: ' ' ‘ ' rW-“L n \‘J I‘.3§““I'I;IE‘WW'"' IL II n.’ IF' I {.1 -II'!31“'§tI'I5 ?,
‘=MIU§“i”“”WIW5 “VII: WIWII III“W“““'Ihfi5m335fifiLQLIK.HILm 5....
I: "I '.v . I ? ‘.' LLILIL‘ ~ .. ‘L L L. d .- . . , ‘ L 5
III: “III“ ”III IIIIIIII I .5. I55 5 ‘ ' I . . 55 I I I. . I I. . -
. :LII'tLv ILALI I": I I ILL‘ ‘I LéL'LL H“: LLI'ILILILELL‘L'L 3L :‘iiiiluf’r LL, ‘ILH' LL LgLL L .I . “33'; rLI‘L’” : L I“? I‘; 111:; I L M, . ‘2'
:‘ ' I‘I 1', ‘1'.“ “l.(t‘:!‘I'IIII‘.I_ I 3‘: L I \ _
35 ‘(Ltt L III!‘I1\ LL | I, L‘ Il¢:.' I I‘M! ,5 .‘H‘ .;.'.'L 92.3%;
J ‘. ‘ V ""1“" .51" L " '5 '1 , ' t . If: IL .‘ L31].
1 :19! 155}; :3”; AW“ .Hl'IE'L'I' 9'11“: 1 ,Ige'LIIfI’L“:‘fiIg. 11‘“;qu I," L...‘{ . “I"
. W “4.1 I 'l 11"“ 11‘ ’13,.“ :‘l" . .
at” I {th‘ | mki’ ‘. II
"i i [I

 

II'“

LLLLQNII'I ~II Ni
ILI I.“ :9” t“IV-IN" '

   

   

   
  
 
 
  
 
  
  
   

     
   
   

     
  
  
  

  

 

        
 
  

    
       

   
   
   

    

      

            
    
   
   

     
   
   
    
       
 
 
  

  
 
 
 
 

wonytv-

cat
JM
. I...)
- n.

u ,

1 042': l L. .L I LIIWI .'. fH‘I- . ‘3' “Ill
MHfi“b¢M%WfigfiFHva fifiWfi'“
‘5 ' ‘“%fi”NJ5bvtaoW5ww=an :1
hi “s.“ h" ' wt)?“ ' I“ I {I ‘1 I4 aqugfii‘ “in.
u E “'1‘ I t‘l' Iffitlirlfd.Lt.".‘.l‘q‘g':,g' {015% "I {nu-u“ 'i' L ‘ :élu'“
l ftpvg ‘ £ {31: "L I “1’ 4’ ”up In; LHUIV I L .LO'I'?='L!‘L§;¢E; ‘ L,
L. . 3L \ {5 L I‘ ‘4! I . I: 'K, _ Q 1.. .. 9. WITH ‘ O. _
5.- :z‘ {"3 !I‘ 5'34.“ = ‘I"I~:I.I.I:I f. .‘t'IIIzJ'I‘Ii- Wu! fins: - 1";3". ‘..
" ”‘5‘ “55* I"! “I1'5'I’I‘I‘WEI‘I515I2451 75715:." SI 5.:.2I'II5II'25-II.-.~.'.II.I::. ._
' ~ - -- o. ' yr“ .0 - .‘ ..-- w”. L.
.‘f I If! .§.:.II.I . »- ...~ I. I ..I III; II ‘ 23 IL IILrt' 5.92531
J. *l -. o'- LLlItzz 4 '31. ". .‘ ‘I‘IUE’ "h 2., . I «An ‘.‘.1 L . .
=*I It. 555‘“ .I‘II 5L ~‘5‘1‘5“‘w~ '5‘ III - I ~. I ' ‘5’”. I. -:~ .
'3 ‘ Ihh I ~"w'IIfi.5JW“J ? ‘in‘ ‘$". 5n52§
“ ' 5“ “ ' ‘I'.|. 2 ‘ I -'< ' ‘1 .m- s :‘.
II ‘1 IIIII“5."‘ ‘1‘: ISII‘;155‘: .ffni;§“!' ., ‘ 2:,. ';¢% _ - {a
u.§a§flmkm$wII%mfiwfl 355$; 51F
k ’ ‘ 1 fil""fh“‘L’n\ 61w?”
:Etfi 55L“4Vflunfif$afih '3 3' L I I1 -¥5
5 .3 qfih y.fiE3Ltg'ihfiy.gvfi. 5" L 4.1:; ngquE.L
3: ‘3 mg; 1:31;}: .Js": :‘x‘gth‘ztr‘v'm "IL-SEIIILE'"‘1"‘;:Y £333 r~:"*.£f‘.£§‘i‘yé £15”; R} If}
I} =2 ~‘§:5'§"~’=~E!i'~25}"‘3 ~ : " I. L55... .- 3'5; 9:215?! ii! I"? 15123;", ;. ‘ J {I} _ . .
’5 ‘ESEEIGN, "2"7‘2 z“ . ‘ . “3:: 2"!” .t::‘(§ I}; 12,?1L 'E‘ICL - L 13““ . _ -|“, . L" . ‘L
r .E ‘28,?“ ‘5‘ ‘1' - . "'1 U' “Uighuiut' N 11‘ Q 'J. | ‘H ‘ 0 ‘1‘“;‘7‘; ‘
I" I 5? LI; I’d“ ‘ . ”,5 w. .- -I r; a: “.i
55 5.2....» I :~ .I -
‘ I ° “Ni .

‘ -a'
‘44

‘ .

 

‘4
q
:-
“VJ-i 5
vi -
'3.- ‘
h
m

‘53:th ‘
I ‘I"“ 5 ..

’f‘né‘ “I5!- ‘r' II ‘15
I ["0 "21b
. “I5
'. .zwwsw c”: I
'L; -.I 42.1%.}. 'I-If? 0.

¢
- ~ .o
«W.

7‘ '
m... ... \1
“-4..

,
. >11“
.‘J”..»«“" 5
‘ . “nu-bu— .A—r. 'I

i

.....

 

’ p . . . '

I 5" . ' ’ J“ ‘
IL L’ ' 7:; M .3!” ‘F‘ 3%. t} .
II', ‘-,‘.. ,.‘) -‘ ' 'II” ESE-E Li‘s
.s I. In... I.’IIUI"’R.I'HI‘I‘ ‘

-"I' “Uh?“ ‘3??? - 5 :. .IIII‘IIIIEL' . ".31, 3913';
1“ WI”:- z'l *II“‘1(,HII$ {WWI 12113;};0'155 ‘ I!”
3“- n .5, 5 w: ~25 “-F'IIN' it"rii'v’I'uI‘fI'ti 5.35.4.
I‘E'E‘EII'M 5:375; I"~I“"5';;:I‘I-.II >;‘5!§3I‘.:?~R31{I‘IISIi‘I-‘ti‘; "
fi‘fiv " ' ‘ ' ‘3 "" ~ ’ ‘5";55595452

 
   

. 0
...£"

-r.-

    
 

' . ‘,,-2
.' ' '5
- I! 'J'. L L

.13 ILL,

2
8“ 3g

 

-
. 5
M.--
. 3.), "S. _.
r“ "223:-
\Ow
I—“'W“'
.

. ’r
"fr“;‘l'ngviaq
. . ‘J IIIIII :55
LL Jihfin I LILLLL. :iiiL LILLLTLML I} Lt
’&:’LE§~:.IID "I’H .1‘";LR'II‘GIAI-I’i-iLWILLA!" ‘

   

       
  
 
     
   
 
      

     
 

   

“3‘5 "- I “I 5.15:. 25145. t
I 29‘ :“3‘ 3th Liz: LE)!

afffig‘a' f 22;? . LN N
L-‘I’QL'LL; 'SE'InI Ki Lax] 3‘5311": {a

 
  
 
 

.,
Law
”-9-
.m
V"'«
c“
4.
Eon:
44‘
:1

M's-v

  
  

 

  
     
     

I. I"
113-" '. I :‘HLL
Sf. ;'- LLLH L;

'~!. W;

"
... 4__-
.A, . .-

.

  
  
 
  

  

\,
«II
Iphyfiflah.. .
.I‘ 6 A
'l‘

.. I
Mi”

  

:h.‘ _ .

  
   
    

. .1 g, mu '
*m' I ‘III’LLII We!» '5,”
l;{ g: I‘ L 3;
l "‘.S’I¥E?.£!dv

    

   
            
  
  
  

 

" - - I H. .I‘ .
D 1o"“¥ ".39: 5
:12; I‘L;Eg:£‘::;::51 I." '3'
“r 1,“; “3”“
ch,"-

. 1b .
' 9“ W?

I -:
.15“

‘5‘
.m-

. .L. 2.
v.1Lto’
””9...

'—

 

m

  

may
mm
o
'3
pp
4 ~v~
~I- u m
.om
"17'”
5 m
- w.-

....

h
m'
m.

”25......

vIeEWaE

GAN STATE UNIVERSITY Ll IBRAR

IIIIIIIIIIIIIII II III IIIIIIIIIII

3 1293 00609 3748

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

LIBRARY
Michigan Stab

L“*IJnhm’sity

 

 

 

This is to certify that the
thesis entitled
The Impact of Changing Socioeconomic Factors
on Migration Patterns in Rwanda
presented by

Jennifer Maria Olson

has been accepted towards fulfillment
of the requirements for

Mas te_r_of_Ar_r.s_ degree in W

\ free-J— ,-// I
I \ \[T'f—iitu

. \

 

 

Major professor

'\ ‘- I \v

Date

 

07639 MS U is an Aflirmatt've Action/Equal Opportunity Institution

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or betore due due.

 

DATE DUE DATE DUE DATE DUE

JUN’Ifl g 2
Z

 

 

 

 

 

K.“

 

 

 

 

 

 

 

 

AUG :2 F', ZUIS

 

 

 

 

 

 

 

,—"I

II— I

MSU Is An Affirmative Action/Equal Opportunity Inltltmion

 

 

 

 

THE IMPACT OF CHANGING SOCIOECONOMIC FACTORS
0N MIGRATION PATTERNS IN RWANDA

By

Jennifer Maria Olson

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF ARTS

Department of Geography

1990

ABSTRACT

THE IMPACT OF CHANGING SOCIOECONOMIC FACTORS
ON MIGRATION PATTERNS IN RWANDA

By

Jennifer Maria Olson

Redistribution of population through migration is a widespread
phenomenon in Africa. It is seen as a response to changes in the
socioeconomic environment at both the origin and destination areas.
Rural-to-rural and rural-to-urban streams have been documented and
studies have shown that the individual characteristics of migrants are
important determinants of the migrant stream. This study examines
internal migration in Rwanda between 1940 and 1980. It demonstrates
that the spatial patterns of migration and the characteristics of
migrants have changed over time. The initiaI stream of rural-to-rural
migration, from the densely populated western Highlands to the savanna
area in the East, has been most important. The distance traveled
increased over time as land near the Highlands became settled. Most
recently, this stream has declined as land in the East has become scarce
and the dominant contemporary pattern is movement to the capital city,
Kigali. Gender, level of education, and position in the life-cycle have
remained important in differentiating between migrants and non-migrants
throughout the period of this study, although education has declined in

importance in influencing migration to Kigali.

Copyright by
Jennifer Maria Olson

1990

AC NON E G M S

I would like to express my appreciation to my advisor, Dr. David J.
Campbell, in recognition of his helpful and stimulating comments on the
nature of change in society over space and time. I would also like to
thank Dr. Daniel C. Clay, also of my committee, without whom this study
would not be possible. Not only did he provide the impetus and
necessary data for the study, but he constantly, if gently, reminded me
of the reality and importance of the subject matter. Thanks also to the
SESA office in the Ministry of Agriculture of Rwanda, especially Serge
Rwamasirabo, for both allowing and encouraging me to use their data.
Finally, I would like to thank my comrade migrant, Christoffel den
Biggelaar, for offering me unquestioning support, thin patience when
need be, and love during the time it took to complete this study.

TA 0 EN

List of Figures .......................
List of Tables ........................

CHAPTERS

I. INTRODUCTION ......................
A. Statement of the Problem ..............
B. The Country .....................

II. REVIEW OF THE LITERATURE ................
.Population Growth ..................
Population Pressure/Out Migration ..........
. Other Socio-Economic Factors, Urbanization .....
Individual Characteristics .............
. Gravity Model ....................
Labor Force Adjustment Model ............
Push-Pull Model ...................
. Migration Phases ..................
. Need for Present Study ...............

Haze-amorous)

III. METHODOLOGY ......................

IV. RESULTS OF DATA ANALYSIS ................

A. Spatial Patterns ..................
Migration Activity Over Time .........
Inter- Prefecture Migration ..........
Population Density/Migration Relationship . . .
The Gravity Model ...............
Changing Distance Over Time ..........

(”#ri

B. Characteristics of Non-Migrants and Migrants
Gender-Specific Migration ...........
2 Educational Levels Affecting Migration

Behavior ....................
3. Farm Sizes and Production Levels

4

t—i
O O

Affecting Migration Behavior ..........
Age and Marital Status of Migrants ......

ii

iv

V. DISCUSSION .

VI . CONCLUSIONS

REFERENCE LIST

OOOOOOOOOOOOOOOOOOOOOO

OOOOOOOOOOOOOOOOOOOOO

Figure
l. The Republic of Rwanda .................
2. Rainfall and Elevation .................
3. Responses to Population Pressure ............
4. Annual Population Growth 1948 to 1978 .........
5. Total and Urban Migration, 1940 to 1980

(Male and Female) ................
6. Migration to Selected Destinations (Male Migrants)
7. Urban and Rural Migration (Male Migrants) .......
8. 1945 to 1961 Rural Inter-Prefecture Migration Flows;

1948 Population Density .............
9. 1962 to 1971 Rural Inter-Prefecture Migration Flows;

1970 Population Density .............
10. 1972 to 1976 Rural Inter-Prefecture Migration Flows
11. 1977 to 1980 Rural Inter-Prefecture Migration Flows

1978 Population Density .............
12. 1977 to 1980 Urban Flows ................
13. Changing Farm Sizes by Prefecture, 1965, 1972, and 1988
14. Rural-Rural Migration During Four Periods .......

15
25

61
63
64

69

71
73

LIST OF TABLES

Table
1. Population of Urban Districts, 1938 to 1978 ...... 34
2. Individual Characteristics of Rural Out-Migrants
Compared to Rural Non-Migrants .......... 37
3. The Hypothesis of the Mobility Transition ....... 47
4. The Shifting Role of Migration Hith Development . . . . 49
5. Changes in Factors of Migration With Development . . . . 51
6. Destinations of Men According to Period of Time
(column percentages) ............... 66
7. Rural-Rural Migration Rates Explained by
Population Density ................ 79, 80
8. Rural-Rural In- and Out-Migration Rates and Population
Density Listed by Prefecture ........... 83
9. Predicting Migration Rates with the
Modified Gravity Model .............. 91, 92
10. ,Average Distance Traveled by Males Migrants Over Time . 98
11. Gender-Specific Migration (column percentages) ..... 102
12. Education Levels of Migrant Groups ........... 106
13. Destinations of Men by Educational Level
(row percentages by period) ........... 109
14. Farm Sizes and Production Level by Household ...... 114
15. Marital Status of Migrants and Non-Migrants Ages 20-29 . 126

16. Changes in Factors of Migration with
Development in Rwanda .............. 147

iv

CHAPTER I
IN ROD CT ON

People migrate in response to economic or other pressures felt at
the home region and for the promise of attractive alternatives
elsewhere. Migration is, therefore, a coping mechanism for changing
circumstances. Past analyses have demonstrated that the direction and
intensity of movements can often be explained with these "push” and
"pull" factors. Previous African studies have often focused on rural
push factors, such as low agricultural prices, and urban pull factors,
such as the promise of high incomes, to help understand the urbanization
process, but they have often neglected to contemplate how these factors
may have changed in a country over time. In Rwanda, a small country of
East Africa (see Figure l), the most important factor causing out—
migration has been believed to be extreme population pressure felt in
the Highlands of the West. People have been streaming from the
Highlands and settling in the savanna area in the East. However, a
closer examination of the factors that have led to out-migration and to
in-migration reveals that the migration pattern is much more complex and
that it has undergone an important transformation reflecting the

socioeconomic development of the country.

 

 

 

   
 
 

Gikongoro Butare i

luau "’M

 

 

 

 

 

 

 

 

Figure l. The Republic of Rwanda

Rwanda has the distinction of having the highest population
density in Africa. Unlike many densely populated countries elsewhere,
however, the population of Rwanda is essentially rural with 93 percent
of the people basing their livelihood on rain-fed farming. This
concentration of agriculturalists on the limited land plus the country’s
3.2 percent birthrate have led to severe problems such as very small and
declining farm sizes and declining food production per capita, but it

has also led to adjustments such as people increasing the intensity of

3
their agricultural practices. Thus, Rwanda illustrates the dynamics of
people’s coping behavior to increasing population pressure--dynamics

that often occur only in certain regions of other African countries.

Along with increasing population pressure, Rwanda has been
experiencing another change, the growth of the non-agricultural sector.
A new economic and social system is being overlaid onto and transforming
the rural society. Although Rwanda, as a small and landlocked country,
has not been the recipient of intensive capital investment and formal
industrial activity is minimal, this has not hindered the development of
high levels of non-farm activity in rural areas and the rapid growth of

informal sector activity in the cities.

Although migration is a coping behavior responding to outside
pressures, in all cases an individual’s willingness to move and
satisfaction with his or her residence is related to personal
characteristics, such as gender, the stage in the life cycle, the level
of education attained, and the amount of land the parents have to
bequeath. These characteristics can help determine the trend of who
migrates, and just as importantly who does not migrate. Previous
studies in Africa have shown that these characteristics differ between
rural- and urban-destined migrants, but few studies have examined if and

how characteristics of migrants change as societies develop.

The importance of forces pushing migrants to leave their home area

and of those attracting migrants to other areas thus appears to change

4

over time. Spatial patterns of migration, and perhaps also the
characteristics of migrants, reflect these changing circumstances. In
the short time of living memory, Rwanda has experienced important
changes in the importance of these factors. The small size of Rwanda
and the relative homogeneity of its people, along with the seemingly
clear migration response to population pressure, therefore makes Rwanda
a good case study of how people migrate in response to changing
circumstances. This study will examine how migration patterns in
Rwanda have reflected these changing factors, and how the results of

migration flows have in turn affected the country.

A. tem f Problem

The objective of this study is to examine the impact over time of
changing socioeconomic factors, such as the increase in population
pressure and the growth of an urban-based economy, on the spatial
pattern of migration and on characteristics of migrants in Rwanda. The
scope of the study is the recent history of Rwanda, from around 1940 to

the present time, with particular emphasis on the latest trends.

Questions addressed ask the importance of these factors on
migration activity. For example, at what level is population density an
important factor of out-migration, and at what level is population
density sufficiently high that a region is no longer perceived as

attractive for in-migrants? Also, what is the relationship of rural-

5
and urban-destined flows and how have they changed over time? And

finally, how have migrant characteristics changed over time?

For the purposes of this study, migration is defined as a
permanent move, or settlement in a new residence for one year or more,
outside the prefecture (the largest administrative unit of Rwanda) of
birth. Temporary migration, or circulation, is not included in this
study but is the subject of a thesis by Innocent Ngenzi in Rwanda. This
study focuses on internal migration between prefectures, and therefore
will not include the large number of Rwandan emigrants that have left to
neighboring countries or the immigrants that have settled in Rwanda;
this is due to data limitations and not because these movements have
been insignificant. The focus of the study is on the movements of men
because most movements by women are due to the custom of female exogamy
and consequent “marriage migration" which has only an indirect relation

to population pressure or other socioeconomic factors.

To summarize, this study is primarily concerned with internal,
permanent migration in Rwanda from around 1940 to the present. The
impact of changing demographic and economic factors on migration

patterns and migrant characteristics will be analyzed.

6
B. The Country

Rwanda is at the heart of the Great Lakes Valley Region in the
western arm of the Great Rift Valley and is composed of a low savanna
region in the East, a Foothills region in the center, and the Highlands
of the Nest. Rainfall distribution is closely associated with altitude
(see Figure 2). The Highlands in the Nest form the Congo-Nile Divide.
They have steep slopes previously covered with dense forest and
relatively fertile soils. Due to their high altitude, the temperatures
are moderate and two adequate rainy seasons per year occur regularly.
The region is, therefore, considered to have high agricultural
potential. The Foothills consist of sharply defined hills which are
often intersected by valleys. The soils are less fertile and rainfall
less abundant, so the agricultural potential is not as great as in the
Highlands. Finally, the savanna in the East is a country of lakes and
plains. The rainy seasons, especially the second, are not as reliable
and the soils much poorer than in the Highlands or Foothills.
Consequently, the agricultural potential is also lower (Prioul and

Sirven 1981, Gotanegre et al 1974).

Rwanda covers an area of 26,338 square kilometers (about the same
size of Maryland) and has an estimated population of 6.2 million. The
Hutu make up 85 percent, the Tutsi 14 percent, and the Twa one percent
of the population. With a population density of 238 people/khf, Rwanda
has the highest population density in Africa. Its population growth

rate is estimated at 3.2 percent annually (Gourou 1990). The Hutu

 

 

§ J
x P’ V
‘5 less than 1000m
a:J - 1000101500!"
IIII nxerhan 2000M
Source: Rossi, "Evolution des Versents et Mice en Velour Agricole eu Rwanda,“ es de G

515 (1984), 25.

 

Figure 2. Rainfall and Elevation

8
people arrived in the area probably near the end of the first
millennium. They settled in the Highlands where their crops of sorghum,
millet, and cowpeas grew well. The pastoral Tutsi reached the area
around the fourteenth or fifteenth century and consolidated their
position around the savanna area to the east of what is now Kigali city
(Jones and Egli 1984). A feudal relationship was eventually established
with the Hutu during "La Grande Expansion” period of 1744 to 1895 under
an ambitious Tutsi "mwami" (king) (Lemarchand 1970). By 1895, the Tutsi
had reached Lake Kivu in the Best by a gradual process of expansion into
Hutu territory and assimilation of Hutu culture and language. The
control of the Tutsi over the Hutu was made uniform across the country
and given official status by first the German colonialists, beginning
1898, but especially by the Belgian colonialists from 1916 to the 19505
(Lemarchand 1970). The Belgians relied on the Tutsi rulers to act as
intermediaries to aid in the collection of taxes and compulsory labor

(Newbury 1980).

These and other forms of repression, as well as changes in the
society, eventually led to revolt and violence in the late 19505. By
the time independence was granted in 1962, a group of Hutu had gained
control and won an overwhelming majority of votes in a U.N. sponsored
election. Grégoire Kayibanda, the leader of the political party which
had demanded Hutu emancipation, became the first president. The worst
Hutu-Tutsi disturbance occurred soon afterwards in 1963 when, in
reaction to an attack by Tutsi emigrants, a massacre of Tutsi led to

more than 10,000 deaths. President Kayibanda held power until a

9
successful coup led by another Hutu, Juvénal Habyarimana, in 1973.
President Habyarimana, who is considered to have a more moderate stand
on the issue of Hutu-Tutsi relations, continues to hold power (Reyntjens

1990).

The economy is limited by the small size and landlocked position
of the country, but it has consistently grown since independence owing
to increases in agricultural production. The agricultural sector
accounted for about 40 percent of the GDP in 1986, and 95 percent of
this is provided by subsistence crops. The annual increase in
agricultural production met the needs of the growing population until
1977, but since then the increases in cultivated land have slowed and
crop yields are declining in may areas due to land degradation. The
industrial and manufacturing sectors, which produce products mostly for
local consumption, contribute 21 and 15 percent to the GDP,
respectively. Coffee provides 82 percent of total export earnings

(Reyntjens 1990).

CHAPTER 11
W

Migration as a symptom of change in socioeconomic systems has been
the subject of a wide variety of academic and applied studies. Some
authors have attempted to explain the forces behind migration whereas
others simply describe the process itself. Two important societal
tensions linked to migration are examined in this chapter--the ”push”
factor of increasing population densities and the ”pull" factor of the
development of the urban sector. Then, important variables in the
individual decision-making process leading to migrant selectivity are
presented. A few migration models are examined for their relevance in
the Rwandan setting: the gravity, the labor force adjustment, and the
push-pull models. Finally, two hypotheses of the pattern of migration
change over time are analyzed and a version of them suggested for

Rwanda.

AW

Population growth and agricultural development have long been
studied by authors interested in the link between the two processes.
One of the most influential thinkers to this day has been Malthus, who
took the bleak viewpoint that ”the power of population is indefinitely

10»

11
greater than the power of earth to produce subsistence for man” (Durham
1979). The earth’s inflexible resources, particularly land, therefore
control the population size by preventing the population from growing
beyond the limit the land can support. Many see this basic relationship
continuing to be the guiding force in the developing world today,
especially in land-poor countries such as Rwanda. Thus follows the
widespread concern over the diminishing supply of arable land per capita
and the espousal of "overpopulation” as the cause of land degradation
(see Blaikie and Brookfield, 1987, for a discussion of the limitations
of this theory). Models of carrying capacity implicitly assume this
static relationship between population and land in computing the maximum
population density a region can sustain without deterioration of the
human or environmental systems at a given level of technology, capital,
and resources (Bernard 1982). This view of strict limits of the ability
of the land to support the population and consequent fears of

”overpopulation" has received much investigation in Rwanda.

