RURAL-URBAN MIGRATION IN SIERRA LEONE:
DETERMINANTS AND POLICY IMPLICATIONS
*Assistant Professor, Department of Agricultural Economics,
Michigan State University, East Lansing, Michigan (formerly Research
Fellow, Department of Agricultural Economics, Njala University College,
Njala, Sierra Leone).
**Lecturer, Njala University College, Njala, Sierra Leone and
currently Ph.D. Candidate, Department of Agricultural Economics and
Rural Sociology, Ohio State University, Columbus, Ohio.
***Graduate Research Assistant, Department of Agricultural Economics,
Michigan State University, East Lansing, Michigan.
TABLE OF CONTENTS
PREFACE . . . .
INTRODUCTION . ....
THEORETICAL SCHEMA OF THE DECISION TO MIGRATE .
THE INTEGRATED METHODOLOGY FOR THE MIGRATION SURVEY .. 11
Features of the Integrated Methodology . .. 11
Rural and Urban Data Collection . .. 11
Tracing of Migrants . . . .. 11
Integration of Migration and Farm
Management Surveys . . .. 12
Complete Coverage of Urban Migration
Streams . . . . .. 12
Simultaneous Analysis of Rural-Rural
and Rural-Urban Migration . .. 12
Multi-disciplinary Research on Migration .... 13
The Sierra Leone Migration Survey in Practice .... 13
Phase 1: Rural Areas . . .. 13
Phase 2: Urban Areas . . .. 16
Phase 3: Rural Areas . . .. 18
Preliminary Analysis of the Sample of Traced
Migrants . . . . . 19
CHARACTERISTICS OF MIGRANTS AND RATES OF MIGRATION .
Definitional--Who Is a Migrant? . . .
Classification of the Rural Population . .
Characteristics of Migrants . . .
Demographic Characteristics . . .
Economic Characteristics . . .
Rates of Migration . . . .
Estimation Procedures . . .
Rates of Rural-Urban Migration . .
Rates of Rural-Rural Migration . .
Summary . . . . . ..
THE PROCESS OF RURAL-URBAN MIGRATION . . .
Migration Decision Making in Rural Areas . .
Moving to Town ......
Settling in Town . . .
Rural-Urban Remittances and Contacts .
Return Migration . . .
Attitudinal Characteristics of Migrants
Summary . . . .
RURAL-URBAN MIGRATION, THE URBAN LABOR MARKET
AND URBAN UNEMPLOYMENT . . .
Method of Analysis . . .
Labor Force Participation . .
Structure of Employment . .
Structure of Urban Earnings . .
Rural-Urban Earnings Differentials .
Urban Unemployment . . .. .
The Rate of Urban Unemployment .
Profile of the Urban Unemployed .
Attitudes and Expectations of the
Unemployed Migrants . .
Summary . . . .
ECONOMETRIC ANALYSIS OF RATES OF MIGRATION
Introduction . . . .
The Model . . . ...
Data and Estimation Procedures ..
Empirical Application of the Model .
Implications of the Analysis . .
SUMMARY AND POLICY IMPLICATIONS .. . .
Summary of Major Empirical Findings
in Sierra Leone .. . . .. . .
Summary of Theoretical and Methodological
Findings . . . . . .
Policy Implications . . . . .
Policies to Raise Rural Incomes . . .
Policies Affecting Urban Incomes . . .
Food Pricing Policies . . .. .
Educational Policies . . .
Distribution of Social Amenities . . .
Policies Affecting Urban Living Costs . .
Policies Affecting Information Flows . .
Policies Directly Controlling Migration . .
BIBLIOGRAPHY . . . . . .
. . 80
. . 81
LIST OF FIGURES
A Schema of the Decision to Migrate . .
Rural Enumeration Areas and Urban
Areas of the Migration Survey . .
LIST OF TABLES
1 Overview of the Sampling Procedure and
Questionnaires Used in the Sierra
Leone Rural-Urban Migration Survey ... 14
2 Distribution of Rural-Urban Migration
Sample by Type of Address, Method
of Tracing and Urban Area . ... 17
3 Distribution by Origin and Destination
of Migrants Traced to Urban Areas
Compared to Migrants Identified
in Rural Sample . . ... 21
4 Characteristics of Migrants Traced to
Urban Areas Compared to Migrants
Identified in Rural Sample . ... 22
5 Urban Groupings, Sizes and Economic
Characteristics . . ... 24
6 Disaggregation of the Rural Population
in Each Region by Nonmigrants, Rural-
Rural Migrants, Urban-Rural Migrants,
and International Migrants . ... 26
7 Education, Age and Sex of Nonmigrants,
Rural-Rural Migrants, Urban-Rural
Migrants and Rural-Urban Migrants ..... 28
8 Characteristics of Rural-Urban and
Urban-Rural Migrants by Urban
Area . . . ... .. 30
9 Education of Rural-Urban Migrants by
Rural Origin and Sex . . ... 31
10 Occupational Distribution of Migrants
and Nonmigrants Ten Years and
Older in the Rural Population ...... 33
11 Rural Per Capita Incomes of Households
with Nonmigrants Compared to House-
holds with Rural-Urban Migrants . .. 35
12 Reasons Given for Rural-Rural and
Rural-Urban Migration . . .. 36
13 Gross Cohort Specific Rates of Rural-
Urban Migration by Sex, Education
and Age for Eight Rural Regions
and Four Urban Centers . . .. 40
16 Rural-Rural Migration--Gross and Net
Aggregate Rates by Origin and
Destination Region . .
Persons Identified as Decision Maker
for Migrants by Type of Migrant
and Age at Migration . . .
Comparison of Incomes Estimated by
Rural Nonmigrants and Urban Migrants
for Four Occupations and Actual Incomes
for Migrants with Those Occupations .
Perceived Wage Rate of Rural Non-
migrants by Migration Inten-
tions and Education . . .
Support in Town, Rural-Urban Remittances
and Property Ownership for Working
Migrants by Income Group and for
Nonworking Migrants . . .
Labor Force Participation of Adult
Migrants by Sex, Education and
Age . . . . ..
Percentage Employed in Large-Scale
and Small-Scale Sectors by Sex
and Education and by Urban Area .
Analysis of Variance of Effects of
Sex, Age, Education, Employer
and Urban Area on Earnings . .
Comparison of Rural and Urban Wage
Rates . . . .
Rates of Urban Unemployment by Age
and Education for Male Migrants
Compared to Unemployment Amongst
All Urban Residents . . .
Aggregate Gross and Net Rates of Rural-
Urban Migration by Sex, Education
and Age for Four Destination
Urban Centers . . .
Ratio of Urban-Rural Migrants to Rural-
Urban Migrants Per Year for Adults
15 to 34 Years Age . .
. . 45
26 Unemployment by Urban Center . ... .77
27 Profile of Urban Unemployed in
Freetown by Education . . .. 78
28 Gross Rural-Urban Migration of
Adults in Sierra Leone:
Ordinary Linear Function . ... .90
This paper has been developed as part of a three year study of rural
employment in tropical Africa financed under a United States Agency for
International Development contract (AID/csd 3625) with Michigan State
University. The research in Sierra Leone was carried out under a sub-
contract to the Department of Agricultural Economics and Extension, Njala
University College, Sierra Leone, under AID/csd 3625. The research pro-
gram at Njala University College was also supported by a grant from the
Rockefeller Foundation and the Population Council--the latter specifi-
cally to cover the field research costs of the migration study reported
in this paper.
This first report on the Sierra Leone migration survey together with
previous African Rural Employment Papers by Derek Byerlee, "Research on
Migration in Africa: Past, Present and Future," and by Sunday M. Essang
and Adewale F. Mabawonku, "Determinants and Impact of Rural-Urban Migra-
tion: A Case Study of Selected Communities in Western Nigeria," have
been developed to specifically address a major objective of the African
Rural Employment Study--that is the determinants and characteristics
of rural out-migration in Africa.
We would like to express appreciation to the many persons who con-
tributed to this study. In Sierra Leone we are grateful to our research
assistants, Ola Roberts and James Kamara; our enumerators and numerous
respondents. At Michigan State University, particular thanks are due
our computer programmer, Linda Buttel, and as always Janet Munn for her
Only a decade ago rural-urban migration was regarded as a necessary
element of rapid economic development. Popular theories and economic
history depicted development as the process of moving labor from agri-
culture to industry with industrialization as the driving force of eco-
nomic growth. Moreover this labor transfer from agriculture to industry
was, and still is, widely equated with movement from rural to urban areas.
The disappointing growth rate of agriculture combined with rapid urbaniza-
tion and high urban unemployment rates has led to a questioning of this
strategy. In particular urbanization has been proceeding much faster
than industrialization and growth in industrial employment has lagged far
behind increases in industrial output.
The magnitude and importance of rural-urban migration in most African
countries including Sierra Leone is increasingly recognized as a problem
by policy makers and planners. At least three dimensions of this problem
can be distinguished: (a) the rate, (b) the concentration and (c) the
composition of migration. The rate of migration may be too high for both
economic and social reasons. Numerous authors (e.g., Eicher, et al.
, Byerlee , Todaro ) have noted various price distor-
tions such as high urban wage rates and low agricultural prices particu-
larly for export crops which act to increase rural-urban income differ-
entials and increase migration Moreover the rapid influx of migrants
into urban areas and the stagnation of employment in urban large-scale
sectors has contributed to high rates of urban unemployment--usually in
excess of 10 percent.
The burden that migration places on the urban labor market is illus-
trated by the case of Freetown, Sierra Leone, which is estimated to be
growing at the relatively modest rate of 5.5 percent annually, while
employment in large-scale sectors is growing at most by 2 percent annually.1
Given that about half of the urban labor force is employed in large-scale
sectors, the implied growth rate of the labor force which must be absorbed
in small-scale sectors or become unemployed is of the order of 10 percent
per year. In addition to these urban problems, high rates of rural-
urban migration deplete rural labor which is a limiting factor to agri-
cultural production [Byerlee and Eicher, 1974]. In Sierra Leone, there
is evidence of a decline in export crops as well as an increase in food
imports corresponding to the "diamond rush" of the 1950s.
The problems created by high rates of migration are compounded by
the concentration of migrants in one or two large cities. As Hance 
notes, most African countries have one "primate" city--usually the capital--
which is also the fastest growing city in the country. As a result urban
problems of housing shortages and unemployment are concentrated in the
largest city. In Sierra Leone, over half of the unemployed reside in
Freetown, the capital city.
The composition of rural-urban migrants is a further dimension of
the rural-urban migration problem. Rural-urban migrants are, on the
average, younger and better educated than the rural population from
which they originate. Since education represents a considerable propor-
1The distinction between small-scale and large-scale sectors follows
Byerlee and Eicher . The literature variously refers to modern
and traditional sectors, formal and informal sectors, etc., to make a
2Byerlee and Tommy  compute that the equivalent growth of the
labor force which must be absorbed in small-scale sectors or become
unemployed for Nairobi and Abidjan are 25 percent and 12 percent respec-
tion of total rural investment in many rural areas, rural-urban migration
embodies a substantial capital transfer to urban areas [Byerlee, 1974;
Essang and Mabawonku, 1974; Schuh, 1976]. This is a particular concern
because capital is a constraint on rural development, yet migrant school-
leavers, the product of this educational investment, form the bulk of
urban unemployment. There are also distortions in the educational system
such as the emphasis on education as a criteria for job hiring even where
education does not increase productivity in that job. In rural areas,
too, the selective migration of younger people increases the age and
the dependency ratio of the rural population intensifying the problem of
rural labor shortages.
Recently there has been concern that the composition of rural-urban
migrants leads to rural income inequalities. Lipton  argues that
since urban migrants depend upon rural relatives for support while look-
ing for a job, only higher income rural households can afford to send
migrants to town. However, if these migrants are successful in their
job search they remit considerable amounts of their wages back to their
rural households thus increasing income disparities in rural areas. A
similar argument would hold if educated migrants originate in higher
income households who can afford to educate their children.
Despite the widespread recognition of rural-urban migration as a
problem in Africa, research on migration has not emphasized policy mea-
sures for dealing with the problem. As we have discussed elsewhere
[Byerlee, 1974], extensive research has been undertaken on migration but
the underlying theory and methodology of this research has been such that
its policy relevance is limited. Research has often been descriptive
in nature leading to a good knowledge of migrants' characteristics and
their processes of migration but little understanding of the determinants
of migration. Numerous studies of migration in Africa have identified
economic motives as dominant in the decision to migrate but only Sabot
, Essang and Mabawonku  and Rempel  have carefully
measured urban incomes and none have measured incomes of rural households
from which migrants originate. As a result reducing rural-urban income
differentials has become a universal panacea for slowing rates of migra-
tion; but as we shall show in this paper, this fails to recognize the
complexity of the migration process.
Part of the reason for these deficiencies in earlier studies stems
from the methodology employed. Many studies (e.g., Beals, Levi and Moses
, Harvey , Mabagunje ) have used census information
which is severely limited by information on current rates of migration
and which is of no value for such important variables as incomes. As
a result conflicting conclusions are often reached from census informa-
Numerous surveys of migration have also been undertaken but these
are usually partial in scope emphasizing either the rural or urban side
(but not both) and selective streams of migrants--most commonly male
adults. The difficulties of using past results of research on migration
in Africa for policy analysis thus stem from both deficiencies with re-
spect to the underlying theoretical framework for analyzing migration
processes and the methodology employed. In light of this background
of previous migration research in Africa, the basic objectives of this
1For example, Mabagunje  in Nigeria finds a negative relation-
ship between migration and regional per capital income while Beals, et al.
 in Ghana finds a positive relationship between the same variables.
study are (a) to develop a theoretical schema of the decision to migrate,
(b) to develop an improved methodology for testing this schema, (c) to
apply this methodology to a comprehensive analysis of rural-urban migra-
tion in Sierra Leone and (d) to formulate policy recommendations for influ-
encing the rate, direction and composition of migration in Sierra Leone.
