International Journal of Statistics and Applied Mathematics 2019; 4(3): 01-10

ISSN: 2456-1452
Maths 2019; 4(3): 01-10
© 2019 Stats & Maths
Statistical analysis of socio-economic determinants of
www.mathsjournal.com internal migration in Somalia using logistic regression
Received: 01-03-2019
Accepted: 04-04-2019 approach
KK Saxena
University of Dodoma, Dodoma,
Tanzania
KK Saxena and Mohamed Abdinur Mohamed
Mohamed Abdinur Mohamed Abstract
Statistician, Directorate of Internal migration refers to the situation in which people prefer to live in another city of the country
National Statistics, Somalia, because of the social, cultural and economic problems in the city they live. Internal migration leads to
Mogadishu imbalances in population distribution, irregular urbanization, unemployment in urban area, inadequate
housing, low paid jobs, and limited access to social services such as health and education facilities.
Owing to this situation, this article analyzed the socio-economic factors affecting internal migration in
Somalia using logistic regression.
This study used secondary data from a cross-sectional population estimation survey conducted in 2014 on
internal migrant households who migrated within Somalia in the last ten years at the time of the survey.
The study found that rural resident households were more likely to migrate than urban resident
households. Household heads, who have no education background, were significantly more likely to
migrate than educated headed households, and unemployed headed households were more likely to
migrate than employed. Sex and land ownership were found not significantly associated with internal
migration. Therefore, developed based interventions are important to improve the socioeconomic status
of the household and are needed to reduce the incidence or prevalence of internal migration.

Keywords: Internal migration, Chi square test, logistic regression

1. Introduction
Somali Arid and Semi-Arid Lands make up more than 80 percent of the country’s land mass
and are prone to extreme weather conditions, including periods of extended drought, highly
erratic rainfall and strong winds. In terms of livestock, low rainfall makes much of the country
suitable only for nomadic herding. During the seasons of drought, pastoralist families move to
the less-affected areas in search of greener pastures for their livestock. Conversely, heavy
rains, mainly in the Eastern Ethiopian highlands, raise the water levels of major rivers, often
resulting in flooding that leads to deaths, displacement, and destruction of livelihoods for the
Somali communities affected. Floods are mainly experienced in the riverine and agro-pastoral
zones along the Shabelle and Jubba rivers (OCHA, 2013) [16].
Internal displacement was estimated around 2 million during the 1990s, with the effects of a
severe drought during the 1992-93 emergency and others fleeing for safer areas in urban
centers in the country (Hammond, 2014) [7]. Furthermore, the famine has resulted in a large
number of rural population migrated to urban cities, particularly Mogadishu, to access
humanitarian assistance (Achour & Lacan, 2012) [1].
Somali community being traditionally pastoralists and agro-pastoralists, the livelihood is
derived from rearing livestock and farming. The Somalia famine of 2011 caused high levels of
forced migration, with around a quarter of the population displaced within the country
(Lindley, 2014) [12], resulting in large-scale mortality, morbidity, and population displacement
and decimated over 70% of the pastoral livestock particularly in the southern regions of
Somalia in 2011-2012 (Ali, 2018) [2]. Internal migration has been increasing in Somalia over
some years due to natural disasters such as frequent droughts or floods. According to World
Bank data for World Development Indicators (WDI) in 2016, the number of internal migrants
Correspondence
KK Saxena
that associated with disasters in Somalia was two times higher than Kenya and Ethiopia
University of Dodoma, Dodoma, combined Figure 1 presents the trends in total internally displaced persons associated with
Tanzania disasters for Somalia, Kenya, and Ethiopia from 2008 to 2016.
~1~
International Journal of Statistics and Applied Mathematics