Populations, however, have outgrown the supposedly inflexible
limits of the land at a tremendous rate not only in developed but in
developing countries. The demographic transition theory was proposed to
explain the population growth cycle; it was first introduced by Harren
Thompson in 1929 (Zelinsky, 1971) and then broadened by Kingsley Davis
in 1963 (Bilsborrow, 1987). The theory has now been accepted as
axiomatic for a process all societies apparently experience after
reaching a certain stage of socioeconomic development. In short, the

transition societies experience is the change from high fertility and

12
mortality rates to an initial decline in mortality rates, followed by a
corresponding decline in fertility rates, and a final stage of uniformly
low mortality and fertility rates. Most African countries are in the
second stage of high fertility rates and declining mortality rates,
which is the stage of the most rapid population growth. Societies pass
from that stage to the next of declining fertility rates for reasons
which appear to vary from one society to another and Africa in
particular has many social characteristics which help maintain high
birth rates (Boserup, 1985). Zelinsky (1971) has used the demographic

transition as an analogy to migratory patterns societies experience.

Malthus’s theory of finite population growth was challenged by
Boserup in 1965 and 1981, when she theorized that increasing population
densities are actually the stimulus for technological modification
allowing more intensive, productive cultivation of the land. According
to her hypotheses, output on a given area responds more generously to
added labor than had been assumed by neo-Malthusians. She shows with
comparisons between traditional agrarian societies a continuum of land
uses from virgin land to land being cropped at shorter and shorter
intervals, and finally to continuous cropping. These changes in
cropping patterns accompany adoption of new technologies which become
possible with an increasing investment of labor-intensive technologies
such as plowing, manuring, terracing, irrigation, and growing fodder for
animals instead of allowing them to graze. Along with the increasing
productivity of the land, however, comes decreasing productivity per

man-hour of labor, which is why a shift in technologies comes only with

13
population pressure. Population pressure also changes land tenure
arrangements, although not as simplistically as first believed in which
first all land becomes fully occupied and finally a small class of
proprietors appears. Rather, Boserup proposes that changes in gathering
and grazing rights occur with eventually private property and buying and

selling of land becoming the dominant feature.

In some cases, pastoralists and farmers practicing long-fallow
techniques move to sparsely-populated areas when population increase
prevents fallow periods of the length to which they are accustomed.
Movements have therefore evened out densities across regions as well as
hastened diffusion of intensification technologies. In areas such as
Africa where people lack knowledge of or access to improved technologies
and inputs, the expansion of food production has been made possible by
settlers using traditional techniques on land previously used for

hunting and gathering, grazing or fallow.

B. o l 0 Pre r 0 Mi n

In Boserup’s theory, the primary response to increased population
densities is intensification of land use, and only if no knowledge
exists of how to increase productivity without exhausting the soil does
starvation or migration ensue (Boserup 1965). She does not consider any
demographic, such as fertility decline or out-migration, responses to

population pressure important (Bilsborrow 1987). Indeed, especially in

14
her 1965 book, she regarded the extensification of cultivation onto
unsettled land as unusual (Grigg 1979). However, as she remarked in her
1985 book, particularly striking in Africa in the past few decades has
been the expansion of cultivation onto new land to increase food
production with land close to areas of high population density exploited
first. The pressure on the land leads to gradual expansion of tilled
land by the hiving-off of communities or groups; many examples of this
population pressure-migration relationship exist in Rwanda, Uganda, and

Kenya (Hance 1970, Connell et al 1976).

Historically, population pressure has led to both land use
intensification and out-migration, with people tending to first exhaust
any readily available agricultural intensification technologies and non-
farm income earning possibilities before resorting to out-migration.
Fertility adjustments seem to come last (Brown and Lawson 1985, Grigg
1980a, Bilsborrow 1987, Davis 1965). These were diagramed by Grigg (see
Figure 3). The figure illustrates three stationary responses to
population increases: the Boserupian intensification of agricultural
production, the development of non-farm income sources, and reduction in
fertility levels. The fourth and mobile response, migration, is the

primary response considered in this study.

In Africa, the adoption of intensive technologies occurred when
populations slowly evolved over time. More recently, however, the
ability of societies to develop new intensive technologies has been

overwhelmed by extremely rapid population growth rates (Campbell and

15

hicveeee 7 Switch to 7
h Mace lncreeee hi0" Regime!
amnion felon I Levee yoelde m w M
if HI iIE 6] J

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I ,
incense -
mucuuueu. ' oma'm
ourPur
Oversees 1'1
A Crene
l .34 e
tredee
. i Incense
23:: -- mm... 0 ”meta" —a W538" -
1 E
I "" 1mm
Rurel 5“ C
urben _ j
CONTROL 5‘, Tmerv
NUMBER “AW
OF BIRTHS negation
T
1 1 n
Limitetion oi . .
1mm within “W .‘9‘ 'm
mm of menuege cehbecv
Source: Grigg, P tion row h nd A rerie Che e- An N r at r 'v , 1980, 46.

 

Figure 3. Responses to Population Pressure

Riddell 1984). Population pressure has given rise to a number of easily
recognized "symptoms of stress” such as the fragmentation of farms,
smaller small farms, reduction of land left fallow, land degradation,
fewer animals, and an increase in landlessness (Grigg 1980a, 1980b).
Finally, "excessive" out-migration to rural and urban areas has followed
(Bilsborrow 1987, Grigg 1980b). Kenya and Uganda and especially Rwanda

provide good examples of response to population pressure.

16

The population-resource context in Kenya has been much studied,
perhaps because of the recent startling population growth rates and
limited supply of arable land. The fertile highland areas around Lake
Victoria and north and east of Nairobi have traditionally had the
highest population densities and have been able until recently to absorb
their own population growth (Bernard 1982). Intensification in the
forms of irrigation and manuring have allowed higher densities to be
supported (Campbell 1984). The ability of the densest regions to absorb
population growth has been called the "sponge effect," but there are
limits to the ability of an area to continuously absorb and retain all

of the rural population (Gould 1985).

With the increase in population, farmers began spilling onto the
marginal and semi-arid lands where rights to land have not been so
guarded (Bernard 1982) and onto mountain slopes, river valleys, and
swamp margins, eventually curtailing the access of grazing land for the
Maasai herders (Campbell 1981). During the dry years of 1972-1976,
Campbell (1984) found that the newly arrived farmers were ill-prepared
for the drought and that the herders were deprived of their drought-
retreat pastures because of the spread of cultivation. These migrants
to settlement and irrigation schemes and the spontaneous "squatters"
occupying medium and marginal lands are numerous; the number of
spontaneous settlers in particular has been a concern of the government
(Mbithi and Barnes 1975). The lack of land does seem to be a driving
force for migration--one study found 70 percent of rural-urban migrants

were landless and had no possibility of inheriting land (Rempel and

17
Todaro 1972) and another study found that population growth rates were
inversely related to population density, with the most densely populated
areas experiencing lowest growth rates due to out-migration (Gould
1985). The scarcity of land even in the settlement schemes and other
dry-land areas has led to a decline of the rural-rural migration pattern
seen in the 19605 and has given way to wage-labor migration to both

rural and urban areas in the 19705 (Gould 1985).

In Kenya, therefore, a process of intensification of land use in
the high density areas was followed by rural-rural out-migration to
lower—density, dry-land areas; this caused some problems of land use
competition. The out-migrants were often those with the least amount of
land. As free land even in the less desirable dry areas became scarce,
however, land-seeking migration was replaced by wage-labor migration to

both rural and urban areas.

Similarly, in-migration to Uganda’s dry lands has become
significant in recent years due to tsetse fly eradication and
infiltration of people from high density regions of central and western
Uganda. Government settlement schemes in the dry lands ultimately have
been exceeded in importance by spontaneous colonization. As farmers
have been "squeezed” out of the wetter high density regions and moved to
the dry lands, they have in turn taken over the better grazing lands of
the pastoralists. Problems have also developed because the newcomers
brought their habits of intensive land cultivation and land tenure

arrangements which lead to fragmentation, habits which had evolved in

18
the high density areas from which they had come. The result has been
land degradation and the threat of further land degradation in rural
areas (Kabera 1985). In contrast, concentrated development and
relatively high incomes in cities along the coast and in certain
strategic areas have created severe regional inequalities that have led
to major flows of population to these urban centers (Gimui 1982).
Therefore, the pattern in Uganda has been similar to that of Kenya. A
large movement of people has occurred from the densely populated
Highlands to the dry lands previously used by pastoralists and this has
caused some problems. More recently, however, with land fragmentation
and other symptoms of population pressure in the dry lands, and with
concentrated development in the urban centers, a large movement of

population to the cities has become very important.

The problem of population pressure has been particularly acute in
Rwanda, with its long history of concentrated settlements and restricted
land area. Although rare in Africa, the high densities of Rwanda are
comparable to those in countries elsewhere in the world. Rwanda is,
however, unusual because of its extremely low level of urbanization
(Bureau National de Recensement 1984). High concentrations of people
were recorded by early missionaries and colonialists (de Lacger 1936,
Lemarchand 1970). Rwanda, and, occasionally, its neighbor Burundi, have
had the highest density in Africa since the earliest censuses (Prioul
and Sirven 1981). This feature of Rwanda has been the focus of many
studies, the most important early one being that of Pierre Gourou in

1953. He documented the remarkably close correlation between high

19
population densities and high to mid-level altitudes, which he felt were
linked because of the ample rainfall, absence of diseases, and high soil
fertility of the Highlands. He also noted the intensive and permanent
nature of agriculture, requiring high levels of labor (Gourou 1953,
1971). Since his study, the population has grown at a rapid rate and is
currently growing at 3.6 percent [Office National de la Population
1985]) and the national population density has more than doubled from 77
people/km’ in 1948 to 188 people/kmz in 1978 (367 people/km’ arable).
However, the population is still not spread evenly across the country:
it remains most dense in the Highland region whereas the Eastern savanna
region is comparatively open (for example Ruhengeri has 313 people/km2

and Kibungo only 88 people/km’ ).

This regional differentiation of densities and the rapid increase
in population have led to several carrying capacity studies in Rwanda.
A bench mark report was written by Préfol and Delepierre (1973, 1975)
who were the first to measure available land and agricultural potential
in regions of Rwanda. They estimated that seven of the ten prefectures
would reach their carrying capacity by the 19805 and that the remaining
three, the Eastern prefectures, by the 19905. This was followed by a
study using a simulation model which predicted similar results but
postponed the breaching of the carrying capacity by about 15 years
(Albert, Crener, and Gagnon 1982). The Futures Group (1981) considered
the country as a whole and predicted the sizes of family farms and food
production per capita using three different population growth rate

projections. All three projections had dismal predictions: the country

20
would soon lose self-sufficiency in food production and the tiny farms
would be unable to support families. The growth of the population has
been a major worry of the government with companion concerns such as how
to supply education and employment (Mukamanzi 1985). 0ft-mentioned
solutions are family planning programs and emigration (Sibomana 1983),
but the willingness of neighboring governments to continually absorb

Rwandan population is unknown.

Although the carrying capacity estimates have been breached in the
highest density areas, the population continues to grow and densities
continue to augment. The responses of the papulation have been much as
Boserup and the other authors would have predicted. The initial and
perhaps most important response has been adoption of intensification
techniques allowing the growing population to be absorbed in
traditionally the most densely populated areas (Prioul 1976). The use
of techniques such as manuring, double cropping, and terracing has a
long history in Rwanda (Lemarchand 1970), and more recently triple- and
inter-cropping have become important (Delepierre 1985). A new response
is the use of nearly all land for permanent cultivation. Fallow periods
and land allotted to fallow have significantly decreased, and the
grazing which occurred on fallow land has become restricted, reducing
the supply of goats and cattle. Similarly, very steep slopes once
reserved for grazing or woodlots are now being cultivated (Prioul 1976).
Short distance migration up the mountainsides have resulted in
widespread deforestation, and migration down to the valleys have

resulted in drainage and cultivation of former swamp land (Gatera 1980,

21
Cambrezy 1981, Prioul 1981). In general, all the micro-variations in
the physical environment are now being exploited in the higher density
areas (Rossi 1984), with a proliferation of new homes adjoining older
households to form small villages (Prioul 1981). The ability of these
areas to absorb additional population increases depends upon increasing
agricultural production, perhaps by draining more swamp land or
developing higher-yielding techniques, or upon expanding non-farm income

earning activities.

Non-agricultural adaptations have already become very important.
The traditionally dispersed settlements now have new homes surrounding
them and groups of non-agricultural buildings are found even in the
rural areas. These new settlements have been termed "bourgs'I (small
market villages) by Prioul (1976, 1981). Half of all farm households
engage in some form of off-farm employment, especially those households
with the smallest landholdings. Although 31 percent of this off-farm
work is in the agricultural sector, usually performed by members of
households with the smallest landholdings as agricultural laborers, 69
percent of the off-farm work is outside of the agricultural sector in
areas such as handicraft production and commerce (Clay, Kayitsinga, and
Kampayana 1989). Short-term labor migration is also common, especially
from the Highlands region which tends to export labor to the East (Clay,
Kayitsinga, and Kampayana 1989). These alternative sources of income
are widespread because the landholdings have become too small to provide

sufficient production to meet the needs of families. Another

22
illustration of this phenomena is that 73 percent of families are net

buyers of beans , the most important protein source (Loveridge 1988).

One of the most visible "symptoms" of population stress, to use
the term of Grigg (1980b), has been the decline in family farm sizes,
from 2.7 hectares in 1965 (Nwafor 1979) to 1.2 hectares in 1984 (SESA
1984). By the year 2000, farms are expected to be only 0.71 hectares
(Delepierre 1985). These sizes vary significantly by region: in
Ruhengeri, a prefecture with one of the highest densities, the average
farm size is only 0.77 hectares, but in Kibungo in the East, farms are
still 1.95 hectares (SESA and ASPAP 1988). A second symptom of
population stress which has become increasingly evident is fragmentation
of landholdings. Farmers in the higher density regions, despite their
small farms, operate 5 to 6 parcels farms, roughly 55 percent more than
farmers in the low-density regions (Clay and Magnani 1987). Another
symptom is that land tenure disputes are becoming increasingly common
(Reintsma 1981) and sale and purchase of land increasingly important in
the higher-density areas where land gained by inheritance can no longer
meet the needs of the family (Clay and Magnani 1987). The number of
land renters has increased, creating a class of renters dependent on the
good will of owners (Contant 1982). Finally, a shift in crop production
has occurred in the denser areas to increasing production of tubers such
as cassava, sweet and white potatoes (Clay and Magnani 1987) which are
more tolerant of poor or declining soils and produce more calories per

land area than higher quality protein crops such as beans or sorghum.

23

As these signs of stress indicate, people in the densest regions
have adopted new agricultural techniques, land-use systems, and non-farm
strategies to cope with the increasing population pressure locally, but
symptoms of stress are becoming increasingly apparent. Like the Kenyans
and Ugandans mentioned earlier, Rwandans have also responded by
migrating out of these higher density areas to drier regions which
traditionally had low densities and were previously the domain of
pastoralists. In the case of Rwanda, this land was found in the low-
altitude eastern part of the country--drier savanna land which before
independence had been reserved as grazing land for the Tutsi ruling
class and was plagued with trypanosomiasis and malaria (Silvestre 1974,
Prioul 1976). After independence, this land came under control of the
new Hutu-dominated government and the old restrictions against
cultivation of pasture lands were lifted (Reintsma 1981, Prioul and
Sirven 1981). Also, a campaign to eradicate tsetse flies was successful
(Prioul 1981). Indeed, the government began to organize and promote
relatively large settlement schemes, the paysannats, in these areas to
relieve the population pressure in the increasingly dense Highlands
region. Large paysannats were opened in the savanna area along the
Akanyaru River in eastern Butare and Gitarama, in southern and eastern
Kigali, and in Kibungo and Byumba (Prioul 1981, Silvestre 1974). All
the paysannats were organized along similar lines: each family received
about 2 hectares of land and agreed to produce a commercial crop such as
coffee or tea (Berry et al 1982). In total, about 80,000 families were
settled in paysahnats (Rwanda 1985), but this represents only about 7
percent of the population. The paysannats were able to absorb only 6.2

24
percent of the population growth during the peak of paysannat activity
between 1957 and 1976 (Cambrezy 1981). Spontaneous migration to these
"pioneer fronts" soon overwhelmed paysannat settlement as a stream of
population left the densely populated areas and headed east to occupy

new lands (Prioul and Sirven 1981).

This expansion of cultivated land was accredited with maintaining
agricultural production increases. Between 1966 to 1983, the area
harvested increased by 3.7 percent and production by 4.3 percent
annually (Delepierre 1985), but these increases have varied tremendously
by region. Meach (1986) compared changes in production, area
cultivated, and population sizes between 1978 and 1982 and found that
prefectures in the Highlands were experiencing increasing intensity of
agricultural production; more and more people were cultivating the land,
and yields per unit of land in most prefectures were still responding
somewhat to these increases in labor. Per capita productivity, on the
other hand, was declining throughout the Highlands, and in Gikongoro and
Kibuye the land productivity significantly worsened signifying land
degradation. Further east in the foothills and denser areas of the
savanna (Gitarama and Kigali), productivity of both land and labor was
still increasing but in general was not very high. The prefectures in
the far East, Kibungo and Byumba, had large increases in the amount of
land going into production but the new land was less productive than
land that had previously been put into cultivation. This would
indicate, therefore, that with no change in agricultural technology,

productivity of labor will continue to decline as increasing numbers of

25
people cultivate the same quantity of land. The implication is that the
people, especially in the Highlands, are experiencing declining income
and are presumably seeking ways to supplement their income with off- or
non-farm employment, or are considering out-migration, following the

path of many others.

The size of this movement of people was estimated by Cambrezy
(1984), who compared regional differences in population growth rates
(see Figure 4). According to his analysis of population growth between
1948 and 1978, the areas of southern and eastern Kigali prefecture,
along the Akanyaru River in Butare and Gitarama, and parts of Kibungo
have had higher population growth rates than the national average of
3.16, suggesting that they were regions of in-migration. The Highland
prefectures, from Butare and Gikongoro north to Ruhengeri, Gisenyi, and
northern Kigali, he identified as net out-migration regions since their
population growth was less than the national average. Some strips of
land along the high-altitude forest, however, had been newly cleared and
settled. Cambrezy compared these varying regional growth rates to
physical characteristics (altitude, slope, and hydrology) and to
population densities and surmised that people in the high density areas
had first cleared nearby forest land and cultivated steep slopes before
drainage of swamps became widespread. Only afterwards, he maintained,
did longer distance movements to lower altitude, lower density regions
occur. His conclusions are similar to those by other authors, who
presume that population pressures and especially the lack of available

land in the Highlands has been behind the general movement to the East

26
(Prioul and Sirven 1981, Gotanegre, Sirven and Prioul 1974) or to Zaire,
Kenya, or Uganda (Nwafor 1979, Gapyisi 1980). No author has yet
followed actual movements by individuals over time to identify the

spatial pattern or size of these movements.

 

 

Source: Cubrezy, Le Surfllement en Guestion, 1984, attachment.

Figure 4. Annual Population Growth 1948 to 1978

From the above discussion, some research questions about the
forces underlying migration and the pattern of movements over time can
be made for Rwanda. One factor that appears to be very important is the
influence of the increases in population density, especially in the
Highlands area. From Boserup and Grigg (see Figure 3), increasing
agricultural intensification could be expected in the Highlands, and

indeed in Rwanda this has been the primary and perhaps most important

27
response. Other expected responses would be an increase in non-farm
income generation, which is also occurring in Rwanda, and a decrease in
fertility, of which little sign is yet apparent (Ministere du Plan
1981). The fourth expected response would be out-migration, which many
authors on Rwanda have assumed to be important. A question to pose,
therefore, is how population pressure is linked to out-migration and in-
migration over time. A hypothesis could therefore be made that regions
experiencing increasing population pressure would simultaneously
experience rising rates of out-migration, and that those regions with
the highest population pressure would experience the highest out-
migration rate. In contrast, areas with low population pressure would
then be expected to experience rising rates of in-migration. Once the
populations of these lower-density areas have reached a certain level
and the densities there become relatively high, they would cease to

attract new in-migrants.

Kenya, Uganda, and Rwanda have demonstrated changes in land use
intensification and migration activity in response to increasing
population densities. All three have been affected by the ten factors
that Bilsborrow (1987) has identified that determine the response of a
society to population growth:

the existing level of living

the availability of untapped, potentially cultivable land
the availability of off-farm rural employment opportunities
the availability of urban employment

the potential for labor-intensive, land-saving technological
changes

the existing crop structure, and its capacity for change
the existing size of the rural p0pulation relative to the
urban population

. the prevailing level of rural fertility and the strength of

@QOU’U

SRO-h

28
factors maintaining its high level

i. the existing size of landholdings and their distribution

j. the institutional structure.
These factors help determine both how societies adjust in their present
location, such as their ability to adopt intensive agricultural
technologies or to earn a non-farm income, as well as the pattern of
out-migration flows, such as the availability of untapped land or of
urban employment. Therefore, the relationship of population density and

migration is important but must be considered in light of other

socioeconomic factors.

C. Other Socioeconomic Factors, Urbanization

 

Other socioeconomic factors are seen as so important by some
authors that they dismiss the population pressure hypotheses. For
example, Lappé and Collins (1978) write that in Africa it is
'colonialism’s cash crops and their continuing legacy, not the pressure
of its population" that is destroying soil resources, and the imposition
of national borders, taxes, and the drilling of wells that have led to
overgrazing. Skinner (1985) reviews historical determinants of
migration in Africa, in particular colonial policies which included
forced labor, forced migration, and taxes, none of which are tied to
population pressure. Since independence, development projects have
often had intentional or unintentional effects on migration (Clarke,
Khogali, and Kosinski 1985). One of the most important factors in the

population-land resource relationship is the distribution of land, and

29
it may be a particularly potent force in explaining migration patterns
(Connell et al 1976, Durham 1979). The social or political structure
may preserve large tracts of land for a small percentage of the
population; this land alienation policy was, for example, firmly

entrenched in Kenya before independence (Campbell 1981).

Indeed, the population pressure dimension of out-migration is
certainly not the sole force affecting population movements in Africa.
The structural transformation process changes the socioeconomic and
personal environment of the potential migrant (Goldscheider 1971), and
new migration patterns evolve reflecting those changes. Many theories
of ”development” have been advanced in which the agricultural and
industrial sectors are on the opposite ends of the modernization pole.
For example, w. Arthur Lewis saw the rural sector as an unlimited
reservoir of labor available at subsistence wages useful for driving
urban industrialization (Lewis 1954). This theory was broadened to
include trade between the sectors and recognition that the agricultural
sector needs to grow so that growth of the industrial sector would not
be choked off (Ranis and Fei 1961). Later, the concept of linkages
between the sectors was developed in which the growth of the
agricultural sector stimulated spin-off growth in the industrial sector

and vice versa (Mellor and Lele 1973).

In general, dual-economy perspectives view third world societies
as consisting of a dynamic, innovative modern sector characterized by

full employment, and a stagnant, declining traditional sector

30
characterized by unemployment. Migration is seen as an equilibrating
mechanism which, through transfer of labor from labor surplus to deficit
sectors, eventually brings about wage equality between the two sectors.
In spatial terms, these have been understood to be the rural-urban
sectors which have been expressed by the core-periphery or heartland-

hinterland model (Brown and Sanders 1981).

In the economic development process, societies eventually
experience a "centripetal stage, with increasing centralization and
primacy” (Pryor 1975) which drives people from the periphery to the
center. Indeed, it is often accepted that an inevitable part of
industrialization is urbanization due to the real or perceived economic
incentives in the urban areas drawing rural migrants. This process has
often been exacerbated by allocations and investments concentrated in
the urban areas by governments and foreign capitalists--the "urban bias"
leading to regional disparities (Lipton 1977, Todaro 1981, Shrestha
1988). Mentioned here is one classical view of the rural-urban
relationship used as a matter of convenience not endorsement; similarly
the concept of "stages" societies pass through is not advocated as
desirable or inevitable (see Simmons 1988, for a discussion of this).
Nevertheless, as Zelinsky (1971) generalized, for many societies there
do seem to be "patterned regularities in the growth of personal mobility
through space-time during recent history, and these regularities
comprise an essential component of the modernization process.“ He saw
the process involving increased rural movements initially and eventually

rapid urbanization. Brown and Sanders (1981) even more explicitly

31
linked industrialization with urbanization. Africa, for example, is the
least urbanized area in the world, but since 1950 it has been going
through the process of urbanization faster than any other continent

(United Nations 1983).

Rwanda has also lately been experiencing this phenomena of
increasing urbanization. Early articles commented on the lack of
attraction that cities held for Rwandans and of their attachment to the
soil. The reasons given for the low level of urbanization included the
isolation of Rwanda from outside influences, the weak transportation
system, the anti-urban policies of the colonialists and missionaries,
and the slow development of commercial activity to supplant households’
near self-sufficiency (Lugan 1976, Sirven 1975). This lack of
attraction of the city meant that Rwanda maintained the lowest level of
urbanization in Africa for many years, estimated at only 3 percent at
late as 1970 (United Nations 1983). Soon after those earlier articles
were written, however, Sirven (1981, 1983) observed that the capital
city, Kigali, had begun to grow rapidly. Kigali grew from a population
of 5,000 at independence in 1963, to 54,200 in 1970, and to 90,000 in
1970 (Prioul and Sirven 1981). The government predicted that the
population would reach 280,000 by 1990 (Rép. Rwandaise, 1977). Kigali
averaged an annual growth rate of 10 percent between 1970 and 1978, but
this varied greatly; the annual growth rate increased from 4.9 percent
in 1971-1972 to 13.1 percent by 1976-1977. In 1977-1978, it was again
around 13 percent. Sirven estimated that with a natural population

growth of 3 percent, in-migration into Kigali between 1977 and 1978 was

32
around 10,000 people annually, representing 7 percent of the country’s
population growth (Sirven 1981). Sirven found that the migrants
preferred to establish themselves in the heart of the city, which was
becoming increasingly densely settled, or as near to their employment as
possible. He also found that half of the city’s population lived in
habitat spontané (squatter settlements) where the government had not
planned or provided for any settlement and where the occupants had no
guarantee of rights of land ownership. Sixty percent of these
inhabitants were renters, which brings up the question of their
permanence in the city, but most were saving money to ”buy" bits of land

even though the transaction is not officially recognized.

