International Journal of Statistics and Applied Mathematics 2019; 4(3): 23-31

ISSN: 2456-1452
Maths 2019; 4(3): 23-31
© 2019 Stats & Maths An analysis of socio-economic and demographic factors
www.mathsjournal.com
Received: 10-03-2019 affecting household main source of income in Somalia
Accepted: 12-04-2019
using binary logistic regression approach
Mohamed Hussein Abdullahi
Senior Statistician, Directorate
of National Statistics of Somalia, Mohamed Hussein Abdullahi
Mogadishu, Somalia
Abstract
This study applied the Binary Logistic Regression model to investigate socio economic and demographic
factors affecting household main source of income. The main objective of this study is to determine the
socio-economic and demographic factors that affect household main source of income in Somalia. This
study used secondary data of cross sectional data targeting households from household survey in 2016.
The results of the four independent variables in the study which are household head education, household
head sex, household residential area and age of household head show that the residence area of
households is the most important factor which determines the household main source of income.
The sex of the household head is revealed to be the second most important variable among the variables
that have an effect on household main source of income. It has been found that the male headed
households are more likely to have a salaried labour source of income than female headed households.
The results show that the opportunity of getting a salaried labour source of income is not the same for
both the educated and the non-educated household heads. The results, also suggest that the age of
household head has a negative relationship with the salaried labour source of income. It is found that an
increase of one year of the age of the household head will decrease the opportunity of the household of
getting a salaried labour source of income.

Keywords: Household main source of income, chi square test, logistic regression, odds ratio

1. Introduction
A household is defined as “a group of persons who share the same living accommodation, who
pool some, or all, of their income and wealth and who consume certain types of goods and
services collectively, mainly housing and food” (UNDESA 2008). Also, this definition
mentions that households have a head and share residence, and that both head and the residents
of the household have characteristics that influence their livelihood. The most important
characteristics are education, age, residential area; rural and urban, and others. These
characteristics have a great effect and impact on the source of the household income.
Household Main Source of Income is the major concern mostly in developing countries. The
statistical reports produced by statistical offices in developing countries and developed
countries illustrated employment conditions and the income sources of the households in their
respective countries. Also the statistical offices inform government and citizens about
livelihood changes. The source of income reflects the livelihood circumstances that surround
any society in the world. The household main source of income has an effect on decision
making, planning, because without knowing it the government and other institutions, will find
it difficult to develop policies towards employment conditions.
The income source of the household measures the living standard differences between Male
and Female Headed Households, where by the income is one of the monetary dimensions for
measuring well-being of households whether in either urban or rural areas or either female or
male household headed. Differences among household living standards may be due to the
Correspondence residence of the household, education level of the household head, the sex of the household
Mohamed Hussein Abdullahi
head or the age of household head. The (UBOS 2014) [12] has published Average Monthly
Senior Statistician, Directorate
of National Statistics of Somalia, Household Income by region in either urban or rural areas. Such classification of households
Mogadishu, Somalia reveals the inequalities between urban and rural areas due to their socio-economic differences.
~23~
International Journal of Statistics and Applied Mathematics

