CHAPTER THREE
METHODOLOGY
3.0: introduction
The study examines the determinants of cocoa production in
the South West Region of Cameroon. The Region is bounded to
the north by the North West Region, south by the Atlantic
Ocean, to the west by Nigeria and in the East by the Litoral
region. The region has a population of covered by cocoa
production (MINADER, 2013). This study covers all the cocoa
producing areas of the six Divisions of the South West Region.
The research design adopted here is across-sectional research
design which includes both survey and ethnographic methods
(Ranjit 1996).
3.1. Research Design
In the study, both quantitative and qualitative approaches were
employed, because it is important to identify and analyze, in
detail, the status and determinants of agricultural productivity,
and household income level. In order to get all the necessary
information on the area under which the research was
conducted, primary sources of information were mainly used.
The main sources of data were mainly households, who
engaged in agricultural activities. In addition to this, community
leaders, government officials, and key informants were also
used as sources of information. The informants that were
considered as key informants include influential persons in the
area, local officials and well-informed people. These people
were more aware of the situation and were expected to give
extra and complementary information regarding Agriculture
Household income level. These sources helped the researcher
to get first hand and relevant information about the real
situation of agriculture. Both quantitative and qualitative data
was collected using questionnaires, and interviews.
Quantitative data was collected through the use of structured
questionnaires from sample households (farmers).
3.2Area of study
The research was carried out in Cameroon's south and east
regions. The Republic of Cameroon's South Western and South-
Central regions make up the South region. The East region
borders it on the east, the Centre region on the north, the
Littoral region on the northwest, the Gulf of Guinea on the west,
and Gabon, and Congo on the south. The South area covers
47,720 square kilometers, making it the country's fourth largest
region. The Beti-Pahuin peoples, such as the Ewondo, Fang, and
Bulu, are the most important ethnic groups. The South has a lot
of industry, with logging, forestry, mining, and offshore oil
drilling being the main sources of revenue. Industrial
agriculture is also vital in the south, with cocoa and rubber
being the most important cash crops. Cattle farming and fishing
are also key economic activities. The great bulk of the
population are subsistence farmers (Nkondjio et al., 2019).
It is the largest and most sparsely populated region in the
country, occupying 109,002 km2 of territory. While the area is
rich in flora, but not fertile for nutrients are seeped out from
these soils. (Maisels et al., 2014). High temperatures (on
average 24°C) and a lack of conventional seasons characterize
the East. Rather, this area experiences a dryness from
December to May which is lengthy, with a short period of
wetness from May to June, and a shorter period of dryness from
July to October. This seasonal variation is crowned with an
unembellished wetness from October to November. The study
area experiences average rainfall of 1500-2000 mm per year,
especially the far east and far north, with lesser rainfall
averages (Maisels et al., 2014).
3.3 Sampling method
Central Statistical Agency (CSA) uses the sample frame as the
standard procedure for data collection for which the slope
containing Enumerations Area (EA) of Jimma and its respective
agricultural households was obtained from the 2007 (1999 E.C)
Population and Housing Census Frame which was used as the
sampling frame for the selection of EAs Primary Sampling Units
(PSU).
As regard to sampling, its objective is to select a set of
elements from a population using Random sampling which
enhances the probability of accomplishing this objective and
also allows for the objective assessment of the reliability of the
sample. Therefore one hundred and fifty (150) households were
selected as sampling size from the study area for which each
element of the population stands an equally opportunity of
being selected; whereas the sample frame was 360. Taking 150
samples from the study area was reasonably achievable and
desirable in line with research fund and that for Woredas Jimma
Zone which comprises of Gomma and Limu. Therefore, this
project was undertaken for a total of two Woredas as the study
site. Based on the 2007 Census conducted by the CSA, Jimma
Zone has a population of approximately 2.5 Million, of whom
50.3% are men and 49.7% are women and with area of
15,568.58 square. Consequently, all sample EAs was selected
from this frame based on the design proposed for the survey.