The economic base of the livelihood of these migrants and the
economic reason for their coming are less easily stated than the
magnitude of their presence. One important fact is that the level of
economy activity is very high for a third world city: 76 percent of the
active population is employed (Sirven 1981). Almost two-thirds of these
are in the tertiary sector, due to a heavy concentration of
administrative activities in Kigali and the high numbers of people
employed in commerce or as domestic servants. The secondary sector was
growing rapidly when Sirven wrote his dissertation, with import-
substitution industries beginning to multiply and construction
activities intense. Much of the growth in employment, however, was due
to the development of the non-formal sector, which grew to fill an
important gap in the economy by absorbing the incoming migrants (Rép.

Rwandaise 1982). One drawing factor for the migrants was probably the

33
difference in income earned between farmers and traders, since the
latter earned up to three times more (Marijsse 1981). Sirven (1981)
felt that Kigali, after having "vegetated" for many years, had become
the motor of economic development of the entire country and that the
influx of migrants was still a relatively healthy phenomenon despite the
growth of the squatter settlements. He generalized that, in contrast to
what had occurred in Europe, in Rwanda urban growth has preceded
industrialization, leading to a conclusion that the development of

employment has been actually a consequence of urban growth.

The rapid growth of Kigali is unique of the cities of Rwanda; no
other city approaches it in terms of economic or population growth (see
Table 1). The economic growth is due to the location of Kigali at the
center of the country and at the hub of the transportation network.
Commerce is accordingly concentrated and most industries and
manufacturing enterprises have chosen to locate in Kigali. Most of the
government administration and all the foreign diplomatic and aid
missions are also based there (Prioul and Sirven 1981). These
activities support a large population and attract increasing numbers of
migrants. Most of the economic growth that absorbs incoming migrants
is, however, in the informal tertiary sector supporting those in the

rest of the economy (Sirven 1983).

Therefore, theories based on the European process of urbanization,
in which urbanization follows industrialization, are not as applicable

in Rwanda where urbanization itself seems to be a stimulus for growth.

34

Table 1. Population of Urban Districts, 1938 to 1978.

 

Urban District 1938 1949 1959 1970 1978
Kigali 3903 12940 4935 57400 117749
Butare 954 11883 3714 8400 21691
Gisenyi 1122 9221 1323 6251 12436
Nyanza 3193 4601 1093 4640 8587
Cyangugu 925 2290 439 3540 7042
Ruhengeri -- -- 939 12500 16025
Gitarama -— -- 58 9350 8534
Rwamagana -- -- -- 4850 5683
Gikongoro -- -- -- 7020 5654
Kibungo -- -- 2440 3860 4081
Kibuye -- -- 34 1670 2764
Byumba -- -- 64 5480 7078
Note: These figures must be viewed with caution due to the

changing definition of urban district boundaries.

Source: ONAPO, 1982.

 

The contrast between the dynamic, innovative modern urban sector
characterized by full employment and the stagnant, declining traditional
sector characterized by unemployment (Brown and Sanders 1981) does,
however, seem to describe the situation in Rwanda fairly well. This
contrast has undoubtedly been the stimulus of the increasing rural-urban
migration, and as the urban sector continues to develop economically,
the city will attract increasing numbers of migrants. A research
question to ask would be how the migration patterns have evolved over
time in Rwanda, i.e., how the importance of the rural-rural and rural-
urban flows have changed in relation to the evolving economic

opportunities in the two sectors.

35

0. Individual Characteristics

Whether the driving force behind migration is population pressure
or industrialization, in all cases personal and structural
characteristics affect the response to place utility. In other words,
an individual’s personal characteristics and class position help
determine his or her ability and desire to migrate based on feelings of
satisfaction with a place (White and Woods 1980). Given a sedentary
population and inducements to move, typically some residents stay and
others leave. This selectivity has been the basis of much research,
beginning with Ravenstein (1889), and the research has focused primarily
on variables such as age, sex, marital status, education, occupation,
and position in the family life cycle (Shaw 1975). The life cycle
factor explains the reasons for many characteristics found; for example,
young adults often lack control over resources, especially land, and
therefore lack commitments to remain in their area and have more of a
tendency to leave (Gugler 1986, Byerlee 1984). Some are also in the
process of continuing their education, for which they may need to move
to another town. In Rwanda, added to these life cycle factors is the
custom of female exogamy, that at marriage the wife moves to join
husband’s family. The husband, in turn, must have his own farm in order
to get married, but if his parents do not have sufficient land to
bequeath, he is obliged to migrate to search for land or to gain

employment.

36

Most studies of characteristics of migrants have examined rural-
urban migrants. In general, in Africa these migrants are young single
males from large families, with more skills and from better-off
households than non-migrants (Zachariah and Condé 1981, Caldwell 1969).
Perhaps the most important selectivity factor, however, is education
since it ”prepares people for urban occupations" and since returns from
investments in education lie in cities (Byerlee 1974, Caldwell 1969).
This conforms to one of Lee’s hypotheses that migrants exhibit [I
characteristics intermediate between those of the origin and destinat'on
(Lee 1966) since the urban population is much better educated than the
rural. Rural-urban migrants’ movements are motivated by a search for
better-paying, non-agricultural jobs (Findley 1981). On the other hand,
rural-urban migration also occurs among the very poor, often landless,
illiterate, and unskilled. The published examples of landlessness or
near landlessness leading to migration are usually related to unequal
land distribution and often come from Latin America (Jones 1980, Thomas
and Minkel 1984, Durham 1979, Buksmall 1980) but some African studies
have also found landlessness or near landlessness an important factor
for out-migration (Goldschmidt 1986, Monsted and Halji 1978). If
opportunities for agricultural wage labor, for example, exist in another
rural area, these less well-off migrants choose to move to rural areas,
but often they perceive cities as their only economic opportunity
(Findley 1981). The migrants to rural areas have been found to be older
than rural-urban migrants with very little education or training and

include members of the whole family (Masser and Gould 1975, Monsted and

Halji 1978, Connell et al 1977).

37

-Migrants Compared to

Table 2. Individual Characteristics of Rural Out

Rural Non-Migrants

.3. .50. ...0E00.:O 8.331.: .a:.0< m:m..0> .0053... 30:53:on .:0E..o_0>0n. .83... 20:05... “00.58

 

 

3.92.0 :o 3:308

.3330 83.. 3o. 3080.83 o...
Eaton...—

.:o.u£ 088 :o 3.
33.8.: .2330 we v.33 o. 0.52

31.: 00.8 o. 3:30 v8.8...
80:033.. :33 39.

8.3::
30.. 8.5.0.0.. 303 .83 :. :5.

30:250.. 3 3.90.5

8.5030
.6 :3 :33 30.3.50: .00... o...
33.8.... 2353:

3080.6:— .03...
-0.— aefi Eur—3:00 09:851.
6.: 30: 0.. 8 .0002 .2233:

.3: 330:0 00:023. 308.502
80. £028

:5 a...» 0:3» 8.8 o. «5...?

25.9.8. :3 .03 305:3 :3...

3:: 303.505 2:35

553:3»: 3:0... .5 3033.0.—
.ao. gestures. :8. oh
.2328 35:63 .53.. oh
0......

09:93. 0:3» 2 5.6.

-3... .8 e8: .3 3.9:... an.

coo—5-4.0....
30.000. «53.0%

0:00.: .03 8.30.60
5:2... 30:033.: :30: cm...

00:0...0..x0

..... 8.3.3.: ..o 8.325
. .. 8.3.3.:— 3. 8.33.0!

............. gmdazmg
2 8.3.... .o 8:98....

.......... mo—Dg> Umfidm

.......... W on: flmdua E 08:

................... $0:
-033: we: 308:8 :35

 

$0.33: .E:§0.Cu< 3...“: 35:353-:0: 5.3 088 30:32.33 0.3.. 03.. o. .30... .......... 3532233000
30:3: E3. 30.96300:— 068
30:03. 3.82 80:03. .3 30:3: 53. .3Em 30:39.03. 0P3). .......... 3:333: .051.
000» 02.9.... 2:: 30.5.
356 3. 3:5: .55.... :25
2:2»
03.: 29.00.59. 00.3... -.E 0.85... 3.. 00...:3. .53: 0:3... 3.0.00: 2 390.0: .02 ............... 00.... 3.6:".
330:: 3.39.0... 5.3008
00302005.. -08 0:5... 32:0: "335...: 053 .339... 0:: SET... .30. .< ................ 8.30:3.
3:3. 05:0 £020 .. .
00.
4:8 0.60:5: .5. 3.360.. 0.35
2:: :03. «0.00:3.
0.5: 8.3.35: 3...:3. 0.5:“... :.o.. >2: 0.53 :5 0.5: .055» .. £3... :. 8.: 0.020.. .032 ............ 383.: ..o x0m
:02: 088
AE3> +°Mv 5030 A30» chin—v u==°> AW§> “NIH—v N==°> ...................... E
330.30.: 53343.! toot-49:: 3.310.033.59
03.50.00 304 09:00.2. =04 09:00.0. b0>

 

 

38

In general, then, out-migrants from rural areas seem to fall into
two separate economic groups: those from relatively wealthy families
who, due to their higher education, can find no challenging employment
in rural areas and choose urban destinations for "strategic“ reasons;
and those from relatively underprivileged groups with no chance of
upward mobility who must move for "survival” reasons and who prefer
rural destinations but often end up in cities (Shrestha 1988,
Goldschmidt 1986, Gugler 1986, Findley 1981). Gould (1985) found that
in Kenya, however, education became less and less of a differentiating
criterion over time for migrants to urban areas as the rural
opportunities for less educated migrants became less attractive. Table
2 summarizes distinctions between these migrant groups. The oft-used
typology of "push" (poor migrants leaving due to conditions in the
village) and "pull" migrants (better-off migrants lured by city
employment) (Lee 1966, Gould 1985) is not as distinct as often suggested
because both types are being pulled by better economic opportunities at
the destination. Nevertheless, the similar distinction made by Findley
and shown in Table 2 between "very selective" potential migrants (those
with education and from families with larger farms), "moderately
selective" potential migrants, and "less or negatively selective”
potential migrants (landless or farmers with very small farms and little
education) is useful. She compares the groups for willingness and
ability to respond to new rural opportunities instead of migrating, but
they can also be compared in the Rwandan context in terms of their

destination and whether their destination has changed over time.

39

How migrant characteristics change over time has not yet been
examined in other African countries, but in Rwanda it may be
hypothesized that originally less selective migrants did circular
migration or migrated permanently in search of free land, and only
afterwards left for the cities due to the lack of rural employment or
available land. These less selective migrants may have little
education, be from families with small farms, and be young men. Those
who migrated to the rural areas are more likely to be married than those
who went to the cities. The very selective migrants, on the other hand,
can be hypothesized to be young, well educated, and from well-off
families; it can be also hypothesized that they are from all areas of
the country, not just those areas experiencing population pressure, and

that they have always migrated to the urban centers.

E. Gravity Model

Turning from individual characteristics to general migration
patterns, one widely tested model is the gravity model based on Newton’s
laws of gravity (White and Hoods 1980) and related to central place
theory (Olsson 1965). In the model, the movement between two places is
related to their masses (the sizes of their populations) and to the
distance between them. Three constants vary from interaction to

interaction, one reflecting the "friction of distance" (the "human

40
space" including the intervening opportunities between the two
populations [Lee 1966, White and Woods 1980]), another the size of the
potential out-migrant pool, and the third an index of attraction of the
destination (usually estimated by the size of the destination).
Commonly, it estimates the flow of migrants moving to the nearest place

which can satisfy their desires (Shaw 1975).

Although the theory has wide applicability and is generally
accepted as a tool for urban-urban movement analysis in more developed
countries (White and Woods 1980), it has little relevance in the rural,
dispersed Rwandan setting where population pressure is assumed to lead
to out-migration and less densely populated agricultural areas are
assumed to act as attractants. The model can be modified to embody
different indexes of potential out-migration and potential in-migration.
Incorporating the concept of land as economic opportunity and therefore
regarding free land as an "indication of attractiveness“ (Masser and
Gould 1975) or as a measure of the "attractive force“ of a destination
(Haggett, Cliff and Frey 1977), a version of the gravity model could be
adopted in which the lack of population in rural areas would be the
attracting force. In other words, the index of attraction would be low
population densities. Conversely, an index of potential out-migration
could be the population pressure felt in that region, measured by
population density. The traditional gravity model’s "friction of
distance" assumption, that the number of migrants falls with increasing
distance, is relevant and it can be hypothesized that migrants choose

the nearest available land.

41

To summarize, a version of the gravity model to estimate flows
between areas using the variables of high population density related to
out-migration and low density to in-migration, and distance as a
moderating factor, could be tested with the rural-rural migration flows
in Rwanda. It is hypothesized that the destinations chosen by rural-
rural migrants are the closest areas with a sufficiently low density to
be perceived as attractive. With time, therefore, rural-rural flows
would cover increasingly longer distances as former destinations gain

density and become less attractive.

F. Leoor Force Adjustment Model

Returning to the individual level, a purely economic model by
Todaro and others (Todaro 1969, Harris and Todaro 1970, Todaro 1981) has
gained much attention. It is based on the assumption that the primary
motive for migration is economic: the existence of job opportunity and
income differentials between the rural and urban sectors. It attempts
to explain observations in Kenya of increasing rural-urban migration
rates despite the existence of positive marginal productivity in
agriculture and high levels of urban unemployment. The distinguishing
feature of the model is the emphasis on perceived rather than observed
incomes differentials and the inclusion of the probability of obtaining

urban employment (Masser and Gould 1975).

42

The original model has been criticized because it does not
consider non-economic factors such as family ties or economic factors
outside the formal sector (Gugle 1984) such as land for subsistence
production as an economic asset or employment in the informal urban
sector. The comparison of rural and urban incomes is also problematic
and not a precise indicator of labor supply and demand (Lawson and Brown
1987). The model abstracted from certain aspects of the economy without
analyzing the socioeconomic structural factors that generate rural-urban
differentials (Bilsborrow, Oberai and Standing 1984). At the individual
level, it assumes all people possess sufficient human capital to qualify
them for modern sector employment (Cole and Sanders 1985). Perhaps one
of the most important restrictions of the model is that it assumes a
simple two-sector rural-urban economy and ignores regional differences

(Masser and Gould 1975) as well as interaction between rural regions.

In terms of Rwanda, these restrictions and omissions limit the
model’s usefulness but certain concepts embodied in the model are useful
in understanding processes in Rwanda. These concepts are the importance
of perceived economic opportunities and the potential migrant’s
consideration of the probability of gaining employment, since in Rwanda
as in Kenya the urban formal sector absorbs far fewer people than the
high in-migration rate would lead one to suspect. Urban migration could
thus be expected to become increasingly important as perceived economic
opportunities in the rural areas diminish, due to declining farm sizes

and lack of available land, and the perceived opportunities in the

43
cities increase with the establishment of formal and informal sector

industries and other employment possibilities.

44
G. -P l M l

The last model presented here is the push-pull model, included
because it is so general that it illuminates factors other models omit,
to wit rural pull factors. Traditionally, it has been used as a device
to list pull factors of urban areas and push factors of rural areas in
an attempt to explain urbanization (Gugler 1985), but the model in its
more general sense includes determinants of rural-rural, urban-rural,
and urban-urban migration as well. The push-pull model is often
implicitly put into practice with the use of regression equations in
which migration flows are dependent upon various indexes of comparative

attractiveness of regions (e.g., Masser and Gould 1975).

Rural push and pull factors were very apparent in the two examples
given above of Kenya and Uganda, in which population pressure in certain
regions led to migration to other less densely populated rural regions.
In those cases, as in Rwanda, the rural pull factor is the availability
of land: the migrants gain an economic asset, land, that can sustain
them in old age and support their children (Gugler 1986). Economic
opportunities, whether in the form of wage rates or availability of
land, guide both rural-urban and rural-rural migration, rendering them
similar in implementation of theoretical models. However, empirical
models written for urbanization do not describe well rural-based
migrations (Brown and Lawson 1985), a pattern has been very important in
Africa. For example, in Nigeria, the predominant form of migration was

from the coast towards underdeveloped land in the Western and Northern

45
Regions (Amin 1974). Similarly, Riddell and Harvey (1972) did not find
conclusive evidence of step-wise migration to urban centers in Sierra
Leone because of the attraction, at times predominant, of rural
opportunities. They note that in many parts of Africa these so-called
deviations from the urbanization pattern may in fact be typical and
dominate the landscape. According to Brown and Lawson (1985), most
rural-directed research that has occurred has tended to focus on narrow
incentives such as land colonization schemes, irrigation, and extension
programs. They conclude that "conventional wisdom that rural-to-urban
flows dominate is not universally applicable" and that urban attraction

is not as important as was traditionally believed.

The importance of rural-directed migration, however, appears to
change over time with development, as predicted by Riddell and Harvey
(1972). Brown and Lawson (1985) cite a study in Ecuador in which 92
percent of highland out-migrants chose rural areas in the 1920 to 1950
period, a figure that gradually declined to 48 percent between 1974 and
1976. Zelinsky (1971) also describes the peaking of colonization
”frontier" movements earlier than the peaking of rural-urban movements.
From this, it can be hypothesized that in Rwanda the pull of the rural
areas may have been originally very important but declined as urban pull

factors became stronger.

mm

This returns us to the previous discussion of the process of
development and its effect on migration patterns, including
urbanization. Several authors have examined how migration patterns
change over time with development; two hypotheses of these authors will
be discussed here, that of Zelinsky’s “hypothesis of mobility
transition" (1971) and Brown and Sanders’ ”development paradigm of

migration” (1981).

Zelinsky reviewed common migration patterns that are seen with
modernization processes and linked them with the demographic transition
theory. In short, he generalized that as a community experiences the
modernization forces that lead to changes in the population as described
by the demographic transition theory, the community simultaneously
passes through phases of mobility transition (see Table 3). During
Phase II, rapid growth in rural population and other associated changes
impel people to adopt a more intensive mode of production or, more
effectively, to out-migrate to form "one great rural stampede." The
rural settlement frontiers at first draw more people and peak earlier
than the urban areas, but this is a the period of great urban growth.
Phase III is much more complex. Indeed, he allows for the possibility
of some communities not reaching this phase of fertility decline and
having a "demographic relapse," but in general the metropolises act as
outposts of modernization and major recipients of migrants as the flow

to the rural frontiers lessens.

Table 3. The Hypothesis of

47
the Mobility Transition

 

Theviteltm

The mobility trenettion

 

l'heeeA: "Wraith-elixir”

(i) Amodeeetelyhightoouite high fertilitypettemthet
tendetofluctueteonly el'qhtly

(a) Mortelity at nearly the eeme level ee fertility on the
everue.butfluctuetingmuchmorefromyeertoyeer
(1) Little. ifeny. long-range neturel increeee ordecreeee

Pheee I: The early teeitee'tienel eeriety

(I) Slight. but eignificent. rice in fertility. which then
remeine feirly conetent et e high level

(8) Mid decline in mortelity

(J) A reletively tepid rete of natural increeee. end thue e
meior growth in eiee of population

Pheee C : The fete mac-d eeriety

(I) A mejor decline in fertility. initially tether alight end
elow. letet quite repid. until enothet elowdown occure ee
fertility epproechee mortality level

(a) A continuing. but elechening. decline in mortelity
(J) A eignificent. but decelerating. neturel increeee. et
reteewellbelowthoeeobeervedduringPheeeB

Pheee D: The duel-ted reciety

(I) The decline in fertility hee termineted. and e eocielly
controlled fertility oecilletee tether unpredtctebly et low to
moderete levele

(3) Mortality ie etebilieed et levele neer or elightly below
fertility with little yeer-to-yeer Wlity

(3) There ie either e light to moderete rate of neturel
increeee or none et ell

PM. E: A lute-er rupee-doomed eeeiety

(I) No pbueible predictione of fertility behaviour ere
eveileble. but it ie likely that birthe will be more carefully
controlled by individuele-end perhepe by new eocio-
politicel meene

(a) A eteble mortelity pettem elightly below preeent levele
eeeme likely. unleee orgenic dieeeeee ere controlled end
lifeepen ie greetly extended

Pheee l: The pee-Mere teedtiewel lonely -

(I) Little genuine reeidentiel migretion end only euch
limited circulation ee ie eenctioned by cuetomery practice
tn lend utilieetion. eociel vieite. commerce. werfere. or
religioue obeervencee

Pheee II: The eerly tremitienel eeriety

(I) Meeeive movement from countryeide to cities. old end
new

(a) Significent movement of rurel folk to colonieetton
frontiere. if lend euttehle for pioneering te eveileble within
country

( 3) Mayor outfiowe of emmrente to eveileble end ettrective
foreign deetinetione

(a) Under certain circumetencee. e emell. but eignificent.
trnmigretion of chilled workere. techniciene. end profee-
eionele from more edvenced perte of the world

(5) Significant growth in verioue hinde of circulation

Pheee Ill: The fete (re-attend ready

(I) Slechentng. but etill mmr. movement from country-
eide to city

(a) Leeeening flow of migrente to colonieetion frontiere
(J) Emigretion on the decline or may have ceeeed
altogether

(e) Further increeeee in circulation. with growth in etruc-

turel complexity

Pheee IV: The educated reet'ety

(I) Reeidentiel mobility hee levelled of! end oecilletee et
e high level

(1) Movement from countryeide to city continuee but to
further reduced in ebeolute end relative terme

(J) Vigoroue movement of migrente from city to city end
within individual urben egglomeretione

(4) If e eettlernent frontier hee pereieted. It ie now etegnent
or ectuelly ntan

(5) Significant net immrgretion of unehilled end eemi-
ehilled workere from reletively underdeveloped We

(6) There may be e eignificent internetionel migretion or
circulation of eltilled end profeeeionel pereone. but direc-
tion and volume of llow depend on epecihc conditione
(7) Vigoroue accelerating circulation. peniculerly the eco-
nomrc end pleeeure-orrenteted. but other veriettce ee well

Pheee V: Allure leper-advanced rainy

(I) Theremeybeedeclinelnlevelofteeidentielmigretion
end e deceleretion in come forme of circulation ee better
communication end delivery eyeteme ere Inetituted

(a) Neerly ell tee-dental migretton my be of the inter-
ur‘ben end tntreurben veriety

(J) Some further immigration of relatively unehilled
lebourfromleeedevelopedereeeiepoeeible

(4) Further ecceleretion in eome current forme of ctr-
culetion end perhepe the inception of new forme

(g) Strict political control of internal ee well ee inter-
netionel movemente mey be impoeed

 

(e)Thephmeeofthenubilnytrmeitionmdtheirrelethiehiptothephueeofthevitdnmeition

 

 

international

     

l
IVV

   

l

I ll Ill
Phase

1

Frontiervvard

 

III III IV V

 

   

Phase

Rural—urban
\

l l I);
I II III IV V

Phase
Urban - urban and ..
intra-urban

  

I II III IV V

Circulation I -

1 I L l
I II III IV v
Phase

Potential migration
absorbed by circulation

 

I II III
Pheec

IVV

Source: Zelinsky, 'I‘hc Hypothesis of the Mobility Transition,“ Gcmohical Review 61, 230-233.

 

48

Many authors have commented on Zelinsky’s hypothesis and lauded
him for providing a holistic perspective incorporating temporal changes
(Pryor 1982). But, a common complaint is that it suggests a unilinear
evolution valid across both time and space--that the Western experience
will be that of everyone (Pryor 1982; Bilsborrow, Oberai, and Standing
1984; Ogden 1984). This should not, however, restrict construction of
models of processes identifying the changing composition of mobility
(Bilsborrow, Oberai, and Standing 1984). Similarly, it is criticized as
being very simplistic and paying too little attention to the factors
explaining changing mobility and how they may be affected by cultural
differences, and too little attention to migrant characteristics (Ogden,

1984, Pryor 1982).

Nasser and Gould (1975) compared the hypothesis of mobility
transition with the situation they found in Uganda (then in the Phase
II) and concluded that in terms of internal migration the hypothesis
somewhat fit Ugandan reality of rapid rural-urban movements but that the
hypothesis fit especially well for the large amount of rural frontier
migration. However, the actual pattern was much more complex than
Zelinsky’s hypothesis with considerable migration into well-established
rural areas and with large variations between regions observed. The
external migration was also more complex; Uganda as a more advanced
country was admitting large numbers of immigrants from Rwanda and other

neighboring countries.

49

Table 4. The Shifting Role of Migration With Development

Ame ..ww=...om 2.53 2.5. 0. 3.8.5.0.". 3.3.2.... 5.3 62.9.3.2 ..o 992...!— 305.3330 a ”9.2.6.7 .8038 28 gm ”08:8

 

.26..- o... 3:22. 2:2» £9... 32;

88. .e 20.93 c.
.28. :80 .c

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

no. .39.. :82...