(Sami Gungor et al. n.d., 2015) [9] found that the probability They have found evidence in the analysis that, as against the
of a household being poor in Turkey decreased as the common perception of agriculture being the dominant source
education level of the household head increased. (Janet of income for farm households, the households earns close to
Currie, 1992) [6] observed that male jobs are paid more than half of their income from non-farm activities. Also they have
female jobs and that and female jobs are much more likely to found that non- farm income is more significant for the
be found in public than the private sector. households at lower end of land distribution. They have
However, the Households of Somalia face many challenges discovered non farm income sources are accessible to a small
concerning access to and availability and quality of the proportion of farm households and have unbalanced effect on
required social services, such as health, education, and income distribution. In addition to that they have found that
opportunity of jobs. Social services need to be delivered nonfarm sources are positively correlated with the total
equitably to all groups of the population, particularly to income. The research fails to indicate the effect of socio-
vulnerable and marginalized groups, in both rural and urban economic and demographic factors that the households earn
communities. Vulnerability and exclusion of households from close to half of their income from non-farm activities.
generating income will be the cause of lack of ability within Fadipe et al. (2014) [5] have examined the determinant of
families and communities to meet required of living standard. income among rural households in Kwara State, Nigeria.
They have used primary data collected by using a
2. Review of literature questionnaire compiled from 90 randomly selected
Household income sources are one of the key priorities for households. They have employed in the data analysis the
each country to stimulate the economic growth and the analytical tools of descriptive statistics and the multiple
improvement of the living standard. This portion highlights regression analysis for the study. They have identified that the
the pertinent empirical literature on the subject investigated. farm size and access to electricity, level of education of the
Mathebula et al. (2017) [7] have estimated household income household head, and sex of the household head were the main
diversification in settlement types of the poorest provinces in determinant of household income in the study area. They have
South Africa – the Eastern Cape, Limpopo and KwaZulu- shown that the farm income is the most significant source of
Natal. They have used the number of income sources, the income for rural households in the study area by making up
number of income earners and the Shannon Diversity Index to 57.9% of total household income. The research fails to show
estimate income diversification in the study provinces. They whether socio-economic and demographic factors has
have used 2010/2011 Income and Expenditure Survey from influence that the farm income is the most significant source
Statistics South Africa and Wave 3 data from the National of income for rural households.
Income Dynamics Study. They have found that the diverse of Talukder (2014) [11] has investigated the determinants of
income in the provinces recommend that targeted policy income and growth in income of rural households in
initiatives aimed at enhancing household income are Bangladesh in the post-liberalization era. He has used
important in these provinces. They have shown that the secondary data sources on both (1985-86) and (2005), by
households in the traditional and urban formal areas applying the ordinary least square (OLS) of regression models
diversified income sources to a greater extent than households for assessing the determinants of both economic and non-
in urban informal and rural formal settlements. The research economic characteristics. He has found that household size
fails to present whether socio-economic and demographic was the only non- economic factor that was statistically
factors have an effect diversification of income sources. significant and positive determinant of household income in
Chamicha (2015) [4], has studied the relationship between both 1985-86 and 2005. In addition to that, he found that
nonfarm activities and rural livelihood in Tanzania. The study endowments, household land area were the largest positive
investigated the factors that permit the rural household to determinant while share of income from wage-salary was the
involve in the nonfarm activities. The findings of the study largest negative determinant of income-growth in 1985-86.
divided the factors that the household’s decision to engage in He also found that the endowments, change in share of
nonfarm activities into two factors which are push and pull income from house rent was the largest positive contributor
factors. The findings showed that the income obtained from and share of income from rice was the largest negative
agriculture activities were used as a startup capital in nonfarm contributor to growth in 2005. The research didn’t show
activities and the income obtained from nonfarm activities whether socio-economic and demographic factors has
were used to finance farm activities. The study found that influence the different income sources.
there is a significant share of the income from nonfarm Ali & IsmaeelRamay (2013) [2], have examined the
activities to the overall household income, and also found that determinants of income and income gap in urban and rural
the income obtained from nonfarm activities were used to areas of Pakistan by using province, literacy, education,
acquire different household needs. They have study concluded occupation, age, gender and marital status as predictors at
that rural nonfarm activities are significant livelihood strategy individual level. They have use their study data for the
for the rural households. The research fails to indicate Household Integrated Economic Survey (HIES) 2010-11
whether socio-economic and demographic factors influenced dataset. They have estimated predictions by applying the
nonfarm activities for the livelihood of rural households. ordinary least square (OLS) method, also used Blinder-
Birthal et al. (2014) [3] has examined farm households’ access Oaxaca decomposition method to analyze the income gap
to different income generating activities, and their impact on between urban and rural Pakistan. They have found that
income distribution by using data from a nationally literacy, education and occupation as the major determinants
representative large scale survey in India. The authors have of income in Pakistan. They also discovered that reading and
used a number of methods to measure economic inequalities. writing skill of individuals has been more important as
They used Gini index which is a widely used measure of compared to the numeracy skill. In addition to that they have
inequality because of its certain desirable properties, such as found that the lower levels of education gets high returns in
Pigou-Dalton transfer sensitivity, mean-independence, rural areas whereas higher levels of education gets more
symmetry, population homogeneity, and decomposability. return in urban areas. Individual characteristics like literacy,