This report details the initial results of our findings from a com-
prehensive study of migration in Sierra Leone. First a theoretical schema
of the decision to migrate is briefly presented and discussed, followed
by a description of the integrated methodology employed in the study and
some preliminary analysis of the representativeness of the sample.
The report then turns to a discussion of the survey results. The
characteristics of migrants and the magnitude and direction of migration
flows are described followed by an analysis of the migration process with
emphasis on migration decision making and intra-urban and rural-urban
income transfers associated with migration. Finally the urban labor mar-
ket in which the migrant participates is analyzed with emphasis on the
structure of urban earnings and unemployment.
The remaining sections of the report integrate the findings from the
descriptive analysis to econometrically estimate the determinants of
rates of migration. This is then used as a basis for a discussion of
policy implications of the study presented in the final section.
THEORETICAL SCHEMA OF THE DECISION TO MIGRATE
In Figure 1 we present a schema for viewing the decision to migrate.
Factors affecting the migration decision can be conveniently segmented
into (a) monetary costs and returns relating to incomes, moving costs and
employment and (b) nonmonetary costs and returns relating to risk, atti-
tudinal characteristics, social ties and expectations. Also a distinc-
tion is made between actual and perceived returns to migration according
to the availability of information on urban life.
The monetary benefits of migration are determined by differences
in rural and urban incomes. Measuring rural incomes to an individual is
difficult where work and income is shared by a household [Knight, 1972].
Nonetheless a useful measure of foregone income is the marginal produc-
tivity of labor which depends on the age and sex of the migrant as well
as a host of other variables such as capital stock and technology.
In urban areas the schema follows Todaro's  expected income
model based on the probability that a migrant will obtain a job in the
large-scale sector with a high wage or alternatively remain unemployed.
The probability that a migrant will be absorbed in the urban traditional
sector with lower wages is however explicitly recognized in this schema.
There are also nonmonetary returns to migration particularly the bene-
fits from improved social amenities such as schools and hospitals and
attainment of higher social status.
Costs of migration include the transport costs of moving, the oppor-
tunity costs of looking for a job in the urban area and the cost of
"setting up house". This latter cost can be greatly reduced by the pre-
sence of friends and relatives in urban areas. Finally there are also
A SCHEMA OF THE DECISION TO MIGRATE
Investment Policies Distribution Media
Income Distribution of Policies
Policies, Etc. Amenities
Income Returns (e.g.,
Social Amenties Information
ision ( Food Returns Expected Perceived
Prices to Present Value Value of
Source: Adapted from Byerlee 119741.
Urban Housing Policies
costs that cannot be readily measured in monetary units particularly the
cost of breaking old and establishing new life styles which is most acute
for older people.
Since educated migrants are of such overriding importance in the
migration stream, we emphasize education in our schema. Education enters
into the migration decision in various ways. First it may increase a
migrant's access to knowledge of urban areas. Second it may enable mi-
grants to derive additional value from urban life styles (and perhaps
devalue rural life styles). Finally and most important there is ample
evidence that despite unemployment the private returns to education are
considerably higher in urban areas compared to rural areas (e.g., Todaro
, Sabot , Hutton ). An important and unresolved issue
is the extent to which education affects the decision to migrate through
each of these three mechanisms.
We would be remiss if we merely accepted education as a given var-
iable in the decision to migrate. It is essential for long run analysis
of migration to understand who gets educated--that is, we need to look
also at the decision to educate. Again a costs-returns framework is a
useful analytical device providing the variation of these costs and re-
turns with individuals is also considered. It is generally true that
the costs of education are relatively lower for high income families be-
cause of their ability to sacrifice present consumption for investment
in education. Thus higher income households invest more in the educa-
tion of their children [Kinyanjui, 1974; Mbilinyi, 1974].
The difference between costs and returns to migration is the ex-
pected present value of migration. However the migration decision is
based on the perceived value of migration which differs from the actual
value according to the information available on the urban labor market.
Although it is generally recognized that informal channels are the most
important sources of information for migrants there is little evidence
on the quality of this information.
The above simplified framework is useful in identifying and explain-
ing various streams of migrants. In general we can distinguish three main
types of migrants: .(1) migrants in the labor force, (2) migrants attend-
ing school and (3) women who migrate for reasons of marriage.
Migrants working or seeking work readily fit the above schema. It
is hypothesized that they perceive that expected benefits of migration
are higher than the costs. These migrants will often be young since
their time horizon for reaping the benefits of migration is longer and
the cost of breaking old and establishing new life styles are less for
young people. Moreover it is convenient to distinguish between the edu-
cated and the uneducated in this stream. The significance of this for
policy purposes is that we hypothesize that uneducated migrants are likely
to conform to the conventional notion that urban migrants originate in
poor rural households and in poor regions of the country, whereas educated
migrants tend to originate in higher income rural households and more
developed sections of the country with long established educational insti-
The decision of migrants to attend school in urban areas also follows
our framework except that the decisions to educate and migrate are taken
simultaneously but still based on perceived long-run costs and returns.
We hypothesize that there are at least three categories of migrant schol-
ars: (1) those who have to leave home to attend school because there
is no school available in the rural area, (2) those who leave because
urban education is perceived to be of higher quality than rural education
and therefore to have higher returns and (3) those who have urban rela-
tives who can support the costs of education in town.
Finally many women migrate for reasons of marriage. There are those
women who are married when they migrate and whose decision to migrate
may be made by the husband. If this is the case, she can be regarded
as a dependent and should not concern us in policy analysis. However,
a second category of women migrate to find a husband in town. This type
of migrant can be readily analyzed within our framework since it can be
presumed that the monetary and nonmonetary benefits of a urban marriage
induce this migration. Unfortunately most surveys of migration in Africa
are based on samples of male migrants and relatively little information
exists on the extent to which women migrate for marriage reasons or al-
ternatively to find work.
In summary, the theoretical schema developed here emphasizes eco-
nomic variables in the decision to migrate although the importance of
many other factors such as risk, expectations and social ties are also
recognized as affecting individual decisions. But to adequately analyze
these motives, the urban labor market must be disaggregated into large-
scale sectors, small-scale sectors and the unemployed. Furthermore it
is essential to disaggregate migration streams by educational level to
capture earnings differentials between rural and urban sectors and with-
in urban sectors.
THE INTEGRATED METHODOLOGY FOR THE MIGRATION SURVEY
Features of the Integrated Methodology
The survey methodology we employed in Sierra Leone was designed
to overcome some of the obstacles to policy analysis inherent in previous
methodologies employed in migration surveys in Africa. Essentially there
are six features in this methodology which lead to the generation of an
integrated set of data on rural-urban migration.
Rural and Urban Data Collection
Exclusive emphasis on studying migration in rural areas or in urban
areas alone gives only one side of the picture. In the Sierra Leone sur-
vey, data were collected in both rural and urban areas and as a result
direct comparisons can be made between rural and urban socio-economic
variables and attitudinal characteristics. Furthermore, expectations of
potential migrants in rural areas can be compared to the reality of ac-
tual migrants in urban areas. Finally both rural-urban migration and
urban-rural migration can be surveyed providing greater insights into the
Tracing of Migrants
The rural and urban data were made more comparable by tracing migrants
from specific locations into urban areas. By focusing on migrants from
given villages or other well defined areas (e.g., census enumeration
areas), the variance of variables describing the rural environment such
as agricultural production systems, incomes, ethnic group, distance, etc.,
is greatly reduced. This may enable a reduction in overall sample size
of urban migrants, and hence a more indepth study of this smaller sample.
Integration of Migration and Farm Management Surveys
The difficulty of obtaining accurate rural income data can be over-
come if a migration survey uses the same sample as a recent or ongoing
farm management or household expenditure survey where economic data are
collected through continuous interviews over a period of time (or even
in a detailed one contact interview). Of course, this presumes that the
sampling method for the farm management survey is appropriate for the
migration survey. In Sierra Leone our migration survey was integrated
with a nationwide farm management survey. The farm management survey
provides information on various measures of rural incomes such as house-
hold incomes, returns to family labor and wages for hired labor.
Complete Coverage of Urban Migration Streams
As shown above migration can be classified into various streams,
such as migrants in the labor force, adult migrants not in the labor
force (primarily housewives and scholars) and children who are sent to
town as wards. Each of these streams was included in our survey to take
into account the various decision makers and motives involved and to
produce a more comprehensive analysis of the migration process than is
afforded by surveys which include only male adults (e.g., Rempel 
Simultaneous Analysis of Rural-Rural and Rural-Urban Migration
The opportunity costs of migrating to urban areas is represented
not only by the alternative of not migrating but also by the possibility
of moving to other rural areas. In Sierra Leone information was also
collected on rural-rural migrants and both rural-rural migration and
rural-urban migration were analyzed.
Multi-disciplinary Research on Migration
Since migration research is in the domain of several disciplines
a fuller understanding of the migration process can be achieved through
involving more than one discipline. In our case we are combining agricul-
tural economics and rural sociology.
The Sierra Leone Migration Survey in Practice
The migration survey was conducted in three phases in 1974/1975
beginning in the rural areas, then moving to urban areas and finally
back to the same rural areas. Details of questionnaires are shown in
Phase 1: Rural Areas
Since one of the features of our migration survey is its integra-
tion with a farm management survey, the rural sample for the migration
survey was based on the sample for a nationwide farm management survey
conducted by Spencer and Byerlee . The country was divided into
eight resource regions shown in Figure 2 reflecting different ecological
zones and hence farming systems. Within each resource region, three
census enumeration areas (E.A.s) were chosen at random with the exclusion
of localities exceeding a population of 2,000 (the former Sierra Leone
definition of an urban area). For the farm management survey, twenty
households were randomly chosen within each enumeration area for a total
sample size of about five hundred households. Each of these households
was visited twice weekly over a cropping year to obtain data on labor
inputs, output, expenditures, remittances and incomes.1
See Spencer and Byerlee  for more details.
OVERVIEW OF THE SAMPLING PROCEDURE AND QUESTIONNAIRES USED IN THE SIERRA LEONE RURAL-URBAN MIGRATION SURVEY
Question- Title of Sampling Sample Frequency Contents of Questionnaire Major Variables
naire Questionnaire Procedure Size of Derived
MG-1 Rural Origin All households in 24 30,000 Once Age, sex, education, fertility, Basic demographic parameters.
questionnaire enumeration areas of persons last place lived, mortality. Population of enumeration
farm level study. Names and addresses of out- areas. Population change.
migrants. Rates of rural-urban migra-
MG-2 Urban All migrants traced 800 Once Detailed information on occu- Urban incomes, unemployment.
migrants into towns 2,000 persons pation, incomes, job search, Rural-urban remittances, etc.
above, support, property, social
participation, the migration
decision, transport, contacts
with home, education, etc.
MG-3 Characteristics All villages in each 100 Once Government, communications, Description of rural envir-
of rural of 24 enumeration villages social amenities, schools, onment.
villages areas. leadership in each village
MG-4 Return Ten persons in each 150 Once Migration history, life in Determinants of return
migrants enumeration area persons town, reasons for returning migration.
who have lived in home
town and returned
MG-5 Out-migrant Heads of households 150 Once Decision making, exchange of Decision making for migration.
households with household mem- persons gifts. Use of remittances.
bers away in town.
MG-6 Nonmigrants Males in each en- 150 Once Migration intentions and per- Determinants of decision to
umeration area, persons ception of urban areas, migrate or not migrate.
15-30 years who
have not left
MG-7 Attitudinal Three migration 110 Once Attitudes to rural and urban Effects of migration on
characteristics streams purposely persons life style, family ties, etc. attidudes.
chosen. Both ur- Occupational prestige.
ban migrants and
MG-8 Unemployment All unemployed 30 Once Details of job-search, Nature and causes of
migrants iden- persons support, expectations. unemployment.
tified in MG-2.
Figure 2. Rural Enumeration Areas and Urban Areas of the Migration Survey
r ." IC
f ,*' Sinkni" ",.
i ,. ,
.. o ... ....., *
.Ka>b^ 5 1 C'
f,- L _-"- E_ .. ; "_/-
1 ) ae.." 3 k e .''
1-.Port 5.. -, L ,.. -
... .., .. brk
-, (. ." M ,, / ..
.i- .* / .r' -al ..
.' ; .... *,H- s/ St--*-n I
8-..,T 1* 2
i * ** *
I la a I II r
!2 6 ,- U -
We urce Regioal Iouedary F
3 0 4 0 so 60Miles a-
EAO I Woof Areas
is. Q0 lie
The first phase of the migration survey was conducted in all house-
holds in each enumeration area (E.A.) including the five hundred selected
households in the farm management study. A census was taken of all peo-
ple in the E.A. to collect data on general demographic characteristics
of the people such as age, sex, education, occupation, etc. At the same
time, data were collected on fertility, mortality and in-migration (see
Table 1). Finally each household was asked to provide the names and
demographic characteristics of persons who had left that household.
Addresses were collected where possible for those who had gone to urban
areas. Together these data enable changes in population in an area
to be explained in terms of births, deaths and in- and out-migration.
Phase 2: Urban Areas
The collection of names and addresses of urban migrants from about
2,500 rural households in the first phase resulted in the names of about
2,000 migrants fifteen years old and above in urban areas. Of these one-
third had gone to Freetown--the capital and main city. Table 2 shows that
we were able to obtain some form of addresses for about half of all mi-
grants although this proportion is considerably lower for migrants in the
diamond mining areas (Kono-Tongo). We had little difficulty locating
migrants because as soon as we had found one or two migrants from a given
village they were able to tell us the whereabouts of other migrants from
that same area. Indeed through this process we located many migrants
1Addresses were obtained from several sources including (a) the
household head, (b) letters written home, (c) school children in the
household who often know the whereabouts of brothers and (d) return
migrants from town.