3:23:88:
22.. on... on... 0E: 0:... cc... 0:...
. . . 0 383:. 2063.262
i/ \. / \l «a 5:22.: 2. 0.0.
{28. :80 .c 9.8..
5.5.52. 9.. .0 5.3.2.2. o... .o 3.98 E89:
.2858 :25 :8ch =2: 2.. 2:. 3.2.2.2.
momma... o >.cc 953.... o >.co 9.38.... 832“. .0905 .8.- o... .c 332“.
:35 o. 3.99 P326. 9.8. 29.5. ace: .3an .0 >55 .coo 2.0598 .399 30.8» 3:3.
:35 ..o 3. =5 88...“. .2. «88.0 =- 3. .593. co 8 =25 .1 3. cooco>u< .2332.
38.25.”... .38. :o...o3 .393 ..o...o.s 2.8.3.55 .852 688.0 a >.co Eeu...ca.m
8e. 2.. 3. 8o. 2.. 8. .53a .3 co. 38.2
.5228...” 28:22: 28:2...5
53.2. 5...»: ”832 8.8.8 3...»: .382 83.2
5956. 8829 .28“ 3o. 2.. 3. .28: .28. a8. 2.. .o. .28. .38: c023.
.59: ca .o :23! o» 29: :5 =0....§> =0...o.s cu EOE .3 :02?» =0...o.s .528» .3226...
85925 20:. o... 3. 88a... .28» 3o. 80:. 88a... .28: no. EOE 3:02.32. £332
228 :5 33:29“ ..o..o. o... 8. o... .0. ..o. cc. 2.. .6. on. .0. 8a.. 268 85...
2 ”My.“ .2528 28: 2...: 28:22: 28:22:. 28:22.0. 28:22". 285.25
5.2.2.2. 8.2.2.2. 5.5.2.2. “88 u 8.2%..de 5:2.
2.. .c .83.“. ”88.9 2.. .o 2.. .o . >363 . a: 3 .
co 3 . . . . cc.
a“: .2053. 3.8: 1.8. .2059. .559. .399 .2929: .25.. 3:2. \ 2330.
.2: :25 ..o 5. ..o 3. =95 =25 z. .0. :25 .>..ow 26:.
c o 28 28:22.... 28:22: a :5 . 25 28:22: . 220 2.3
38.3 38.3 38.3 38.3
383 8.5.85 8 to
28 o ”:2. 0.33.2.» 5.03 5.8... 259. on m
.. 358.2. c Boos. 3.5.3 cc- 2320
20:23:. .252... co: 88... 52.9 52. 2.5:.- \E.e.:oc .23. .5 c253 3 .5 U258 3
.0 Ceca—at 63:35:50 60229: 00. 88: .220 6:0
.223... . cozauacw 23.5.2.6 3.5.22.8

 

 

 

 

cat-3.5 .o 80.8”.

 

88.3 3253.260

 

 

 

50

Ten years after Zelinsky’s hypothesis was published, Brown and
Sanders (1981) produced their “development paradigm of migration."
Their theoretical framework is based on the processes that underlie
migration, in particular the individual decision-making model of Brown
and Moore (1970) and Mobogunje’s (1970) systems approach, and they
identify structural characteristics of society that they feel change
with development and influence migration patterns (see Table 5). The
result is structured much like Zelinsky’s hypothesis with some additions
and deletions; new is inclusion of push and pull variables, migration
chain effects, and wage and job opportunity differentials. The major
difference, though, is that instead of making an analogy to the
demographic transition phases, Brown and Sanders consider phases of
"development” (defined as industrialization); they also restrict their
migration patterns to rural-urban movements (ironic since Brown later
wrote on the neglected phenomenon of rural-destined migration [Brown and
Lawson 1985]). Similarly, their wage and job opportunity differentials
do not include agricultural earnings and their push variables are all
rural-based whereas their pull variables are all urban-based.
Nevertheless, the inclusion of the change over time of opportunity
differentials and push-pull forces is useful. Many of the same
criticisms of Zelinsky’s hypothesis, namely that the definition of
phases assumes that all regions will follow a pattern experienced by the
West, can be made about Brown and Sanders’ paradigm. This should not

preclude testing of the general framework, which has not yet been done.

51

Table 5. Changes in Factors of Migration with Development

.28.. 5.28 .2. .52.. .2. :5. 2.2.3 22. 8:28,. 82:8

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

cemueulus uo eucaoeh

 

aceuuoaeu nun» 3-: .umclun Aueouaiz
.umcwwe euuuw~ eeneeuo
zacwaoot acolo>oa :ua .ucecueunuaim -e nan veuoeuue auowoom
uceum non ac and can vouuouue can «a vooce>u<
82.0...2u.n2. :-: -:.cu.uc.u:-: a ...a. ...=u.. 3.. u .....
acoeo>oe 3-x oemnmaa noune~o -o can uuo uuo "no you
"~- scan "aim cocoon .ano umcwanusiz -oz iuouuon uuoi-os nowumcnu abouoom
acoflo>08 .vo ofiuu«~.oiuou .oonneno «no new aceua noon uou now one" now iuonao nonoMuweceuh
mum maficoxoean dance scan "at: .uwclan ouannoua iMmethauziz .uucumn .umcumn .uwcumn nosey oaoa
cocoauoaeu a. con acne»
lawmaouocm >~3o~u 2-x .coimnoz ”2-: >uwooaoo uuoiuouuon ecu uuo uuo «no: zuowoom
comma-nu nemauueo nouo:\n90a"=-¢ uuo "do: uouuon noon couooa ~e:o.:«n=onh
teen quucOuu 0: at: nine" .naueu logo-en menu uuoa noon non you uou uceo ecu outta:
ooceunmv neon goal aaeflo sou: "mix "ousnoouo .noo v:e~ 5:. ”an: .umu.n .umcuan lunacumn .uuclun
Avono~o>ev do: :0: com «o oceano-
nowuwo. :-¢ o.uam. :oe ucaoa humoueom mun-n ecu nommono haemoom
.uoi-o= ”at: one. "coca noun:\onon ”aim unmoon , can no u ~ocomuwncouh
Inmxeoe one" can mix nsueu condone ”ens-noun oce~ and nag can go Ame-o non acne” unuem
ooceanwu uuosn o>mnwoe Boom ”mum sewuenaooa cuneameeeuaia .umcume u uncle Dawson. necteun
asuonom ea a non no acoeuoe co.»
nconeou nucouumlncoc nuancoucm on: nae-n shoe uou unmsoon
«newuuuoo no nemoon vce nuceuual use“ unannouocu none:\n50n"=i¢ aceom no u acne. aces“ nuouuom
«on acole>oa oIoe :ooauon cauuenu ”oneeuocn teen iu.cu«nc. queen numcuwncm reovoun «enamuuveuh
.co«:e~l«fl o~uuw~ icououuwv onuuwa ceauensnoo enneuueeeuznm >ue>
unaccucu deacon anemicoz fluem
canueuuuz uo nowuomuouoeucgo aucunu. coon: Hausa
:uouuom ~euuoam wanna“: sham aqua uneaucououumo huucsuuonoo o«locoou
enough
uGOIQOAODoo

 

 

 

 

52

Aspects of the process of change over time incorporated in both
Zelinsky’s and Brown and Sanders’ works are relevant in Rwanda. Their
works, however, lack explicit reference to certain factors of migration
very important in Rwanda. Therefore, from the foregoing analysis of
population pressure, rural frontier movements, urbanization, and other
components of migration in Rwanda, and incorporating facets from the
migration models, Table 4 was created within the Zelinsky/Brown and
Sanders framework. This table contains additions to their works,
including farm opportunities in the economic opportunity differential
columns, rural pull factors, and migrant characteristics. The columns
of economic opportunity differentials present the growth of non-farm
activities in rural areas and of formal and informal sector
opportunities in the cities while farming opportunities decline, and the
columns of push-pull factors reflect the strengthening push of
population pressure in some and the draw of new land in other rural
areas with the simultaneous growth of urban opportunities. The two
right-hand columns, describing the changes in migrant characteristics
and in spatial patterns of migration flows, summarize the research

questions that will be examined in this study in the Rwandan context.

The hypothesized changes of migrant characteristics are drawn from
the studies discussed above which indicate rural-rural migrants tend to
be similar to non-migrants or to be from smaller farms. Rural-urban
migrants may be initially better educated, but they later include less
well-educated people from poorer families. The spatial patterns

proposed are very similar to Zelinsky’s, with rural-rural flows being

53
important early and rural—urban migration rapidly increasing later.
Rural-rural flows are from more densely populated regions to less
densely populated areas, reflecting the "push” factor of population
pressure and the "pull” factor of available land. The urban-destined
flows, on the other hand, are not as affected by the rural push factors.
An intermediate stage, middle transitional, was added to better observe
the evolving pattern. Rwanda appears to be currently about in this
middle transitional phase. The period before independence corresponds
approximately to the traditional phase and the 1962 to 1972 period to

the early transitional phase.

I. N o Pr

This study will investigate many of the theories and assumptions
of migration and development literature that have not yet been
explicitly examined with actual migration flows. One of these is the
population pressure/out—migration hypothesis that has been accepted as a
general rule but has not yet been systematically tested to determine the
importance of population pressure in explaining out-migration rates.
This study tests this relationship, and compares how it has changed over
time as other factors developed. The other side of the equation is the
allure of areas that are not experiencing population pressure, those
that have the attraction of available land. The ability of available
land as an economic opportunity to attract in-migration has also not

been tested. This study examines how strongly rural areas with

54
available land draw in-migrants. It also looks at how the in-migration
rates change over time with increases in population pressure in those
receiving areas and with the development of alternative economic
opportunities in the urban sector. A modified version of the gravity
model analyzing the power of these origin and destination attributes and
containing the factor of distance decay to explain rural migration flows
is tested. Another question not yet explicitly examined in an African
country is the changes in characteristics of rural and urban migrants
over time. In general, then, this study is unique in its comparison of
rural-rural and rural-urban migration flows over time, and in its
examination of the socioeconomic factors behind these spatial patterns

in a developing country.

CHAPTER III
M T ODOLDGY

To examine spatial migration patterns and individual
characteristics of migrant and non-migrant groups in Rwanda, information
acquired from several data sources was analyzed with SPSS PC Plus
Version 3. From these analyses, various maps, graphs, and tables were
drawn up to illustrate patterns. This chapter provides a general
description of the data sources and the procedures used to handle the

data in order to illustrate and test the significance of these patterns.

The relationship between population densities and migration
patterns was examined using population and density data obtained from
the Rwandan Government surveys of 1948 and 1970 (source: Prioul and
Sirven 1981), the census of 1978 (source: Bureau National de
Recensement 1984) and an analysis of data from the 1981 Rwandan

Demographic Survey (source: Min. du Plan 1981).

The 1981 Demographic Survey data was gathered following the 1978
census by the Ministry of Planning to estimate the effectiveness of the
1978 census and to provide additional information on factors concerning
the country’s birthrate (Min. du Plan 1987). As such, several
additional questions were included which made this study of migration
patterns possible, although migration was not the primary objective of

the survey. The answers from a nationwide sample of 100,000 people out

543

55
of the population of 5.3 million were recorded, and the weights used by
the Ministry for the sample to represent the population are also used in
this study.’ Especially important for migration analysis, the survey
included questions on birthplace, current residence, and year of move to
the current residence. The question on location of birth allowed an
answer of within the secteur, commune, or prefecture of current
residence, or if outside the prefecture of current residence, the
prefecture or country in which the respondent was born. The current
residence information was also at the secteur, commune, and prefecture
levels. The answers to these questions were used to draw up an inter-
prefecture migration matrix and to separate rural- and urban-destined
migrants (similar to the Masser and Gould 1975 study). Such a mobility
matrix does not include intervening moves by the migrant, so it may
underestimate actual total movement. Also, of course, the survey does
not contain those that emigrated from the country or those that had

died, so again movements may be underestimated.

For the purposes of this study, the definition of a "migrant” is a
person 15 years old or older who has been living outside his or her
prefecture of birth for one year or longer. Since this study

concentrated primarily on economic factors of migration, in most cases

 

’ In some statistical analyses, non-weighted data was used. Little
difference between weighted and non-weighted data was found in the
results. For example, the mean years of education received for long-
distance rural migrants in 1977-80 was 2.85 for the unweighted sample
and 2.78 for the weighted sample; the difference between them was not
significant, nor was any difference found in the results of t-tests
comparing migrant groups and/or non-migrants.

55
only men were examined. As illustrated in the upcoming section on
gender-specific migration, most women migrate due to other factors. The
bulk of the analysis concerns, therefore, the 36,526 adult males who
were surveyed, of whom 4,271 had migrated out of the prefecture in which
they were born. The 4990 women migrants are also included in some of
the analyses, in particular in projections of future migration and in

the section on gender-specific migration patterns.

From the survey question determining the year each person moved to
his or her current residence, categories of time periods were made (as
suggested by Tabah 1970). The migration maps and other analyses are
based on five time periods: from 1945 to 1961 (pre-independence), from
1962 to 1971 (accelerating rural-rural migration), from 1972 to 1976
(transition from primarily rural-rural to rural-urban migration), and
from 1977 to 1980 (rise in importance of rural-urban migration). The
number of migrants in these periods was divided by the appropriate
number of years to obtain a ”floating" average. This method of
obtaining annual averages allows a general appreciation of migration
patterns and reduces irregularities in the data; it may underestimate
migration rates but is useful in cases examining broad patterns, such as
this study (United Nations 1970). Retrospective data in itself may have
problems of recall error and be less precise than following individual
migrants through time, but is considered to be well adapted to
longitudinal studies and can be useful in illustrating migration
patterns (Haeringer 1969, Gould 1976). Especially in Rwanda, the

question of the year of movement gives a more accurate sense of

57
migration over time than age groupings because in the most recent time

period it was found that a wide spectrum of age groups had migrated.

In- and out-migration and migration stream rates are all based on
prefecture-level data since these are the most reliable. The in-
migration rates were calculated using the following formula (in which
Im, the in-migration rate, is equal to Mn the number of in-migrants,
divided by pp the destination area’s population, multiplied by 100):

m. - < "/9. > * k-

Similarly, the out-migration rate, m“ is equal to the number of out-
migrants, M” divided by the origin area’s population, p" multiplied by
100:

"I. - ( M/p. ) * k.

The migration stream between two areas was calculated according to the
following (in which my, the migration stream, is equal to My, the number
of migrants between two areas, divided by p" the population of the area
of origin, multiplied by 100):

m,-(M,/p.)*k
(United Nations 1970). Male migrants were the target population so only
their movements were included in the calculation of the rates. Also,
the rural-urban migrants were separated from the rural-rural migrants to

better evaluate the rural migration patterns.

The distances migrants traveled from their areas of origin to
their destinations were considered for comparison of trends over time

and as a variable in the gravity model. Distance was measured as the

58
number of kilometers between the centroid of the prefecture of origin
and the centroid of the commune of destination, the level of detail the
data provides. The centroids were determined by using the center of the

smallest possible rectangle drawn around the area (Monmonier 1982).

The migration maps and graphs, Figures 5 through 12, were produced
with data from the Ministry of Planning’s 1981 Demographic Survey. The
movements illustrated in the maps were drawn from an inter-prefecture
matrix of the net number of migrants’ between prefectures who had
migrated during the various time periods. Urban migrants, or migrants
to the two cities, Kigali and Butare, were separated from the rural

migrants in producing the maps.

The Demographic Survey questions determining an individual’s age,
marital status, educational level, and economic activity allowed
comparisons between migrants and non-migrants. Another important source
of data for the examination of individual characteristics was the Non-
Farm Strategies Survey performed by Service des Enquétes et des
Statistiques Agricoles (SESA) and co-funded by the Rwandan Ministry of
Agriculture, Livestock, and Forestry, and USAID (SESA 1988). The Non-
Farm Strategies Survey is part of an ongoing agricultural survey
analysis project based within SESA. It consisted of a random survey of

1019 rural households during the three months starting in July 1988. A

 

’ The net number of migrants was calculated as the difference
between the number of in-migrants and out-migrants in a prefecture
(United Nations 1970). -

59
team of SESA field staff supervisors was engaged to carry out the

interviews, which often required several repeat visits.

The questionnaires were designed to gather information from
various members of the households--husbands, wives, and adult children--
on topics ranging from demographic characteristics of household members
and engagement in non- and off-farm employment to aspirations and
opinions of parents and adult children regarding the future of young
people in farming. The survey also gathered detailed data on
agricultural production and farm size operated. Of particular
importance to this migration analysis were questions on where the head
of household was born and, if different from the present residence, why
he or she had moved. Also important were questions on where and why

adult children who were living outside the parental household had moved.

In summary, two basic data sources were used for this study: the
Demographic Survey gathered in 1981 by the Ministry of Planning and the
Non-Farm Strategies Survey of 1988 gathered by SESA of the Ministry of
Agriculture, Livestock, and Forestry. Most of the spatial and some of
the individual characteristics data came from the former, whereas the
latter supplied most of the individual characteristics data. The data
analysis was performed with the SPSS PC Plus Version 3 statistical
analysis program, and the original maps, graphs and tables in this study

are the results of this analysis.

CHAPTER IV
R U T F ATA ANA S S

In order to analyze the impact of changing demographic and
economic factors on migration patterns and migrant characteristics, this
chapter first presents the spatial pattern of migration over time and
compares it to evolving social and economic factors, in particular
population pressure, important in migration activity. Secondly, the
pattern of changing characteristics of migrants and non-migrants is

surveyed and compared to economic and social factors.

A. Spatial Pattern

The factors affecting migration changed over time in Rwanda, and
these changes were reflected in an evolving migratory pattern. This
section inspects these variations over space and time by first comparing
the number of migrants going to different destinations over time, and
secondly, with the use of maps, examining the spatial pattern of
movements and comparing them with changes in regional population
density, a factor expected to have an important influence on migrant

activity.

60

61
1. Migration Activity Over Time

The spatial pattern of migration will be first investigated by

comparing the number of migrants to various destinations over time. A

general illustration of total national migration activity is shown in

 

 

 

  

3m flliiiillIlliilrliiirirrriiiiiiiititiITii
em— 4
In
4.:
c
2
mmooo— -
OH
2
‘0.
012000— L
L.
m
.o
E
z 6MB
8‘

 

50 55 60 65 0
Voar

 

45

 

0 Urban Migrants + Total Migrants

 

Figure 5. Total and Urban Migration, 1940 to 1980 (Male and Female)

62
Figure 5. This displays the gradual increase in internal migration

activity around independence in 1962, a large peak in the early 19705, a

decline, and then finally a spurt in the late 19705. The general trend,

though, is of rapid growth of migration activity in Rwanda at an average

annual rate of 12 percent since independence, a much faster growth rate

than the population increase during the same period.

The peaks and troughs can be better understood by breaking down

the total migration into a few of the more important destinations over

time (see Figure 6). In this second figure and in the ones following,

only the number of male migrants are included to better illustrate the

economic factors of migration. The peaks in the numbers of migrants in

Figure 6 represent a concentration of in-migration to certain

destinations during relatively short periods of time. For example,

Gitarama was an important destination in the 1940s when it was the

nearest locale with free land for the migrants from the high density

areas .

After independence, the land in the East that had previously been

reserved as Tutsi pastoral grazing land was opened up for settlement

(Prioul and Sirven 1980). Increasing densities in the Nest had led to

very small farm sizes and lack of farm land for young families, so

migration activity soared and rural Kigali prefecture became the primary
destination. It experienced a large influx of migrants which began in
the early 19605, peaked in 1972, then fell sharply (notice the sudden

surge of in-migrants in rural Kigali illustrated in Figure 6). Much of

63

 

 

 

 

5” l l l l l l T T l I l
m .—
n
‘5’
IN —-
2 am
a
E
a. 2m -—4
O
z.
38 m } —
j; 9 ' :
' 1‘ ‘
l

J

2:27;?“ ,_.___,«___._ ._.__ ~ 1 .
6.1. 64 67 70 73 76 79

 

a K
1940 43 46 49 52 55 58
a-yoar Intervals .
+ mum D Gitarama x Riyal! City

0 Kigali Prefecture

 

Figure 6. Migration to Selected Destinations (Male Migrants)

this wave was made of migrants going to the government settlement

schemes, the pgxsannats which the government was actively promoting in
rural southern Kigali prefecture along the Akanyaru River. For example,

the comune in the Southwest corner, Ngenda, experienced more than
Other land that was

10,000 in-migrants during its peak year of 1970.

not in settlement schemes also received many in-migrants, peaking only a
This indicates that

few years after the settlement scheme areas.

 

 

64
although the schemes may have been the attraction for many people, they
were soon full and migrants settled in the surrounding areas. The sharp
decline after 1972 was due partially to government announcements that
the settlement schemes in Kigali prefecture were full (Rwamasirabo 1988)
as well as the sudden increase in population densities of the area and

relative scarcity of free land.

 

 

 

 

 

 

mm TWTTTYV'jjir‘VVTY'Y‘YYr'jYWIVYY‘fllT‘VIYUI
%}
8408—- _(
In
3;
s.
m Hm
m —- —-i
E
m
d.
o 1
{-0
381m_ -1
g_LaJcLaA4LLAJcL+AJLLA4uLaauupaauazfildrf?‘ 11:1 (III Le

 

 

40 48 50 55 60 65 70 75 N
Veer
0 Urban Migrants + Mal Migrants

 

Figure 7. Urban and Rural Migration (Male Migrants)

65

The decline in migration did not last and a new growth spurt began
around 1975. This last period for which data exists, however, is
composed of a very different set of destinations. Migration to rural
southern Kigali prefecture began to slowly rebound as areas of
relatively low density began again to absorb new migrants, and migration
further east to Kibungo began to climb as migrants passed over the
increasingly crowded rural Kigali prefecture. Migration to Kibungo
overtook migration to rural Kigali prefecture in 1980, since it was now
the area with the most possibilities for attaining land. The capital
city, Kigali, however, became the destination receiving the highest
number of migrants and having the fastest rate of growth, with an annual
average of around 20 percent in the early 19705 and 27 percent in the
late 19705. It had begun to grow rapidly when migration to rural Kigali
prefecture declined in 1971 and within seven years had overtaken all
other destinations. Migration to Kigali city alone made up 40 percent
of all national migration in the years 1978 to 1980, and each year the
proportion of urban-destined migrants increased and rural-destined

migrants decreased (see Figure 7).

From these different migration waves, four periods were
delimitated to compare movements over time; these periods are used in
the migration maps and in much of the following analysis. They are:
pre-independence from around 1940 to 1961; the period of expanding

rural-rural migration from 1962 to 1971; the intermediate period of the

66
decline of rural-rural movement in 1972 to 1976; and finally the period

of rapidly increasing urban flows from 1977 to 1980.

Table 6 summarizes much of previous discussion. It shows the

general trend of migration of men during the four time periods (non-

migrants were placed in an appropriate period by matching their age

Table 6. Destinations of Men According to Period of Time (column

 

 

percentages)
Period of Time
1 4 - 9 196 - 971 7 - 1977-1980
non-migrant 67 57 57 50
within pre- 22 16 ll 10
fecture

to rural 10 22 l9 14
area in another

prefecture

Butare city 1 1 2 3
Kigali city 0 4 12 23
TOTAL 100 100 100 100

Data source: gm“ Qéflgramigu; Post-Censitairg (Kigali: Min. du Plan, 1981).

 

67
group with that of the migrants). Clear from the table is how migration
has become more prevalent with time, as Zelinsky (1971) hypothesized,
and how it has shifted from primarily short distance movements within
the prefecture of birth, to longer distance moves to rural areas, and
finally urban migration to the capital city Kigali. The reasons for

this shift will be explored in the following sections.

2. Inter-Prefecture Migration

In the previous section, changes were seen over time in the choice
of destinations and in the intensity of flows. In this section, the
origin as well as the destination of flows will be examined in the
framework of the time periods defined earlier to present the changing
spatial pattern of migration. The pattern will be compared to the
changing nature of economic and other forces affecting migration
activity. One factor especially, that of population density, will be
considered as a potentially important factor in both out- and in-
migration; density is used to gauge population pressure in very high
density areas and availability of land in low density areas. The change
in choice of destinations is related to the decline in availability of
land as former migrants and population increases "saturated" the earlier
destinations. This is illustrated by changes in the population density
and by maps of net migration flow between Rwandan prefectures (Figures 8

through 12).

68

In the earliest period considered here, from around 1945 to
independence in 1962, the number of migrants between the prefectures was
small (see Table 6). Most movements were short distance within
prefectures since land was still often available locally; the settlement
of nearby forest, fallow or marsh land absorbed much of the population
increase. A general trend had begun of movement out of the Highland
prefectures Ruhengeri and Gikongoro to the nearby foothills prefectures,
such as Gitarama, Byumba, and Butare. Population densities were already
comparatively high in the Highlands area except in the high altitude
forests of, for example, Gikongoro. Nevertheless, much of the migration
pattern is not explained by density, such as the movement from rural
Kigali prefecture and Kibuye to Gitarama. This may have a historical
political explanation since two important political centers of the Tutsi

regime were then in Butare and Gitarama (Lemarchand 1970).

The next period, from 1962 to 1971, experienced rapidly increasing
population with corresponding increases in densities and a sharp
increase in rural-rural migration. In this period, high density
prefectures such as Butare and Ruhengeri adjoined prefectures with
comparatively low densities such as Kigali and Byumba (see Figure 9) and
migration patterns were closely tied to these density differences. The
Highlands prefectures of Ruhengeri and Gikongoro had large numbers of
out-migrants as in the previous period; however, prefectures in the
Foothills which had been receiving migrants, such as Butare, Byumba, and
Gitarama, now experienced net out-migration as their densities increased

and available land became scarce. Gitarama was then being passed over

69

 

 

   
    
 

Byumba

Kibungo

lOO-l14
125- ISO

[:1
I
I

 

Figure 8. 1945 to 1961 Rural Inter-Prefecture Migration Flows;
1

948 Population Density.

70
by some migrants from the Highlands as they sought land in unprecedented
numbers in the savanna of Kigali prefecture. Out-migration from the
prefectures in the extreme Hest, such as from Gisenyi, Kibuye, and
Cyangugu, does not appear in the Demographic Survey data but other
sources such as the Non-Farm Strategy Survey indicate that emigration to
Zaire was occurring. In general, then, this period was one of high out-
migration from the Highlands and increasingly from the Foothills, with
migrants beginning to pass over the Foothills to settle in large numbers

in the savanna immediately to the east, Kigali prefecture.

