environmental Protection Agency, Washington DC.. USA.
Muchena, P. and A. Iglesias. (1995). ' ‘Vulnerability o f Maize Yields to Climate Change in 
Different Farming Sectors in Zimbabwe”  In, Climate Change and Agriculture: 
Analysis o f  Potential International Impacts (229-240). ASA  Special Publication No 
59, American Society o f Agronomy, Madison, Wisconsin, USA.
Nordhaus, W. (1991). “Too Slow or Not too Slow: The Economics o f the Greenhouse 
Effect” , Economic Journal, 10 1: 920-937.
Nordaus William D. (1993). “ Climate and Economic Development: Climates Past and 
Climates Change Future”  Proceedings o f  the World Bank Annual Conference on 
Development Economics. The World Bank, Washinton, D.C.,USA.
Ofori-Sarpong, E. (1985). “ The Nature o f Rainfall and Soil Moisture in the Northeastern Part 
o f Ghana During the 1975 -  1977 Drought” , Geografiska Annaler, 67: 177 -  186.
Ofori-Sarpong, E. (2001). ‘ Impact o f Climate Change on Agriculture and Fanners' Coping 
Strategies in the Upper East Region o f Ghana” , West African Journal o f  Applied 
Ecology, 2: 21-35 .
Ontoyin, Y. (1993). A Comparative Study o f  Temperature as a Climate Indicator fo r  Periods 
1931-1960 and 1961-1990. WRRI, CSIR, Accra, Ghana.
Owusu, G. (2002). “Climatic Variability and Crop Production in the Forest-Savannah 
Ecotone: A Case Study o f Sekyere-West District" B.A Thesis Presented to the 
Department o f Geography, University o f Ghana, Legon, Accra, Ghana.
87
Pagan, A.R. and MLR. Wickens (1989). “A Survey o f Some Recent Econometric Methods” , 
Economic Journal, 99: 962-1025.
Pardey, P.G. and N.M. Beintema (2001). ’'Slow Magic: Agricultural R&D a Century After 
Mendel” , Food Policy Report, International Food Policy Research Institute, 
Washington, D.C., USA.
Parry, M. L. (1990). “Climate Change and World Agriculture” Earthscan Publications, 
London, U.K.
Parry, M.L. (ed.) (2000). “Assessment o f Potential Effects and Adaptations for Climate 
Change in Europe”  The Europe Acacia Project. University o f East Anglia,U.K. and 
European Commission.
Pearce, D., W. Cline, A. Achanta, S. Fankhauser. R. Pachauri. R. Tol and P. Vellinga 
(1996). “The Social Cost o f Climate Change: Greenhouse Damage and the Benefits o f 
Control” . In, J. Bruce, H. Lee, E. Haites, (eds.), Climate Change 1995: Economic and 
Social Dimensions o f  Climate Change, Cambridge University Press, Cambridge, U.K. 
Rasmussen, K. (1999). “ Land Degradation in the Sahel-Sudan: the Conceptual Basis” .
Danish Journal o f  Geography, Special Issue, 2: 15 1-159 .
Reilly, J., W. Baethgen, F.E. Chege, S.C. van de Geijn, L. Erda, A. Igiesias, G. Kenny, D. 
Patterson. J. Rogasik, R. Rotter, C. Rosenzweig, W. Sombroek, J. Westbrook, D. 
Bachelet, M. Brklacich, U. Dammgen, M. Howden, R.J.V. Joyce, P.D. Lingren. D. 
Schimmelpfennig, U. Singh, O. Sirotenko, and E. Wheaton (1996). "Agriculture in a 
Changing Climate: Impacts and Adaptation” In, Climate Change 1995: Impacts, 
Adaptations and Mitigation o f  Climate Change: Scientific-Technical Analyses (pp 
427-467). Contribution o f Working Group II to the Second Assessment Report o f the 
Intergovernmental Panel on Climate Change, R.T. Wa.jon, M.C. Zinyowera, and
88
R.H. Moss (eds.). Cambridge University Press, Cambridge, U.K. and New York, NY, 
USA.
Ribot, J.C., A. Njam and G. Watson (1996). “ Climate Variation, Vulnerability and 
Sustainable Development in the Semi-arid Tropics” . In, J.C. Ribot. R. A. Magalhaes 
and S. S. Panagides (eds.). Climate Variability, Climate Change and Social 
Vulnerability in the Semi-Arid. Tropics (13-51). University Press, Cambridge, U.K.
Rosenberg, N.J. and P.R. Crosson (1990). Processes fo r  Identifying Regional Influences o f  
the Responses to Increasing Atmospheric C02 and Climate Change: the MINK 
Project. An Overview. Resources for the Future. Washington, DC., USA.
Rosenzweig, C.. and D. Liverman (1992). “ Predicted Effects o f Climate Change on 
Agriculture: A  Comparison o f Temperate and Tropical Regions". In, S. K. Majumdar 
(ed.). Global Climate Change: Implications, Challenges and Mitigation Measures 
(342-61). The Pennsylvania Academy of Sciences, PA., USA.
Rouanet, G. (1992). Maize. The Tropical Agriculturalist Series. MacMillan. London, U.K.
Sanghi, A., R. Mendelsohn and A. Dinar (1998). 'The Climate Sensitivity o f Indian 
Agriculture" In, A. Dinar, R. Mendelsohn, R. Evenson, J. Parikh, A. Sanghi, K. 
Kumar. J. McKinsey, and S. Lonergan (eds.), 'Measuring the Impact o f  Climate 
Change on Indian Agriculture. World Bank Technical Paper 402, Washington DC. 
USA.
Schelling, T. (1992). “ Some Economics o f Global Warming” , American Economic Review, 
82: 1-14 .
Schulze, R., G. Kiker and R.P. Kunz (1996). "Global Climate Change and Agricultural 
Productivity in Southern Africa: Thought for Food and Food for Thought”  In. T. 
Downing (ed.), Climate Change and World Food Security. North Atlantic Treaty
89
Organisation ASI Series, Volume 137, Springer-Verlag, Berlin and Heidelberg, 
Germany.
Takyi, S.K. (1970). “ Effects o f Land Preparation on Maize Yields” , Ghana J. Agric. S c i 3: 
15 1-154 .
UNEP/ OECD/ IEA/ IPCC, (1995) IPCC Guidelines fo r  National Greenhouse Gas 
Inventories. Bracknell, U.K..
van der Geest, K. (2002). “  Vulnerability and Responses to Climate Variability and Change 
Among Rural Households in Northwest Ghana' M.A. Thesis, University o f 
Amsterdam, Faculty o f Social and Behavioural Sciences, C.M. Kan Institute (Human 
Geography and Planning) Impact o f Climate Change on Drylands with a Focus on 
West Africa (ICCD). Amsterdam, The Netherlands.
Watson, R. T. (2001). “The Challenge o f Climate Change: Poor Fanners at Risk” In, CGIAR 
Annual Report 2000. Consultative Group on International Agricultural Research 
(CGIAR) Secretariat, The World Bank, Washington. D. C., USA.
World Bank (1996). Toward Environmentally Sustainable Development in Sub-Saharan 
Africa: A World Bank Agenda. The World Bank, Washington, DC., USA.
Zekri, S. (2002). "Economic Impacts o f Climate Change on the Tunisian Agricultural 
Sector”  A Paper Presented at the Workshop on Regional Climate, Water and 
Agriculture: Impacts on and Adaptations o f Agro-Ecological Systems in Africa. Cape 
Town, South Africa, 4-7 December.
90