~24~
International Journal of Statistics and Applied Mathematics

education, occupation and marital status have been found as 2.1 Conceptual framework
the major determinants of income gap. However, this study The result of Household Main Source of Income, in of terms
will use a secondary data to see if the results will be that the labour source and other source is generally is a function of a
education as the major determinant of income of the number of variables which require grasping the associations
households as this research found. of the variable with the source of household income. Source
Yadollahi et al. (2013) [13] have determined the effect of of income has interrelation with socioeconomic and
demography variables that is, age, gender, level of education, demographic characteristics of the human being and the
and occupation, on family economic status in three economic development. The factors affecting household main
dimensions, income, expenditure, and ownership of physical source of income can be classified under three broad
asset, by using data of 390 households Kerman city of Iran. categories which are demographic & Social, Economic and
They have used to analyze the data Pearson product moment Education variables as follows:
correlation, Chi-square, Spearman rho, and independent A. Demographic & Social variables
sample t-test. They have discovered that the demography B. Education variables are:
variables have effect on family economic status, and the C. Economic variables
dimensions, but this research will examine the effect of socio-
economic and demographic variables on the household main However, conceptual frame work chart is presented in Figure
sources of income. 1, for this study.

Source: Author’s Construction

Fig 1: Conceptual Framework Chart

3. Data Coverage of the Research Frequency Survey (SHFS). Thus, this study is attempting to
The Republic of Somalia is situated in the Horn of Africa. Its determine the extent of the relationships between the
boundaries are defined by the Gulf of Aden and Djibouti in dependent and independent variables using the data obtained
the north, Indian Ocean in the east, the Federal Democratic by the method used by the Directorate of Notational
Republic of Ethiopia in the west, and the Kenyan Republic in Statistics’ Somali.
the south.
The study based on the secondary data which covers Somalis 5. Variables
living in urban and rural areas. This ad hoc approach was to The independent variables are household head (HHH)
create the missing sample frame aimed to ensure education, which is a dummy variable derived to categorize
representativeness of the covered population but has technical the education variable into educated household head and non-
limitations which could not allow collecting data from the educated household head. Second variable is sex of household
nomadic population. head in two categories, male household head and female
The Somali High Frequency Survey (SHFS) Survey in 2016 household head. Third variable is residential area of
was designed to provide indicators and data needed to household which has two categories rural and urban areas and
measure, monitor and analyse living standards and poverty in fourth variable is the age of household head which is
Somalia. The survey used a comprehensive questionnaire to continuous. The dependent variable is household main source
collect data on the number of households and persons living of income which has two categories, salaried labour source of
in the selected areas in the sample. The survey comprised the income and other sources of income.
following three data collection methods:
 A representative household survey (HHS) Table 1: Definition and coding of study variables
 A Market Survey (MS) for prices of food and non-food Variable
items Dependent variable (DV) Code label
Type
 A Currency Exchange Rate Survey (CERS) Salaried Labour
Main source of household
Categorical Source = 1
income
This research has used data from the representative household Other Sources = 0
survey. Independent Variable Variable
Code label
(IV) Type
4. Research Design Educated = 1
Household head education Categorical
The aim of the study is to find out the effect of some socio- Non-educated = 0
economic and demographic variables on the household main Male = 1
Household head sex Categorical
Female = 0
source of income. The study has used secondary data which is
Urban = 1
obtained from the data collected in 2016 by the Directorate of Household residential area Categorical
Rural = 0
Notational Statistics’ Somali (DNS) with close collaboration Age of the household head Continuous N/A
with World Bank and the Survey is called Somali High
~25~
International Journal of Statistics and Applied Mathematics

6. Theoretical Model 𝑥2 = HHH sex and k = Male, as

Logistic regression is a popular and useful statistical method
in modeling categorical dependent variable and independent 𝑒 𝛼+𝛽2𝑥2
variables. The logistic regression is a mathematical modeling 𝑝(𝑥2 ) = (3.4)
1 + 𝑒 𝛼+𝛽2𝑥2
approach used to investigate the relationship between the
independent variable and dependent variable as dichotomous The probability that the main source of income of the
variables. household in the urban residential area will be a salaried
labour source of income, holding household head’s education
1 if the ith household gets labour source of inocome
level, household head’s sex, and household head’s age
constant, is computed for,
Yi =