Author's computation by World Bank data

Fig 1: Internal migrants by selected countries

The migration studies have always been considered important are found generally in more developed countries, while rural
topics over many years. However, the empirical work on to urban flows are much more typical in developing countries
internal migration is limited as far as Somalia is concerned. (Todaro, 1980) [18]. Various empirical studies on internal
Previous studies on this topic in the context of Somalia have migration of population were conducted in developing
largely ignored the role of socio-economic factors affecting countries.
internal migration, and hence their results are subject to (Melesse & Nachimuthu, 2017) [14] identified that less
misspecification bias. According to the Somalia National productivity of agriculture, environmental degradation, poor
Development Plan (NDP2017/19) mission statement, the social services, and insufficient of land in rural areas as the
government planned to regularize migration and make a safe major push factors of migration. The authors found some
and successful enterprise for migrants that may benefit effect of pull factors that drew rural societies to urban settings
themselves and the country. The internally displaced people such as education, health services, job opportunities, and
bring more pressure to the urban cities and will limit access to other urban services. (Sharma, Kone, Liu, & Mattoo, 2016)
[17]
basic services such as clean water, education or health, as well , conducted research on internal borders and migration and
as inadequate access to food. found that part of the state borders were critical impediments
Therefore, there is a need to investigate socio-economic to internal mobility. They found that the impact of state
factors affecting internal migration in Somalia in order to borders differs by education, age and reason for migration, it
reduce internal pressure for people to migrate to urban cities. is constantly large and significant. They have stressed the role
The current study addresses this shortcoming by extending the of the state borders as significant impediments to internal
analysis to clearly address internal migration by using logistic mobility. But the word mobility is deferring from internal
regression analysis in order to execute an in-depth migration because mobility can be a permanent or temporary
investigation of the research topic. Furthermore, since limited change of place of residence, while internal migration is a
empirical studies are available for Somalia on internal permanent change of place residence. Therefore, the current
migration at household level, this study provides in-depth research was concentrated on internal migration as per
statistical analyses of data on internal migrants in Somalia in described by demographers and population scholars.
respect to socioeconomic and demographic perspective. (Jorgji, 2015) [10], conducted an empirical examination of
determinants of "why do women migrate internally" and
2. Review of literature discovered the traditional significant variables influencing
There is no single theory widely accepted by social scientists migration include mobility by age, level of education, and
when it comes to internal migration phenomenon as research work status. (Kanwal, Naveed, & Khan, 2015) [11] conducted a
on migration is primarily interdisciplinary. However, study about socio-economic determinants of rural-urban
researchers have revealed that the tendency of young persons migration by using logistic regression to estimate the impact
to migrate could be consistently higher than other age groups of socio-economic determinants on rural to urban migration.
when the area of origin is rural (Greenwood & Hunt, They found that there is a positive and significant relationship
2003). This type of migration from rural settlements to urban between sex, age, and employment, whereas marital status
areas is almost always permanent. Urban to rural migration was negatively related to rural-urban migration.
also exist; this kind of migration known as back to the land (Tsogtsaikhan, 2014) [19] studied the behavior of internal
movement, where urban residents decide to leave their migration and estimated the influencing factors of the
congested places to reside in rural areas where they can have migrant’s decision to return and estimated the duration of the
a better quality life (Halfacree, 2006) [6]. Such migration flows internal migration. (Awuse & Tandoh-offin, 2014), conducted
~2~
International Journal of Statistics and Applied Mathematics

an investigation of the influence on internal migration on model and found that there was a higher probability of
development and found that there is a positive relationship migration for the households that have better-educated
between internal migration and financial advancement in light members. It also revealed that the main factors for migration
of the fact that they make accessible and open doors for was poverty. (Haider, 2010) [5] examined rural-urban
access to prompt and adaptable work constraints. (Msigwa, migration and found that the migration process has been
2013) [15] and (Islam, Rahman, & Hossain, 2013) [9] studied driven by economic and social.
that sex, age, head of household, household income,
household family size, marital status, level of education, and 2.1 Conceptual framework
skill level were significant and considered to be the The decision of household to migrate from one region to
determinants of migration in Tanzania. (Ishtiaque & Ullah, another within the country is normally influenced by many
2013) [8] studied on the status of rural-urban migration and variables which needed to analyze in order to minimize the
found that people migrate to cities mostly for economic negative impacts of internal migration. In Figure-2 a
reasons. (Amare, Hohfeld, Jitsuchon, & Waibel, 2012) [3], conceptual frame work has been presented, which was
investigated the effects of rural to urban migration on operationalized in this study in order to achieve the objectives
economic development by using probit logistic regression of the study.