DISTRIBUTION OF RURAL-URBAN MIGRATION SAMPLE BY TYPE OF ADDRESS,
METHOD OF TRACING AND URBAN AREA
Urban Areas (By Size)
Percent urban migrants identi-
a. Name and address 52 27
b. Name only 48 73
Total 100 100
Percent traced through:
a. Name and address from rural
b. Name only from rural sample
c. Referral from other urban
migrants (not identified in
________________________________________ .4 S
Sample Identified in Rur;
-Sample Located in Urban
A v -*
who were not originally identified in the rural survey increasing the
total number of migrants by over a third (see Table 2).1
Migrants who were traced and located were interviewed to obtain
indepth information on jobs, migration history, initial support in town,
remittances, expectations, plans to return home and socio-cultural fac-
tors (see Table 1). The incomes of these migrants were obtained using
separate forms for wage and salary earners, self-employed traders and
workers in small industries and the unemployed. Incomes for the self-
employed which are particularly difficult to estimate are being checked
against incomes estimated separately in a small industries survey con-
ducted by Liedholm and Chuta . Overall, we traced and interviewed
over eight hundred migrants in sixteen urban areas.
Phase 3: Rural Areas
The final phase of the study involved a return to the same rural
areas to interview three groups of rural people.
Heads of Out-migrant Households. Heads of households from which
migrants have left for urban areas were interviewed to supplement the in-
terviews with migrants in urban areas. This was important since in many
cases these household heads have been heavily involved in the migration
decision of a household member. For example, the decision of school
children or wards to migrate at an early age is almost entirely made by
the rural household head. Thus the household head was interviewed to
determine the motives and reasons for sending or encouraging someone to
live in town. At the same time estimates of remittances of migrants and
Enumerators were paid a bonus of Le .20 to Le .25 in lieu of over-
night allowances, etc., for every migrant located and interviewed (le 1.00
= U.S. $1.10).
the extent to which these remittances were invested in agriculture and
other businesses were obtained.
Return Migrants. Phase 1 of the survey indicated that for every
three rural-urban migrants there were about two urban-rural migrants, many
of whom were return migrants. Hence of particular interest to us are
the determinants and consequences of return migration. A sample of urban-
rural migrants was interviewed to obtain information on their stay in
town, their reasons for returning and the impact that migration has had
on their rural social and economic status.
Nonmigrants. Nonmigrants in rural areas were interviewed to under-
stand why people do not migrate. Nonmigrants may be classified as those
not intending to migrate and those intending to migrate. In both cases
expectations of urban incomes and jobs were measured to determine the gap,
if any, between rural expectations and urban reality. The sample of non-
migrants was weighted toward those most likely to migrate, i.e., male,
young and educated persons.
Preliminary Analysis of the Sample of Traced Migrants
If rural areas are sampled randomly and all migrants identified are
traced into town the urban sample will also be random. However because of
time constraints it was not possible to trace all migrants and possible
biases in the urban sample may result if some groups of migrants are more
easily traced than others. Prior to our analysis of the data we have
The sampling for all three questionnaires in Phase 3 was drawn
such that selected farm management households were included in the sample
if they fitted one or more of the categories: out-migrant households,
return migrants and non-migrants. For these selected households accur-
ate income data are available. For other households a short questionnaire
on total output of crops was administered. This was converted to house-
hold income through correlations derived from the farm management survey.
run some checks on sample bias by comparing the characteristics of urban
migrants identified by rural residents in Phase 1 of the survey, with the
characteristics of migrants actually traced into urban areas. Table 3
gives a distribution of both samples by origin and destination. In gen-
eral there is good correspondence between the two samples although the
traced sample is clearly underrepresented in Kono in the diamond mining
areas where we had few addresses. In Table 4 some general demographic
characteristics of the two samples are compared. In the case of the per-
centage male and the average age in each sample there is a very good
correspondence in nearly all cases. However our traced sample has a con-
sistently higher level of education than the rural sample. Reasons for
this include (a) higher success in tracing scholars in the town of Bo
and Kenema (see Table 4), (b) the concentration of our good enumerators
in the better educated southern part of the country leading to higher
success in tracing and (c) likely understatement of the education of
absent migrants by rural household heads, particularly for scholars who
have acquired education in town. Overall we do not view this bias as
serious since in any event urban incomes are estimated and analyzed for
each educational subgroup. In addition the tracing provides several advan-
tages which outweigh this possible disadvantage. For example we obtained
excellent cooperation in urban areas when migrants learned we had visited
their home area and obtained their name and address (and sometimes mes-
sages for the migrants) from a relative. This cooperation was important
to obtaining accurate data on sensitive variables, such as income.
DISTRIBUTION BY ORIGIN AND DESTINATION OF MIGRANTS TRACED TO URBAN AREAS
COMPARED TO MIGRANTS IDENTIFIED IN RURAL SAMPLE
Urban Destination (j)
(By Size) Over 100,000- 20,000-100,000 2,000- Rural
200,000 200,000 20,000 Region
Rural Freetown Kono Bo Kenema Makeni All Small
S 2.4 1.1 0.0 0.0 .2 .7 4.4
3.6 1.1 .3 .1 .3 1.1 6.5
2. Southern 6.5 .5 2.4 .1 0.0 2.8 12.3
Coast 3.2 1.1 1.4 .4 .2 2.6 8.9
3. Northern 4.0 2.7 .i 0.0 1.1 1.3 9.2
Plains 7.5 5.6 .6 .5 1.6 4.4 20.2
4. Riverain 3.0 .5 1.8 .4 .2 2.7 8.6
Grasslands 1.5 .9 1.0 .4 .1 1.8 5.7
5. B s 13.1 1.0 1.3 0.0 2.3 2.2 19.9
9.2 1.6 .5 .4 1.8 1.9 15.4
6. M B 1.5 4.2 0.0 6.3 .2 1.5 13.7
1.9 3.9 .8 4.5 .3 3.6 15.0
7. Northern 3.4 1.9 0.0 0.0 .8 2.3 8.4
Plateau 2.4 5.5 .1 .4 .6 2.8 11.8
8. Southern 6.5 5.8 5.8 2.2 0.0 3.8 24.1
Plains 2.9 3.8 3.7 1.7 .3 3.6 16.0
Total 40.5 17.3 11.0 8.8 5.0 17.3 100.0
32.4 23.5 8.4 8.4 5.2 21.8 100.0
Key: Upper left corner:
Lower right corner:
Migrants traced from rural region, i, to urban area, j, as percent of all mi-
grants traced (total 825).
Migrants identified in urban area, j, by survey in rural region, i, as per-
cent of all migrants identified in rural sample survey (total 1,900).
CHARACTERISTICS OF MIGRANTS TRACED TO URBAN AREAS COMPARED
TO MIGRANTS IDENTIFIED IN RURAL SAMPLE
Urban Area (By Size)
Over 100,000- 20,000-100,000 2,000- All
200,000 200,000 20,000 Urban
Freetown Kono Bo Kenema Makeni All
Rural sample 64 61 58 54 59 52 59
Urban sample 68 62 59 77 65 55 65
Rural sample 28.2 25.9 25.8 22.9 27.7 28.5 27.3
Urban sample 29.0 28.9 27.8 24.0 26.5 28.7 28.3
Average education (years)
Rural sample 3.5 2.0 4.7 4.6 3.6 3.4 3.3
Urban Sample 4.9 3.6 7.2 7.2 5.1 4.2 5.0
Scholars as percent all
Rural sample 13 5 22 27 17 17 14
Urban sample 12 4 30 42 24 25 18
aSample of urban migrants identified by interview in rural areas.
Sample of urban migrants traced to urban areas.
CHARACTERISTICS OF MIGRANTS AND RATES OF MIGRATION
We now turn to a presentation of the results of our Sierra Leone
migration survey beginning with a description of migrants' characteris-
tics and estimation of migration rates. However before proceeding with
this analysis we divert briefly to establish an operational definition
of categories of migrants used in this study.
Definitional--Who is a Migrant?
Migrants for the purpose of this study were defined on the basis
of both space and time dimensions. To qualify as a migrant an individual
must have crossed a chiefdom boundary, or moved to an urban area within
that chiefdom. In crossing a chiefdom boundary a migrant was classified
as a rural-rural migrant if he or she moved to another rural location.
Rural locations were defined as any location with less than 2,000 per-
sons--the size limit officially used in Sierra Leone. A rural-rural
migrant was defined as an intraregional migrant if he or she moves to
an area inside the same resource region and an interregional migrant
if he or she moves across a resource region boundary. Alternatively a
migrant was classified as a rural-urban (or urban-rural) migrant if he
or she moved to (or from) an urban area--i.e., towns above 2,000 persons.
In much of the following analysis towns are grouped by size as shown in
Table 5 with each group having characteristics related to its economic
base. Finally migrants were classified as international migrants if they
had moved across a national boundary--in this case mainly to and from
Guinea and Liberia.
The chiefdom is the basic unit of local government in Sierra Leone.
URBAN GROUPINGS, SIZES AND ECONOMIC CHARACTERISTICS
Groups Towns Estimated Total Economic
Population Population Characteristics
Size of Towns in Groups
Freetown Freetown 275,000 275,000 Capital city
and main commer-
cial and indus-
Kono All towns 100,000+ 110,000 Main diamond
in Kono mining area
Medium Bo 20,000- 100,000 Provincial cap-
towns Kenema 50,000 itals, educa-
Makeni tional services
and some indus-
Small Bonthe Less 130,000 Some district
towns Rokupr than capitals, large-
Segbwema 20,000 ly commercial
Kabala centers for
etc. rural areas
In the time dimension, a migrant must have resided in an area for
longer than six months to be considered a migrant to that area. This
eliminated the problem of classifying people visiting towns and school
children returning home at vacation time as migrants. For a migrant
who had left his place of birth and moved to another area and then re-
turned home again he must have resided in that place for six months or
more and have returned for six months or more to be considered a migrant.
An individual who satisfied these criteria was defined as a return mi-
grant since he had returned to his home area after a period of residence
In summary a migrant was defined as a person who had moved across
a chiefdom boundary for at least six months. A nonmigrant was defined
as an individual who had resided in his chiefdom of birth all his life
or who had not resided elsewhere for more than six months.
Classification of the Rural Population
Using the above definitions, the rural population was divided into
various groups--nonmigrants, rural-rural migrants, urban-rural migrants
and international migrants. Table 6 shows the disaggregation of the
rural population for each rural region. Nonmigrants consistently com-
prise about two-thirds of the rural population. Rural-rural and urban-
rural migrants are about equal in importance and together contribute
about 25 percent of the rural population. Each of these groups is divid-
ed into return migrants and migrants born elsewhere. Return migrants form
about half of all urban-rural migrants but a very small proportion of
rural-rural migrants. International migrants are generally unimportant
except in Region 7 which borders with Guinea and shares several ethnic
OF THE RURAL POPULATION IN EACH REGION BY NONMIGRANTS, RURAL-RURAL MIGRANTS,
URBAN-RURAL MIGRANTS AND INTERNATIONAL MIGRANTS
Migrant Category Percent of Rural Population in Each Regionb
1 2 3 4 5 6 7 8 All
Scarcies Southern Northern Riverain Boli- Moa Northern Northern Rural
Coast Plains Grasslands lands Basin Plateau Plains Areas
Nonmigrants 77 62 76 71 73 66 64 70 69
Rural-rural migrants 11 26 15 21 11 16 6 15 13
Return migrants 1 7 1 3 4 1 0 1 2
in other rural
areas 10 19 14 18 7 15 6 14 11
Urban-rural migrants 9 11 9 7 15 16 5 14 11
Return migrants 1 5 3 4 5 6 1 6 4
in other rural
areas 2 2 2 0 3 2 0 2 2
in urban areas 6 4 4 3 7 8 4 6 5
migrants 2 1 0 1 1 2 25 1 7
population 100 100 100 100 100 100 100 100 100
aThe rural population base used here excludes people who have resided in the area enumerated for
less than six months and hence fall outside the definition of both nonmigrants and migrants.
bsee Figure 2 for location of regions.
groups in Guinea. For this reason international migrants will be ignored
in further analysis.
Rural-rural migrants and urban-rural migrants shown in Table 6 are
in-migrants to that region. The opposite streams of migrants are of
course rural-rural out-migrants and rural-urban out-migrants. Since
we had a nationwide rural sample rural-rural out-migrants to one region
are rural-rural in-migrants to another region and hence in the follow-
ing discussion only rural-rural in-migrants are analyzed.
Characteristics of Migrants
Table 7 summarizes the education, age and sex characteristics of
various groups of migrants. In general rural-rural migrants have char-
acteristics resembling very closely that of the rural population as a
whole which in turn is dominated by nonmigrants (see Table 6). However,
the breakdown of rural-rural migrants into return migrants and migrants
born elsewhere reveals that return migrants are substantially older and
tend to be predominantly male. Urban-rural migrants, on the other hand,
have a higher level of education and also contain a higher proportion
of males. These characteristics are most pronounced for the return mi-
grants who as in the case of return rural-rural migrants are also much
older than other groups in the population.
The higher level of education and percentage of males among urban-rural
migrants is a reflection of these characteristics among rural-urban out-
migrants. Nearly half of all adult rural-urban migrants have some educa-
tion at the time of migration as opposed to only 10 percent for the rural
adult population as a whole (Table 7). It is significant that although
EDUCATION, AGE AND SEX OF NONMIGRANTS, RURAL-RURAL MIGRANTS, URBAN-RURAL MIGRANTS
AND RURAL-URBAN MIGRANTS
Type of Migrant
+I _ _
None Pri- Second-
population 90 8 2 .44 40 16 13 30 25.1 47
Rural-urban migrants 55 12 33 2.82 28 41 20 10 17.5 54
SOURCE: Migration survey, Phase 1.
aAge and education are computed
15 years old and above.
for the year when migration occurred.