During the next period, 1972 to 1976, migration was still
predominately rural-rural but migration to the capital city had begun
(see Figure 10). The migrants continued to come primarily from the high
density prefectures of Ruhengeri, Gikongoro, and Butare, with Gisenyi
now appearing in the data. The Foothills were no longer an important
destination and Kigali prefecture was still the primary target of all
migrants despite the sharp downturn of migration into Kigali prefecture
following government announcements that free land in the paysannats was
no longer available (Rwamasirabo 1988). For the first time, some
migrants are moving past Kigali prefecture to settle in the far eastern

prefecture, Kibungo, which had very low densities.

Migrants to the capital city, Kigali, came in any numbers only
from Butare and Gitarama (about 150 migrants per year), but started to
trickle in from other prefectures, making rural-urban a total of 17

percent of all migration. Although the second largest city, Butare, had

71

 

    
   
  

Bvumba

”‘58

Bulave
Average net migrants/m:

  

— 50-99
IIII im-u9
- 250-999

Byumba

   

Kibungo

  

ZOO-299
300-349

E]
as
II
II

 

Figure 9. 1962 to 1971 Rural Inter-Prefecture Migration Flows;
1970 Population Density.

72

been about the same size as the capital city at independence (ONAPO
1982), it attracted very few migrants in comparison with the capital,
with Kigali city eventually becoming the administrative, commercial,
industrial, and service center of the country (ONAPO 1982). This is
also seen in Table 10 which shows that although both cities had similar
numbers of in-migrants before independence, by the 1977-1980 period,
annual migration to Kigali city had increased to 3077 migrants compared

with 186 migrants to Butare city.

73

 

 

 

Figure 10. 1972 to 1976 Rural Inter-Prefecture Migration Flows.

74

The latest period, from 1977 to 1980, was one of increasing
intensity of movement and of a diversification of both origins and
destinations. Major flows to the city were perhaps the most important
change; urban migration made up 40 percent of all migration with a sharp
increase each year and by 1980 urban and rural migration of men reached
the same level. The origins of the urban migrants included all
prefectures with major flows from the East as well as from the Highlands
of the Nest; indeed, the pattern did not vary with population density
but was uniform across the country (see Figure 12). The city of
preference was overwhelmingly Kigali, with only five percent of all
migration destined to Butare. Of all migration into the prefecture of

Kigali, 77 percent was now urban-directed.

Rural migration continued from west to east and from high to low
density areas (see Figure 11), but a striking phenomena is the
importance of the flows to Kibungo, especially considering the distance;
Kigali prefecture attracted only 20 percent more rural migrants than
Kibungo. This switch in destinations followed a doubling of Kigali
prefecture densities (see figure 5, 1978 density) from 111 people/km’ in
1970 to 220 people/kn? in 1978, whereas Kibungo had a comparatively low
88 people/kn? in 1978 (Prioul and Sirven 1981).

75

 

    
 

  

Ruhenue
Byumba

£3.

       
    
 

Average net mgr' midyear
- TOO-249
- 250-499

Kibungo

- 400-499

 

Figure 11. 1977 to 1980 Rural Inter-Prefecture Flows;
1978 Population Density.

76

 

 

Averaoenaw
- lOO- 249

   

 

Figure 12. 1977 to 1980 Urban Flows.

77

From the pattern observed in the maps and from the graph of
migration into Kigali and other prefectures, it appears that once
regions obtain a certain population density, they cease to attract new
migrants. This was especially apparent in Kigali prefecture: in the
1972 to 1976 period it was the primary destination of all rural
migration, but by 1977 to 1980, in-migration declined as density
increased and migrants bypassed Kigali to head for Kibungo. To estimate
the density at which a region ceases to attract new migrants, graphs
were drawn comparing the change in the number of in-migrants (source:
1981 Demographic Survey data) and the density of individual communes in
Kigali and Kibungo between the years 1967 and 1980 (sources: Prioul and
Sirven 1981 and the Bureau National de Recensement 1984). It was found
that, for example, the southern Kigali communes of Mgenda and Kanzenze
were major destinations of migrants until a sharp decline after 1972

when densities reached around 70 people/kmfl

More recently, however, this point at which migrants choose other
destinations has changed. For example, migration into Birenga commune
in Kibungo climbed and then fell after 1978 when its density reached
about 150 people/km’. Migrants then headed further east to Rusumo.
Similarly, in 1980, Ngenda, Kanzenze, and Bicumbi in southern Kigali
continued to each attract about 1250 migrants per year, although their
densities had reached about 200 people/kmf. By comparison, higher
density communes such as Kanombe and Butamwa in Kigali with densities

above 250 people/kn? did not attract new migrants. To summarize, the

78
point at which the density of an area causes it to lose its status as a
major destination of migrants has changed from around 70 people/kn? in
1970 to 150 people/km? in 1978. Nevertheless, areas with densities up
to 200 people/kn? continued to attract new migrants if at only moderate

levels.

3. Population Density/Migration Relationship

From the above discussion, it appears that differences in
population densities have been closely associated with rural in- and
out-migration, with people leaving high density areas and moving into
other relatively low density areas. In this analysis, the strength of
the relationship between density and migration is tested, with
particular emphasis upon the importance of high densities as a "push"
factor contributing to out-migration and of low densities as a "pull”
factor contributing to in-migration. The densities of the prefectures
were included in linear regression models to test their prediction of
three migration rates: the out-migration rate, the in-migration rate,
and the migration stream between prefectures. The migration stream
measures the rate of movement between two prefectures or the probability
of moving from a given origin to a given destination (United Nations
1970). The strength of the relationship between densities and migration
was found to be different for each time period, but in total, origin and
destination densities statistically account for about 44 percent of the

variation in the rural out-migration rate (see Table 7).

79
Table 7. Rural-Rural Migration Rates Explained by Population Densities

 

variables: x, is the origin density
x, is the destination density

All equations are significant at the .0000 level using the F statistic.

Period: 1248

l. Out-Migration Rate = constant + (-).473x,
st. error 1.762
significance .0000

Adjusted R2 = .218.

2. The in-migration rate and migration stream equations are not
significant.
MM.
1. Out-Migration Rate - constant + .792x,-+ (-).083x,
st. error 1.258 1.41
significance: .0000 .0020

Adjusted R’ = .646.

2. In-Migration Rate - constant + (-).578x,-+ .307x,
st. error 4.907 4.358
significance .0000 .0000

Adjusted R’ . .464.
3. Migration Stream . constant + .689x -+ (-).128x,
st. error .017 .019

significance .0000 .0001
Adjusted R2 . .508.

(continued next page)

 

80
Table 7 continued

 

P i d: 97
1. Out-Migration Rate = constant + .543x,-+ (-).122x,
st. error 1.834 1.459
significance .0000 .0010

Adjusted R’ . .314.

2. In-Migration Rate = constant + (-).640x,
st. error 9.691
significance .0000

Adjusted R’ = .409.

3. The migration stream equation has a low R5

All Periods Combined

1. Out-Migration Rate - constant + .658x,-+ (-).238x,
st. error 9.177 1.003
significance .0000 .0000

Adjusted R’ .. .439.

2. In-Migration Rate - constant + (-).419x,-+ .381x,
st. error 2.732 2.450
significance .0000 .0000

Adjusted R’ =- .269.

3. Migration Stream - constant + .460x,-+ (-).241x,
st. error 9.802 .011
significance .0000 .0000

Adjusted R2 = .233.

 

81

In addition, the spatial pattern differs between time periods. In
1948, little migration took place. 0f the movements that did occur, the
relationship between origin densities and out-migration was very weak.
Unexpectedly, low densities were associated with high out-migration
rates, accounting for 22 percent of out-migration. The rate of in-
migration and the migration stream were also not significantly
associated with the density variables. In 1948, therefore, of the
little inter-prefecture migration that took place, most out-migration
was not associated with a lack of land nor in-migration with seeking of
land; other factors were linked to movements such as the settlement of
Tutsi in the Highlands (Lemarchand 1970) and the gathering of people to

political centers in Butare and Gitarama.

The situation was very different in 1970 when high population
densities clearly contributed to high rates of rural out-migration; high
densities at the origin alone explain 65 percent of the variation in
rates of rural out-migration. Low densities were also attractive to in-
migrants, although the predicatory power of low density/high in-
migration is not as strong as for the high density/high out-migration
relationship. Low destination densities explain 37 percent of the
variation in rural in-migration with high densities at the origin adding
9 percent to the model. The third rate tested, the migration stream
between two areas, is also best explained by high densities at the
origin, accounting for 49 percent of the variation in the migration
stream rate. Low densities at the destination add only about 1 percent

explanatory power to the equation. Therefore, in 1970, high origin

82
densities are very important in explaining both out-migration rates and
the migration stream between regions, with low destination densities
adding little to the explanation. The rate of in-migration, on the
other hand, is best explained by low destination densities attracting

migrants.

The situation in 1978 again changed. In this period, rural-rural
migration accounted for only 60 percent of all migration since rural-
urban migration had become widespread. For this reason as well as the
fact that densities reached relatively high levels throughout the
country and free land was limited, density variables alone do not
explain rural-rural migration as well as in 1970. The most important
difference is that the rate of rural-rural out-migration is not as well
explained by high origin densities: only 30 percent of the variation of
the out-migration rates is thus explained. In contrast, the rate of
rural in-migration is still 41 percent explained by low destination
densities attracting migrants. Curiously, the rural migration stream
between regions is not significantly explained by density variables,
although high origin and low destination densities are weakly correlated
with the migration stream (correlation coefficients of .12 and -.11,
significant at the .01 level). This indicates that density variables
alone, although still important in explaining in- and out-migration in
regions with relatively high or low densities, are no longer useful
predictors of movements between any two given prefectures. The density

pattern of the country had changed from one of extremes within short

83
distances to one of variations of medium to high density throughout the

country.

Table 8. Rural—Rural In- and Out-Migration Rates and Population
Densities
Listed by Prefecture

 

 

1948 1970 1978

ln’ Out: Denei tY' in Out Density in Out Density
Butare 1.35 O .35 143 0.64 9.42 305 1.32 9.41 321
Byuba 1.66 0.76 41 0.81 1.43 95 0.96 1.23 120
WWW“ 0.53 0.32 78 0.47 0.26 145 0.19 0.20 191
Gikongoro 0.50 2.00 80 0.50 3.11 134 0.18 2.33 182
Gisenyi 0.48 0.39 119 0.70 1.05 271 0.62 1.03 286
Gitarama 1.87 0.88 113 0.93 1.32 202 1.29 0.97 297
Kibungo 0.67 0.55 37 1.72 0.32 52 6.59 0.20 88
Kibuye 0.45 1.48 94 0.41 0.71 187 0.25 0.53 252
Kigali 1.02 2.45 57 21.21 1.17 111 5.46 0.71 220
Rmengeri 0.30 0.87 133 0.22 6.01 255 0.35 5.05 313
IRMA T7 150 188
t The in-migration rate of rural in-migrants (data source: WM”

(Kigali: Min. d1 Plan, 1981).

2 The out-migration rate of rural-destined out-migrants (data source: W
gaitairg (Kigali: Min. du Plan, 1981).

3 Persons per aware kilometer (data source: Prioul and Sirven, W, 1981).

 

These migration/density patterns are apparent at the prefecture or
regional scale, but variations and exceptions to the patterns appear as
well (see Table 8). For example, high out-migration rates are expected
to be associated with high densities in the last two periods, but this

differs between regions. Although the prefectures Butare, Ruhengeri,

84
and Gikongoro do conform well to the high density/high out-migration
hypothesis, the western prefectures Gisenyi, Kibuye, and Cyangugu do
not. These latter prefectures experienced high densities but not the
corresponding high out-migration rates; this is reflected in the
residuals of the regression model which show that actual migration from
Gisenyi, Kibuye, Gitarama, and Cyangugu was less than expected. This is
probably due to limitations of the migration data which only contain
internal Rwanda migration and exclude probable emigration flows westward
into Zaire.’ Emigration from Ruhengeri has probably also been
important, but land within Rwanda was closer for Ruhengeri than for the
far western prefectures so that out-migration from Ruhengeri to Kigali
and other prefectures is reflected in the data. Another anomaly is the
relatively high out-migration rate of Byumba although its densities
appear low. This is due to the inclusion of a large amount of low-
potential or non-arable land in the computation of the prefecture
density; excluding non-arable land, densities are much higher (234
people/km’ in 1978). Byumba is divided between high altitude, high
density areas in the western half of the prefecture, which has produced
many out-migrants, and low altitude, low density areas in the eastern
half (not all of which is park land). The eastern half has not absorbed

all the out-migrants from the western half because of very poor soil and

 

’ Before independence, emigration to Zaire was encouraged by the
Belgian authorities and approximately 128,700 people migrated. Since
independence, emigration to Zaire has been clandestine and uncounted
(ONAPO 1982, Prioul 1976). The data presented here, however, suggests
the existence of a steady out-flow from the western prefectures into

aire.

85
dry conditions (Lemarchand, 1982, Prioul 1976), and many people have

been forced to migrate outside their prefecture.

For every prefecture, the rural-rural out-migration rate declined
between 1970 and 1978 although densities increased at a very rapid rate.
This is another reflection of the decline in importance of rural-rural
flows while rural-urban movements grew. The total out-migration rate
(summing rural- and urban-destined migrants) did increase between 1970
and 1978 for almost all prefectures. The result was that by 1978 all
prefectures had a total out-migration rate greater than 2.50 except
Kigali and Kibungo, reflecting the pressure of increased densities.
These out-migration rates provide evidence that a greater pr0portion of
Rwandans are moving from their home prefectures and that general

mobility has increased, much as Zelinsky hypothesized (Zelinsky 1971).

The regression analysis using density variables was not as
successful in predicting nationwide in-migration in the 1970 and 1978
periods, and the prefecture data illustrates why this is so. In-
migration was strongly focused on Kigali prefecture in 1970: Kigali
experienced extremely high in-migration rates with a doubling of
population in 8 years. In-migration in 1978 was similarly focused on
the two prefectures Kigali and Kibungo, although other prefectures also
contained comparatively low densities. The residuals confirm the
deviation from the expected, with much higher levels of in-migration in

1970 expected in Kibungo, Byumba, Gikongoro, and other prefectures. In

86
1978, the same pattern of under prediction of many prefectures except
Kibungo and Kigali appeared. Factors other than density played an
important role in the migrant decision-making process, including
distance and the knowledge of land availability. In 1970, the
government was promoting the paysannat settlement schemes in Kigali
prefecture and that focused potential migrant attention on the southern
region of Kigali. Land not part of a paysgnnat was also quickly settled
in the same area as migrants continued to pour in. By the 1978 period,
the government had announced the end of available 231533311 land and
rural migration to Kigali had slowed, but new paysannat and other land
was free in Kibungo. Prefectures such as Byumba, Gikongoro, and
Cyangugu may not have high densities, but much of their land has low
potential (poor soils and dry land in Byumba, and high altitudes in
Gikongoro and Cyangugu) and so did not attract many migrants from other
prefectures. Therefore, although migrants did move in a unidirectional
flow towards low-density areas, contributing factors such as
announcements of settlement scheme land focused in-migration to specific

regions.

Kigali prefecture stands apart with its doubling of population
density in the 8 years between 1970 and 1978, but Kibungo also underwent
a 69 percent increase. The growth in population is linked to their
status of being the major recipients of migrants during that period. In
contrast, the three prefectures with the highest densities, Butare,
Ruhengeri, and Gisenyi, also had the slowest growth of population and

the highest total out-migration rates, suggesting that out-migration

87
helped slow their growth rates. Most of the other prefectures, such as
Byumba, Cyangugu, Gikongoro, and Kibuye, experienced between 25 and 35
percent increases in population during those 8 years. This population
growth signifies that although high densities contributed to high out-
migration rates, the out-migration was not massive enough to prevent a
rapid growth in population. Many people there adapted to the
consequences of increasing densities without resorting to permanent out—
migration, by means such as earning non- or off-farm income including
short-term migration to the East, and/or changing their agricultural

practices (Clay, Kayitsinga, and Kampayana 1989).

In conclusion, the relationship between density and rural
migration has changed from one period to another. In the earliest
period examined, the little inter-prefecture migration that took place
was not associated with the "push" of high density or the "pull" of low
density but with other social and political factors. In contrast, the
first post-independence period did experience the expected phenomena of
high origin densities associated with high rates of rural-rural out-
migration and low destination densities with high rates of in-migration.
The strongest predictor of migrant behavior was high origin densities,
which not only statistically explained much of the out-migration
activity, but also explained the migration stream between regions. The
ability of relative differences in densities to statistically explain
rural migration again changed in the latest period, in which urban
centers had begun to attract increasing numbers of people from

throughout the country so rural-rural out-migration accounted for only

88
part of all out-migration. The ability of density variables to account
for out-migration, or the migration stream between rural areas,
had,therefore, seriously diminished. Low densities at rural
destinations continued to be important predictors of in-migration, but
they had never been as powerful a predictor as high densities at the

origin explaining out-migration.

4. The Gravity Model

The same basic assumption of high densities contributing to out-
migration and low densities to in-migration is tested here with the
addition of the influence of distance on migrant behavior. To
investigate this relationship, a version of a well-tested model of

spatial interaction, the gravity model, was tested.

The traditional gravity model predicts movements between two
points or centers using the mass of their populations divided by the
distance between them (Woods 1979). In Rwanda, however, the population
is not concentrated in centers but is scattered across a rural setting,
so that the population density of a two-dimensional area is a better
indicator of the mass of population than the size of population at a
point location. Therefore, in the version tested here, the population

density of areas (prefectures) is used as the mass variable.

89

The traditional gravity model uses population size as an indicator
of the scale of generation and attraction of interaction between
centers; inherent in the model is the assumption that the size of the
origin approximates the number of potential out-migrants and that the
size of the destination reflects the attractiveness of the center or
represents an indication of opportunities for the potential in-migrant
(Haynes and Fotheringham 1984). In Rwanda, the first assumption that
population mass would provide an indication of the number of out-
migrants is relevant; indeed, as seen above, high population densities
are associated with high out-migration rates. In contrast, unlike the
assumption in the traditional gravity model that the population size of
the destination indicates attractiveness, in Rwanda the attraction of an
area for agriculturalists is related to a lack of population, or to low
densities, where the possibilities of attaining land is higher. Rural
movements in Rwanda are therefore associated with high densities as a
”push” factor and low densities a "pull" factor, resulting in a
unidirectional force of gravity. The influence of distance upon
migration patterns is assumed to be the same as in the original gravity
model, namely that the force of the attraction is constrained by the
distance between the centers. The flows between two places would be
expected to decline in proportion to distance (Haynes and Fotheringham

1984).

In other words, the traditional gravity model predicts the rate of
migration between two centers as the sum of their populations divided by

the distance between them; in this version, the density at the

90
destination is subtracted from the density of the origin and the result

is divided by the distance between them:
Model One: (origin density - destination density) / distance.

In the interest of obtaining a best fit of the model, a second version
similar to the first was tried in which the weight of the origin density

was doubled:
Model TWO: ((origin density * 2) - destination density) / distance.

Each time period was examined separately to investigate how the ability

of the models to predict migrant behavior changed over time.

91
Table 9. Predicting Migration Rates with the Modified Gravity Model

 

variables: M, is Gravity Model One:
(origin density - destination density) / distance.

M2 is Gravity Model Two:
((origin density * 2) - destination density) / distance.

All equations are significant at the .0000 level using the F statistic.
Period: 1948

1. Out-Migration Rate = constant - .382M,
5t. error .043
significance .000

Adjusted R’ is .139.

2. Out-Migration Rate . constant - .421M,
st. error .029
significance .000

Adjusted R’ is .172.

The in-migration rates and migration stream were not significantly
explained by the model.

Period: 1970

1. Out-Migration Rate - constant + .700M,
st. error .045
significance .000

Adjusted R’ is .489.
2. Out-Migration Rate - constant + .788M,
st. error .025

significance .000

Adjusted R’ is .621.

(continued next page)

 

92

Table 9 continued

 

3. In-Migration Rate = 8.869 + .320M,
st. error .107
significance .000

Adjusted R’ is .344.

4. In—Migration Rate . constant + .320M,
st. error .107
significance .000

Adjusted R’ is .101.

5. Migration Stream . constant + .687M,
st. error .519
significance .000

Adjusted R’ is .472.

6. Migration Stream . constant + .782M2
st. error .287
significance .000

Adjusted R’ is .612.
For the 1978 period, neither gravity model explains migration rates

better than the density or distance variables alone in a regression
equation.

 

As seen in the previous discussion, the influence of density on
migration patterns in 1948 is very weak, so it is not surprising that
neither version of the gravity model explains the early migration
patterns very well. The in-migration rate and the migration stream

between areas are not significantly explained by the models, and the

93
out-migration rate is only explained by 14 and 17 percent by the Models

One and Two respectively (see Table 9).

For the 1970 period, however, the density variables are much more
important in explaining the migration pattern and the performance of the
models is much improved. The models are better able to predict the in-
migration rates and migration stream between areas than any of the
variables singly. In contrast, the out-migration rates are so heavily
influenced by high origin densities that the model does not add

explanatory power to the one variable origin densities.

In 1970, the migration stream between prefectures is explained
especially well by Gravity Model Two: 61 percent of the rate of
migration between any two prefectures is statistically explained by this
model. This is higher than the ability of any one of the variables in
the model, which account for 49 percent of the movements. Model One,
without the double-weighting of the origin density, only explains 47
percent of the variation of the stream. The ability of the model to
perform well for predicting migration streams is not surprising since
the original model was designed to estimate streams between centers: the
inverse relationship of population and distance variables seems to be
similar for streams between urban centers and, in this case, between
rural areas. Model Two performs better than Model One because high
origin densities are significantly more important in predicting migrant

activity between prefectures than low densities at the destination; the

94
role of distance moderating the lure of extreme low density areas is

also captured in the model.

The residuals from Gravity Model Two explaining the migration
stream were similar to the residuals seen earlier in the
density/migration rate equations. Again, migration from the western
prefectures Gisenyi, Gitarama, Kibuye, Cyangugu, and Ruhengeri was less
than the model predicted, perhaps due to uncounted emigration (see
footnote 3 page ?). On the other hand, Byumba and Butare had more out-
migration than expected due to low densities but unfertile land in
Byumba and perhaps to compensate for high levels of in-migration to
Butare. Butare had more in-migrants than expected possibly because of
some paysannats then opening along the Akanyaru River. The convergence
of migrants to Kigali prefecture to the virtual exclusion of all other
prefectures resulted in many prefectures, such as Ruhengeri, Gikongoro,
and Byumba, having far fewer than expected in-migrants. This tremendous
influx into Kigali prefecture was also tied to the opening up of the

paysannats and their surrounding land for settlement.

The out-migration rate, as seen earlier, is heavily affected by
high origin densities and Model Two does not add much explanatory power
to this one variable: they both explain about 62 percent of out-
migration. Model One explains less, only 50 percent. This indicates
that the influence of distance and the lure of low density were not
important moderating factors on the rate at which people left an area.

In areas of high density, people left according to the density of their

95
home region and were not as influenced by the distance they had to

travel or whether their destination had the lowest possible density.

Although in-migration rates are heavily affected by low densities
attractive to migrants, Model One is able to explain in-migration better
than any density variable. Low destination densities alone explain only
24 percent of in-migration rates, but Model One explains 34 percent.

The improvement is due to the effect of distance: unlike the out-
migration rates, the in-migration rate is greatly influenced by the
distance migrants must travel moderating the lure of areas with low
densities. Model One is still not a strong predictive tool for in-
migration rates, indicating that other factors, such as the perception
of land or other economic opportunities not strictly related to low
density, also played an important role in the migrant decision making
process. The pattern of the residuals sustains the proposition of the
importance of mitigating factors. Migration into almost all prefectures
except Kigali was over-estimated, especially for the lower density
prefectures Gikongoro, Kibungo, and Byumba, and that into Kigali was
much higher than expected, due presumably to the government’s promotion

of land and infrastructure development in the new Kigali paysannats.

The situation in 1978 changes the performance of the model again.
As seen in the discussion above, the development of the attraction of
urban centers for migrants reduced the importance of density variables
in explaining rural-rural out-migration or migration streams; the models

based on these density variables are therefore also not able to predict

96
the rates well. Similarly, the role of distance in explaining migration
behavior has changed. During the 1978 period, rural-rural migrants had
to travel long distances in order to reach areas with available land
(the correlation coefficient of destination density and distance is -
.532 significant at the .001 level). Therefore, for the first time,
high rates of in-migration are associated with long distances
(correlation coefficient of .478, significant at the .001 level). This
situation is different from that assumed by the gravity model, that
migration rates decline with distance. Therefore, the assumptions of
the gravity model do not apply as well to this period and the result is
that the models do not perform any better than the density or distance

variables singly in a regression equation in explaining migration rates.

In summary, the modified gravity models using population densities
instead of size performed relatively well for the 1970 period when
migration patterns were generally rural-rural movements from high to low
density areas and covering the shortest possible distances. Model Two
performed especially well in explaining migration streams and Model One
in explaining in-migration rates. In the 1948 and 1978 periods,
however, the models did not predict migration rates well. This is
explained by various factors: in 1948 migrant behavior was not yet
driven by land-seeking; in 1978 rural-urban migration had become
widespread, complicating the rural-rural pattern; and again in 1978
migration rates no longer declined with distance as assumed by the model

since available land was then far away.