Fig 2: Conceptual Framework

3. Methodology for both economic and noneconomic reasons, irrespective of

The Federal Republic of Somalia is located in the Horn of their previous district of residence. This study only looks at
Africa and its boundaries are the Gulf of Aden to the north, household heads who migrated within the last ten years at the
Djibouti to the north-west, Ethiopia to the west, Kenya to the time of PESS survey 2014, since it was difficult to trace the
south-west, and the Indian Ocean to the east. Somali has the motivations for those who migrated over last ten years ago as
longest coastline on Africa’s mainland and the Middle East, significant changes in their individual characteristics as well
its coastline is more than 3,333 kilometers in length. as the characteristics of their destination places.
According to the population estimation survey 2014, Somali The data were obtained from the population estimation survey
population estimated around 12.3 million inhabitants. The while its sampling frame were comprised of defined clusters
total fertility rate of Somalia is 6.08 children born per woman of enumeration areas for the urban areas, settlements for rural
(2014 estimates) and the annual growth rate of 1.75% per areas, camps for IDPs and water points for nomadic areas.
annum with a median age of 17.7 years. These were the area primary sampling units;
The sample employed in the study has been limited to the
households, who have migrated with the regions of Somalia
~3~
International Journal of Statistics and Applied Mathematics
Table 1: Sample selection

No. Frame PSU (primary sampling units) Sample selected

1 Urban 6750 868
2 Rural 6519 1104
3 IDP 107 28
4 water point 5332 735
Source: Author’s own construction by PESS, 2014 data

The overall sample size selected was 2,735 PSU and the migration, urban to rural migration, rural to rural migration,
selection within the strata was based on probability and urban to urban migration as a dependent variable. If a
proportional to size (PPS) taking into account the measures of household migrated from rural to urban, the dependent
size (PESS, 2014). The analysis of this study included 14212 variable was considered as “1”. On the other hand, dependent
households who internally migrated from or to urban, rural, or variable took the value of “0” when the household migrated
IDP in the last ten years prior to the survey. from urban to rural, rural to rural, or urban to urban.
Additionally, study used seven different explanatory variables
3.1 Variables in this study so as to determine the factors that affect internal
Since the aim of the study was to analyze the socio-economic migration. The details of these variables have been given in
factors affecting internal migration, this study used the Table-2.
number of household internal migration i.e. rural to urban

Table 2: Measurement of the variables

Dependent variable Codes Description
The possible outcome of internal migration is either: 1 = Rural to urban migration and 0 =
Internal Migration Internal_migration
Otherwise
Independent Variables Codes Description
Sex sex Whether the household is headed by a male or female: 1 = Male and 2 = Female
Age Hhage Age of the household head (years)
Household size Hhsize Number of family members in the household
Type of the previous Household place of residence. 1= Rural(include nomadic) and 2 =
Residence resid
Urban(Include IDP settlement)
Educational background of the household heads categorized into;
Education edu
1 = None (Have no educational background) and 2 = Educated(at least primary education)
Household head employment status.
Employment status empl
1 = Employed and 2 = Unemployed
Land ownership of the household
Land ownership land
1= Owns land and 2= Does not own land
Source: Author’s construction

To examine the association between the dependent variable households by grouping into two categories according to rural
(internal migration) and independent categorical variables; to urban migrants and other migrants (rural to rural, urban to
sex, place of residence, education, employment status, and rural, and urban to urban). The response variable y is internal
land ownership, the chi-square test of independence was migration. If household migrated from rural to urban then (y =
conducted before the logistic regression analysis to check the 1) and if the household migrated but not rural to urban then (y
association between these independent categorical variables to = 0). Before using logistic regression, Chi-square test was
the dependent variable. If the probability of the presence of conducted to check the association between the independent
the characteristic of interest is π , the probability of absence of categorical variables and the dependent variable.
the characteristic of interest is 1-π . The effect of independent 3.2 Handling missing data
variables is usually explained in terms of odds since logistic For missing data, the imputation technique was applied using
regression calculates the probability of an event occurring R package “Amelia1”. It is a visual way to check for missing
over the probability of an event not occurring. Since the data, which enables to plot all the variables in the study by
outcome of logistic regression is binary, the dependent needs grouping into missing and observed with their respective
to be transformed by using logit transformation which gives percentage in the data. Figure-3. demonstrates that there were
the following: 1% of missing values and 99% of non-missing observations
for all variables. It also illustrates that only HHSIZE_NEW
  