Education is for persons
bTotal rural population includes nonmigrants, rural-rural migrants and urban-rural migrants.
urban-rural return migrants have a higher level of education than the
rural population, they have only about half the number of years of educa-
tion as those leaving for town despite the fact that many migrants acquire
further education while in town. Return migration is selective of older
persons with little education.
Consistent with other migration surveys in Africa, young people domi-
nate in the rural-urban migration stream. Youths aged 15 to 24 years
comprise 41 percent of all rural-urban migrants and the mean age is only
The characteristics of rural-urban and urban-rural migrants are fur-
ther disaggregated by urban areas in Table 8. Medium size towns which
consist of Bo, Kenema and Makeni attract the youngest migrants and migrants
with the highest average education. To a large extent this reflects the
substantial proportion of scholars migrating to these towns. Freetown
also receives migrants with a relatively higher education while migrants
to Kono have a significantly lower education reflecting the dominance of
self-employment in diamond mining which does not require educational skills.
The larger urban centers attract a higher proportion of males than
medium and smaller towns. Nonetheless the statistic of 58 percent male
migrants to Freetown or Kono, is not unduly high when compared to statis-
tics from other countries, particularly Kenya where males comprise about
70 percent of the migrants to Nairobi.
In Sierra Leone the education of rural-urban migrants is highly re-
gion and sex specific. Table 9 shows that for the southern regions (2,
4, 6, 8) almost three-quarters of male migrants have some secondary school-
ing while for the northern regions (1, 3, 5, 7) only about one-quarter
CHARACTERISTICS OF RURAL-URBAN AND URBAN-RURAL
MIGRANTS BY URBAN AREAa
Migrants Urban Areas All Urban
Freetown Kono Medium Small
Number years of education
Rural-urban migrants *2.87 1.76 3.81 2.89 2.82
Urban-rural migrants 1.47 .82 1.58 1.04 1.23
Rural-urban migrants 18.1 18.8 15.6 17.4 17.5
Urban-rural migrants 23.9 23.0 23.5 23.7 23.5
Rural-urban migrants 58 58 49 54 54
Urban-rural migrants 55 66 55 50 53
SOURCE: Migration survey, Phase 1.
aAge and education are computed for the year migration occurred;
education is for persons 15 years old and above.
EDUCATION OF RURAL-URBAN MIGRANTS BY RURAL ORIGIN AND SEXa
2. Southern Coast
3. Northern Plains
4. Riverain Grasslands
6. Moa Basin
7. Northern Plateau
8. Southern Plains
All rural regions
No Primary Secondary
74 3 23
26 12 62
65 10 25
18 18 64
72 2 26
16 8 76
71 9 20
12 17 71
44 10 46
No Primary Secondary
- (Percent Distribution) --
-87- 4 9
46 15 39
77 12 11
61 19 20
98 2 0
60 18 22
84 13 3
60 20 20
70 14 16
aEducation of adults 15 years and above.
have secondary schooling. Education of females is much lower but follows
a similar regional pattern.
In addition to age, sex and educational characteristics it is in-
structive to note the occupation of migrants and nonmigrants in the rural
population. A higher proportion of rural-rural migrants are in nonfarm
occupations such as small industries (tailors, carpenters, blacksmiths),
small-scale trading and services and government jobs than is true of non-
migrants or the rural population as a whole (Table 10). This dominance
of nonfarm occupation is even more pronounced for urban-rural migrants.
Almost 20 percent of urban-rural adult migrants have a nonfarm occupation
compared to less than 5 percent for nonmigrants. These results indicate
that persons with nonfarm occupations are more mobile perhaps in part due
to lack of necessity for land and in part because many serve apprentice-
ships in town where apprenticeship fees are lower [Liedholm and Chuta,
An important hypothesis arising from our theoretical schema is that
uneducated rural-urban migrants originate in poorer rural households while
educated migrants originate in higher income households who have the re-
sources to educate their children. For a subsample of five hundred rural
households we obtained accurate data on household income in an associated
farm management survey by Spencer and Byerlee . Rural per capital
incomes were computed for adult migrants and nonmigrants in the age cate-
gory 15 to 35 years in which most migration takes place. The results,
1Occupations reported here are the stated primary occupation of rural
people. In practice the occupation may change from season to season (see
Liedholm and Chuta ).
OCCUPATIONAL DISTRIBUTION OF MIGRANTS AND NONMIGRANTS TEN YEARS
AND OLDER IN THE RURAL POPULATION
Total rural population 66.7 3.4
Trade Govern- Housewives
--- (Percent Distribution)
2.7 .7 14.7
3.9 2.1 16.7
7.6 3.7 15.5
4.0 1.6 15.5
5.6 3.2 100
aIncludes unemployed, Arabic scholars, religious workers,
reported in Table 11 do indeed support our hypothesis since incomes of
rural households are significantly lower for households with uneducated
male migrants and significantly higher for households with educated male
migrants. (Both differences are significant at the 5 percent level.)
Educated female migrants also originate in higher income rural house-
holds but uneducated females originate in households with average incomes.
This is probably in part because (as we show below) most uneducated females
migrate for reasons such as marriage rather than to seek a higher pay-
ing job. The fact that educated migrants originate in higher income
households is strongly underlined by the fact that migrants under 15 years
of age sent for schooling in town originated in households with per capital
incomes 68 percent above the average.
Differences between migrant and nonmigrant household incomes arise
in part out of a tendency for uneducated migrants to originate in some-
what poorer regions and villages and educated migrants to originate in
higher income regions and villages. However the differences in house-
hold incomes by type of migrant persist even at the village level where
households with male uneducated migrants had average incomes 8 percent
below average incomes for that village and households with male educated
migrants had incomes 6 percent above average incomes for that village.
These differences are not large in part because incomes within a village
tend to be evenly distributed.1
Finally the reasons for migration are shown in Table 12. Although
reasons for rural-urban migration will be considered in more detail in
It is possible that lower per capital household income of house-
holds with uneducated migrants is in part the result of the migration
since older persons are left behind. This is the subject of ongoing
RURAL PER CAPITAL INCOMES OF HOUSEHOLDS WITH NONMIGRANTS
COMPARED TO HOUSEHOLDS WITH RURAL-URBAN MIGRANTSa
Type of Migrantb
2. Uneducated rural-urban
Per Person Per Year)
aFor rural-urban migrants
from which migrants originate.
incomes refer to the rural household
Incomes exclude rural-urban remittances.
Includes only adults aged 15 to 35 years old.
CDifferences between all male groups and between nonmigrant and
educated female migrants are significant at the 5 percent level.
REASONS GIVEN FOR RURAL-RURAL AND RURAL-URBAN MIGRATION
Migrants Work Marry Schooling Warda Other Total
aChildren sent away for upbringing.
a later section the comparison of reasons for rural-rural and rural-urban
migrants shows considerable similarities in both cases. Significantly
only about a quarter of migrants leave for work related reasons. Marriage
is equally important for rural-rural migrants while schooling is the rea-
son given for over one-quarter of rural-urban migrants. This underscores
the limitations of surveys which focus only on male migrants in the labor
Rates of Migration
Rates of both rural-urban and rural-rural migration were computed
from our demographic survey in rural areas. Persons who had left the area
enumerated were identified and the year they departed recorded. Likewise
persons residing in the area enumerated at the time of the survey were
asked their last place of residence and the years they lived in their pre-
sent residence. Rates of migration were computed from the number who
had moved in and out of the area each year using the last five years as
a base. Two deficiencies are inherent in this approach. First even
though our total sample included 30,000 persons it was necessary to use
the last five years rather than the last year to provide a large enough
sample for measuring origin-destination specific migration rates. Hence
there is some recall lapse which tends to underestimate in- and out-migration
by about 25 percent. It is also possible that the recall lapse is less
For rural-rural migrants, work related reasons include farming.
Recall lapse was estimated by fitting the function, mt = m e
to the cumulative average migration rate where mt is the migration rate
estimated for t, mo is the migration rate corrected for recall lapse,
k is a constant and t is time [Som, 1968].
for certain groups of out-migrants, particularly those who have been
successful in town. Second there is likely to be a better reporting of
in-migrants who are resident at the time of the survey than out-migrants
who are absent.1 For these reasons the absolute value of both gross and
net out-migration are probably underestimated although we believe the re-
lative magnitudes of our estimates are valid.
In estimating migration rates two measures are employed. First the
aggregate rate of migration, mijk, is defined as the number of persons in
the kth age, sex, education cohort, Mijk, migrating from origin i to
destination j per thousand of the rural population N. in i. That is,
mijk = Mijk x 1,000/Ni. Second, we computed cohort-specific rates of mi-
gration, mjk by expressing the migration rate as the rate per thousand
of that specific age, sex, education cohort in the rural population, where
m. = M. x 1,000/N and N. is the number of the k age, sex, educa-
ijk ijk ik Ik
tion cohort in the rural population.
These two measures--the aggregate rate and the cohort specific rate--
are both useful in analyzing migration streams. Aggregate rates are a
measure of the number of persons in a specific cohort migrating while
cohort specific migration rates measure the propensity to migrate. For
example in a given area the propensity for educated persons to migrate--
as measured by the cohort specific rate--may be high but the number of
educated persons migrating as measured by the aggregate rate may be low
simply because there are very few educated persons in that rural popula-
tion. It should also be noted that aggregate rates are additive over
Evidence that this is the case is obtained for rural-rural migrants
where the number of rural-rural out-migrants should equal the number of
rural-rural in-migrants because we had a nationwide sample. In fact, we
found that in-migrants outnumbered out-migrants by about 50 percent.
cohorts (k) and destinations (j) but cohort specific rates are only addi-
tive over destinations (j).
Finally we estimated both gross and net migration flows. Aggregate
net migration rates were computed from gross rates by the equation mjk =
(Mijk Mik )/Ni] x 1,000 where M. is the number of persons of the kth
cohort migrating from i to j and M.ik is the number of persons of the kth
cohort migrating from j to i. Cohort specific net migration rates were
similarly estimated. Gross rates are, of course, a measure of the total
movement of people while net migration rates are an indicator of changes
in population size and structure.
Rates of Rural-Urban Migration
Gross cohort-specific rates of rural-urban migration measuring the
propensity to migrate for twelve age, sex and education cohorts are shown
in Table 13. Here migrants are divided into three age groups--15 years
and younger, 15 to 34 years and 35 years and older--and two educational
levels--the uneducated with less than five years of schooling and the
educated with five years or more of schooling. Both age and education have
marked effects on the propensity to migrate to urban areas. Consequently
the 15 to 34 year age group has the highest propensity and the over 34
year age group the lowest propensity to migrate for both sexes and both
educational levels. Likewise the propensity to migrate for educated per-
sons is consistently five to ten times higher than those without educa-
tion for all ages and sexes. On the other hand, sex has relatively little
effect on the propensity to migrate although there is a slight tendency
for educated females to have a lower propensity to migrate compared to males
in the same age cohort.
GROSS COHORT SPECIFIC RATES OF RURAL-URBAN MIGRATION BY SEX, EDUCATION AND
AGE FOR EIGHT RURAL REGIONS AND FOUR URBAN CENTERSa
Rural Regions Sex
and Male Female
Uneducated Educated Uneducated Educatec
<15 15-34 >34 <15 15-34 >34 <15 15-34 >34 <15 15-34 >34
By Rural Origin (Rate Per Thousand)
1. Scarcies 1.6 15.8 8.8 22.2 145.5 n.a. 11.0 9.4 3.3 100.0 100.0 n.a.
2. Southern Coast 5.1 10.5 1.9 55.6 134.9 16.7 16.1 7.7 2.8 46.2 87.0 n.a.
3. Northern Plains 3.8 37.6 6.5 23.5 248.6 75.0 5.7 14.3 3.2 120.0 428.6 n.a.
4. Riverain Grasslands 6.4 5.2 1.9 54.5 116.3 n.a. 11.9 9.2 2.1 55.6 146.7 n.a.
5. Bolilands 4.7 30.2 4.2 12.1 85.0 44.4 13.2 16.6 4.7 100.0 22.2 n.a.
6. Moa Basin 8.0 12.7 1.3 55.8 170.5 23.1 15.4 11.4 3.3 25.0 98.0 n.a.
7. Northern Plateau 5.8 3.0 3.0 133.3 107.1 50.0 3.9 11.8 3.1 n.a. 72.7 n.a.
8. Southern Plains 10.0 22.7 2.8 33.3 154.1 85.1 14.6 21.8 3.8 61.6 108.8 n.a.
By Urban Centere
Freetown .7 4.4 1.2 21.7 43.5 20.5 2.1 2.3 1.0 14.0 28.7 n.a.
Kono 1.3 10.5 .9 2.3 23.2 5.6 1.8 5.5 .7 n.a. 18.2 5.7
Medium Townsd 2.6 4.5 .3 14.5 46.2 8.2 4.6 3.9 .8 25.4 44.8 11.3
Small Towns 1.9 3.4 1.0 23.7 37.0 10.8 2.4 2.1 .9 9.2 34.3 22.0
All Rural-Urban 6.4 22.9 3.4 62.1 149.9 45.1 10.9 13.7 3.3 49.6 125.9 39.0
Cohort specific rates of rural-urban migration are computed as the number of rural-urban migrants
per year of a particular age, sex, education cohort per thousands persons of that cohort in the rural
bThe number of educated migrants in the age category 35 years and above is sometimes too small to
estimate a cohort specific migration rate.
Computed from all rural regions weighted by population for each rural region.
dMedium size towns are Bo, Kenema and Makeni.
NOTE: n.a. = not available because sample too small for estimation.
Overall there are substantial differences in cohort-specific migra-
tion rates by rural region of origin and urban centers of destination.