97
5. Changing Distance Over Time

As mentioned in the above discussion, in the past rural-rural
migrants were able to find available land nearby and their preference
for moving short distances helped explain migration rates. Hith time,
migrants were forced to travel longer distances east in order to find
available land. As these distances became longer, the distance to the
capital city, Kigali, became shorter in comparison and Kigali city
became an intervening opportunity between rural origin and rural
destination areas. Kigali city eventually became the most important
economic opportunity in the country attracting migrants nationwide.
Butare city, in contrast, attracted very few migrants following
independence although it was close to high density areas within Butare
and Gikongoro prefectures. Comparing the distances traveled to both
rural and urban areas and how that has changed over time illuminates one
factor behind the decline of rural-rural migration and the emergence of

rural-urban migration.

For the country as a whole, the average distance traveled outside
one’s home prefecture to rural areas increased from 44 kilometers before
1962 to 80 kilometers in the late 19705, with each time period having
significantly longer distances than the previous one (see Table 10).

The average distance traveled to the capital city Kigali, on the other
hand, remained around 66 kilometers. In the year 1972, the distance to
rural areas overtook the distance to Kigali, but not until 1980 did the

momentum of increasing migration to Kigali result in Kigali receiving

98
more migrants than rural areas.. This dramatic increase in the number of
in-migrants to Kigali reflects the concentration of economic growth in
Kigali. In contrast, Butare city maintained a low level of in-

migration.

Table 10. Average Distance Traveled by Male Migrants Over Time

 

 

PEle KIGALI CITY BUTARE CITY ML DESTINATIGIS

Distance ln-Migrants Distance ln-Migrants Distance In-Migrants
Totgl Am Ave' Totgl Ann Ave Total—Am

m

1945 to 1961 66 168 11 43 297 19 44 12802 800

1962 to 1971 68 1554 155 54 315 32 53 38691 3869

1972 to 1976 65 3717 743 53 216 43 76 16624 3325

1977 to 1980 69 12306 3077 65 747 186 80 17973 4493

8 The ml average of in-nigrants.

Data source: Engusss_2ém2acaehigus_£2:s;£soaisa1:e (Kigali: Hin- du Plan. 1981)-

 

This shift in the choice of migrants from heading to rural areas
to Kigali city occurred over a number of years and had different impacts
on the in-migration to rural prefectures. The prefectures that had been
important receivers of rural migrants include Gitarama, Kigali, and
Kibungo. Gitarama, very near the densely populated Highland region,
received many migrants in the early years when migrants needed to travel
only an average of 45 kilometers to reach land there. As land became

scarce in Gitarama and the number of out-migrants grew, they passed over

99
Gitarama and chose rural destinations further away, especially in rural
Kigali prefecture. By the 19605, Kigali was the closest prefecture with
available land for the densely populated western prefectures and it
received a sharp increase in numbers of in-migrants until 1970 when in-
migration declined suddenly. In this same year, average distances to
rural Kigali prefecture became longer than the average distance to the
capital city and so the city became an intervening opportunity between
the rural origin and the rural destination areas. In the next few
years, migration to the city gained momentum and began the climb up the
"S' curve. Nevertheless, rural Kibungo in the East then began to
experience a steady if small increase in in-migration although migrants
had to travel long distances in order to reach it: the distance
traveled by migrants to Kibungo increased from an average of 59
kilometers before 1962 when most in-migrants were from rural Kigali or
Byumba to 103 kilometers in the late 19705 when migrants also came from
the far western prefectures. This increase in distance is a reflection
of the decline of available land elsewhere, forcing rural migrants from
all over the country to travel to the less desirable extreme East, where
rainfall is much more irregular and land less fertile. It is also a
reflection of the perseverance of the desire of migrants to obtain their

own land and remain in agriculture.

In summary, the distance migrants have moved to find available
land has greatly lengthened over time and surpassed the average distance
traveled to Kigali city in 1972. Most migrants initially continued to

choose rural destinations in the same prefecture as Kigali city or in

100
the prefecture further east, but migration into the city soon began to
soar, reflecting the increasing importance of the city as an intervening
opportunity. Nevertheless, a small number of migrants still continued

to travel past the city to settle in the distant East, Kibungo.

lOl

mumsiummmmmm

The previous discussion focused on the spatial and temporal
patterns of migration at the regional, or prefecture, level. At the
individual and family level, however, comparable patterns of migration
are evident. Some of these patterns reflect similar factors affecting
migration activity at both regional and individual levels, such as
population pressure reflected by regional population densities and by
family farm sizes. In examining individual or family characteristics, a
better understanding of who migrates, where that person migrates, and

the reason for migrating can be realized.

In this chapter, the same category of individual characteristics
used by Findley in her table (see Table 2) will be used for comparing
Rwandan migrant groups with those of others in the literature. These
include gender, educational level, land ownership, marital status, and

age at time of migration.

1. Gender-Specific Migration

One of the most common distinctions made between migrants is
gender since often men and women have different reasons for migrating
and different migration patterns. This is true in Rwanda as well

because most women migrate to their husbands’ farm at marriage and only

102
later may migrate with their husbands for economic reasons. Therefore,
a greater proportion of women are migrants than men, but their movements
are usually short distance. This is reflected in the Demographic Survey
data in Table 11 which shows that consistently about 70 percent of women

have migrated, but predominately within the prefecture they were born.

Table 11. Gender-Specific Migration (column percentages)

 

Before 1962 1952-1971 1972-1976 1977-1980
men Men when MW

 

 

Destination:

non-migrant 3O 67 28 57 27 57 31 50
within pre- 53 22 48 16 44 11 40 10
fecture

rural area 17 10 23 22 26 19 24 14
in another

prefecture

Butare city 0 l 0 1 1 2 0 3
Kigali city 0 0 l 4 2 12 5 23

Data Net: Wm (Kigali: Min. ch "-0. 1941)-

 

103

Very few women have migrated to Kigali city in comparison with
men. During the period 1977 to 1980, for example, 40 percent of all
migrant men went to the capital city but only 2 percent of migrant women
went there. These urban women made up only 4 percent of the total
rural-urban stream. The low numbers of women urban migrants suggests
that many men in the city may not be permanent migrants. The men may be
working for only a few years in the city before marriage or the men may
be leaving their wives in the rural areas. This is consistent with the
literature from other African countries and conforms to information on
the predominance of single migrant men in Kigali presented later in this
chapter. Of the women migrants in Kigali, 30 percent are single; they
are almost all young and living in the household of a relative or
someone else. Of the recent migrants who work, two-thirds of the single
women are domestic servants, but two-thirds of the married women are
small business owners. This is consistent with Findley’s table (Table
2) of rural-urban migrants being predominately men with some single

women doing domestic service jobs.

The Non-Farm Strategies Survey information also reflects the
pattern of short-distance migration of women for marriage reasons.
Among the daughters who had left the home farm, 88 percent had left to
marry and the remainder had left usually not for economic but for other
personal reasons. The vast majority, 86 percent, had moved within their
own prefecture or to a neighboring commune in another prefecture. The
picture for migrant female heads of household is almost identical: 80

percent said they had migrated for marriage reasons and 85 percent had

104
stayed within their prefecture or gone to a neighboring commune in an
adjoining prefecture. Only 30 percent of all female heads of household
were non-migrants, versus 74 percent of male heads of household. Of the
migrant male heads of household, less than one percent said marriage was

the reason they had migrated and 86 percent cited the need to seek land.

These figures illustrate that men and women have very different
reasons for migrating and different patterns of migration. A small
proportion of men migrate and those that do migrate leave their
prefecture for economic reasons but a large proportion of women migrate
short distances for marriage reasons. These findings are consistent
with much of the literature that women often migrate short distances to
the household of their husbands, but in Rwanda the proportion of women
that migrate is higher than in other African countries (Zachariah and
Condé 1981, Masser and Gould 1975). Since this study concentrates on
economic factors leading to migration, in most analyses only the

movements of men are examined.

2. Educational Levels Affecting Migration Behavior

A second characteristic which is an important factor in the
migrant decision making process is the level of education an individual
has attained. This characteristic helps determine who migrates and what
destination they choose. Studies mentioned in the foregoing survey of

literature had concluded that the effect of education on migrant

105
activity in Africa is that people with high levels of education tend to
move to the cities, those with less education but still more than the
average go to frontier rural areas, and those with the least amount of

education tend to remain in their rural areas.

To examine this premise in the case of Rwanda, non-migrants and
migrant men with different destinations are compared over time by the
level of education they attained (see Tables 12 and 13). The levels of
education shown in Tables 12 and 13 are grouped answers from the
Demographic Survey of the number of years completed in primary,
secondary, and superior level education. The system of primary
education has changed, previously 6 years was the maximum but since 1978
supplementary years allowing the student to concentrate on a technical
subject have been introduced. The secondary level is approximately the
same as the American high school and the superior level is equivalent to
the American college or university. The university is located in Butare
city, so Butare attracts professors as well as students with high levels
of education. Butare, however, has fewer administrative, service, and

industrial activities than Kigali.

From Table 12, the changing importance of education on migrant
patterns can be observed. The difference in levels of education between
non-migrants and both short and long distance rural migrants is
negligible, they all have very few years of education. Only in the pre-
independence period did rural migrant men have significantly more years

of education than non-migrants, but since independence, the two groups

106
Table 12. . Education Levels of Migrant Groups

 

Education Levels by Migrant Group
(column percentages)

1945-1961 1962-1971 1972-1976 1977-1980
Migrant non rural urban non rural urban non rural urban non rural
urban grouo--> A B C D E, A QT C D .E, A A! C DAA,E .A Q C

 

£4

Mo education 77 79 72 41 39 52 59 53 8 12 46 49 46 12 13 42 40 44 8
14

Primary 1-5 19 18 19 26 15 39 33 29 14 29 34 31 31 11 24 38 32 30 9
30

Primary 6 + 3 3 6 15 31 6 6 14 22 27 15 15 16 21 32 16 19 19 21
29
Secondary O 1 1 13 15 2 1 2 38 23 2 3 S 26 20 2 4 5 28
18
Superior 0 1 2 5 0 1 1 2 18 1O 2 1 2 30 11 2 5 3 33
11

A: Non-migrants

8: Rural migrants within the prefecture of their birth
C: Rural migrants out of the prefecture of their birth
0: Urban migrants to Butare city

E: Urban migrants to Kigali city

Migrant groups:

Mean Years of Education Received

 

 

 

Rural Migrants Urban Migrants
Period Non-Migrant Within-Pref Out Pref Butare. .Kigali TOTAL
1945-1961 .9 .9 1.3 3.8 3.5 1.0
1962-1971 2.0 1.7 2.1 7.1 5.7 2.7
1972-1976 2.5 2.6 2.6 7.8 5. 3.2
1977-1980 2.6 2.7 2.8 9.0 5.6 3.4
TOTAL 2.1 2.1 2.2 7.6 5.5 2.5

Data source: Eggggge Démograghiggg Post-Censitairg (Kigali: Min. du Plan, 1981).

 

107
have had no significant difference in years of schooling attained
(tested with two sample t-tests), i.e., a few years of primary schooling
have not led to increased rural migration. During each time period, a
gradual increase in national levels of education has been experienced,
in particular, an increase in primary schooling and a decline in the
percentage of the population that has attained no education. Therefore,
increasingly non-migrants and rural migrants have attained some
education; in the last period almost 60 percent have a few years of
primary schooling. These results indicate that, throughout the years,
education has not been a distinguishing factor between men who have not
left their home area and those who have migrated permanently long or

short distances to rural areas.

In contrast, the role of education has been very important in
urban migration, and urban migrants have formed a class apart. In the
pre-independence period, too few people had attained education past
primary school and too few had migrated to urban areas to make any
testable conclusions on the relationship of migration and education. In
the post-independence periods, however, urban migrants have had
significantly higher levels of education than the other groups.
Initially, very few migrants to the city were men with low educational
levels; the migrants to the cities had an average of 4 years more
education than the rest of the population. Migrants to Butare in
particular were well educated--consistently over half had secondary
schooling because of the national university located there. Kigali

never quite reached that level but even so has had few migrants with no

108

education or incomplete primary schooling. The level of migrants to
Kigali, however, has changed. The proportion of migrants with no
education and incomplete primary schooling has slowly increased despite
improvements in the nationwide levels. This is reflected in the
basically unchanged or slightly decreasing average school years of
Kigali migrants, a contrast to the rising national and Butare average.
Separating out the numerous foreigners who have migrated to Kigali,
mostly from Zaire, Burundi, and Uganda, the pattern of declining
educational levels of Rwandan migrants to Kigali is even more striking:

the average years of schooling has declined from a high of 6.0 in 1962-
1971 to 5.7 in 1972-1976, and then to 5.3 in 1977-1980. The conclusion
to be drawn from this finding is that the selective nature of Kigali
migrants has changed, they increasingly represent the broad spectrum of
society. This may be a reflection of the lack of rural alternatives, in
both their home and other rural areas, for those with less education,
and a growth in economic opportunities in Kigali city, especially in the

informal sector.

These changes in decision-making are very clear in Table 13, which
illustrates the change over time in the destinations of men with
different educational levels. During all time periods, longer distances
are associated with higher education and there is a general increase in
distance over time. During the pre-independence period, there were few
people with secondary or superior education and few urban migrants.

Most of those with lower educational levels migrated only within their

own prefecture. By 1962 to 1971, more migrants left their prefecture to

109
settle in distant rural areas than moved within their prefecture, and
those who had completed primary school made this leap to distant rural

areas earlier and in greater proportions than those with less education.

Table 13. Destinations of Men by Education Levels (row percentages by

 

 

period)
1945-1961 1962-1971 1972-1976 1977-1980
Migrant non rural urban non rural urban non rural urban non rural
ban
Egouo--> A Ag: C D E A B C D E, A 8 C D E A, E C D
5

Mo education 68 22 1O 0 O 58 17 24 O 1 63 11 21 1 4 61 11 18 1
9
Primary 1-5 68 20 11 1 0 64 14 19 O 3 62 9 19 1 9 58 9 13 1
20
Primary 6 + 59 18 17 4 2 41 10 35 3 11 50 8 17 3 22 41 9 14 4
33

Secondary 35 27 18 15 6 37 5 29 15 24 24 5 19 10 42 12 5 10 14
60
Superior 36 14 41 9 0 25 7 29 15 24 26 4 13 19 39 15 9 7 21
48

Migrant groups: A: Mon-migrants
8: Rural migrants within the prefecture of their birth
C: Rural migrants out of the prefecture of their birth
0: Urban migrants to Butare city
E: Urban migrants to Kigali city

Data source: gggggte ngggraphiggg Pgst-gensitairg (Kigali: Min. du Plan, 1981).

 

110
The distant rural areas were attractive even for men with secondary or
superior education until the 1972-1976 period when the rural areas lost
their relative attractiveness and over half of these educated men
migrated to the cities, especially Kigali. Kigali then began to attract
even less well-educated men; for example, of all those with 1 to 5 years
of primary schooling, the percentage migrating to Kigali jumped from 9
percent in 1972-1976 to 20 percent in 1977-1980. This trend explains
why the level of education of Kigali migrants has not improved

proportionately with the rest of the country.

Table 13 also illustrates a clear pattern of increasing propensity
to migrate with increased education. Men in the first two levels, with
no education and with 1 to 5 years of primary education, did not behave
very differently; even in the latest period about 60 percent remained
non-migrants. Of those with 6 or more years of primary schooling, about
10 percent more migrated, usually to urban areas. The point which
clearly differentiated non- or rural migrants from urban migrants,
though, was secondary schooling. Only a small percentage of men with
secondary schooling remained at home; in the last period, 41 percent of
primary school leavers but only 12 percent of those with secondary
schooling did not migrate. Most with secondary schooling left rural
areas altogether--60 percent went to Kigali city in 1977-1980. Men with
the highest level of education behaved similarly, rarely remaining in
rural areas and usually heading for Kigali or Butare. The reason for
this difference is linked to the fact that secondary schools are located

in urban centers and that almost all jobs for well educated people are

111
also located in urban centers. As mentioned above, this educational
division between non—migrants or rural migrants versus urban migrants is
breaking down with time, with more and more of those with less education

migrating and more frequently choosing to go to the capital city.

This pattern is reconfirmed by information in the Non-Farm
Strategies Survey, which asked why sons of household heads migrated,
when they migrated, and where they had migrated. Until the mid-19705,
sons with no education or incomplete primary education most often
migrated to seek land or for marriage reasons, since men need to own
farmland in order to get married so often must migrate to obtain land.
Therefore, they usually migrated to rural areas. Beginning around 1980,
however, even those sons with very little schooling migrated not to seek
land but to seek work, often as a laborers. Those with secondary
education or higher left to study, or if older to work, in one of the

cities.

These findings are somewhat different from those expected from the
literature. Although the literature suggests that rural migrants have a
somewhat higher level of education than non-migrants, if still low, in
Rwanda little difference in educational levels between the two groups
was found and this similarity did not change with time. Migrants to the
city have been better educated than those to rural areas, as expected,
with secondary education the turning point at which most recipients
migrate to the city. With time this pattern changed and less well-

educated migrants began to head to the city, increasing the proportion

112
of ”less selective” urban migrants, as Findley (1981) termed them.
These ”less selective” urban migrants, with at the most a few years of
primary schooling, have in fact even less education than Findley

presumed.

3. Farm Sizes and Production Levels Affecting Migration Behavior

In the section examining the association between population
densities and migration at the prefecture level, a strong relationship
was found between high densities and high rates of out-migration. This
may reflect a situation at the family level in which increasing
population density has led to a decline of farm sizes to the point at
which many farms are too small to support a family and out-migration has
become an important means of finding alternative sources of income.
First, a short regional comparison of the sizes and production levels of
farms is presented. Secondly, to examine the effect of farm sizes at
the family farm level, farm size and production data are compared
between non-migrant and different migrant groups.‘ Comparisons are then
made between rural migrant and non-migrant heads of households, and the
advice that the heads of household would give to their children to
migrate is compared by size and wealth of farm. Finally, the migrant

behavior of the sons of the heads of household is examined.

 

‘ Data from the Non-Farm Strategies Survey is used for this

analysis (SESA 1988). T-tests are used, with alpha equal to .05, to
test the significance of differences between migrant and non-migrant
groups.

113

 

 

Phastauwes
n) b
I l

 

.\\\\\\\\\\\ \\\\\\\\\\\\\\\\\‘

 

5
, I
I I
5 5
I l‘
I "
I "
:2 5‘ ' 'I ..
55 5. 7 ’
5- 5' ,: 5
', 'I ,, 'o‘ l‘ 5 I
I I ’3 ’0 I r, I I
I; l- 'e l- r ” l'
’V I ' ’:.. 4. ’1 "-'. ’_ I -o
1 — ’ ’-' ’L’.’ a. '4.-. e ’. ’1’. I: ’5'.
I». 5:. I". ‘o ’2'. 5. ’. 5-2: I. 5"
53:0 [.3 51:6 :0 III. ‘3». g. I”. I”: 5";
53:3. 53: Iii: :1 59:: 53:: 53:3 51:: 5 :23 5":
8- 4'»:- 9: 5.: .: 5: 4'. 4: cc 6'. w!
00-. ‘ fl 0'8 . a o
g : .. s. .. L >- >~ - ~
ca ca '- ¢= . . a: 3 a o
a ‘- a 5 a o . a o c-
: ‘ w-O a >* c . ”'7 c s
an ca-0 8 we CI . “'0 8 C
an a: an
.1955 "1972 BM

 

Figure 13. Changing Farm Sizes by Prefecture, 1965, 1972, and 1988.

A rough indication of regional variations in population pressure
is obtained by a comparison of comparing farm sizes. Figure 13
illustrates average farm sizes at the prefecture level from two
different sources. The data for the years 1965 and 1972 are from the
Géographje dg Rwanda (Gotanegre, Sirven, and Prioul 1974) and were
calculated by dividing an estimate of the number of families into the
hectares of arable land for each prefecture. The farm sizes in Byumba,
Cyangugu, Gikongoro, and Kibungo were probably overestimated, especially
for 1965, because at that time they had relatively large expanses of
uncultivated savanna, marsh or forest land. The 1988 data originate

from the Non-Farm Strategies Survey (SESA 1988) and were derived from

114

ground measurements of fields so is probably more accurate. The data

Table 14. Farm Sizes and Production Level by Household

 

 

Farm Size (ha) Value of Production
lgtgl Per §on Rw fra O r
Butare 1.11 0.57 491 628
am 1 . 26 0 .64 605 776
Cyangugu 0.95 0.48 395 550
Gikongoro 1.41 0.71 357 597
Gisenyi 0.70 0.32 413 581
Gitarama 1.34 0.67 622 751
Kibungo 1.95 0.91 977 1150
noun. L45 mas an ' nu
Kigali 1.45 0.68 667 839
Ruhengeri 0.77 0.31 486 752
AMANDA 1.22 0.59 543 732

Data source: Mon-Farm Strategies Survey, 1988 (SESA).

 

can be used for comparative purposes, however, and illustrate the
dynamics of changing farm sizes. Some prefectures that have experienced
high densities and small farms for many years, such as Ruhengeri and
Kibuye, show little change because people have out-migrated or otherwise
adjusted to the population growth. Other prefectures that still had
fallow, forest or marsh land to settle in 1965, such as Cyangugu,
Gikongoro, Butare, Gisenyi, and Byumba, experienced rapidly declining
farm sizes as the older farms were divided among the sons. The most
striking change is the decline of farm sizes of the savanna prefectures,

especially Kigali. The farms in Kigali are no longer significantly

115
larger than those in other traditionally high density prefectures, such
as Gikongoro, Gitarama or Kibuye. Kibungo still has significantly
larger farms than the other prefectures, with an average of 1.95
hectares, but this is probably declining due to continuing in-migration
and division of the large farms between the children of the original
settlers. Kibungo also has by far the highest production per household,
especially in terms of Rwandan francs (see Table 14). This is probably
due to the large farms and high degree of banana production, an
important cash crop. The high production helps explain the attraction
of Kibungo for permanent migrants that this study has documented and for
temporary labor migrants as seen in the study of Clay, Kayitsinga, and

Kampayana (1989).

As mentioned in the section on gender-specific migration, 86
percent of the rural male heads of household who had migrated had done
so to search for land. Those who migrated to search for land now have
much larger and more productive farms than those who did not migrate:
the former group has an average 1.97 hectares of land, whereas the later
has only 1.13 hectares of land. This means that the migrants have more
land to pass on to their sons (all sons, but not daughters, inherit
equally from their parents at their marriages); the migrants have an
average of 0.88 hectares per son and the non-migrants 0.54 hectares per
son. Production, converted to kilocalorie and monetary value
measurements, is also significantly higher for the migrant household,
but the proportional difference is not quite as great as for farm size,

probably because most migrants are located in the East where land is not

116
as intensively cultivated nor as fertile. For example, the production
of the migrant household was the equivalent of 80,600 Rwandan francs
(about $785 US), but of the non-migrant household was only 50,600
Rwandan francs (about 5502). These figures suggest that those who had
migrated to seek land were very successful in their search; not only do
they have more land and higher production themselves, but their sons
will also inherit much more land than sons of non-migrants. Even the
sons of migrants, however, will not enjoy large farms and for everyone
the problem of very small farms will intensify; the national average

hectares available per son is only 0.59.

The heads of households were asked whether they felt they had
enough land for their children to inherit, and if not, what advice they
would offer their children. Only 21 percent of the answering households
felt they had enough land for their children, mostly older household
heads that may have already divided up their land. They operated an
average farm size of 1.51 hectares with 0.86 hectares available per son.
The 0.86 hectares per son is below the current national average farm
size of 1.13 hectares, indicating that even the children of those
parents who feel they have enough land would receive less than the

current national average.

Those parents who said they do not have enough land for their
children, the remaining 79 percent of the answering households, operated
an average 0.52 hectares per son. The advice these parents said they

would give to the children regarding earning a living varies depending

117
upon the size and wealth of the farm. In general, about half of the
parents would tell their children to stay in agriculture and not
necessarily to migrate; this advice includes saving money to buy land,
working for neighboring farmers, renting land, and forming cooperatives.
On the other hand, one third would suggest to their children to leave
agriculture; they suggested learning non-agricultural professions,
becoming small business owners, or studying. Since most non-
agricultural employment is outside the rural areas, these children would
most likely move to urban centers. Some heads of household said
directly that migration was the most important advice they would give to
their children; some said they would advise migration to seek land (8

percent) and others migration to seek work (5 percent).

This last group of household heads seeing no hope in their local
situation and advising migration was significantly poorer than the other
groups, with smaller farms and lower production. Their average farm
size was 0.88 hectares, or 0.34 hectares per son, and production
equivalent to 40,600 Rwandan francs. Most were located in six
prefectures: Gisenyi, Ruhengeri, Butare, Kigali, Kibuye, and Cyangugu.
The percentage of households advising migration varied from highs of 23
percent in Ruhengeri and Gisenyi and around 20 percent in Butare,
Kibuye, and Kigali to almost none in Byumba, Gikongoro, Gitarama, and
Kibungo. This group, although from small farms, did have marginally
larger farms than the very poorest group which would advise agricultural
labor on neighboring farms rather than migration. The households which,

on the other hand, would advise their children to save money to buy or

118
rent land locally were in the same range as those which would advise
migration; they did not have larger farms but operated significantly
more land for each son to inherit than those which would advise their
children to migrate to seek land. Whereas the latter operated 0.36
hectares per son, those which would like their sons to save money to buy

additional land already operated 0.54 hectares per son.