logit  y   log  odds   log    0  1X1  2 X 2  .......  k X k   has missing values. After replacing the missing values, each
 1  
variable in the dataset has 14212 observations with no
missing values. This cleaned data then subjected for the
where π is the probability of an event occurring .The binary descriptive and inferential analysis.
logistic model was used to analyze internally migrated

~4~
International Journal of Statistics and Applied Mathematics

Source: Author’s computation

Fig 3: Plotting missing values

4. Results and discussion region was Benadir (43.4%), followed by Wagooyi Galbeed,
4.1 Distribution of internal migrants by region Lower Shabelle, Mudug, and Lower Juba, (8.2%), (7.4%),
The distribution of household internal migrants by region are (6.9%), and (6.6%) respectively.
presented in Fig-4 the most frequently observed category

Source: Author's computation

Fig 4: Internal migrant households by region

~5~
International Journal of Statistics and Applied Mathematics

While, Hiraan, Bakool, Sool, and Middle Jubba have 4.2 Distribution of internal migration according to
observed the lowest migration (0.9%, 0.5%, 0.4%, and 0.3%) demographic variables
respectively. Seventy percent of the households belonged to Table 3 present the distribution of household head age and
other migration, while thirty percent of the households were household size. The average household age was found to be
migrated from rural to urban. 39 years with standard deviation of 13 years and the average
household size was found to be 7 members with standard
deviation of 2.694.
.
Table 3: Summary statistics for age and household size
Variable N Min Max Mean SD SEM Skewness Kurtosis
hhage 14212 15 95 39.48 12.849 0.108 0.897 0.862
hhsize 14212 1 23 6.55 2.694 0.023 0.790 2.114
Source: Author’s calculation

Table 4 presents the distribution of internal migrants by the while rural households were (20.2%). 60.8% household heads
sex of the household head, household head previous had no education background compared to those who have an
residence, household head education, household head education background or at least attended primary school
employment status and household head land ownership. 39.2%. For other migration, non-educated household heads
It is clear that migration of male-headed households were were 54.9% while educated households were 45.1%. For rural
more (75.9%) than female-headed households (24.1%). For to urban migration, non-educated household heads were more
other migration, male-headed households were more (76.2%) 75.0% compared to educated households 25.0%. 65.1% of the
compared to female-headed households (23.8%). For rural to internal migrant household heads were employed and 34.9%
urban migration, male-headed households were (75.2%) while of the household heads were unemployed. For other
female-headed households were (24.8%). 85.9% migrant migration, the employed headed households were 66.6%
households had urban place of residence while 14.1% while 33.4% were unemployed headed households. For rural
belonged to rural. For other migration, urban households were to urban migration, 61.6% headed households were employed
more (88.5%) compared to rural households (11.5%). For compared to 38.4% of the unemployed headed households.
rural to urban migration, urban households were (79.8%)

Table 4: Distribution of internal migration according to demographic variables

Internal Migration
Variable Other migration Rural to urban Total
Male 76.2% 75.2% 75.9%
sex of the household heads
Female 23.8% 24.8% 24.1%
Household’s previous residence Urban 88.5% 79.8% 85.9%
Rural 11.5% 20.2% 14.1%
Educated 45.1 25.0 39.2
household head education
Non educated 54.9 75.0 60.8
Employed 66.6 61.6 65.1
household head employment status
Unemployed 33.4 38.4 34.9
Own Land 48.7 48.5 48.6
household land ownership
No land 51.3 51.5 51.4