As observed earlier uneducated migrants have a high propensity to migrate
to Kono while educated migrants tend toward Freetown and medium size towns.
Aggregate gross rates of migration shown in Table 13 follow a simi-
lar pattern to cohort specific rates except that the female uneducated
are more important and female educated migrants less important than males
becuase females have a much lower level of education. However, aggre-
gate net migration rates also shown in Table 14 reveal several points of
interest. First for uneducated migrants of both sexes, net rates for per-
sons 34 years and older are negative indicating that the urban-rural flow
exceeds the rural-urban flow. For males this urban-rural flow is so large
that the net rate of migration for uneducated males of all ages is nega-
tive. For educated persons, however, the net flow is always positive,
even for those above 34 years of age. In fact, educated males 15 to 34
years comprise almost exactly half of all net rural-urban migration.
A second interesting finding of Table 14 is that the most important
destination in terms of net flows to urban areas is Kono. For example,
the net migration rate for all people to Kono is 2.12 compared to 1.45
to Freetown. In fact, using (a) net rates computed here, (b) approximate
urban population figures of Table 5, (c) urban natural growth rate of 2.5
percent and (d) allowing for the underestimation bias against out-migration
reported previously, we can compute rough population growth rates for Free-
town of 4.5 percent; Kono, 9.0 percent; medium towns, 5.1 percent and small
Bear in mind, however, that we believe our out-migration figures
are an underestimate as discussed earlier.
AGGREGATE GROSS AND NET RATES OF RURAL-URBAN MIGRATION BY SEX, EDUCATION
AND AGE FOR FOUR DESTINATION URBAN CENTERSa
Urban Centers Sex Total
Males Females All
Uneducated Educated Uneducated Educated
<15 15-34 >34 <15 15-34 >34 <15 15-34 >34 <15 15-34 >34
Gross Migration Rates
Freetown .13 .49 .15 .09 .77 .09 .39 .41 .13 .04 .17 0 2.88
Kono .26 1.11 .12 .03 .47 .04 .33 1.04 .09 .01 .15 .01 3.67
Medium Towns .50 .42 .04 .19 1.17 .07 .82 .71 .12 .13 .43 .02 4.62
Small Towns .38 .36 .14 .08 .57 .09 .43 .37 .14 .05 .20 .05 2.86
All Urban Centers 1.27 2.38 .45 .40 2.98 .30 1.97 2.52 .48 .23 .96 .07 14.01
Net Migration Ratesc
Freetown -.08 .27 -.04 .05 .66 .07 .20 .18 -.02 .03 .14 -.01 1.45
Kono .03 .70 -.22 .02 .40 .02 .17 .80 .03 .01 .13 .01 2.12
Medium Townsb -.12 -.05 -.42 .12 .83 -.04 .31 -.02 -.10 .05 .26 0 .82
Small Towns -.03 .04 -.20 .06 .46 .06 .05 -.19 -.10 .05 .15 .03 .38
All Urban Centers -.20 .97 -.88 .24 2.35 .12 .73 .77 -.19 .15 .68 .03 4.77
Total all ages -.13 2.71 -- 1.31 I- .86 + 4.77
Total all ages
levels 2.58 + 2.17 4.77
aAggregate rates of migration are computed as the number of migrants for a given age, sex
and education cohort per thousand total rural population.
bMedium towns are Bo, Kenema and Makeni. Small towns have less than 10,000 population.
cNet rates of migration are computed by subtracting the rate of urban-rural migration
from the rate of rural-urban migration.
towns, 3.5 percent. These growth rates are consistent with estimated
growth rates for these centers.
Finally even casual inspection of Table 14 indicates that the differ-
ence between net migration and gross migration is largest for uneducated
groups and for smaller towns. For example, gross migration is largest
for medium size towns but when net rates are computed medium towns receive
only a small proportion of the net flow of migrants. In Table 15 a mea-
sure of this difference, the ratio of urban-rural migrants to rural-urban
migrants is computed. Without exception this ratio is higher for unedu-
cated migrants than educated migrants. This is expected since return
migrants are likely to be less educated and move more freely between rural
and urban occupations with a relatively low differential in pay. In
addition the ratio is highest for small towns and least for large towns.
This implies that migration to the large towns of Kono and Freetown is
relatively permanent whereas migration to smaller towns is much more
circular in nature with more return migration. There is then consider-
able mobility of rural people, particularly uneducated, to and from
small towns usually over short distances.
Rates of Rural-Rural Migration
Gross and net aggregate migration rates for rural-rural migration
are reported in Table 16. Again gross migration rates indicate signi-
ficant flows of migrants for some regions although intraregional flows
often dominate. However, when net migration flows are computed the impact
on population changes is usually quite small. Regions 2 and 3, the South-
ern Coast and Northern Plains, are the major out-migration areas while
Region 1, the Scarcies Area, is the main recipient. The determinants
of the magnitude of these flows will be analyzed later in this report.
RATIO OF URBAN-RURAL MIGRANTS TO RURAL-URBAN MIGRANTS PER YEAR
FOR ADULTS 15 TO 34 YEARS AGE
Towns Males Females
Uneducated Educated Uneducated Educated
Large towns: Freetown, Kono .39 .14 .32 .16
Medium and small towns 1.01 .26 1.19 .35
RURAL-RURAL MIGRATION--GROSS AND NET AGGREGATE RATES
BY ORIGIN AND DESTINATION REGION
Region Destination Region
Scarcies Southern Northern Riverain Boli- Moa Northern Southern Total
Coast Plains Grass- lands Basin Plateau Plains Rate
1 2 3 4 5 6 7 8 Desti-
2. Southern Coast
3. Northern Plains
4. Riverain Grasslands
6. Moa Basin
7. Northern Plateau
8. Southern Plains
2. Southern Coast
3. Northern Plains
4. Riverain Grasslands
6. Moa Basin
7. Northern Plateau
8. Southern Plains
Gross Migration Rates
3.5 .4 .3
1.5 -- 1.6
.2 -- 3.7
-- .1 .2
.2 .1 1.7
Net Migration Rates
--6 -.3 .
2.6 -.3 .3
aRate per thousand of origin population.
A final observation is that rural-rural migration is relatively unim-
portant compared with rural-urban migration. Our data indicate that only
about 12,500 persons or 0.5 percent of the rural population change rural resi-
dence in a year, compared to some 50,000 or about 2.0 percent of the total
population who change residence between rural and urban areas each year.
The methodology employed in our survey allows a disaggregation of
migration streams into various categories--nonmigrants, rural-rural, rural-urban
and urban-rural migrants. The finding that rural-urban migrants are young,
well educated and with a higher percentage of males is consistent with evi-
dence from other African countries [Rempel, 1971; Caldwell, 1969]. Also the
propensity to migrate is several times higher for educated persons and is also
higher for young adults 15 to 34 years old--but does not appear to differ by
sex. Furthermore in Sierra Leone there is a clear north-south dichotomy with
the southern regions producing the bulk of the educated migrants and the north-
ern regions producing most of the uneducated migrants. An important finding
was that uneducated male migrants originate in poorer rural households while
educated migrants originate in higher income rural households. The necessity
of disaggregating migration streams by educational level is clearly demon-
strated by these results.
Some important differences were noted between rural-rural and rural-
urban migration. Rural-rural migrants do not differ significantly in age,
sex and educational characteristics from the rural population as a whole.
Moreover in absolute numbers rural-rural migration is much less than rural-
urban migration and is largely confined to intraregional migration over short
Our survey provides some of the most detailed information available
in Africa on urban-rural migration. About half of urban-rural migrants are
migrants returning home. These return migrants are generally older than the
rural population as a whole. Return migrants also have a low level of educa-
tion compared to migrants who leave for urban areas. As a result the net flow
of uneducated males to urban areas is negative while educated males comprise
about half of net rural-urban flows. Also substantial back and forth mobility
exists between rural areas and small and medium urban towns as measured by
gross migration rates but migration to the large towns of Kono and Freetown
is more permanent with less return migration.
Finally a brief examination of the rural-urban migration streams
shows that migrants seeking work, housewives and scholars are about equally
important, each group comprising about 25 percent of the total number of rural-
urban migrants. These figures underscore the need to disaggregate migration
streams and not stereotype all migration as "labor" migration.
THE PROCESS OF RURAL-URBAN MIGRATION
Rural-urban migration will be examined in this section with respect
to the sequential processes of (a) decision making in rural areas, (b) mov-
ing to town, (c) settling in town and entry into the labor market, (d) main-
taining ties with rural areas particularly through remittances and (e) re-
turning home again and re-entry into rural society.
Migration Decision Making in Rural Areas
Our survey revealed two aspects of rural-urban migration that were
important in migration decision making in Sierra Leone. First only a
minority of rural-urban migrants initially leave home to obtain work.
Migration for marriage and schooling are equally important as migration
for finding work. Secondly migrants leave home at a relatively young age.
In our sample, male migrants without education left home at an average
age of 18 years and educated migrants left at the age of 12 years. As a
result the decision to migrate is more often made by persons other than
the migrant--usually the head of the household--as seen in Table 17. Even
for migrants seeking to work in town almost half the decisions were made
by a parent at home or a relative in town.
Almost all educated migrants initially moved to an urban area to
attend school. Typically an educated migrant had attended school for
11 years of which 5 years were at home and 6 years were in an urban area.
Ninety percent of all migrants with education had attended a school in an
urban area. Of these who had completed school in town, only 27 percent
were working in the same town in which they attended school indicating
substantial mobility among educated persons.
PERSONS IDENTIFIED AS DECISION MAKER FOR MIGRANTS BY TYPE OF MIGRANT AND AGE AT MIGRATION
Type of Migrant
Age at Migration
1. Below 15 years
2. 15-24 years
3. Over 24 years
I 1 7
n Spouse Total
Since the household head was largely responsible for the decision
to send children to school in town we asked why they had chosen a school
in town rather than a rural school. Fifty-six percent made this decision
because there was a relative or friend in town who could help pay fees.
Thirty percent claimed that urban schools were better while 11 percent
responded that there was no school in the vicinity of their villages.
Most women gave marriage as the reason for their migration. In 20
percent of the cases the woman accompanied her husband who was moving
to town. Another 20 percent moved to town seeking a husband while most
moved to town to marry a man already in town.
Migrants who left home to seek work were primarily interested in
obtaining a higher paying job than farming, although a more interesting
job and improved social life were also mentioned. Eighty percent of un-
educated migrants and 93 percent of educated migrants in town felt they
were earning more than was possible at home. Similar beliefs were ex-
pressed by nonmigrants in rural areas although only 60 percent of non-
migrants believed that a city job would pay more.
Migrants, however, are aware of the difficulty of obtaining a job
before they leave rural areas. Among nonmigrants who were intending to
migrate only 15 percent with no education were certain they would obtain
a job. Those with education were more confident with 40 percent certain
they would obtain a job.
Job information is provided by relatives and friends in town for
two-thirds of all migrants while visits to town and friends and relatives
at home provide information to others. An effort was made to measure
the quality of this information by asking a comparable group of urban
migrants and rural nonmigrants the earnings of four occupations--government
clerk, policeman, medical doctor and driver. Results shown in Table 18
show that there is no consistent evidence that rural potential migrants
lack information about urban occupations. In fact, the difference between
perceived incomes and the actual incomes of migrants in town with that
occupation, is negligible except for a government clerk which nonmigrants
ranked much higher and which is the only occupation to show a statisti-
cally significant difference between rural and urban persons. It is appar-
ent, however, that the variance of the estimates of rural persons was
higher than urban migrants indicating that rural people as a whole do not
have unduly high perceptions of urban earnings although there is wide
variation in these perceptions.
Further evidence of rural perceptions is provided by an interview
with young adult male nonmigrants in rural areas--the group with the
highest propensity to migrate. Each person was asked to state his future
migration intentions and to estimate his earnings if he were to move to
town. The comparison of perceived earnings of nonmigrants disaggregated
by migration intentions with actual earnings of migrants already in town
is shown in Table 19. For both levels of education, intending migrants
had higher perceptions of urban earnings than nonintending migrants with
this difference being larger for educated persons. Furthermore intend-
ing migrants in both cases had perceived earnings higher than migrants
in town were actually receiving. There is therefore some evidence that
migrants who leave home have somewhat higher perceptions of urban earn-
ings than are realistic.
Finally among young male rural residents who had no intention of
migrating we found that most had some contacts in town, had in fact visit-
ed town and most believed that their earnings could be increased by
COMPARISON OF INCOMES ESTIMATED BY RURAL NONMIGRANTS
AND URBAN MIGRANTS FOR FOUR OCCUPATIONS AND ACTUAL
INCOMES FOR MIGRANTS WITH THOSE OCCUPATIONS
NOTE: n.a. not available.
aDifferences between rural nonmigrants and urban migrants are
not statistically significant at the 5 percent level except for clerks.
Le 1.00 = $1.10.
PERCEIVED WAGE RATE OF RURAL
NONMIGRANTS BY MIGRATION INTENTIONS AND EDUCATIONa
_ _ _ I
Perceived Wage Ratesb
Actual Wage of
with Same Age---
(Le./Month) (Le./Month) (Le./Month)-
Mean 38 52 42 36
aSample includes only adult males, 15 years to 30 years of age.
Difference between perceived wage rates of intending and not intending migrants significant
at 5 percent level for educated migrants.
migrating. We, therefore, asked these nonmigrants why they did not intend
to move to town. The most important reason given was the need to support
parents and family, suggesting that factors such as kinship ties are im-
portant in the decision not to migrate.
Moving to Town
As Sierra Leone is a small country most rural-urban migration covers
a relatively short distance averaging only about one hundred miles. Be-
cause of this short distance and because over two-thirds move without
dependents the average cost of moving to town is only Le 2.30 and the
cost is nearly always less than Le 10.