In contrast, no difference in size or production level exists
between those which would advise their children to save money to buy or
rent land and those which would advise their children to leave
agriculture and learn a non-agricultural profession. Regional
differences help to explain the different advice: the prefectures
Kigali, Gitarama, and Butare have very high percentages of people saying
they would advise learning a non-agricultural profession, and these
prefectures are nearest to the large urban centers, Kigali and Butare
cities. On the other hand, prefectures further from the cities,
Gisenyi, Gikongoro, Cyangugu, and Byumba, have a large percentage of
people giving advice to save money to buy land. Those advising non-
agricultural careers to their children also tend to be young, in their
205 and 305, whereas the older heads of household still see agriculture
as a promising career. The last important difference between those
advising non-agricultural solutions and those advising buying or renting
more land is the educational levels of their sons: the sons of those
who had given advice to learn non-agricultural professions had
significantly higher educational levels, an average of about 3 years

more education. Therefore, the parents had already supported and

119
encouraged the education of their sons because they felt a non-

agricultural career was preferable.

In summary, it seems that the families with very small farms see
no local opportunities for their children and would advise them directly
to migrate. Those without sufficient land for their children but not in
the very poorest category would advise their children either to stay in
agriculture locally and buy or rent additional land, or to leave
agriculture altogether. One difference between these groups is
regional: those near the urban centers tend to advise non-agricultural
solutions, and those further away advocate answers within farming.

Other differences are the age of the head of household, with younger
people tending to see non-agricultural careers as more promising than
agricultural careers, and finally the difference in the level of
education the sons already have, with the sons of those parents advising
non-agricultural professions having an average of 3 years more
schooling. Together, the two groups giving advice to leave their
locality, i.e., the poorest who advise migration and those who advise
shifting into non-agricultural employment, account for about half of all
households. The other half is made up of families who feel either that
they have enough land for their children to inherit, or who see local,
agricultural solutions. This is consistent with the findings in the
first section of increasing percentages of people migrating with time
and indicates that in the future this trend will intensify. It is also

consistent with the trend towards increasing urbanization with a full

120
third of households giving advice to their children to leave

agriculture.

The behavior of the sons may, of course, be very different from
the advice the parents give. Many of the same trends found in the
parents’ advice is nonetheless reflected in the sons’ behavior,
including choosing agricultural or non-agricultural paths. According to
the literature, the amount of land a family has influences the migrant
behavior of the family; for example, ”selective” urban migrants are
thought to come from large to medium-sized farms, and "less selective"
rural migrants from small or landless farms (Findley 1981). In Rwanda,
however, little difference was found between the sizes of farms of the
families with sons who had migrated short distances to rural areas and
those with sons who had migrated either long distances to rural areas or
to urban centers. The only clear indication of the influence of farm
size on migrant destinations is that the sons from the very poorest
farms, those of less than half a hectare, tended to remain very close,

within the commune.

Although farm size was not found to be a significant influence of
the destination of migrant sons, distinguishing individual
characteristics, especially differences in educational levels, were
found between the sons who went to different destinations. For example,
the urban migrants and long-distance rural migrants had significantly
higher educational levels than those who remained within their

prefecture. Differences between the heads of households were also

121
important, with younger heads of household and households near the
cities more likely to advise non-agricultural, urban-based, careers for
their children. Therefore, in Rwanda, the advice of a household head on
the destination of a migrant son is not as influenced by farm size as
much as by other household and individual characteristics of the

migrant.

In Rwanda, the reason a migrant son leaves is linked to the size
and production levels of his family farm; those from the largest farms
tend to migrate for either study or marriage reasons whereas those from
smaller farms leave to seek land or work. Formerly, migration to seek
land was more common, but more recently migration to seek work is the
norm. This shift could be due to declining availability of free land
and the increasing attraction of non-agricultural opportunities in the

urban centers.

The job the migrant son assumes is also often linked to his
family's farm size; an example is the difference in average family farm
size between sons who become agricultural laborers or household servants
and those who take other jobs. Agricultural laborers and domestics come
from significantly smaller and less productive farms than sons who
become farmers themselves, with only an average of .32 hectares
available for the sons who become agricultural laborers and .22 hectares
for the sons who become domestic servants. Their family farms were too
small or poor for them to inherit, buy, or rent sufficient land to

become farmers. The sons who were working as agricultural laborers and

122

domestics had recently left the home, on the average two years previous
to the survey, so it is likely that those jobs are a temporary solution
for a son from this situation of near landlessness. Craftsmen, in
contrast, come from somewhat similar sized farms as those sons who had
become farmers, but the craftsmen have a higher level of education than
the farmers. Civil servants and small business owners also come from
similar sized farms but have attained higher levels of education; the
difference between civil servants and small business owners is that the

civil servants are better educated and older.

Therefore, those sons performing agricultural labor or acting as
domestic servants come from very small farms, have little education, and
tend to have been gone only a few years. This indicates that these jobs
may be temporary solutions for young men from poor farms or that this
situation of near landlessness forcing the sons to migrate is a
relatively new phenomenon. On the other hand, students, civil servants,
small business owners, and craftsmen come from somewhat similarly sized
or larger family farms than sons who remained in farming, but they have

varying degrees of higher levels of education.

In summary, although the size of the family farm is not directly
related to the destination of the migrant son, it has many indirect
effects. These include whether the parents feel they have enough land
for their children to inherit and what the parents would suggest their
children do in the case that they do not have enough land. The sons of

the families which feel they have enough land either did not migrate or

123
migrated very near to the family farm. With time, however, this
situation has become less common. More recently, sons migrated to seek
land and went to settle rural areas increasingly distant, but the latest
trend is that sons have left to seek work either in various rural areas
or in an urban center. The sons from very land-poor families have
migrated to perform agricultural labor or to act as domestic servants,
but other occupations seem to be tied to the educational level of the
individual rather than the size of the family farm per se. Finally, a
regional comparison of farm sizes and production levels shows that the
disparities between large farms in the East and very small farms
elsewhere have diminished to uniformly sized small to medium farms. The
far East, however, still has significantly larger farms and higher
production, which helps explain results presented earlier on the

attraction of migrants to the East.

4. Age and Marital Status of Migrants

Migration in Africa often occurs at a certain stage in the life-
cycle, according to the literature: around marriage for rural migrants
as they begin a new family and start a new farm, and earlier for single
young men going to cities to earn money (Zachariah and Condé 1981).
Therefore, the rural migrant is expected to be older, at least 30, and
married, whereas the urban migrant is expected to be much younger,
between 15 to 25 for "very selective," or 15 to 30 for ”less selective"

migrants, and unmarried or migrating alone (Findley 1981). To examine

124
this in Rwanda, the age at time of migration was compared between rural
and urban migrants, and between different educational levels and
professions of migrants to the city.’ ‘The marital status of non-
migrants, rural migrants, and urban migrants between the ages of 20 and
29 was then compared to see whether indeed a larger proportion of the

urban migrants was unmarried than that of the other groups.

The age at time of migration does indeed differ between migrant
groups, but the difference is not as great as expected from the
literature. The average migration age for all rural-rural migrants was
28 and this has not changed significantly over time. A wide standard
deviation, 11, around that figure suggests, however, that permanent
migration to rural areas does not depend strictly upon reaching a
certain age. Migration to urban areas, on the other hand, occurred
significantly younger than for migration to rural areas, on the average
at age 25, with little difference between cities and little change over
time. The standard deviation, 8, is smaller than for rural migration,
suggesting that this migration behavior is much more tied to reaching

the age group of the early to mid-twenties.

Differences in age at time of migration to urban centers do exist
between professions; for example, those performing domestic service
migrated much younger, around age 22, than any other group. Professions

such as government administration and educators had last migrated to the

 

’ Only those who had migrated when they were 15 years or older were
considered in order to exclude migration of children with their parents.

125
city around age 29, whereas merchants, construction workers, craftsmen,
and industrial workers migrated around age 26. Similarly, the higher
the educational level of a migrant, the older he was at time of last
migration to the cityu‘ Men with less than secondary schooling
migrated around age 24 and those with higher levels of education at
around age 26. The age at time of first migration to the city for
highly educated men may have occurred when they were much younger and
came to study; data from the Non-Farm Strategies Survey confirms this,

with the average age at time of migration for students only 16.

To summarize the differences in age at time of migration between
groups, therefore, the rural-rural migrants leave at a significantly
older age than the rural-urban migrants but the difference between them
is not as large as expected. The urban migrants with little education
or those who perform jobs requiring little education, such as domestic
servants, migrate much younger than other, better-educated migrants
working as, for example, educators or government administrators. This
latter group may have originally migrated much younger to study, at

around age 16.

Similarly, the marital status of urban migrants is very different
from that of rural migrants or of non-migrants. As seen in Table 15, of

those men in their twenties who migrated to Kigali City, 70 percent were

 

‘ The Post-Census Survey data only contains the year, and therefore
the age, at time of last move. Therefore, for migrants who have moved
more than once it overestimates the age of first migration. This is
likely to be the case for those who migrated to attain schooling.

126

Table 15. Marital Status of Migrants and Non-Migrants Ages 20-29
(column percentages)

 

 

Non-Migrant Short-Distance Long-Distance Butare City Kigali
City
Marital Status (rural) Rural Migrant’ Rural Migrant2 Urban Migrant Urban
Migrant
Single 50 52 37 61 70
Married (one wife) 48 46 60 38 29
Married (polygamous) 2 2 3 0 0
widowed/Separated] 1 1 1 1 0
Divorced

’ Rural-rural migrant within the prefecture of birth.
Rural-rural migrant outside of prefecture of birth.

Data source: figgggte ngggraghiggg ngt-ggggitaire (Kigali: Min. du Plan, 1981).

 

not yet married, compared to 37 percent of long-distance rural migrants
and 50 percent of non-migrants. One reason for the dissimilarity
between the groups was illuminated in Non-Farm Strategy Survey data
discussed earlier, in which the reason for migration of sons of heads of
household differed between rural and urban areas: most migrants to
rural areas did so for marriage reasons or to seek land, so were
probably married or preparing for marriage, whereas most migrants to
urban centers did so to seek work, often to save money for marriage at a
later date. Again, similar to the differences of age at time of
migration, differences of marital status exist between professions; for
example, domestic servants tend to be unmarried whereas the more skilled

workers, such as those in construction, are more often married.

127

Although this study attempted to exclude circular migration by
examining only men who had been living in their new residence for more
than one year, the high incidence of young, single men migrating to the
cities brings into question the permanence of their stay. Many of them
may have come to the city to earn money for marriage and may later
return to live permanently in rural areas, as suggested by Findley
(1981). On the other hand, the steady increase in numbers of urban
migrants seen in Figures 5 and 6 suggests that many of the earlier urban
migrants did remain in the cities, and therefore many of the more recent
urban migrants can also be expected to stay. What proportion will
remain in the cities cannot be estimated from this data and will depend

upon the relative opportunities in the rural and urban settings.

Therefore, the pattern seen in the literature of older, married
men migrating to rural areas and younger, single men migrating to urban
centers is generally the case in Rwanda. A wide variation in age at
time of migration to rural areas, and high incidence of unmarried men
migrating short distances to rural areas, however, indicates that many
exceptions exist to this general rule. The urban migrants are also a
diverse group, with differences in age at time of migration and marital

status seen at various educational levels and professions.

V. DI C 5 ON

What has just been presented of the migration patterns in Rwanda
will now be reflected upon in light of the theories of migration and
development introduced in Chapter II. This discussion will around the
time periods and the factors important for explaining migration activity
during each phase as presented in the table "Changes in Factors of
Migration with Development" (Table 5). At the end of this chapter, a
revised version of this table incorporating the results of this study is

presented.

The settlement and migration patterns in Rwanda have been the
object of past research and analysis because of seemingly simple but
powerful forces which appear to have influenced the patterns (Gourou
1953, Prioul 1981, Cambrezy I984). The early observers had already
noted the unusually high concentration of settlements confined to the
Western Highlands region, a region which possessed high agricultural
potential with ample rainfall and fertile soils and which was
comparatively healthy with moderate temperatures and few diseases
(Gourou 1953). Around the time of independence, changing circumstances
resulted in intense migration activity which eventually redistributed
the population of the country. This study examines the importance of

one of those circumstances which had been assumed to be the most

128

129
important causal factor of out-migration from the Highlands, population

pressure (Cambrezy 1984, Prioul 1981).

The population pressure literature places migration as one of a
series of strategies that people use to cope with the problems related
to population pressure, but usually only after other strategies have
been adopted and the problems remain (Grigg 1980a, Bilsborrow 1987).
The responses to population pressure in Rwanda have been similar to
those that Grigg (1980a) suggested are common elsewhere. The earliest
and perhaps most important of these responses in Rwanda has been a
classical Boserupian intensification of land use (Boserup 1965), with
the agricultural system in the Highlands and Foothills moving to double
and triple cropping, reduction in land allotted to fallow, cultivation
of all nearby land, and switch to production of high-caloric crops such
as tubers. The result has been an increase in yields from the land, but
at the expense of declining yields from labor and possible land
degradation (Meach 1986). As an extension of this process, substantial
short-distance movement occurred in the periods before and soon after
independence as nearby forest, fallow, and swamp land was converted to
cultivation. The second response, also expected from the literature,
has been growth of non- and off-farm employment; half of all farm
households are now involved in these activities. Those from very small
farms tend to perform agricultural labor, which makes up 31 percent of
non- or off-farm employment. The remaining 69 percent is work outside
the agricultural sector which is performed by people from farms of all

sizes and often includes the better-educated (Clay, Kayitsinga, and

130
Kampayana 1989). A third response the literature would lead us to
expect is change in fertility patterns, but there is little evidence of
this occurring in Rwanda. Finally, the fourth population pressure

coping strategy the literature suggests is long distance migration.

This study examines this population pressure/out-migration
relationship by testing the importance of population pressure, as
measured by regional population density, in explaining out-migration
rates during three time periods. For each time period, the relationship
differed. During the first, around 1948, migration and population
pressure were not closely related and out-migration was tied to other
factors such as political movements and labor migration associated with
the colonial system. The densities at that time, although unusually
high for Africa, were apparently not at the point at which people could
not adjust locally. By 1970, the situation had radically changed. The
population densities had doubled in those 22 years to become a
recognized national problem and symptoms of ”population stress“ (Grigg
1980a) appeared, such as very small fragmented landholdings, a decline
in numbers of livestock, land degradation, and declines in labor
productivity. Also, the post-independence government had opened up new
land in the East, previously reserved for pastoralists, to be cultivated
by the farmers of the Highlands. The result was a massive departure
from the densely populated Highlands and later from the Foothills to the
Eastern Savanna. Long-distance rural-rural migration changed from being
a rare occurrence before independence to an increasingly common event,

growing 30 percent annually in the late 19605. A regression analysis

131
shows that the single variable high population densities at the origin
explains 64 percent of all variation in rates of out-migration during

this 1970 period.

The densities in the Highlands and Foothills continued to augment,
even as the densities of important receiving areas such as Kigali
prefecture doubled in less than eight years from the influx of migrants.
Therefore, although out-migration may have helped reduce the rate of
population growth in these densely populated areas, it did not ”solve"
the population growth problem. The ability of the densely populated
regions to continue to absorb population increases is similar to what
Gould called the "sponge effect" in Kenya (Gould 1985). As comparisons
between carrying capacity studies in Rwanda show, estimates made thus
far on the "limit” of the land to support a certain density of people
have been breached as the population continues to grow. Adoption of the
range of coping strategies mentioned above has changed the assumptions
made in past Rwandan carrying capacity studies, i.e., that people were
dependent on their own static agricultural production for food and would
not develop or adopt different agricultural technologies or sources of
non-farm income. The dependence of people in Rwanda solely on their own
agricultural production is no longer the case. New carrying capacity
models must incorporate coping strategies such as new agricultural
technologies, local non- and off-farm employment, and short- and long-
term migration. The ability of the rural population to adopt these
alternative strategies depends on many factors, some regional, such as

the growth of the local economy, the agricultural potential, and the

132
proximity to urban or other rural employment. Therefore, the tenet of
Malthus that the land is a finite resource which can support only a
certain number of people cannot be accepted at face value in Rwanda,
where people have quickly adopted new agricultural and non-agricultural

sustaining strategies.

An offshoot of the question of the relevance of carrying capacity
in understanding out-migration is the importance of population pressure
in receiving areas in the attraction of new in-migrants. In Rwanda, the
primary destination choice of in-migrants changed over time as the
population densities of the earlier choices swelled. Rural-rural
migrants initially went to areas near their home, but succeeding
migrants traveled further and further east, passing over the
destinations of previous migrants. Communes in the East often
experienced rapid in-migration for a few years, and then were passed
over. The explanation of this behavior is not simplistic. The common
density at which communes ceased to attract increasing numbers of in-
migrants, the ”saturation density," is linked to time period; around
1970, for example, the figure was around 70 people/knf, but by 1978, it
had apparently increased to about 150. Similarly, areas which had
previously lost their status as major destinations began again to
attract new migrants by 1980 when their densities were around 200
people/kmfi Therefore, perceptions of the "saturation point" of
potential destinations changed with time as densities throughout the

country rose. In their choice of destinations, migrants appeared to

133
consider the relative merits of several rural locations in terms of

possibilities of gaining land.

The predominant out-flow during this period was to other rural
areas, areas which offered free and available land for agricultural
settlement. The attracting force was, therefore, the availability of
economic opportunities in the form of land. Although the initial
impetus for in-migration to the Eastern Savanna may have come from the
government, which was promoting newly organized settlement schemes ("leg
paysannats"), spontaneous migration soon overwhelmed the paysannats.
People colonized land surrounding the paysannats and other frontier land
as they continually moved further east (see Figure 14). This pattern of
movement from densely-populated Highlands to drier, more marginal, land
previously used by pastoralists is similar to movements seen in Uganda

and Kenya.

The predominance and scale of the rural-rural flood in Rwanda is
probably not unusual in Africa, but although its presence has been
noted, rural-rural migration has rarely been explicitly examined. For
example, the study of Riddell and Harvey (1972) of step-wise migration
to urban areas in Sierra Leone had inconclusive results because of the
attraction, at times predominant, of rural-rural movements. This lack
of attention to rural-rural movements is surprising considering that
most increases in food production in Africa have been attributed to the
large increase in cultivated land (65 percent increase between 1950 and

1976 [Grigg 1980]).

134

 

 

1945 - 1961

 

1972 - 1976 1977 - 1980

 

Figure 14 Rural-Rural Migration During Four Periods

135

The predominance of rural-rural movements in Rwanda is a
reflection of the strength of the attraction of rural areas for
migrants, or of rural pull factors in the push-pull model. In Rwanda,
the pull factor that drew migrants was the availability of land for
settlement; land offers potential economic opportunities as do jobs or
wages, which are the usual measure of attraction in migration models.
To measure the importance of land in determining in-migration,
population density of the receiving areas was used as a surrogate
measure of the availability of free land. During the 1970 period, low
population densities did account for around 37 percent of the variation
of in-migration rates. Another factor, that of distance, was also
important in the decision-making process of the migrants; instead of
heading for the area with the smallest possible density, migrants chose
to settle in nearer areas with densities still low, if not the lowest of
the country. This conforms to the general understanding of migrant
behavior, of distance moderating economic opportunities, but new is the

explicit examination of land as an important pull factor.

In the 1978 period examined, the distribution of regional
densities had changed from one of extreme contrasts within a small area
to one of only relatively more populated areas over a uniformly dense
surface. The intense rural-rural migration of the previous 15 years had
redistributed the population of the country so that free and available
land even in the East was less common. Increasing densities in the
traditionally dense Highlands had led to growing population pressure and

”symptoms" of population pressure such as near landlessness and reliance

136
on agricultural labor for low wages by the land-poor, were becoming
increasingly evident. These changes reduced the ability of migrants to
find available productive land in the East or elsewhere and made rural-
rural migration much less economically promising or attractive.
Nevertheless, rural-rural migration to the East remained important and
low densities at destination sites still explained 41 percent of rural
in-migration rates. The high rates of rural-rural migration of the
previous 1970 period had declined by 1978, however, and the more
uniformly high densities throughout the country significantly reduced
the ability of variations in population density to explain rural-rural
out-migration rates. Another significant reason rural-rural migration
in 1978 had become more difficult to predict with tests of population
pressure is that a new trend of migration, migration to cities, had

gained great importance.

Before proceeding with a discussion of urban migration, a rural-
rural migration model is considered which incorporates the ability of
population pressure to explain out-migration, available land to explain
in-migration, and distance as a moderating force. These three factors
are presumed to be behind the unidirectional flow of rural-rural
migrants from high to nearby low density areas. The model is based on
the gravity model, but in this version, the population density of the
destination is a surrogate for attractiveness, the density at the origin
is an indication of potential out-migration, and the physical distance

between the two prefectures is a measure of distance decay. Two

137
versions were tested: one with equal weight given to the population

variables, and one with the origin population variable weighted double.

The models performed differently in the three time periods
analyzed, reflecting the differences in the importance of population
pressure, the attractiveness of land, and distance decay in any rural-
rural migration activity during each period. For example, in 1948,
neither model was very successful in explaining migration rates;
population pressure was not an important factor of long distance out-
migration. At that time, people in the dense areas were still able to
cope with the situation without resorting to high levels of out-
migration. In 1970, however, the models were much more successful in
predicting migration rates, especially out-migration rates from
prefectures and the migration stream rates between prefectures. During
this period, the densities had increased in the Highlands to the point
at which many people had adopted out-migration as a coping strategy, and
the predominant destination of these migrants was to nearby rural areas
to seek available land. At this time, the assumptions of the models met
the actual conditions. The results of the model support the p0pulation
pressure/out-migration hypothesis, but also support the hypothesis of
the importance of available land in predicting the destination of rural-
rural migrants, along with the moderating effect of distance. Of the
three factors incorporated in the models, the most important for
explaining out-migration is population pressure. The interaction
between prefectures is best explained by population pressure and

distance and in-migration by availability of land and distance. By the

138

1978 period, the situation in the country had changed and the models do
not fit the actual movements as well. The variations of density between
prefectures were less pronounced and the migration rates less tied to
regional density comparisons. Also, rural-rural migration in this last
period constituted only 60 percent of all migration, since rural-urban
migration had gained substantial importance, and this complicated the
migration pattern. Therefore, the performance of these versions of the
gravity model varied with each time period, as the relative importance

of the forces underlying migration patterns changed.

By 1978, one change not directly related to population pressure
had begun to influence potential migrants in their choice of migrating
and their choice of destinations--the growth of non-agricultural
opportunities in the cities, especially in the capital city. Much as
has happened in other African countries, and as predicted by dual-sector
economists and others, Rwanda experienced sudden and rapid increases in
rural-urban migration. Another similarity Rwanda has with many other
developing countries is that in-migration and other measures of growth
have been concentrated in one city of the country, in this case the
capital, Kigali. In-migration to Kigali grew at a 22 percent annual
increase from 1970 to 1978; the number of migrants increased from less
than 300 before 1970 to 6,000 per annum in 1978. This growth is much
faster than the 10 percent estimated by Prioul (1981) for that period,
and at that rate the proportion of the national population living in
Kigali could be expected to dramatically increase. This rate is so high

due in part to the extremely low level of migration into Kigali before

139
1970. The numbers of migrants into,Kigali will probably augment each
year, but the rate will probably decline in the 19905 to around a 15
percent annual increase, and by the 20005 be close to Prioul’s 10
percent estimate, which is approximately what the nation was
experiencing in total migration was from 1962 to 1980. Even so, this
translates into large numbers of in-migrants. Migration to Butare and
other prefecture capitals has been relatively modest, although they have

also experienced some acceleration in migration in recent years.

The concentration of migration and economic activity in Kigali can
be expected to continue, due to cumulative causation and agglomeration
economy factors, rendering Rwanda similar to many other African
countries with a growing and dynamic primate city. The implications of
this differential growth pattern are potentially serious, considering
the constricted land area around Kigali suitable for housing
construction or industry and the concentration of social ills often
associated with large urban squatter populations. Encouragement of
migration to secondary urban centers, for example to Butare or other
prefecture capitals, by the government may have limited success,
however, since the few existing economic opportunities are now so
focused in Kigali. Experience in other countries has shown that the re-
channeling of migration is very difficult without offering the
attraction of relatively better economic opportunities in alternative
areas. The problem may not become one of excessive migration into
Kigali but of poor conditions for migrants in Kigali and limited

opportunities elsewhere.

140

due to the unplanned but the ability of the government to effectively
channel migration elsewhere may be limited due to small distances and

ease of travel between areas in the country

The recent expansion of urban-destined migration was apparently
unrelated to growth in the industrial sector or any other sector in the
formal economy, but to a wide range of opportunities in the informal
sector which did not require high levels of education (Prioul 1976).
Therefore, the urbanization process was not tied to “industrialization"
per se, as many authors presumed, or even to a concentration of services
and economic growth activities by the government that could lead to a
reference of urban bias. Also, non-agricultural income-earning
activities were very important, if not highly remunerative, for people
remaining in rural areas, so the urban centers did not have a monopoly
on the growth of non-agricultural sector activities, contrary to what
many dual-sector theorists would presume (Lewis 1954). Instead, the
growth of the capital city was spontaneous and service-sector based; the
informal economy seemed able to absorb most migrants as they arrived.
The in-migration seemed to be a reflection of perceived urban economic
opportunities, if not actual industrial or other formal sector jobs, as
described by Todaro (1969); in a broader sense, it is a reflection of
the choice of perceived economic opportunities in the cities and in

various rural areas.

141
Therefore, by 1978, a new influence on potential migrants had
become important: the attraction of urban-based economic opportunities.
This altered the spatial pattern of migration: 40 percent of all
migration activity became rural to urban and the predominant destination
was the capital city. Migrants came from all regions of the country and
regional out-migration rates were not significantly related to any

measure of population pressure; the draw of the city was universal.