The results show that there were 51.4% of the internal migration) and independent variables- sex of household head,
migrant household heads, who did not own land as compared household place of residence, household head education,
to 48.6% who owned land. For other migration, the household head employment status, and land ownership of the
percentage of the household heads, who did not own land was household. The results are presented in table- 5 which show
51.3% compared to 48.7% who owned land. For rural to that there is no association between sex and internal migration
(
urban migration, the percentage of the household heads, who 2
= 1.710, df =1 with p = 0.191). Regarding the
did not own land were 51.5% while those who owned land
were 48.5%. association between the household place of residence and
internal migration, it is concluded that there is an association
4.3 Test of the Associations between migration and between internal migration and household place of residence
(
2
various Categorical Variables = 181.131, df =1 with p < .001).
Chi-square test of independence was used to examine the
association between the dependent variable (internal

~6~
International Journal of Statistics and Applied Mathematics
Table 5: Test of the associations for the categorical variables
Migration
Variable Chi square d.f. p-value
Male
Sex 1.710 1 0.191
Female
Urban
Place of residence 181.1311 1 0.00
Rural
Educational status Educated 500.828 1 0.00
Uneducated
Employed
Employment status 32.772 1 0.00
Unemployed
Land ownership Owns land 0.57 1 0.811
Does not own land
Source: Author’s computation

Regarding the association between household head education there is no dependency between land ownership and
internal migration (  = 0.057, df =1, and p = .811).
and internal migration, it is concluded that there is an 2
association between internal migration and household
(  = 500.828, df =1 with p < .001).
2
education 4.4 Model Diagnostics
Regarding the association between household head 4.4.1 Absence of multicollinearity
employment status and internal migration, it was found Table- 6 presents the results of the variance inflation factor
that there is a dependency between employment and for the variables. VIF (Variance Inflation Factor) were
internal migration (  = 32.772, df =1 with p < .001).
2 calculated to examine the presence of multicollinearity among
predictors. If the VIF value is of 10 or higher, there is an
Regarding the association between household land evidence of multicollinearity and further investigations may
ownership and internal migration, it was found that be done on the data. The VIF values for all the variable are
less than 10 implying there is no evidence of
multicollinearity.

Table 6: Variance inflation factors for the predictors

Variables hhage hhsize resi empl edu
VIF 1.13 1.10 1.01 1.08 1.07
Source: Author’s calculation

4.4.2 Hosmer-Lemeshow test (HL test) The HL statistic assumes sampling adequacy, with a rule of
The Hosmer Lemeshow test evaluates whether observed event thumb being enough cases so that 95% of cells (typically, 10
rates match expected event rates in subgroups of the model decile groups times 2 outcome categories =20 cells) have an
population. The hypotheses were framed as follows expected frequency >5. HL statistic has a significance of
Ho : There is no significant difference between the observed 0.558 which means that it is not statistically significant and
therefore our model is quite a good fit. Table-7 presents the
and model predicted values.
results of the Hosmer-Lemeshow test.
H1 : There is significant difference between the observed and
model predicted values.

Table 7: Hosmer-Lemeshow Test

Chi square value d.f. Significance
6.801 8 0.558
Source: Author’s computation

4.4.3 Receiver operating characteristic (ROC) curve under the curve (AUC) for the diagnostic ability of binary
ROC curve is used to see how any predictive model can classifiers. AUC value should be between 0.5 and 1.0. It
distinguish between the True positive rate and false positive measures the predictive accuracy of a model. An AUC equal
rate. R software has been used by employing Random Over- to 0.5 indicates a random classification model and therefore
Sampling Examples (ROSE) R package. The package should be greater than 0.5 for a model to be acceptable. The
provides functions to deal with binary classification problems Area under the Curve for this study was 0.641. Therefore, the
in the presence of imbalanced classes to compute the area model was considered reasonably well.