There is considerable mobility of migrants after leaving home. The
average migrant resided in two other locations for six months or more
before arriving at his present destination, one of which was an urban lo-
cation. Educated migrants exhibit more mobility so that by the age of
twenty-five they have lived in, on an average, two other urban centers
besides their present urban residence.
Settling in Town
Our survey showed that the prior presence of relatives and friends
in town is almost essential to a migrant's successful adjustment to town
life. Almost 90 percent of migrants were initially supported by rela-
tives and friends in town. The remainder either obtained a job immediately
or had some initial savings for support. On the average a migrant was
supported through food, lodgings and sometimes money for one and a half
years on arriving in town. Nearly all of this support was provided by
urban relatives, most of whom are themselves migrants of an earlier period.
Only apprentices received significant support from other than relatives--
in this case their instructor.
The importance of this support of new migrants underscores the sub-
stantial intra-urban income transfers among migrants. In an effort to
learn who was giving and receiving support we asked each migrant to value
the food, lodging and cash gifts he gave or received to or from an adult
who was not a parent or spouse or child of the migrant.
The results reported in Table 20 show a clear division between work-
ing migrants who are providing support and nonworking migrants including
scholars and the unemployed, who are receiving support. Working migrants
on an average "transfer" Le 9.50 or about 17 percent of their income to
support relatives and friends in town. The amount transferred increases
absolutely (but not proportionally) with the income of the migrant so
that the top 5 percent in the income distribution support up to three
persons at a value of Le 30 per month.
Those who received support are predominantly scholars, apprentices
and the unemployed. Scholars receive support of about Le 16 per month
which is higher than other groups because of the cost of school fees and
books. Significantly migrants as a whole have a net intra-urban income
transfer of almost zero indicating that migrants as a group do not depend
on urban nonmigrants for support.
New migrants seeking a job require support during the period of job
search. Migrants who are currently employed on an average reported a ten
month period to obtain their first job. However, many migrants, parti-
cularly those in the lowest income group, continue to receive some
support for some time after obtaining a job. Furthermore the importance
SUPPORT IN TOWN, RURAL-URBAN REMITTANCES AND PROPERTY OWNERSHIP FOR WORKING MIGRANTS
BY INCOME GROUP AND FOR NONWORKING MIGRANTS
Working Not Working All
Income (Le./Month)a All Unemployed House- Scholars Appren- Other
Income wives tices
<32 32-50 50-90 90-150 150+ Groups
Support in Town
Value given (Le./month) 7.2 10.8 18.1 24.2 30.6 12.9 3.6 2.8 .7 .5 2.9 7.0
Value received (Le./month) 4.5 3.3 4.0 -- 3.4 12.4 2.5 16.4 16.7 4.9 6.2
Net value given (Le./month) 2.7 7.5 12.1 24.2 30.6 9.5 -8.8 .3 -15.7 -16.2 -2.0 -.8
Percent giving 41 48 64 77 70 55 20 17 -- 20 28
Percent receiving 29 21 16 -- 20 55 18 80 82 30 37
Value given (Le./month) 1.6 2.8 3.0 6.1 12.0 3.1 .5 .8 .2 .1 .8 1.5
Value received (Le./month) .7 1.0 1.3 2.3 1.9 1.9 1.0 .9 1.4 .5 1.2 1.1
Net value given (Le./month) .9 1.8 1.7 3.8 10.1 1.2 -.5 -.1 -1.2 -.4 -.4 .4
Percent giving 77 80 86 100 90 82 44 62 23 41 43 57
Percent receiving 56 69 63 -- 50 66 63 63 73 41 52 64
Property at Home
Percent males owning 36 43 46 46 60 45 27 n.a. 23 29 39 30
Property income (Le./Mo.) .4 1.3 3.7 3.1 33.3 3.8 .6 1.2 4.5 2.9
aMigrants' incomes are distributed as follows: 25 perce
less than 90 Le./month and 95 percent less than 150 Le./month.
nt less than 32 Le./month, 50 percent less than 50Le./month, 90 percent
of relatives and friends is again underscored by the fact that two-thirds
of working migrants obtained their first job through a relative or friend.
Rural-Urban Remittances and Contacts
The remittances of income by urban migrants to rural areas has been
widely noted (but rarely measured) in Africa. Our survey shows that re-
mittances follow a similar pattern to intra-urban income transfers in
the form of support (Table 20). The working population remits about
Le 3.10 (about 5 percent of their earnings) to rural areas each month.
However this same group receives Le 1.90 per month so that the net trans-
fer to rural areas is only Le 1.20 per month. Both gross and net urban-
rural remittances increase with urban incomes. Urban-rural.remittances
are largely cash with some imported items such as clothing, while rural-
urban remittances are largely food.
Among the nonworking urban migrants, there is a net transfer from
rural to urban areas. These transfers are largest for scholars and the
unemployed where they could be considered a form of support by rural peo-
ple of their relatives in town. However this form of support to scholars
and the unemployed is almost negligible compared to support received from
relatives in town.
When all working and nonworking migrants are considered together
there is still a small net transfer of income to rural areas of about
40 cents per month or Le 5.00 per year. In our interviews with rural
households we obtained a figure of net remittances received of Le 2.00
per year. The difference in these two figures suggests that migrants
send money to more than one rural household. Most cash remittances re-
ceived by rural households were used for consumption purposes although
about one-third was used for hiring labor and small amounts for equipment,
school fees and medical expenses.
In addition to remittances, migrants also maintained contacts with
their home area in other ways. Visits home for vacation and special pur-
poses were frequent and averaged about one visit per year among our sam-
ple. Significantly too, migrants tended to acquire property at home--
more so than in the town in which they lived. About half of all working
migrants owned property in their village, such as land, tree crops and
houses (Table 20). They also received small incomes from ownership of
that property, particularly migrants in the highest income group. In addi-
tion over 90 percent of all migrants in town stated that they had access
to land in their village so that acquiring land is not an obstacle to
migrants returning home.
The importance of return migration was noted in the previous sec-
tion. When we asked urban migrants about their intentions to return home
about 65 percent stated they planned to return home although few were
very definite about when they would do so. The intentions to return home
were strongest among uneducated migrants and older migrants. For exam-
ple, only 54 percent of youths 15 to 25 with secondary schooling planned
to return while 86 percent of migrants above 45 without education planned
It is also likely that some of the difference is due to rural per-
sons understating their receipts and urban migrants overstating their gifts.
Three primary reasons were given by urban migrants for planning to
return home. First, about one-third wished to retire in their home vil-
lage. Second, another third wished to return for economic reasons believ-
ing that farming was at least as profitable as their urban job. Finally
about one-quarter felt that they may not receive support in town in the
long run and would return. When return migrants were interviewed in rural
areas over half gave reasons relating to problems in obtaining a job or
support from urban relatives suggesting that economic hardship is more
important than retirement as a motive for return migration. In fact,
25 percent of return migrants who sought jobs were unsuccessful and re-
turned without working in town.
As noted earlier, return migrants are older and with lower education
than those who leave for urban areas. On an average our return migrants
had spent fourteen years in town and had typically left at the age of
18 years and returned at the age of 33 years.
Return migrants are of potential significance to rural communities
if they bring money or new ideas acquired in town to that community.
However, our interviews with return migrants would indicate that this
role is relatively minor. Only 20 percent of return migrants had made in-
vestments in property while in town compared to a third of migrants who
were currently residing in town who had made investments in property.
On returning home most brought cash averaging about Le 32 for each return
migrant of which about Le 8 was spent in farming and the remainder con-
sumed. Some 13 percent of migrants had undergone an apprenticeship re-
flecting the fact that many of the skills for small rural industries--
tailoring, carpentry and blacksmithing--are acquired in urban areas
[Liedholm and Chuta, 1976]. Another 10 percent had acquired some educa-
tion in town but as noted previously most educated persons do not return
to rural areas. Finally almost one-third of return migrants felt that
they had not benefitted in any way from their stay in town.
Attitudinal Characteristics of Migrants
Throughout our interviews with various categories of migrants we tried
to gain a perspective on attitudes toward rural and urban residences.
Here we briefly note some of the attitudinal characteristics toward so-
cial amenities that may have a bearing on the migration decision. Both
migrants and nonmigrants attached considerable importance to social amen-
ities such as school, medical facilities and utilities in town. About
40 percent of the urban households but none of the rural households in
our sample had electricity and piped water. Both rural and urban respon-
dents cited these as important advantages of urban residence. Likewise
educational facilities in towns were considered advantages and both rural
and urban respondents felt that rural schools even when available provided
less opportunity for a good education.
When urgan migrants were asked to list disadvantages of urban living
the overwhelming response was the high cost of living in urban areas. Of
course, this was to some extent expected since it was a period of rapid
price inflation. However, among rural persons who were intending to mi-
grate, 40 percent could not think of any disadvantages of urban living
suggesting that their attitudes are changed by the experience of living
In examining the process of rural-urban migration in this section,
we have highlighted migration decision making, urban support and rural-
urban contacts through remittances and return migration. Because most
migrants leave home at a very early age decision making by parents or
other members of the rural household is more important than by the migrants
themselves. This underscores the need to conduct rural-urban migration
surveys in rural areas.
Through interviews with potential migrants in rural areas we obtained
information on rural perceptions of urban opportunities--a deficiency of
most earlier migration research in Africa. Rural nonmigrants do not appear
to have unduly high perceptions of urban wages or job opportunities. How-
ever, perceptions do vary quite widely with individuals and it was shown
that rural people intending to migrate have higher income expectations
than nonintending migrants. These income expectations of intending mi-
grants are also higher than actually realized by urban migrants in town
suggesting that high income expectations do play some role in the deci-
sion to migrate.
A particularly important part of the migration process is the support
given by friends and relatives in town. It was shown that working migrants
are transferring about 17 percent of their earnings to support nonwork-
ing scholars and the unemployed. This intra-urban transfer of income
enables migrants to acquire an education or undergo an average of one
year's job search. Significantly migrants as a group seem to be "self-
sufficient" and do not depend on urban nonmigrants or rural households
for support. In addition relatives and friends are important in helping
new migrants obtain a job.
The importance of intra-urban income transfers is in contrast to
the relatively small rural-urban remittances observed in our sample.
Whereas Johnson and Whitelaw  observe in Kenya that 20 percent of
urban wages are remitted to rural areas the comparable figure for Sierra
Leone for working migrants is only 5.5 percent or Le 3 per month. Net
urban-rural remittances are a good deal smaller--about Le 5 per year--
since rural people also send remittances to urban areas and in the case
of nonworking scholars and the unemployed, these remittances exceed urban-
rural remittances. The most likely explanation for this difference between
Kenya and Sierra Leone is the practice in Kenya of maintaining a wife and
family in rural areas.
We conclude then that intra-urban income transfers are much more
important than urban-rural income transfers in migration in Sierra Leone.
This evidence does not support Lipton's  thesis discussed earlier
that migrants originate in higher income rural households who support
their job search and who after the migrant is employed receive substan-
tial remittances further increasing rural income inequalities.
Finally return migration is numerically important and also contributes
some skills, particularly in small-scale industry, to rural communities;
However, migrants largely return for reasons of economic hardship and
therefore contribute little capital to rural areas. The relatively easy
access to land enjoyed by migrants even when away in town probably in
large part explains the substantial back and forth migration between rural
and urban areas existing in Sierra Leone.
RURAL-URBAN MIGRATION, THE URBAN LABOR MARKET
AND URBAN UNEMPLOYMENT
Method of Analysis
An important aspect of migration to urban areas is the participation
and remuneration of migrants in the urban labor market. In this section
adult migrants 15 years and older are analyzed with respect to (a) par-
ticipation in the labor force (i.e., those working or seeking work),
(b) employment structure, (c) earnings and (d) unemployment. In this
analysis the effects of migrants' sex, age, town of residence, education
and employer are considered. Because the sample is relatively small, var-
ious aggregations are used in this analysis. Two basic age groups are
used--those between 15 and 24 and those 25 years or older. Towns are
aggregated into four size categories as in earlier sections. With respect
to education, migrants were classified as educated if they had completed
more than four years of formal education and the remainder were treated
as uneducated.1 Finally the migrant's employer was disaggregated by large-
scale and small-scale sectors where small-scale sectors consist of firms
employing less than ten persons. Large-scale sectors are further disaggre-
gated into the government sector, including public corporations and semi-
government agencies, and large private industrial and commercial firms.
Migrants employed in small-scale sectors are further disaggregated by
wage earners and self-employed.
In interpreting the results, particular caution must be exercised
for female migrants since the sample size is quite small as a result of
1In fact the educated male migrants in our sample are overwhelming-
ly secondary school-leavers since in Sierra Leone a very high proportion
of male scholars who complete primary school enter (but do not necessar-
ily complete) secondary school.
(a) the dominance of males in rural-urban migration and (b) the low fe-
male participation in the urban labor force. However, because statistical
techniques do point up significant sex differences some results are re-
ported for female migrants.
Labor Force Participation
Labor force participation rates for eight age, sex and education
cohorts are given in Table 21. Seventy-five percent of adult male mi-
grants are in the labor force. The remaining one-quarter are largely in
the 15 to 25 year age category where 56 percent of educated migrants
are still attending school or in the case of uneducated migrants 23 per-
cent are acquiring skills through apprenticeship.
Among female migrants, however, only a quarter are in the labor force.
This proportion rises with both age and education but still remains sub-
stantially lower than for males. These low participation rates are in con-
trast to the important contribution of women in rural occupations, par-
ticularly farming [Spencer, 1976]. Moreover as a result of the substan-
tial number of scholars and housewives not in the labor force overall
labor force participation rates for urban households are lower than rural
households and hence earnings for those who work will have to be higher
to offset the reduced number of workers.
Structure of Employment
The government is the dominant employer of migrants in our sample,
employing half of all migrants who currently hold a job (Table 22). Self-
employment in the small-scale sectors is also important. In contrast
wage employment in both small and large private firms together accounts
for only 20 percent of total employment.