𝑥3 = 𝐻𝐻 𝑟𝑒𝑠𝑖𝑑𝑒𝑛𝑡𝑖𝑎𝑙 𝑎𝑟𝑒𝑎 𝑎𝑛𝑑 𝑘 = 𝑈𝑟𝑏𝑎𝑛

0 If the ith household gets other sources of income

𝑒 𝛼+𝛽3𝑥3
The essential aim of the study analysis is to describe the way 𝑝(𝑥3 ) = (3.5)
that household main source of income varies by considering 1 + 𝑒 𝛼+𝛽3𝑥3
household head’s education, household head’s sex, residential
Probability that the main source of income of the household
area of household, and the household head’s age. Consider
head of a specified age will be a salaried labour source of
first the case where the response variable 𝑌𝑖 is binary,
income, holding household head’s education level, household
assuming only two outcome values, 1 or 0. This may be
head’s sex and household residential area constant, is
defined as:
computed for,
The distribution of dependent Y is specified by probabilities
𝑝(𝑌 = 1) = 𝜋of success and 𝑝(𝑌 = 0) = (1 − 𝜋) of failure,
𝑥4 = HHH age and 𝑘 = 𝑎𝑔𝑒 𝑖𝑛 𝑦𝑒𝑎𝑟𝑠
and its mean is𝐸(𝑌) = 𝜋. The n independent observations, the
number of successes have the binomial distribution specified 𝑒 𝛼+𝛽4𝑥4
by the index n and parameter π. The formula is shown below 𝑝(𝑥4 ) = (3.6)
1 + 𝑒 𝛼+𝛽4𝑥4
in equation (3.1). So, each binary observation is a binomial
variate with n = 1, (Agresti 1996, p.68) [1]. 8. Descriptive Analyses
Descriptive statistics as a terminology is defined as Methods
𝑛!
𝑝(𝑦) = 𝜋 𝑦 (1 − 𝜋)𝑛−𝑦, of organizing, summarizing, and presenting data. Descriptive
(𝑛
𝑦! − 𝑦)! statistics as computed quantities are very important in
researches before proceeding to the model, because they
y = 0,1 ,2, ....., n (3.1) simply enable presentation of data in a summary way, to
allow simpler interpretation of the data. To report summary
7. Associated probabilities of the variables findings on the socio-economic and demographic profile of
Let x1, x2, x3, x4 denote the variables of household head’s households in the research, simple descriptive statistics such
education level, household head’s sex, household residential as frequencies and percentages were generated.
area, and household head’s age, respectively. Then the
associated probabilities of the variables are computed from 9. Chi-Square Test of Independence
the equation: The outcome and independent variables are both categorical
variables. In order to test the relationship between the
𝑒 𝛼+𝛽1𝑥1+𝛽2𝑥2+𝛽3𝑥3+𝛽4𝑥4 outcome and the individual independent variable, the Pearson
𝑝(𝑥𝑖 ) = (3.2)
1 + 𝑒 𝛼+𝛽1𝑥1+𝛽2𝑥2+𝛽3𝑥3+𝛽4𝑥4 chi-square test for independence will be used by examining
for a statistically significant relationship between two
Probabilities of individual categories are then calculated by categorical variables at a time.
considering:
𝑥𝑖 = 𝑘, k = 1 and 0, and i = 1, 2, 3 and 4, 10. Binary logistic Regression
Binary Logit Model or Logistic Regression model is most
for one variable holding the other three variables constant. useful when the response variable is not continuous but has
The probability that the household main source of income of only two possible outcomes (dichotomous), 1 or 0. This
the educated household head will be a salaried labour source model is typically used when predicting an event which has
of income holding household head’s sex, household two possible outcomes. Since the probability of an event must
residential area, and household’s age constant, is then lie between 0 and 1, it is impractical to model probabilities
computed for: with linear regression techniques, because the linear
regression model allows the dependent variable to take values
𝑥1 = HHH education level and greater than 1 or less than 0. The logistic regression model is
k = Educated, as: a type of generalized linear model that extends the linear
regression model by linking the range of real numbers to the 0
𝑒 𝛼+𝛽1𝑥1 or 1 range (Agresti 1996, p.70) [1].
𝑝(𝑥1 ) = (3.3)
1 + 𝑒 𝛼+𝛽1𝑥1
11. Results
The probability that the household main source of income of a 11.1 Household Head’s Education Level distribution
male headed household will be a salaried labour source of The survey collected data on household’s education level. The
income, holding household head’s education level, household household heads or representative respondents were asked by
residential area and household head’s age a constant, is the enumerators of Somali High Frequency survey (SHFS) the
computed for:
~26~
International Journal of Statistics and Applied Mathematics