~7~
International Journal of Statistics and Applied Mathematics

Source: Author's computation

Fig 5: Receiver operating characteristic curve

4.4.4 Results for the Binary Logistic Regression Model was calculated to examine the model fit, where values greater
Table 8 presents the summary of the results of the logistic than 0.2 are indicative of models with excellent fit (Louviere,
regression. The results reveals that age of the household head, Hensher, & Swait, 2000) [13]. The McFadden pseudo-R-
household place of residence, household head education, and squared value calculated for this model was found to be
household head employment had a significant effect on the 0.169, hence the model considered well fitted.
odds of the internal migration. McFadden's pseudo-R-squared

Table 8: Binary logistics regression output

Variable Odds Ratio 95% confidence Interval Pr(>|z|)
household head’s age Age 1.01 1.00 1.01 <0.0001
Household size Household size 0.99 0.97 1.00 0.1126
Place of residence Urban (reference) 1.00
Rural 1.95 1.76 2.16 <0.0001
Educated (Reference) 1.00
Education status
Uneducated 2.22 2.03 2.42 <0.0001
Employed (Reference) 1.00
Employment status
Unemployed 1.17 1.08 1.28 0.0002
Source: Authors computation

The result from Table 8 indicates that the age of the place of residence, were 1.95 times more likely to have a
household head was significant. This indicated that the chance of being internally migrated than the urban household
increase in age of the household head, the odds of migration residence. The probability, that the internal migration of the
from rural to urban migration increases by approximately 1% household in a rural area will be migrating to urban is 0.63 (if
and hence it can be concluded that the age of the household we hold age of the household head, household size,
has an effect on internal migration. educational status, and employment status as constant) is
The probability that the internal migration of the household given as
head of a specified age will be migrating from rural to urban,
e 1.827  0.667(1)
holding household size, place of residence, educational status,
and employment status as constant is given by (X3  1)   0.63
1  e 1.827  0.667(1) .
e 1.827  0.007(1)
(X1  1)   0.14 The results for household head education reveals that the non-
1  e 1.827  0.007(1) . educated household heads had odds 122% than educated
household heads on internal migration. The null hypothesis
The result reveals that household size was not significant at 5 has been rejected since there is not enough statistical evidence
percent level of significance. This may have an effect on to accept the null hypothesis. Therefore, households who were
internal migration and less likely to influence internal having no educational background had 2.22 times higher the
migration. chance of internal migration than having at least primary
The results for place of residence indicates that the rural education.
household residents had more than urban households on The probability that the internal migration of the non-
internal migration. The null hypothesis was rejected and educated headed household will be migrated to urban holding
hence it can be concluded that households at their previous
~8~
International Journal of Statistics and Applied Mathematics

age of the household head, household size, place of residence, e 1.827  0.160(1)
and employment status as constant is (X5  1)   0.38
1  e 1.827  0.160(1) .
e 1.827  0.795(1)
(X 4  1)   0.71 Logistic regression output is presented below:
1  e 1.827  0.795(1) > Intmigration<-glm (Internal_migration ~ hhage + hhsize +
resid + edu+ empl, data = imigration, family =
The results indicates that the unemployed headed households "binomial"("logit"))
had 17% more likely to migrate to urban compared to > summary (Intmigration)
employed headed households. The null hypothesis was Call:
rejected since the statistical evidence wasn’t enough to keep glm (formula = Internal_migration ~ hhage + hhsize + resid +
the null hypothesis. The probability that the internal migration edu+ empl, family = binomial ("logit"), data = imigration)
of the unemployed headed household will be migrated to Binary Logistic Regression Results:
urban holding age of the household head, household size, Model fitted by Standard Maximum Likelihood
place of residence, and household head education as constant
is;

Deviance Residuals
Min 1Q Median 3Q Max
-1.3526 -0.8671 -0.6191 1.2386 1.9728
Coefficients Estimate Std. Error z value Pr(>|z|)
(Intercept) -1.826958 0.075824 -24.095 < 2e-16 ***
hhage 0.007281 0.001602 4.544 5.53e-06 ***
hhsize -0.012029 0.007581 -1.587 0.112588
resif(rural) 0.667209 0.051695 12.907 < 2e-16 ***
edu(noneducated) 0.795383 0.045002 17.674 < 2e-16 ***
empl(unemployed) 0.159763 0.042613 3.749 0.000177 ***
Significance codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘’ 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 17251 on 14211 degrees of freedom
Residual deviance: 16570 on 14206 degrees of freedom
AIC: 16582
Number of Fisher Scoring iterations: 4
p < .001, df = 5, McFadden R2 = 0.1688684
95% Confidence Interval
> exp(confint(Intmigration))
Waiting for profiling to be done...