LABOR FORCE PARTICIPATION OF ADULT MIGRANTS BY SEX, EDUCATION AND AGE
Labor Force Sex
Education All Education All
Uneducated Educated Uneducated Educated
__Age -- -- Age
15-24 25+ 15-24 25+ 15-24 25+ 15-24 25+
Wage employed 33 54 25 85 51 -- 2 11 33 6
Self-employed 16 29 2 5 13 13 21 2 19 14
Unemployed 19 10 14 6 11 4 5 6 5
Total in the
labor force 68 93 41 96 75 17 28 19 52 25
Housewives -- -- -- 78 65 35 33 59
Scholars -- 56 1 20 -- 45 -- 12
Apprentices 23 2 2 1 3 1 -
Others 9 5 -- 2 2 4 6 1 14 4
Total not in
the labor force 32 7 58 4 25 83 71 81 47 75
Total 100 100 100 100 100 100 100 100 100 100
PERCENTAGE EMPLOYED IN LARGE-SCALE AND SMALL-SCALE SECTORS BY SEX
AND EDUCATION AND BY URBAN AREA
By Sex and Education By Urban Area All
Males Females Migrants
Unedu- Edu- All Unedu- Edu- All Free- Kono Med- Small
cated cated Males cated cated Females town ium Towns
Government sector 40 73 57 7 48 20 63 13 55 64 52
Large private firms 9 16 13 0 14 5 12 27 6 9 12
sector 49 89 70 7 62 25 75 40 61 73 63
employed 14 4 9 0 10 3 7 10 9 0 8
employed 37 7 21 93 29 72 18 51 30 27 28
sectors 51 11 30 93 39 75 25 60 39 27 36
Total 100 100 100 100 100 100 100 100 100 100 100
aIncludes local government.
The division of employment between small and large-scale sectors
differs significantly with education and sex. Over half of the employed
male migrants without education are employed in small-scale sectors but
almost all educated migrants are employed in large-scale sectors. Female
migrants with and without education have a stronger tendency than males
to be self-employed in small-scale sectors. This reflects to a large ex-
tent the dominance of women in food trading activities.
The structure of employment is quite uniform across urban centers
with the exception of Kono where diamond mining increases the share of
both large private firms, in this case the National Diamond Mining Com-
pany, and small-scale self-employment comprised of diamond diggers.
Structure of Urban Earnings
The structure of earnings of urban migrants is important in deter-
mining migration flows but at the same time serious problems occur in
the estimation of earnings. Earnings in large-scale sectors are generally
easiest to determine. However, fringe benefits such as housing and allow-
ances can be quite important. In our survey these extra benefits were
estimated and added to reported income. For migrants self-employed in
small-scale sectors two methods were used to estimate incomes. First the mi-
grant was asked to state his earnings in a normal month after subtract-
ing all his business costs except his labor. Second, for the week prior
to the interview migrants were asked to recall their transactions. For
small-scale industries respondents were asked to recall all cash transac-
tions for purchased inputs and sales. For traders we recorded wholesale
purchases of commodities, the time to sell their stock and their buying
and selling prices. An estimate of income for the previous week could
then be computed. In most cases, this second measure was used but where
this was unsatisfactory because of missing information or because the pre-
vious week's activity was abnormal, the first measure (i.e., the stated
income) was employed. Finally a high proportion of migrants in Kono were
diamond diggers whose incomes are particularly difficult to measure--in
part because of the illegal nature of much mining. Interpretation of
their incomes must therefore be treated cautiously.
Analysis of variance procedures were used to analyze the effects
of age, sex, education, employer, rural origin and urban centers on
earnings of urban migrants. Results of this analysis are shown in table
23 where the independent effects of sex, age, education, employer and lo-
cation are reported relative to the average income of all migrants. This
analysis demonstrates a wide gap between male and female incomes even
when allowance is made for the different education and employment status
of females. This parallels a similar observation that female wage rates
are lower than male wage rates in rural areas [Spencer and Byerlee, 1976].
However when self-employed persons are excluded from this analysis, sex
is no longer statistically significant. This can be explained by the fact
that many women are engaged in self-employed trading activities on a
part-time basis and receive very low monthly earnings.
Age is also a significant determinant of urban earnings. This is
expected as migrants acquire more skills and capital the longer they stay
on the job. Education has generally the largest effect on urban earnings.
A person with five or more years of education can expect to earn about
50 percent more than his uneducated counterpart.
In the case of self-employed traders and artisans, earnings in-
clude returns to capital.
ANALYSIS OF VARIANCE OF EFFECTS OF SEX, AGE,
EMPLOYER AND URBAN AREA ON EARNINGS
Effect Due To: Percentage Change Significance
from Mean Incomea Level
25 Years and Above
Less Than 5 Years
Five Years and More
Large Private Firms
Small Private Firms
5. Urban Center
aMean income of all migrants = Le 56.37.
Even after allowing for age, sex and education the type of employer
has a significant effect on migrants' earnings. In particular for wage
earners, large-scale private firms pay the highest wage--substantially
higher than the government. At the same time small-scale sectors pay
a wage significantly lower than the government. This is evidence of a
dual labor market with small-scale sectors paying a competitive wage
below the government and large-scale wage structure.
Self-employed workers in the small-scale sectors in our sample re-
ceived earnings above other sectors for two reasons. First, their earn-
ings include returns to capital as well as labor which in the case of
traders and small-scale industries are an important component of earn-
ings. Second this self-employed category includes diamond diggers in
Kono who sometimes have high incomes. It should also be noted that earn-
ings for the self-employed had the highest variance reflecting the hetero-
geneity of composition of this category.
The size of the urban center had some effect on the earnings of mi-
grants with earnings in large towns being above earnings in small towns.
However neither the magnitude nor significance of this effect is as large
as for other variables such as age and education. Only when the effect
of employer is omitted from the analysis does urban location become sig-
nificant. That is, earnings differences between location are largely
due to the differential structure of employment rather than wage differ-
ences per se.
The above analysis treating each effect separately is only rele-
vant if higher order interactions are not important. For example, it
could be hypothesized that there is interaction between age and educa-
tion with education having a larger effect with age. In fact all two-way
interactions were not statistically significant and the only interac-
tion that was not negligible was between education and urban size.
This reflects the fact that educated migrants to Kono received a very
small differential in earnings as a result of education.
Rural-Urban Earnings Differentials
The difficulties of comparing rural and urban earnings are well re-
cognized [Knight, 1972; Collier, 1976]. In comparing rural and urban
incomes here we compare directly the actual wage rate per hour worked
in rural and urban areas. Rural wage rates were derived from the daily
wage observations from a farm management survey reported in Spencer
and Byerlee  where all payments in kind were converted to mone-
tary values and the wage per hour computed from the observation of the
number of hours worked. Urban wage rates were computed from the migra-
tion survey using the hours worked in the week preceding the interviews.
Comparison of these wage rates is given in Table 24. Wage rates
for uneducated migrants in urban sectors are on the average about Le 0.25
per hour or about three times higher than the wage rates of Le .08 per
hour in rural areas. The lowest paying urban sector--the small-scale
sector--has wages above the average rural wage rate but only slightly
above the rural wage rate in the region with the highest wage rate (i.e.,
the Scarcies region). In all cases, of course, educated migrants have
a wage rate higher than uneducated migrants.
1Significant only at the 27 percent level.
COMPARISON OF RURAL AND URBAN WAGE RATES
Rural Areas Urban Areas
Region Wage Employer No Educated
(Le./Hr.) (Le./Hr.) (Le./Hr.)
1. Scarcies .13 Government .19 .35
2. Southern Private large-
coast .08 scale sector .38 .37
3. Northern Small-scale
plains .07 sector .15 .21
4. Riverain .08 Average urban
wagea .25 .35
5. Bolilands .07
6. Moa basin .08 Expected wage
6. Moa basin .08 o y 15
of youth 15
to 24 given
7. Northern to 24 given
plateau .08 probability
. Southern mentb .11 .18
rural wage .08
aAverage over all employers and all age cohorts.
wage for youths 15 to 24 years of age multiplied by
employment for that age and education group.
A more relevant measure of urban wages is the expected wage of young
male migrants between 15 and 24 years taking into account the probabil-
ity that they will be unemployed. That is, the expected wage is computed
as W= (1-Uk)Wk where Uk and Wk are the unemployment rate and average
wage respectively for young male migrants. The wage rate was computed
as the average for all migrants in both small and large-scale sectors
while unemployment rates were derived from data presented in the next
section. The expected wage for uneducated migrants is only marginally
higher than the average rural wage rate and lower than or equal to the
wage rate in two rural regions. Educated migrants still maintain a
considerable wage differential over all rural regions.
These results suggest that over the long term a migrant in an urban
job can earn a considerably higher wage rate in urban areas compared
to rural areas. However in the short term given the lower wage rates
and the high unemployment rates, young uneducated migrants stand to gain
These results must be qualified by at least two factors. First
there is a cost of living differential between rural and urban areas
partly because the basic consumption item is food which includes a mar-
keting margin in urban areas. Secondly, the wage rate is not necessar-
ily the best measure for comparison since urban persons work a larger num-
ber of hours per year than rural persons due to the agricultural slack
season. Thus Spencer and Byerlee  find that rural men work about
1,400 hours per year compared to urban migrants in our sample who worked
over 2,000 hours per year. Migrants may move to urban areas not only
for a higher wage but also to have the opportunity to work longer hours
than is possible in rural areas.
The relationship between unemployment and migration is important
both because unemployment is a central variable of the well-known Todaro
model of migration and its derivatives and because urban unemployment
is aggravated by the influx of new migrants. In this section we brief-
ly examine urban unemployment rates, draw a profile of the unemployed
migrant and his job search and examine his attitudes and expectations
with respect to obtaining a job.
The Rate of Urban Unemployment
The overall rate of male unemployment of migrants in our sample was
14.7 percent (see Table 25) which is slightly higher, but very comparable
to the 13.9 percent figure for all urban residents which can be derived
from the household surveys of the Central Statistics Office [1967-1971].1
However, when migrants are disaggregated by age and education in Table
25 it is found that this unemployment rate rises to 33 percent for young
migrants in the 15 to 24 years age group. In fact, the marked difference
between age groups is common to both educated and uneducated migrants.
For the young age group the educated migrants have a higher unemployment
rate but not significantly so.
The Central Statistics Office surveys provide only a breakdown by
age and by education separately but even these estimates shown in Table
25 are surprisingly consistent with our survey--despite our relatively
small sample size. One implication of this consistency is that the
Our sample shows the rate of female unemployment is 20 percent--
somewhat higher than males. However, the number of females in the labor
force is too small to make a further disaggregation of female unemploy-
RATES OF URBAN UNEMPLOYMENT BY AGE AND EDUCATION
FOR MALE MIGRANTS COMPARED TO UNEMPLOYMENT
AMONGST ALL URBAN RESIDENTS
Migrants 33 9 14.7
persons 30 9 -- 13.9
Central Office of Statistics [1967-1971].
unemployment rates of migrants are similar to the urban population as a
whole although there may be some initial adjustments. Thus for Freetown
the Central Statistics survey computed a rate of unemployment of migrants
in the first year of residence in Freetown of 19.6 percent compared to
17.3 percent for our survey of migrants (of whom a third are new migrants)
and 15.5 percent for all urban residents.
The unemployment rate also varies substantially with urban areas.
The largest urban areas tend to have the largest unemployment rate as
shown in Table 26. In absolute numbers half of all unemployed persons
reside in Freetown.
Profile of the Urban Unemployed
Although the rate of unemployment in our sample differs more with
age than with education, since most young urban migrants are also educa-
ted the dominant group numerically in our sample are young, educated males
who make up 44 percent of the unemployed. Older male adults with no edu-
cation constitute another 29 percent of the unemployed. In Freetown a
special interview was conducted with each unemployed migrant to determine
his length of unemployment, job search activities, etc., as well as his
attitudes and expectations. Although this sample is quite small (forty)
some important attributes of these unemployed migrants emerge. These are
reported in Table 27 disaggregated by education.
Contrary to the image that unemployed migrants are new arrivals in
town, only one-third of our unemployment sample were new migrants in town.
However, among educated migrants 83 percent were seeking their first job--
that is they were "school-leavers". Over half of these school-leavers
had attended school in Freetown and therefore were not new migrants.
UNEMPLOYMENT BY URBAN CENTER
275,000 110,000 20,000- 2,000- All
100,000 20,000 Towns
Freetown Kono Medium Small
migrantsa 17.3 16.8 12.3 10.3 14.7
residents 15.5 11.6 12.2 n.a. 13.9
NOTE: n.a. = not available.
aSORCE: Migration survey.
SOURCE: Migration survey.
Central Office of Statistics [1967-1970].
PROFILE OF URBAN UNEMPLOYED IN FREETOWN BY EDUCATION
Employment and Job Search
Percent new migrants 29 36 32
Percent seeking first job 36 83 62
Years unemployed 1.0 1.1 1.1
Percent registered employment
exchange 13 50 38
Percent seeking casual work 18 19 19
Number of job applications per
month .6 1.6 1.2
Job search expenses per week
(Leone) .92 1.14 1.04
Current household income
(Leone per month) 25 62 45
Attitudes and Expectations
Expected wage (Leone per
month) 39 49
Actual wage for employed
migrants of comparable
age and education 38 44 -
Minimum acceptable wage
(Leone per month) 35 39
Percent more than half certain
of job 55 85 71
Percent risk takers 21 44 36
Years unemployed--risk takers .3 .5 .4
Years unemployed--risk neutral .5 -- .5
Years unemployed--risk averters 1.3 1.6 1.5
Total income of all working household members.
Risk attitudes measured by choice between secure job and
uncertain job with same expected earnings.