3.4 Data types and source
The data for this research came from the panel cultivated
Cameroon Rural Household Socioeconomic Survey (CRHSS),
conducted by Central Statistical Agency (CSA) with the alliance
of the World Bank Living Standard Measurement Study (ISLM)
team. The survey was conducted for the year 2013/14 in
Cameroon as full sample coverage at regional. The primary
data for the research contained two part namely agriculture,
and households‟ diverse socioeconomic characteristics. The
agriculture part includes cultivated land area, the types of input
used, crop production and Livestock production of the farm
households. The Household socio-economic characteristic
comprises of household borrowing, lending, from where they
borrow and to whom they lend, food consumption items, the
income from different source including remittance and source
of their income, educational status of the household members
and household demographic characteristics etc. This
socioeconomic survey focused on the number of household
assets, the loan amount of money in cash or in kind that the
household received, time of repayment, if the household repay
on time or not and the amount to be paid back and the reasons
for those who do not get loan.
The research also utilised secondary data sources, which
covers in period of 1991 to 2020 in which majority of the data
on Crop production, fertilizer, agricultural land, tractor, labor
and export are obtained.
The source of the data acquired from World Bank and relevant
thesis and journals. Data for this study is time series data in
which observations are made on the basis of several variables
over a certain period and are usually organized sequentially
when the time horizon, such as annually, is different.
The study was collected annually for all the variables such as
crop production, fertilizer, agricultural land, tractor, labor and
export, quantitative data will also be analyzed using trend
analysis and descriptive statistics including standard deviation,
mean, kurtosis and skewness.
Table 3.1 Summarizes the variables used in the study, their
measurements, data sources and hypothesis.
3.5 Analytical tool and model Specification
The study used both descriptive statistical tools and
econometric models. Descriptive statistics like percentages,
ratios, mean values, Standard Deviation, and frequencies were
used to examine demographic and socio-economic features of
respondents. In addition, the micro-econometric technique,
multiple regressions model, was applied to identify the main
socio-economic determinants of agricultural productivity. The
data which was collected through household survey were
entered, manipulated and analyzed using STATA software. In
addition MS-Excel was used to supplement the software.
3.5.1 Model Specification:
The previous studies that were conducted in different part of
the world used various approaches to examine determinants of
agricultural productivity. For instance, Camelia Burja (2012)
used Data Envelopment Analysis (DEA) method; Meseret
Urgecha Kussa (2012) Cobb Douglas stochastic frontier
analysis. Moreover, agricultural productivity can be measured
in aggregate or partial. Though a conventional agricultural
productivity index (Total factor productivity), which is a
measure of output divided by a measure of inputs used in the
production and the best indicator of agricultural productivity. It
is very difficult to construct TFP measures since it is often
difficult to key value inputs where markets are not well-
functioning. Hence, many studies used partial factor
productivity (PFP) to derive agricultural productivity. Partial
factor productivity measure divides physical output by a given
physical factor input, like land, labor, etc.
This study used multiple linear regression models to examine
factors affecting agricultural productivity in the study area. To
measure agricultural productivity, partial factor productivity
was used. As partial factor productivity cultivated total land
was taken as physical input. Variations in land productivity
arose from differences in socio-economic variables,
demographic variables, technology or variations in others
(unmeasured inputs). After seriously reviewing previous
literature including potential variables that affect productivity,
the used regression model is specified as follows:
Where:
Dependent variable: Agricultural productivity expressed as
Crop value per hectare of total cultivated land. It expressed as
logarithm functional form.
Independent variable: Fertilizers, agricultural land, Tractor,
Labor productivity, Export, are coefficient of each variables
that measure by how much crop value per hectare of land
changes when the given variable changes by a unit. The
ordinary least square estimation techniques were based on
some basic assumptions like homoscedasticity of error term
variance, error term should be normally distributed, there
should be no serious correlation among explanatory variables
(no serious problems of multicollinearity), and no problem of
autocorrelation. Therefore, it was mandatory to conduct various
diagnostic tests to apply ordinary least square estimation
techniques. Based on this, the study used Breusch- Pagan
Godfrey test for heteroscedasticity, which were used on the
residuals to determine the OLS assumption on the error term.
The Ramsey RESET test for functional form was conducted to
check for the correct specification of the error-term. The
variance inflation factor (VIF) test was conducted to check for
existence of multicollinearity problem.