2.5% 97.5%
(Intercept) 0.1386136 0.1865945
Age_years 1.0041461 1.0104745
HHSIZE_NEW 0.9734371 1.0028034
Place_residence 1.7607520 2.1563193
Education 2.0287977 2.4202328
Employment 1.0791535 1.2753551

5. Discussion of the Findings who found that marital status, sex, and family size were
In the following, we discuss the findings from the specific significant in determining internal migration.
objectives of this study and compared them with previous It has been found that place of residence has a significant
studies on internal migration. effect on internal migration. More specifically, households
The findings from this research were consistent with the living in rural areas are more likely to migrate than urban
current literature on internal migration. Among demographic households. The findings from this research were consistent
variables, age was significant in predicting internal migration. with the current literature on internal migration
These findings were consistent with the results of (Sharma et (Tsogtsaikhan, 2014) [19], There were several reasons that
al., 2016) [17], (Kanwal et al., 2015) [11], (Tsogtsaikhan, 2014) motivated rural households to migrate their place of origin to
[19]
, and (Msigwa, 2013) [15]. Among the other studies that cities permanently; due to population growth with shortage of
explored the demographic variables of internal migration, agricultural land resources in rural areas, frequent droughts,
(Msigwa, 2013) [15] found additional demographic variables loss of livelihood in both livestock and farmers in rural areas
such as gender, marital status, and household size to be in Somalia for the last decades, and improving stability of
significant in predicting the chance of internal migration, also destination cities.
(Kanwal et al., 2015) [11] as well. The effect of household head’s education on internal
Household size was not found significant at 5% level but it migration was found to be significant in predicting internal
has a negative effect on internal migration. Sex of the migration. This study found that there is a positive
household head and land ownership were found not to be relationship between internal migrations with households
affecting internal migration. This outcome is different from having no educational background. In other words, non-
the results of (Tsogtsaikhan, 2014) [19] and (Msigwa, 2013) [15] educated headed households living in rural areas without
employment are more likely to migrate to urban places than
~9~
International Journal of Statistics and Applied Mathematics