question “what is the education level of household head?” households of socio-economic survey in 2002 was 87.4%
Respondents were supposed to choose one among the listed male headed, and 12.6% female headed.
responses:
i) Complete secondary 11.3 Household Residential area distribution
ii) Complete primary but in complete secondary Figure 3 below displays the distribution of household heads
iii) Incomplete primary by type of residential area and sex. The rural communities
iv) No education there are no more difference in the numbers and percentage of
v) Other education households headed by male and female, the difference is
0.97%. While urban communities are almost same to the rural
vi) University
communities, the observed difference in urban areas is 1.65%.
In addition to that the counts of female headed household in
The distribution of the household head’s education level is
urban communities has the highest observation at 1,566
presented in Figure 2 below. The results in Figure 2 show the
female headed households and 1,503 male headed households
percentage distribution of household head Education at all
in the urban, while in the rural 388 are female headed and
levels of education. It is noted that 55.59 % of household
351male headed.
heads are non-educated and 44.41% of household heads are
educated, computing 10.48% of household heads who have
complete secondary education, 13.58% who have incomplete
secondary education, and 9.85% who have incomplete
primary education, and 8.9% who have university education,
and 1.6% who have other education.

Source: Constructed by author using SHFS data

Fig 3: Percentage and counts of household heads by type of
Source: Constructed by author using SHFS data residential area and sex
Fig 2: Percentage distribution of household head’s education level
11.4 Distribution of household Main Source of income
For the purpose of this study only two categories of education The survey collected data on household of all income sources.
were adopted, educated and non-educated. As detailed above Household heads or representatives of household head were
the educated combines the rest of the other education levels in asked by the enumerators of Somali High Frequency survey
the above list, except no education which is labeled as non- (SHFS) the question; “What is the main source of income for
educated. the household?” Respondents were supposed to select one
among the listed responses:
11.2 Sex of the household head distribution i) Family assistance within country
The household heads by sex are presented in Table 2. The ii) NGO
results indicate that 51.3% of the households are female iii) None
headed while 48.7% are male headed. iv) Other small family business
v) Pension
Table 2: Household head’s Sex vi) Remittance from abroad
vii) Revenue from sales of asset
Valid Cumulative
Frequency Percent viii) Salaried labour source
Percent Percent
Female 1954 51.3 51.3 51.3 ix) Saving, interest or other investment
Valid Male 1854 48.7 48.7 100.0 x) Trade in domestic goods/products
Total 3808 100.0 100.0 xi) Trade in foreign goods/products (export or import)
Source: Calculated by author using SHFS data
The distribution of the household main source of income is
However, the pattern of household headship is inconsistent presented in Figure 4 below.
with other previous surveys. The Population Estimation
Survey in 2014 showed that the households headed by men
were 81.3%, and households headed by women were 18.7%.
Furthermore, the distribution of households by sex of head of

~27~
International Journal of Statistics and Applied Mathematics

are less than 18 years, although the constitution of the country

recognizes any person 15 years and above of age as an adult.
The youth headed household that is with heads aged between
(18-34) years represent 40.63% among observed household
heads. The middle age headed a household that is heads aged
between (35-64) years represent 53.62% of the households,
while 5.121% of the households have heads aged 65 years and
above.
It is noted that, the pattern of household headship age is
almost consistent with other previous surveys. The Population
Estimation Survey in 2014 showed that the households
headed by individuals of age less than 18 years were 0.2%,
and the youth headed households were 35.4%. Furthermore,
the middle aged (35-64) years headed households were
55.7%, while the retired household heads were 8.6%.