The collected data through household survey was entered,
manipulated and analyzed using STATA 13 software. In addition
MS-Excel was used to supplement the software. Tables were
used to describe the data result. The expected relationship
between dependent and independent variables used for this
study were summarized in the Table 3.2 below.
Path regression as well as the Principal Component analysis
(PCA) has been adopted in this study. Path analysis is the
extension of multiple regressions. It forms the bases of
structural equation modeling. This approach to data analysis
was adopted because it is easy to understand and to test for
relationship between sets of exogenous and endogenous
variables. This method of estimation through the maximum
likelihood is equally useful in testing for counterintuitive
reasons.
3.5.1.1 The conceptual model
The Barnum-Squire farm household model developed by
Barnum and Squire (1970) provides a framework for generating
predictions on how farm households respond to changes in
output and domestic variables (family size and structure) and
market variables (output prices, input prices, wage rates and
technology) among others. The production function in this
model refers to farm output which can be traded not just as
home production for direct use. This model is very vital in this
study given that cocoa output in the South West Region and
Cameroon in general is mainly for sale. The Barnum-Squire
production function is stated as:
Y = f (A, L,
V)..................................................................................................
(3.1)
From the production function above, Y is the output, A is land
under cultivation, Lis the total labour input (both household and
hired) and Vis other variable inputs used in production.
This study therefore augments this production function by
adding some of the variables deemed necessary in this work
that the original Banum-Squire model did not take into
consideration. Hence we present an augmented Banum-Squire
model thus:
AGP(CPI)= f
(-------------------------------------------------------------------------------------
(3.2)
Where CCP = cocoa production
Zr= Socio-economic variables (household size, price, farm size
(land under cultivation), education of household head, and
government assistance, chemical spray an: agricultural
land( AGLND),Fertilizer( FZER), Export(Ex), Tractor (TR), Labor
productivity (LP)
From here the study conducted a pre-test analysis known as
principal component analysis (PCA); a Varimax method of
extraction used in computing the communalities of the items.
Communality captures the percentage of variability in cocoa
production. That is, the extent to which each item can explain
the variation in cocoa output. The extraction of the
communalities shows that each of the items retained for further
statistical investigation in this study can at least explained
more than 50% of the variance in the regression factor scores.
This is therefore the reason why some of the observed
variables in our conceptual model will no longer appear in our
econometric model below.
3.5.1.2 Cobb-Douglas Production Function
The Cobb-Douglas function type is homogeneous of degree E$i.
The return to scale parameter, or function coefficient, is equal
to the sum of the $ values at each input, provided all inputs are
explicitly treated as variables. The $ values represent the
elasticity of production with relation to the corresponding input
and are constants (Debertin, 2012).
Returns to scale show the proportionate change in output when
all components experience the same proportional change.
Cobb–Douglas production function is a specific kind of
production function that is often used to depict the technical
connection between the quantity of two or more inputs
(typically physical capital and labor) and the quantity of output
that may be generated by those inputs.
Cobb-Douglas production function states that returns to scale
equal labor and capital output elasticities: α + β. If α + β = 1
you may argue that returns to scale are constant. It indicates
that doubling the quantity of capital and labor would results in
double the output. This is what Paul Douglas saw when he
originally established the function from the available United
States industry data (Manias, 2022).
3.2 Model used in the analysis
Model specification is derived from theoretical framework as
follows;
CrP= β0 + β1Frt+ β2 AgricLND + β3Tr + β4 L + β5EXP+ ε
Where:
Frt – Fertilizer, AgricLND – Agricultural land, Tr –Tractors , L –
Labor, EXP – Export β0 – Constant, β1, β2, β3, β4 and β5 are
Regression coefficients.
ε - error term.
3.6 imitation of the study
Lack of existing research on the issue that would have
permitted further analysis, In fact, the previous studies forms a
separate section and generally occupies the second section of
the framework of scientific inquiry, and is of course closely
related to it.
The prior studies contribute to the improvement and
enrichment of scientific research and provide the researcher
with a solid basis and essential information about the research
carried out.In this way, the researcher should be familiar with
all the information related to their research. In addition, the
previous studies also help the researcher not to repeat the
areas already studied an