educated household heads. The findings from this research from Thailand. ADB Economics Working Paper Series,
were consistent with the current literature on internal 2012, 309).
migration (Kanwal et al., 2015) [11], (Ishtiaque & Ullah, 2013) 4. Greenwood MJ, Hunt GL. The Early History of
[8]
, (Msigwa, 2013) [15], and (Tsogtsaikhan, 2014) [19]. Migration Research Revisited, 2003.
For the effect of the household head employment status on 5. Haider SKU. Factors of Migration on Urban Bangladesh :
internal migration, it was found that unemployed headed An Empirical Study of Poor Migrants in Rajshahi City,
households, who live in rural areas with no educational 2010; 30(2):307-323.
background, are more likely to be migrated to the urban cities 6. Halfacree K. From dropping out to leading on? British
than employed headed households. The findings were counter-cultural back-to-the-land in a changing rurality.
consistent with the current literature on internal migration. Progress in Human Geography, 2006; 30(3):309-336.
Among the economic variables, employment was included in https://doi.org/10.1191/0309132506ph609oa
this analysis and was significant in determining internal 7. Hammond L. Somali refugee displacements in the near
migration. These findings were consistent with the results of region: Analysis and Recommendations Paper for the
(Ishtiaque & Ullah, 2013) [8], (Kanwal et al., 2015) [11], and UNHCR Global Initiative on Somali Refugees, 2014, 1-
(Tsogtsaikhan, 2014) [19]. 22.
The major push factor that encourages people to leave their 8. Ishtiaque A, Ullah MS. The influence of factors of
homes in the rural areas is the recurrence of droughts in migration on the migration status of rural-urban migrants
Somalia, which demoralize agricultural activities and cause in Dhaka, Bangladesh. Human Geographies, 2013.
food insecurity. In the case of urban desirability, there is a https://doi.org/10.5719/hgeo.2013.72.45
common unanimity that the most leading factor is better 9. Islam R, Rahman O, Hossain S. Factors Affecting on
employment opportunity in the urban areas. Migration to Chapai-Nawabganj City in Bangladesh :
Multivariate Approach, 2013; 3(6):273-277.
6. Conclusion https://doi.org/10.5923/j.economics.20130306.05
When an individual or household is not satisfied with the 10. Jorgji E. Why do women migrate internally ? Empirical
situation in which he is, then it may fall under the categories analysis of, 2015, (2).
of either forced or voluntary migration. The decision of 11. Kanwal H, Naveed TA, Khan MA. Socio-Economic
migration depends upon variables like socioeconomic Determinants of Rural-Urban Migration in Pakistan,
differences between the rural settings and the urban areas as 2015; 4(3):72-85.
well as the availability of the information about employment 12. Lindley A. Questioning “drought displacement”:
in the destinations, but there are some challenges that these environment, politics and migration in Somalia. Forced
migrants face, when they arrive in the cities, like Migration Review, 2014; (45):39-43. Retrieved from
unemployment or low paid jobs which sometimes forced https://search.proquest.com/docview/1503672920?accou
some of the migrants to return to their place of origin. ntid=9883%5Cnhttp://whel-rimo.
The current study revealed that the decision of a rural hosted.exlibrisgroup.com/openurl/44WHELF_CAR/44W
household in Somalia to move for work or education to the HELF_CAR_services_page?genre=article&atitle=Questi
urban area is strongly related to household characteristics. oning+%27drought+displacement%27%3A+environment
Generally, it is rural households with lower socioeconomic %2C+politics+and+migration+
status, who are most likely to migrate to the urban settings 13. Louviere JJ, Hensher DA, Swait JD. Stated choice
and there seem to be strong push factors of migration such as methods: analysis and applications. Cambridge
poor access to social services, less productivity of agriculture, University Press, 2000, 1-418.
environmental degradation and insufficient of land in rural https://doi.org/10.1002/jae.701
areas at a district or regional levels. There are some pull 14. Melesse B, Nachimuthu K. A review on Causes and
factors that attract people to migrate to the major cities like Consequences of Rural- Urban Migration in Ethiopia.
Mogadishu, the capital city of Somalia which has been found 2017; 7(4):37-42.
to be a number one destination for internal migrants, this is 15. Msigwa RE. Determinants of internal migration in
due to the availability of essential amenities in the city. Kazakhstan, 2013, 4.
The study revealed that household heads had migrated to the 16. OCHA. Key facts and figures for the Somalia Common
city with a view to giving their children a better education Humanitarian Fund in 2013. response? Conflict and
since there is a good quality of education and a wide range of Health. 2013; 7(1):1-5. https://doi.org/10.1186/1752-
facilities are available in the urban area which encouraged 1505-7-22
them to migrate. Study results also showed that migrants were 17. Sharma S, Kone ZL, Liu MY, Mattoo A. Internal Borders
mainly economic migrants, who moved to seek employment and Migration in India, 2016.
opportunities. 18. Todaro M. Internal Migration in Developing Countries:
A Survey of Population and Economic Change in
7. References Developing Countries. (ed. V. Richard A. Easterlin, Ed.)
1. Achour M, Lacan N. Drought in Somalia: a migration (Vol. I). University of Chicago Press. 1980. Retrieved
crisis. The State of Environmental Migration 2011, N° from http://www.nber.org/chapters/c9668
06/12, 2012. 19. Tsogtsaikhan B. Determinants of internal migration in.
2. Ali M. The formation of the Somali National Disaster 2014.
Management Policy. Retrieved from 20. UNDESA. United Nations Department of Economic and
http://adesoafrica.org/what-we-do/blogs/the-formation- Social Affairs. Global Sustainable Development Report,
of-the-somali-national-disaster-management-policy-/, 2013.
2018.
3. Amare M, Hohfeld L, Jitsuchon S, Waibel H. Rural–
Urban Migration and Employment Quality: A Case Study

~10~