Source: Constructed by author using SHFS data

Fig 4: Distribution household Main Source of income

Results in Figure 4. indicate that households that receive

income from salaried labour constituted 35.87% of the total of
income sources and the rest combined source of income,
which is now labeled as “other sources” of income accounted
64.13%. This category is composed mainly of none out of the
mentioned sources of 19.93%, income from remittances
sources of 16.31%, income sources, income from domestic
business of 12.47%, income from family assistance within
country of 12%. Other sources of income in this category are
export, import, saving, revenue from sales of asset, pension
and aid which accounted for the remaining sources of income
3.42 % among household main sources of income.
As hinted earlier in section 3.3.2, for the purpose of this
study, only two categories of income sources were adopted,
the salaried labour source and the other sources. The other Source: Constructed by author using SHFS data
sources of income combined the rest of the other sources of Fig 5: Household heads by age
household income as detailed above.
11.6 Relationship between dependent variable and the
11.5 Distribution of household heads by age independent variables in the model
Figure 5 presents the age distribution of the household head’s 11.6.1 Relationship between household main source of
for the total population in the study. The Figure shows the income and household head’s education level
distribution various age brackets. The results show that out of
all households 0.63% is child-headed, that is, with heads who

Table 3: Chi-Square tests of Household Main source of income and household head’s education level
Value df Asymp. Sig. (2-sided) Exact Sig. (2-sided) Exact Sig. (1-sided)
Pearson Chi-Square 67.040a 1 .000
Continuity Correction 66.484 1 .000
Likelihood Ratio 66.918 1 .000
Fisher's Exact Test .000 .000
Linear-by-Linear Association 67.022 1 .000
N of Valid Cases 3808
Source: Calculated by author using SHFS data

The relationship between the household main source of 11.6.2 Relationship between household main source of
income and the household head’s education level is confirmed income and household head’s sex
by the Pearson chi-square test results presented in Table 3. Table 4 show relationship between household main source of
The Pearson Chi-Square test statistic is statistically significant income and head’s sex. The relationship is confirmed by the
at p<0.000. Thus, there is a relationship between the Pearson chi-square test results presented in Table 4. The large
household head’s education level and household main source computed value of the Pearson Chi-Square test statistic of
of income. 81.970 and its small significance level p<0.000 indicate that
there is a relationship between household head’s sex and
household main source of income.

~28~
International Journal of Statistics and Applied Mathematics
Table 4: Chi-Square tests of household main source of income by household head
Value df Asymp. Sig. (2-sided) Exact Sig. (2-sided) Exact Sig. (1-sided)
Pearson Chi-Square 81.970a 1 .000
Continuity Correction 81.359 1 .000
Likelihood Ratio 82.245 1 .000
Fisher's Exact Test .000 .000
N of Valid Cases 3808
Source: Calculated by author using SHFS data

116.3 Association between household Residential area and test statistic of 71.670 and its small significance level
household Main source of income p<0.000 indicate that there is a relationship between the
Table 5 show the relationship between household residence household’s residential area and the household main source of
area and household main source of income. The relationship income.
is confirmed by the Pearson chi square test results presented
in Table 4 below. The large computed Pearson Chi-Square

Table 5: Chi-Square tests of household main source of income by household residential area
Value df Asymp. Sig. (2-sided) Exact Sig. (2-sided) Exact Sig. (1-sided)
Pearson Chi-Square 71.670a 1 0.000
Continuity Correction 70.949 1 0.000
Likelihood Ratio 75.928 1 0.000
Fisher's Exact Test 0.000 0.000
Linear-by-Linear Association 71.651 1 0.000
N of Valid Cases 3808
Source: Calculated by author using SHFS data

11.7 Binary logistic Regression Analyses at p<0.000. The interpretation of this odds ratio is that the
11.7.1 Preliminary analyses of the model odds ratio of salaried labour source is (1/0.559) =1.789 less
Results presented in Table 6 of variables in the baseline likely that the other sources of income. These results show
equation show that the intercept only model only the predictions before the addition of the independent
has𝑙𝑛(𝑜𝑑𝑑𝑠 𝑟𝑎𝑡𝑖𝑜) = −0.581. Thus, the predicted odds ratio variables in the model.
of this expression is 𝐸𝑥𝑝(𝐵) = 0.559 which is significant

Table 6: Variables in the baseline equation

B S.E. Wald df Sig. Exp(B)
Step 0 Constant -.581 .034 295.630 1 .000 .559
Source: Calculated by author using SHFS data