Thus the most important group of unemployed are the young school-leavers
who had not worked before.
Both educated and uneducated unemployed migrants had on the average
been unemployed for about one year. This compares with nine years for the
average time period for an employed migrant to obtain a job. A few migrants,
however, reported being unemployed for up to five years.
The survey of unemployed migrants revealed that they were in general
quite active in searching for a job. Most reported undertaking job search
activities, such as inquiry, request through relatives, applications, etc.,
several times per week. In all, the costs of this activity in transport,
influence, etc., are not insignificant amounting to about one leone per week.
Very few unemployed migrants reported to be seeking or doing casual work.
Most felt that their chances of obtaining casual labor on a daily basis
were too small. Significantly, less than half of our sample--particularly
uneducated migrants--were currently registered with the employment exchange.
This suggests that the use of registered unemployed figures from the em-
ployment exchange to measure unemployment is quite unreliable. The corre-
spondence obtained by Levi  between the number registered as unem-
ployed and the number of unemployed derived from surveys is possibly in
part due to employed persons seeking to change jobs through the exchange.
Finally there is a very pronounced difference between the educated
and uneducated with respect to the income of the households in which the
unemployed reside. Given that the average household income in Freetown
is about Le 50 per month [Central Statistics Office, 1967], the esti-
mates from our survey show that,the educated migrants reside in households
Average household income of Le 45 in 1967 adjusted for 11 percent
with above average incomes of Le 62 per month. The uneducated on the other
hand live in quite poor households earning an average of only Le 25 per
month. This difference is due in large part to the fact that the educated
unemployed are supported in households by other educated migrants working
at a relatively high pay.
Attitudes and Expectations of the Unemployed Migrants
The unemployed migrants were asked various questions about their ex-
pectations concerning a job. The expected wage of the job they were seek-
ing was slightly higher than the average wage of working migrants in Free-
town in a comparable age and education category (Table 27). However, all
migrants were willing to accept a job with an income below that average.
Thus, the unemployed would seem to be quite well informed about the urban
labor market. Educated migrants seemed more confident that they could ob-
tain a job with 85 percent reporting that they were certain or fairly
certain of obtaining the job they were seeking.
An experimental question was asked of all unemployed migrants to
measure their risk attitudes. The hypothetical question was posed where-
by a migrant had to choose between (a) a job paying his minimum accept-
able salary and (b) a job paying twice that salary but with a training
period after which he must take an exam with only half a chance of passing.
The expected wage in both cases is the same but the second job is risky
as opposed to the secure first job. On the basis of their response migrants
were classified as risk takers, risk averters and risk neutral. Educated
migrants were more likely to be risk takers possibly reflecting the fact
1Households in which the head is unemployed and which receive no
income are included in this average.
that they live in higher income households. The most interesting find-
ing is that risk takers had been unemployed less than six months while
risk averters had been unemployed for one and one-half years. It would
appear that migrants generally begin their job search with higher aspira-
tions holding out for a good job but as the period of unemployment length-
ens they are willing to revise these aspirations downward.
An analysis of the employment and earnings of migrants provides use-
ful insights into the urban labor market in which migrants participate.
Female labor force participation in our sample.is quite low (30 percent)
compared to rural areas. Moreover, females of both education levels tend
to participate in the small-scale sectors. Males on the other hand par-
ticularly those with education are employed in large-scale sectors where
the government is the dominant employer.
As expected education is one of the most important determinants of
urban earnings. We also found evidence of a dual urban labor market where
large-scale sectors--private and government--pay a wage considerably above
the wage in small-scale sectors. In fact, wage differences between urban
areas could largely be explained by the differences in composition of employ-
ment between urban areas.
Migrants who obtain a job, receive in the long run a wage substan-
tially above rural wages although this difference is not large if the
migrant is employed in small-scale sectors. In the short run, however,
given the probability of unemployment, the expected wage of an uneducated
migrant is very little higher than rural wages. This implies that for
uneducated labor, the rural and urban labor markets are quite competitive.
There is, however, still a substantial differential in rural and urban
wages for educated persons. This helps explain the back and forth mobil-
ity of uneducated migrants between rural and urban areas noted earlier.
Unemployment rates for migrants are particularly high averaging 33
percent for young, educated males. However rates of unemployment for
migrants are very comparable to unemployment rates among nonmigrant urban
residents. Numerically the most important group of unemployed are school-
leavers who have not previously worked and who are concentrated in Free-
Although unemployment and poverty are widely equated, our survey
indicates that this applies only for unemployed persons without education.
The educated unemployed are largely supported by relatives with well pay-
ing jobs and in fact reside in households with above average incomes.
The unemployed in our sample had been without work for an average
of one year. However, evidence was obtained that migrants, particularly
school-leavers, are initially risk takers willing to wait for a job con-
sistent with their above average expectations of earnings rather than
take the first job available. These results lead us to conclude that
urban unemployment is not a critical problem partly because many unem-
ployed are not suffering from poverty and partly because an element of
voluntary unemployment is present as migrants wait for the "right" job.
However there is a considerable cost of unemployment associated with the
loss of on-the-job skill acquisition.
ECONOMETRIC ANALYSIS OF RATES OF MIGRATION
From a policy perspective it is not cnly necessary to know who mi-
grates but to understand factors determining the rate of migration. The
elasticity of migration rates to such variables as rural and urban wage
rates is clearly an important consideration in formulating migration
Econometric analysis of migration rates is now a standard part of re-
search on migration. However, several problems are inherent in past ana-
lyses of this type in developing countries. First migration is often
estimated from birthplace information in census data (e.g., Beals, Levy
and Moses , Sahota , Adams  and Greenwood ).
The use of these data is questionable since migration which has occurred
over a long period of time is related to present economic variables which
in themselves are a function of past migration flows. Second, most ana-
lyses of migration have focused on interregional migration which includes
both rural-rural and rural-urban migration (e.g., Beals, Levy and Moses
, Sahota ). Although a few studies have delineated rural-
urban migration for separate analysis we are not aware of any analysis
which examines both rural-urban and rural-rural migration and examines
possible differences in structural and behavioral characteristics. Fur-
thermore we have noted that migration rates depend markedly on education.
Although this has been observed in other studies the education variable
has been very superficially included--usually by using average levels
of education for the origin and destination regions. For example, studies
in Egypt by Greenwood [1969, 1971], in Ghana by Beals, Levy and Moses
, in Brazil by Sahota  and in Columbia by Schultz  reach
quite inconsistent conclusions regarding the effects on migration of edu-
cation in origin and destination areas. Two recent studies by Levy and
Wadycki  and Barnum and Sabot  have disaggregated the popu-
lation by education and found structural differences in migration rates
by educational level which cannot be explained by the effect of education
on earnings differentials. Finally measurement of rural incomes is a
universal difficulty of almost all analyses of migration. Often proxy
variables are included such as regional per capital income (e.g., Sabot
[19671 or even per capital food production [Levi, 1973].
In the following analysis some of these deficiencies in earlier
analyses are overcome through data collected specifically for the purpose
of analyzing migration rates. This survey data was used to compute
education specific rates of migration for the last five years as dis-
cussed earlier in this report. Migration rates were analyzed for both
rural-urban and rural-rural migration. Rural-urban migration rates are
analyzed by two educational subgroups using education specific urban wage
and unemployment rates. Finally rural wages are obtained from a sample
of 25,000 wage observations obtained in a farm management survey.
The objective of the analysis is to quantify the effects of several
variables on migration rates from specific rural destinations to specific
rural and urban destinations. The model builds upon our earlier theore-
tical framework in which costs and benefits of migration are the major
determining factors of migration. However, since the objective is to ex-
plain aggregate rates of migration and not individual decisions to migrate
variables employed in the model are those that are characteristic of
particular rural and urban locations and not variables such as age, sex,
urban social ties, etc., which are important in individual decisions
but which are not location specific. These latter variables are being
included in ongoing micro-analyses on the decision to migrate. Further-
more in analyzing aggregate migration rates scholars are specifically
excluded since other variables such as the location and quality of schools
are probably more important than variables such as wages used to explain
migration of the working population. Finally we include both males and
females in computing migration rates. Because the most important rea-
son for female migration is marriage usually to a male from the same rural
area, female migration is highly correlated to male migration. In fact,
in our sample the correlation coefficient between male and female migra-
tion from specific origins to specific definitions was 0.78 for unedu-
cated migrants and 0.87 for educated migrants. For these reasons our
model is formulated in terms of variables which are more relevant to
male migrants who are largely in the labor force. However since persons
in the labor force provide the economic base for other nonworking migrants,
particularly housewives from the same area as shown by the above corre-
lations, the model is used to explain total migration (excluding scholars).
The variables of the rural-urban migration model are given by:
mjk = f (Wi, Wjk Ujk. Pj, Dij, e)
where mj = the cohort specific gross rate of adult migration for
Sthe kth educational cohort from rural origin i to urban
Wi = average daily agricultural wage of adult males in rural
Wjk, Ujk = average monthly income and percentage unemployed
respectively for the kth educational cohort of male
migrants in the jth urban center
P. = population size of the jth urban area
D.. = the road distance in miles between the main center
3 of rural region i to urban center j
e = random error
and i = 1, 2,...8, corresponding to the eight rural resource
regions of Figure 1
j = 1, 2,...5, corresponding to the five urban centers
above 20,000 population--Freetown, Kono, Bo, Kenema
k = 1, 2, representing two educational cohorts--less than
five years education and greater than five years edu-
Some comments on the specification of the variables and the hypothe-
sized relationships are in order. The measure of rural income used here
is wage rate rather than household income. This measure of rural income
was chosen because (a) it was shown that an active and competitive rural
labor market exists [Spencer and Byerlee, 1976] and (b) given this com-
petitive market and dominance of household rather than individual deci-
sion making this wage rate should be a close approximation of the supply
price of labor [Knight, 1972].1 Furthermore since females have a low
participation rate in the urban labor market, male wage rates were used.
However, the same rural wage rate was used for both educational cohorts
on the assumption that educated persons receive the same wage rate in
traditional farming activities as those without education.
In the case of individual decision making the relevant income is
the value of the average product if income is shared among household mem-
Urban wage rates were estimated from wage rates of all working ur-
ban migrants analyzed in the previous section. The urban wage is then
the weighted average of wage rates in the large-scale and small-scale
sectors for each urban destination area. The inclusion of urban unemploy-
ment as an explanatory variable, of course, follows the Todaro 
model of migration where it is hypothesized that high unemployment rates
tend to reduce migration.
The size of the urban area is included to represent a number of fac-
tors such as a larger labor market with possibly more perceived oppor-
tunities and also urban amenities (i.e., "bright lights"). Distance is
also a proxy variable for a number of costs associated with moving includ-
ing (a) the economic cost of moving and (b) the social costs of leaving
home which become greater the longer the move and the more cultural or
ethnic differences between home and town. Also distance is likely to
be a factor in determining available information.
The model for rural-rural migration is essentially similar. However
since education is considerably less significant in rural-rural migration
we did not disaggregate by education. Also unemployment is not concep-
tually meaningful in rural areas and hence is not included in the analy-
sis. Finally an ethnic dummy variable was used to test the hypothesis that
rural-rural migrants will move to areas with the same ethnic group to
facilitate social adjustment and access to land.
Data and Estimation Procedures
All data with the exception of urban unemployment and urban size
were obtained from our survey information. Although urban unemployment
data are available from our sample, the sample was too small to estimate
education specific unemployment rates for the medium size towns of Bo,
Makeni and Kenema. Unemployment data were derived from the urban house-
hold survey of the Central Office of Statistics [1967-1971] which we have
previously shown to be highly consistent on a national basis with our own
unemployment data. Also our sample size prevented us from estimating
reliable wage rate data for the small towns (less than 20,000) and hence
they were excluded from the analysis.
Migration rates can be both gross and net as defined earlier. From
a policy perspective both flows are important. Net flows are an indica-
tor of overall rates of urbanization. However it has been previously es-
tablished that return migration is dominated by older persons and hence
gross flows are a better indicator of those entering the urban labor force--
particularly the young who constitute the bulk of the unemployed. A further
important factor is the extent to which variations in net migration are
the result of variations in gross out-migration or of variations in gross
in-migration. In fact in our data the correlation coefficient between net
migration and gross out-migration from rural areas is .891 while the corre-
lation between net migration and gross in-migration is only -.14. Hence
the bulk of variation in net migration from rural areas is due to varia-
tions in gross out-migration, a conclusion similar to Beale's  obser-
1For subgroups of the migration streams the correlations are slightly
lower. The correlation between net and gross migration for uneducated
migrants is .68 and for educated migrants is .87.
vations on net and gross migration flows in areas of the United States
with a net out-migration rate. For these reasons and because net rates
are more unreliable since they include residual errors in estimating rural-
urban and urban-rural migration rates, we analyze gross out-migration
The estimation procedure employed was ordinary least squares regres-
sion. Both linear and log-log functions were tried but linear functions
consistently improved the estimation ability and hence are reported here.
To test if there is any significant difference between the behavior
of educated and uneducated migrants, data for both types of migrants were
pooled and the following linear relationship was fitted:
mijk = b + b + b Wi +b EWi +b Wijk + bEWjk + b Ujk
+ b7EUjk + bP + bEP + b0Di + b EDij + e,
where all variables except E are as defined previously. Following Barnum
and Sabot , E is a dummy variable for education such that E = 0 for
an observation on uneducated migration and E = 1 for educated migration.
The coefficient on these interaction terms indicates whether migration
response differs significantly for educated and uneducated migration
Empirical Application of the Model
Table 28 contains the estimated relationships for rural-urban migra-
tion by educational subgroups. The first figure below each coefficient
is the "t" statistic while the second figure is the elasticity calculated
at the mean value of the variables. Up to three equations are reported
for each group. First is the standard linear form on all variables in
the model. In the case of educated migration, however, strong multicolli-