Results presented in Table 7 of the independent variables not variables have statistically significant influences on the
yet in the model give an indication of whether or not each dependent variable of household main source of income.
additional independent variable in the model improves the
model performance. The all variables are significant Table 8: Omnibus Tests of the Model Coefficients
(p<0.000), and thus inclusion of these variables in the model Chi-square df Sig.
would add to the predictive power of the model.
Step 202.163 4 .000
Table 7: Inclusion of the variable in the model Step 1 Block 202.163 4 .000
Model 202.163 4 .000
Score df Sig.
Source: Calculated by author using SHFS data
HHH_Education (1) 67.040 1 .000
HHH_Sex (1) 81.970 1 .000
Variables 11.8 Statistical tests on individual parameters in the model
Step 0 HH_Residential Area (1) 71.670 1 .000
HHH_Age 26.037 1 .000 The model estimation results are presented in Table 9 below.
Overall Statistics 193.820 4 .000 In SPSS the estimates of the coefficients of the model are
Source: Calculated by author using SHFS data given in column ‘B’ in the ‘Variables in the Equation table.
The standard way of interpreting a ‘B’ in logistic regression is
11.7.2 Goodness-of-fit of the Model using the conversion of it to an odds ratio using the
11.7.3 Omnibus Tests corresponding exp(B) value. Wald statistic and associated
The Omnibus Tests of Model Coefficients in the Table 8 significance value provide an indication of the significance of
examine the hypothesis of whether or not there is a each estimated coefficient in the equation. The Wald statistic
statistically significant impact of the independent variables on has a chi-square distribution. Also the table gives a
the prediction of the dependent variable. Thus, in this case the confidence interval of the exp(B) column.
p-values are p<0.001, which means that independent

~29~
International Journal of Statistics and Applied Mathematics
Table 9: Variables in the Equation
95% C.I.for EXP(B)
B S.E. Wald df Sig. Exp(B)
Lower Upper
HHH_Education (1) 0.296 0.073 16.368 1 0.000 1.344 1.165 1.551
HHH_Sex (1) 0.585 0.072 65.590 1 0.000 1.794 1.558 2.067
Step 1 HH_Residenatial area (1) 0.717 0.098 53.885 1 0.000 2.049 1.692 2.482
HHH_Age -0.013 0.003 18.496 1 0.000 0.987 0.982 0.993
Constant -1.128 0.151 55.770 1 0.000 0.324
Source: Calculated by author using SHFS data

The application of the Wald test is to take the significance 𝑒 -1.128 +0.585(1)
values less than 0.05, to reject the null hypothesis that 𝑝(𝑥2 = 1) = = 0.37
1 + 𝑒 −1.632+0.585(1)
independent variable doesn’t have a significant contribution
to the relationship and instead accept the alternative The Wald statistic of household residence area variable is
hypothesis that the independent variable has a significant significant (p<0.000) in Table 9, and the odds ratio of an
contribution to the relationship. In this case, all independent urban household getting a salaried labour source of income
variables contribute significantly to the relationship because than other sources of income is 2.049. Therefore, an urban
p<0.000 for all independent variables. household is 2.049 times more likely to have a chance of
The Exp(B) column in Table 9 presents the extent to which getting a salaried labour source of income, than a rural
changing to the corresponding measure influences the odds household. The probability that an urban household will have
ratio. Exp(𝐵) is interpreted in terms of the change in odds. If a chance of getting a salaried labour source of income is 0.40,
the value in this column exceeds 1, it means the odds of an which is calculated using equation (3.5), That is,
outcome occurring increase; if the figure is less than 1, any
change in the predictor variable leads to a drop in the odds of 𝑒 -1.128 + 0.717 (1)
the outcome occurring. 𝑝(𝑥3 = 1) = = 0.40
1 + 𝑒 -1.128 + 0.717 (1)
The Exp(B) value associated with household head’s education
level is 1.334 with 95% confidence interval [1.165, 1.551]. The Wald statistic of household age variable is significant at
Interpretation of the coefficient associated with household (p<0.000) in table 9. The coefficient of household age is (-
head education level is that, if level of education of the 0.013), which means that the odds of a household’s main
household head change from non-educated to educated, the source of income being a salaried labour source decreases by
opportunity that household head will get a labour source of 0.013 for each year age increase of the household head
income as main source of income will increase by 0.296. increase, holding other variables constant. A better way of
However, when the variables are categorical the most useful interpreting this is by using the odds ratio which is included in
interpretation is given in terms of the odds ratio which shows the Exp(B) column in Table 4. The odds ratio is 0.987 with
point of comparison between leading and the latent variable. 95% confidence interval [0.982, 0.993]. The interpretation of
The odds ratio of an educated household head getting a this is that the household head is (1/0.987) =1.013 less likely
salaried source of income rather than any other source of to get a salaried labour source of income than other sources of
income is 1.334. Therefore, the educated household head is income for each year age increase. The probability that the
1.334 times more likely to have an opportunity of getting a main source of income of the household head of a specified
salaried labour source of income than the non-educated age will be a salaried labour source of income is 0.24, which
household head. The probability that the household headed by is calculated using equation (3.6), that is,
the educated household head will get a salaried labour source
of income is 0.30, which is calculated using the equation (3.3) 𝑒 -1.128 − 0.013
of section 3.3.3.1, that is, 𝑝(𝑥4 ) = = 0.24
1 + 𝑒 -1.128 − 0.013
𝑒 -1.128 +0.296(1) 12. References
𝑝(𝑥1 = 1) = = 0.30
1 + 𝑒 -1.128 +0.296(1) 1. Agresti A. An Introduction to Categorical Data Analysis,
Pharmaceutical Statistics, 1996. doi: 10.1002/pst.339.
The Wald statistic of household head’s sex variable is 2. Ali L, Ismaeel Ramay D. Analysis of the Determinants of
significant (p<0.000) in Table 9, and the Exp(B) value Income and Income Gap between Urban and Rural
associated with household sex is 1.794 with 95% confidence Pakistan, 2013; 5(1). Available at: http://journal-
interval [1.558, 2.067]. Interpretation of the coefficient archieves32.webs.com/858-885.pdf (Accessed: 26 March
associated with household head sex is that the opportunity 2017).
that a male household head will get a labour source of income 3. Birthal PS et al. Income Sources of Farm Households in
as the main source of income will increase by 0.585. The odds India: Determinants, Distributional Consequences and
ratio of a male headed household getting a salaried labour Policy Implications, Agricultural Economics Research
source of income than other sources of income is 1.794. Review. 2014; 27(1):37-48. doi: 10.5958/j.0974-
Therefore, a male household head is 1.794 times more likely 0279.27.1.003.
to get an opportunity a salaried labour source of income than 4. Chamicha S. Nonfarm Activities and Rural Livelihood in
a female household head. The probability that the household Tanzania The case of Njombe District, 2015. Available
headed by a male will get the opportunity of a salaried labour at:https://www.google.com/#q=Nonfarm+Activities+and
source of income is 0.37, which is calculated using equation +Rural+Livelihood+in+Tanzania+The+case+of+Njombe
(3.4), that is, +District&*.
5. Fadipe AEA, Adenuga AH, Lawal A. Analysis of Income
Determinants among Rural Households in kwarastate,
~30~
International Journal of Statistics and Applied Mathematics

Nigeria. Trakia Journal of Sciences. 2014; 12(4):400-404.

doi: 10.15547/tjs.2014.04.010.
6. Janet Currie, RC. Male Jobs, Female Jobs, and Gender
Gaps in Benefigs coverge, 1992; 4106. Available at:
https://www.princeton.edu/~jcurrie/publications/Male_jo
bs_female_jobs.pdf (Accessed: 28 January 2017).
7. Mathebula J, et al. Estimation of household income
diversification in South Africa: A case study of three
provinces. South African Journal of Science S Afr J Sci.
2017; 113113(1212):2016-73. Available at:
http://www.sajs.co.za (Accessed: 2 July 2017).
8. Nations U. (no date) Principles Recommendations
Population Housing Censuses.
9. Sami Gungor M, et al. The Impact of Household Heads’
Education Levels on the Poverty Risk: The Evidence
from Turkey, 2015. doi: 10.12738/estp.2015.2.2354.
10. SNA. SNA, 2008. Available at: https://unstats.un.org/
unsd/nationalaccount/docs/sna2008.pdf.
11. Talukder D. Assessing Determinants of Income of Rural
Households in Bangladesh: A Regression Analysis.
Journal of Applied Economics and Business Research
JAEBR. 2014; 4(2):80-106.
12. Uganda Bureau of Statistics. Uganda National Household
Survey 2012/2013, Uganda National Household Survey,
2014. doi: UGA-UBOS-UBOS-2012-v01.
13. Yadollahi M, Haj Paim L, Taboli H. African Journal of
Business Management Head of household characteristics
and family economic status (FES) among Kerman
households, Iran, 2013; 7(19):1903-1913. doi:
10.5897/AJBM10.1234.

~31~