Before the sudden increase in urban migration, most migrants to
the city had come from a select, well-educated group, whereas rural
migrants and non-migrants had little differentiation of educational
levels. This changed after 1970 and the percentage of urban in-migrants
with less than complete primary school steadily increased, lowering the
average number of years of education of in-migrants. This shift was
concomitant with the decline in rural opportunities for less well-
educated men as available land became scarce, and reflects the wide
rural/urban income differentials for off-farm activities. Since around
1975, the income-maximizing potential migrant has faced little choice
between rural and urban destinations, even if he has had little
education. If these trends continue, the average educational level of
urban migrants will continue to decline and the ability of new migrants
to integrate into the formal economic sector and be productive in, for

example, an industry, will diminish.

Another change has been the increasing and now pervasive tendency

of men with secondary school or higher to migrate to the city since

142
nearly all employment opportunitiesffor the well-educated are located in
the cities; this has repercussions on the rural areas from whence they
came. The rural areas have lost the very people in whom they have
invested most and who have potentially the most to offer, but they may

be somewhat compensated by remittances from the urban areas.

Besides the level of education, a characteristic found in the
literature which often differentiates between non-migrants, rural
migrants and urban migrants is the ownership status or size of the
parents’ farm. Those from the "selective” class that go to the city are
expected to come from large farms or, if of the "less selective” class,
from smaller farms, whereas those going to rural areas are expected to
be mostly landless (Findley 1981). In the microfundia situation of
Rwanda, very small family farms are the norm and three-quarters of the
families feel they do not have enough land for their children to
inherit. Therefore, the question becomes one of choosing between
various non-agricultural options, most of which would entail leaving the
rural areas or gaining additional land through purchase, renting or
migration. In general, it is only the extremely land-poor families who
explicitly mentioned migration to seek land or work as the best option,
whereas those families from very small farms with a bit more land for
each son prefer that their sons remain and attempt to gain additional
land locally. There is little difference in farm size between this
latter group and those advising non-agricultural careers for their
children, but these children had been better prepared for such careers

with a higher level of education. Actual migration behavior by the

143
sons, however, is difficult to tie to parents’ farm size, except that
sons from the very poorest farms tended to migrate only very short
distances. Therefore, in Rwanda the pervasive situation of very small
farms and insufficient land for children to inherit has meant that the
large majority of families must consider options for their children
beyond earning a living from the family land, and it is not possible to
predict by the size of the families’ farms which alternative option they

consider best for their children.

Similarly, in Rwanda less difference than might have been expected
from the literature was found between urban and rural migrants in age at
time of migration. Rural-rural migrants were older than rural-urban
migrants, but only by a few years. More important age differences were
found between urban migrants performing different jobs: those in jobs
requiring little or no education migrated younger than the others. The
marital status of migrant groups was, however, very similar to what was
expected from the literature, with rural migrants having a strong
tendency to be married, but urban migrants not. This indicates that
migrants to urban areas may be in a different stage in their life cycle
than the rural migrants; whereas the rural migrants are setting up their
own families in their chosen permanent location, the urban migrants are
still preparing for or are delaying entering this stage. This brings up
the question of the male urban migrants’ permanency in the cities; since
only 2 percent of migrants to the city are women, many men will
presumably return to rural areas for marriage.' On the other hand, the

data indicate that many urban migrants are remaining beyond two or three

144
years and that the numbers who are coming and staying are rapidly
increasing, so one consequence may be a rapid increase in the numbers of

single women or women head of household in rural areas.

The adjustments that have occurred during the transition from
rural-rural to rural-urban migration brings up the question of the
decline of distance decay on migrant activity over time. The physical
distance between places has become much easier to traverse as roads have
been laid and eventually paved. Although Rwanda at independence had
very few all-weather roads, by the 19805 it had the highest density of
paved roads per square kilometer of any country in Africa (A10 1987) so
that traveling from one end of the country to the other became a trip of
a few hours. Therefore, although the actual distance rural-rural
migrants traveled steadily increased as they sought land further and
further east, the relative distance shrunk. The cultural distance
between places also changed over time; before independence, a clear
demarcation between ethnic groups was maintained in the Eastern Savanna,
which was reserved for the Tutsi pastoralists. After independence,
these barriers were broken down. Once the initial group of cultivators
from the Highlands settled the savanna land, other cultivators must have
found the cultural distance to the savanna much less daunting.
Similarly, the cities were initially the domain of the better-educated
and others linked to the formal economy. As opportunities in the cities
grew for less-well educated migrants and the average educational level
of in-migrants declined, the rural, less well-educated migrant would

have found the life of the city less distant or forbidding.

145

The shift in Rwanda of the relative attractiveness of rural and
urban destinations, and the resultant decline of rural-rural migration
and growth of rural-urban migration, was similar to that described by
Zelinsky (1971) in his "hypothesis of mobility transition." His
generalization that rural-rural frontiersward migration would peak early
and rural-urban migration would peak afterwards reflects the situation
in Rwanda. Also, Zelinsky’s linking of the "vital transition"
(demographic transition) and the ”mobility transition" phases is
interesting considering that one of the important forces of out-
migration in Rwanda has been population pressure. He ignores, however,
the common factors behind the two transitions, namely integration into a
wider society. In Rwanda, this has meant an introduction to new
medicines and a decline in death rates, leading to an increase in
population growth, as well as changes in political, environmental, and
socioeconomic factors encouraging first rural-rural migration and later

rural-urban migration.

The "development paradigm of migration" of Brown and Sanders
(1981) incorporates other factors underlying migration and considers how
the importance of various factors changes over time, but unfortunately
misses some of the most basic of the processes that have been important
in Rwanda and probably elsewhere. The strongest force in their model
acting on migration is industrialization and the resultant rural/urban
income differentials, so their model is confined to rural-urban and

urban-urban migration. This has only recently started to become

146
important in Rwanda and then not because of industrialization per se but
from growth in the informal urban sector. Brown and Sanders did not
explicitly include the force of population growth and responses to
population pressure, and left out entirely any rural pull factors such
as land. Therefore, although their general framework incorporating
various factors was useful, in Rwanda the factors that were very

important were not included.

A revised table, Table 16, incorporates these missing factors from
Zelinsky and Brown and Sanders and encompasses the general findings from
this study. Those findings that differ from expected following the
literature review are printed in italics. The table is structured
around factors important to migration patterns in Rwanda. For example,
the population pressure ”push” out of the older rural settlements
occurred after a period of increases in local land use intensity.
Similarly, the "pull” of other less-populated rural areas was very
apparent, and the expected massive rural-rural migration did ensue.
Rural-rural migration then declined as rural farm opportunities in the
form of available land became very limited throughout the country. The
forces behind the rapid increase in rural-urban migration included rural
push factors, such as population pressure, as well as rural/urban income
differentials and job and wage opportunities in the formal and informal
sectors of the cities. These forces encouraging rural-urban migration

are now predominant in Rwanda and will probably intensify in the future.

I47
Changes in Factors of Migration with Development in Rwanda

Table 16,

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

acmuuoaaw mum> :1: «mafia» mua>u=1=
.cmnumu oHuSMH mommamo
mamCMHoou bcoam>oa =-m .bcucuuum”=-m Han can unbeaten monsoon
annum goo co Hen you noboocma non co noocusuc
unnumcmzmmmcm m-m -cmaummcanm-m a Hanan .cmcumm nae a sauna
ucmsm>os 21m m>wmmmi museum .uo Hum mom who who Ham mom
awn scum ”31m hauooo .Qmm acmowacuMmuaum Hams sumuumn mamuwCSu hummoom
ucmam>oa no mHuuMH.mEumm .mmmmmno and mom mmmH you you .Acmfi &5& income Hm:o«uwm:aua
m-m acmcugoaam Hanan scum ”m-m .cmcumm uncommon ucaoacmzmmmcmum-m .Mmcuam .amcumm .cwnuwm. bosom own;
oocmuuomaw :« cos munch 30H momma
mcflmmouocw mason =um .umIHHm: ”21m .mmomHQ mucuumuumn you who hemN coma >u0woom
:3 Hanan moman\mnon "=-m 203308 sun: moon“ encamommcaae
mama nmwucoum cu mam maumm oou manna ummfl .Ammfi.hok mom hex «emu nah manta:
oocacmmo used can: ”deem scum "m-m uncommoaa .808 name sea "m-m .mwmwm .cmcumm -muwnuwm .uwzuwm
AanoHo>mu uoh uo: moo mo ucmflumm
mmmuwov :um mauuwa :08 maso> huwoumom HHmam you mmmmmao hummoom
.uo-HHma "=-m name Hnoofl momaz\mnon ”=-m Humoom moo co a encamummcuae
mewxmmm mama mom mum magma umHHmam "musmmouo mama flan new mom «0 Hanan mom ucmcM hanmm
mocmumfio unocm m>mmmms some ”mum :0wumasooa mHanwm>mumum .mwcuwm N Hanan madnoum uaououm
asummom ma a non «0 u:03flmm came
mcommmu museummsnco: AufimcmuCM mm: Hanan >um> now inflamed
Macmuwfioa no Hmwoom can meadow“: mama mcwmmmuocw mmwm3\mnonu=sm memo“ no N acme“ annum hummoom
you acmam>oa maom cmozumn c0wumwu "mmmmuOCM mama unacwmeM Hanan um«:flwm:w iaoumua amcowuwnmuh
_coMbaumma maubmm -cmumccmo masses camouflaged manaemmsaum-m sums
amauomcH Hmauom Bummicoz Baum
sewumumw: mo mumumwamuomamco Afimusuv cane: Hausa
cumuumm Hmwumam acmumH: swam Hasm mamaucmummmMo zuwcsquQEO awiocoom
3
6683059

acmuawum: no awoaoam

 

 

 

 

 

148

Other aspects of the factors of migrations are slightly different
from expected and as listed on the original Table 5; they have been
altered for Table 16. For example, instead of farm opportunities being
significant only for the poor once land became scarce in the "Middle
Transitional Society,” as the literature predicts, farm opportunities
remained important for those who were rich enough to retain sufficient
land. Those who were relatively land-poor had to search for alternative
opportunities in the local non-farm sector or by migrating to either
rural or urban areas. Similarly, non-farm opportunities in the rural
areas were not just for the less well-off, as was expected, but also for
those with education or skills that could take advantage of
opportunities. The original table indicated a rather rapid decline in
importance of the informal urban sector, but this may not occur as fast
as indicated, considering the very small formal sector in Kigali. One
of the differentiating migrant characteristics mentioned in the table,
that rural-rural migrants are from small farms, could not be
substantiated with our data due to the relative uniformity of small

farms in Rwanda.

In summary, the spatial pattern of migration in Rwanda has
followed the expected pattern. The rural-rural movements were forecast
from what had been known of the various push, pull, and economic
opportunity differentials important in Rwanda, if the scale and rapidity
of movements were unusual compared to what most studies have indicated
in other African countries. The ability to predict the actual pattern

with a rural-rural migration model was successful only during the period

149
when the factors influencing migration were clear, but for other periods
in which the situation was more complex the model was less successful.
The rural-urban movement also began as expected from the literature.
The reasons behind the movement are not as closely tied to
industrialization or another job-creating force as expected, but to
growth in the urban economy apparently arising from the influx of new
population. The individual characteristics of gender and position in
the life-cycle have remained important in differentiating between non-
migrants and migrants to various destinations. One change that has
occurred is the role of education; the barrier of needing a high level
of education to enter the urban economy has changed and now migrants to
the city are no longer only those with secondary education or higher but

from the mainstream of society.

VI- W

The pattern of migration in Rwanda has constantly evolved,
changing shape and intensity, during the slice of Rwandan history that
this study examined. Migration has been a dynamic process reflecting
the changing importance of social, political, economic, and
environmental forces. The migration flows themselves have then affected

these forces, creating an interwoven set of processes.

During the first time period that this study examined, migration
activity was tied to various political and social factors, and the
resultant movements were small in scale and dispersed. The settlement
pattern was then closely associated with favorable environmental factors
for agricultural production in the western part of the country. After
independence, increasing population pressure resulted in changing
economic circumstances, such as a rapid decline in farm sizes and
available land per person. One response was a high rate of out-
migration from the areas experiencing the most pressure. The
destination of these migrants was influenced by political factors; the
government was interested in settling land previously used for pastoral
activities so it promoted organized settlement schemes in the East. The
availability of this and other land in the East was an important
economic incentive for the in-migrants, and the unidirectional and

intense flow was very focused from the high density areas to the

150

151
settlement scheme region. This large influx of migrants moving
constantly further and further east indeed settled this land, and the
population of the rural areas in the East suddenly soared. The result
was a redistribution of the country’s population so that the East also
began to experience a decline in economic opportunities as productive
land available for new in-migrants or for local young people became
increasingly scarce. In-migration to the rural East then began to
decline, reflecting the change in economic opportunities there. Around
the same time, new economic opportunities had begun to develop in the
capital city, Kigali, especially in the informal sector. Therefore,
although previous in-migration flows to Kigali had been very small and
confined to the well-educated, now the number of in-migrants suddenly
multiplied and the migrants represented a much broader spectrum of
society. These urban economic opportunities attracted migrants not just
from the traditional out-migration regions of the West, but from
throughout the country. This new development created another migration
pattern in the country and has changed the economic and social character

of Kigali.

The role of migration in adjusting to population pressure and the
lack of rural economic opportunities is in the process of change since
the bulk of migration flows are no longer permanent moves by families
seeking land but increasing by waves of single young men seeking
employment in the city. Therefore, not only are the socioeconomic
processes that influence migration activity evolving, but the impact of

migration in the rural and urban areas is changing.

152

In summary, the factors affecting migration in Rwanda have been
different for each time period, and migration patterns have evolved
reflecting those changes. One consequence of realizing the difficulty
of separating the migration outcome from the evolving social and
economic processes is that prediction of future migration becomes very
difficult without fully understanding what the future dynamics will be.
As seen in this study, interpreting even past migration patterns using
current models and hypotheses has had limited success due to the

complexity of the evolving situation.

The framework for examining the evolution in migration patterns
was drawn from works of Zelinsky (1971) and Brown and Sanders (1981).
Their structure of "migration phases“ was useful for the time-frame
analysis of this study, and spatial migration patterns in Rwanda have
broadly followed those generalized by the authors, with the exception of
rural-rural migration not having been predicted by Brown and Sanders.
The forces important in the changing patterns in Rwanda are, however,
different than those emphasized by the authors, as indeed the importance
of forces affecting migration are probably different in each country.
The usefulness of this study, as the first case study to explicitly test
the migration phases framework, is to illuminate the importance of
incorporating rural coping mechanisms often overlooked in traditional
migration analyses. These mechanisms provide opportunities that are

important in moderating the rate of out-migration and in determining the

153
destination of migrants, such as the exploitation of local non-farm

opportunities and rural opportunities in distant regions.

The assumption made in this study and many other studies is of a
dual-sector economy corresponding to the agricultural and industrial
sectors clearly divided between the rural and urban areas. This
assumption, however, has become less valid over time in Rwanda as the
two economic sectors have become increasingly spatially mixed. Half of
the people in rural areas are earning part or all or their income with
off- or non-farm activities, and "urban" functions such as commerce and
small-scale industry are developing outside of the urban economic
system. A wave of young men are migrating from the rural areas to work
in the informal sector of the city for a few years, blurring the
distinction between the rural and urban informal sectors. The country
is experiencing increasing "urbanization,” both in the cities with a
wider segment of society incorporated into urban centers as well as
outside the cities with urban functions performed in rural households,
market towns, and along roads. Therefore, future studies need to
incorporate a broader outlook and include the fluid interrelationships
between rural and urban areas, and between agricultural and non-
agricultural sectors, instead of examining them only as separate

entities.

The population pressure/out-migration hypothesis and the diagram
of Grigg (1980b) of responses to population pressure (Figure 3) was

found to be a useful framework for examining the relationship between

154
rural forces such as declining farm sizes due to “population stress" and
migration as one of the coping mechanisms adopted. Even in the
"classical” case of densely-populated Rwanda, though, the population
pressure/out-migration relationship was of overriding importance only
during a restricted time period. This limitation in examining
population growth as the impetus for change is very similar to Zelinsky
linking migration patterns to phases of the “vital transition“ or to
Boserup explaining changes in adaptation of agricultural technology to
population growth, and so suffers from the same constraints in applying
the hypotheses to actual situations. Therefore, the hypothesis must be
included as only one part of a larger analytic framework that examines

other forces affecting rural and urban society.

Similar to the limited ability of a population density variable to
predict out-migration over time, density variables were found to have
only limited usefulness in understanding in-migration over time or
changes in a region’s carrying capacity. Predictions of in-migration
using population densities underestimated actual in-migration as the
”saturation density” of rural origin regions increased over time. The
limits predicted by carrying capacity models were also breached as
people adjusted locally to the changing circumstances. Similarly,
predictions of urban migration based on the number of opportunities in
the city or the number of young people with a certain education level
would be hard to justify since, as this study has shown, thresholds

change through time as people adjust to ongoing processes.

155

Future research could expand the understanding of migration as a
coping mechanism adopted in response to changing circumstances. An
examination of migration as one of a series of strategies adopted in an
area could provide a better understanding of the sequence of adoption.
One important aspect that this study did not address is the relationship
between long-term migration and circular or short-term migration to
rural and urban areas in this coping process. Since migration into
rural areas had begun to increase again near the end of the study
period, rural-rural migration seems to be regaining importance in this
process. It is not known whether this is a new wave of settlers
searching for remaining bits of free land, or of agricultural laborers
forming a group of near-landless migrant laborers. Similarly, little is
known of the growing rural-urban migrant population, whether the young
men will remain in the city for only a few years, become circular

migrants, or remain permanent city dwellers.

This study also did not examine the effects of migration. For
example, not yet well understood are the dynamics of the growth of the
informal economy in absorbing incoming migrants in urban areas and
retaining people in rural areas. The numbers of in-migrants to Kigali
are presumably changing the economic, social, and physical character of
the city. Similarly, the effect of the migrant settlers from the
Highlands to the East may not only have altered the social, political,
and economic system, but affected the physical environment. The impact
of out-migration of young men from the rural areas also requires study;

these impacts could include an increase in the number of single women or

156
women head of households remaining in the rural areas. Future research
could also examine the effect of out-migration on population pressure
stresses felt by the remaining people. Finally, one important aspect of
migration not covered in this study is the repercussions of the large
movements of people into and out of Rwanda due to political

disturbances.

WISHES

AID (Agency for International Development). 1987. Cddnt[y_0ayalddmadt
t m Y .Washington, DC: Agency for
International Development.

Albert, T., Maxime A. Crener, and G. Gagnon. 1982. flddg1g_dg_§1md1atidn

Qdmo-Econdmjgda pad: la Buanda. Ottawa, Canada: Institut de
Développement International et de Cooperation, Université
d’Ottawa.

Amin, Samir.1974. Introduction. In e r

ed. Samir Amin, 65-126. London: Oxford University Press.

Bernard, Frank E. 1982. Rural Population Pressure and Redistribution in
Kenya. In Badjstrjbgtjdn of Edpdlatjda jn Afrjga, ed. John 1.
Clarke and Leszek A. Kosinski, 150-56. Guildford, Surry: Biddles
Ltd.

Berry, Eileen, Leonard Berry, David J. Campbell, and David C. Major.
1982. 1" 00- R' on 81 :1 ' 0 i . - 1' 0 1 0
Zaira. Worcester, Massachusetts: International Development
Program, Clark University.

Bilsborrow, Richard E. 1987. Population Pressure and Agricultural
Development in Developing Countries: A Conceptual Framework and

Recent Evidence. EQElQ_D§!§lQDM§nl 15, no. 2: 183-203.

Bilsborrow, Richard E., A. S. Oberai, and Guy Standing. 1984.M19:a119n
- s n . uu . 1 i ° '1- in- 1r _
i ' .London: International Labour Organization,
Croom Helm Ltd.

Blaikie, Piers and Harold Brookfield. 1987. Ladd_Qagrada11dn_and
Sdgjaty. London: Methuen.

Boserup, Ester. 1965. [h a Candjtjdna df Ag riddltdral Gro nth; lh a

W London:
George Allen & Unwin Ltd.

. 1981. Edddlatjdn and Tadhadlogigal ghaaga: A Study at Lang-
.Chicago: The University of Chicago Press.

157

158

. 1985. Economic and Demographic Interrelationships in Sub-

Saharan Africa Eg2ulat1gn_and_Dexelgnment_Bexiew 11 no 3
(September): 383- 397.

Brown, Lawrence A. and Victoria A. Lawson. 1985. Rural-Destined
Migration in Third World Settings: A Neglected Phenomenon?

32919031_§Iflfii§§ 19.5: 415-432.

Brown, Lawrence A. and E. G. Moore. 1970. The Intra- Urban Migration

Process: A Perspective. figngnafiaka_Annalan 528: 1-13.

Brown, Lawrence A. and Rickie L. Sanders. 1981. Toward a Development
Paradigm of Migration, with Particular Reference to Third World

Settings. In Migration Oegision Making; Multjdjaginljnary

00 ea ‘6 0 1C 0] V ‘ 00's lo I a 00 no
Cnnn1_1a§, ed. Gordon F. De Jong and Robert W. Gardner, 149-85.
New York: Pergamon Press.

Buksmann, P. E. 1980. Migration and Land as Aspects of Underdevelopment:
An Approach to Agricultural Migration in Bolivia. In Ina
, ed. Paul White and Robert Woods,
108-128. London: Longman Group Limited.

Bureau National de Recensement, République Rwandaise.1984.3gggnaangnt
l .Kigali:
Bureau National de Recensement.

Caldwell, John C. 1969.
finana_§_ln_n§. London: C. Hurst & Co.

Cambrezy, Luc. 1981. Conquéte des Marais au Rwanda et Dynamique de la
Population. £1ndaa_3nnala§ 83 (Juillet-Septembre): 45-67.

. 1984.1 n 1
E5919gia_des_Uigratigns_au_Buanda. Parisz ORSTOM

Campbell, David J. 1981. Land-Use Competition at the Margins of the
Rangelands: An Issue in Development Strategies for Semi-Arid
Areas. In E1ann1ng_A£nidan_Qayaldnnant, ed. Glen Norcliffe and Tom
Pinfold, 39-61. Boulder, Colorado: Westview Press.

. 1984. Response to Drought among Farmers and Herders in
Southern Kajiado District, Kenya. unman_£§nldgy 12, no. 1: 35-64.

Campbell, David J. and James C. Riddell. 1984. Social and Economic
Change and the Intensity of Land Use in the Mandara Mountains

Region of North Cameroon. I1idsshriIt_xg2r_£sgngmisshe_en_§gsiala
Gangnafia 75, no. 5: 335-343.

Clarke, John I, Mustafa Khogali and Leszek A. Kosinski. I985. Edndlatidn
and_0exelgoment_2rgiests_in_Africa. Cambridge. U.K.: Cambridge

University Press.

159

Clay, Daniel C. and Y.M.J. Dejaegher. 1987. Agro-Ecological Zones: The
Development of a Regional Classification Scheme for Rwanda.

IIQEISUlLUEQ 5 no. 4: 153-159.

Clay, Daniel C., Jean Kayitsinga, and Theobald Kampayana. 1989. L’Emnlni
na a . ASPAP/0A1 Report No. 74. Kigali:
Service des Enquétes et des Statistiques Agricoles, Min. de
l’Agriculture, de l’Elevage et des Foréts, Rép. Rwandaise.

Cole, William E. and Richard D. Sanders. 1985. Internal Migration and

Urban Employment in the Third World. Angnidan_£ddndm1d_figy1an 75
no. 3: 481-494.

Connell, John, Biplab Dasgupta, Roy Laishley, and Michael Lipton. 1976.
r t f o r Ar : h v c o l S .
Delhi: Oxford University Press.

Contant, Rodolf B. 1982. Ra or o v rn m b i u
w ndai ° 5 em a o d R r ol u Rwa
The Hague, Netherlands: International Service for National
Agricultural Research.

Delepierre, G. 1985. Evolution de la Production Vivriere et les Besoins

d’Intensification. In ConLe Bendg dd Er anjan Saninaira Natinnal
l. -r l . ., .- o Rwaio. . -.

1285. Ministere de l’ Agriculture, de l’ Elevage, et des Foréts,
56-86. Kigali, Rwanda: Rép. Rwandaise.

 

Durham, William H. 1979.59anc11y an nd Survival in Cantna! America;
c i 'n f S . Stanford: Stanford

University Press.

Findley, Sally E. 1981. Rural Development Programmes: Planned Versus
Actual Migration Outcomes. In Population Djagrjnggjon Pdligjaa jn
Qayelonment Plannjng. United Nations, Department of International
Economic and Social Affairs, Population Studies No. 75: 144-168.
New York: United Nations.

Gapyisi, E. 1980. Note sur le Probleme Démographique au Rwanda. Ballgtin
Agniddla_dd_flnanda (Juillet): 118-121.

Gatera, F. 1980. Accroissement Démographique et Déboisement au Rwanda.
Wilda (13): 28-30.

Gimui, Kiboma. 1982. Spatial Inequalities and Population Redistribution
with Reference to Uganda. 1n
Afnida, ed. John 1. Clarke and Leszek A. Kosinski, 146-149.
Guildford, Surrey: Biddles Ltd.

160

Goldscheider, Calvin. 1971. EQDUIQLl dn, undgrnizagi nn , and §ddjal
§LIU£10£§- Boston: Little, BrOwn and Company.

Goldschmidt, Armin M.F. 1986. - i a
l- rban Mi rati in en : A c -Ps h l ic l

in Hard Diagrigg (Mt; Kenya). University of the Saar, Federal

Republic of Germany. Saarbrficken: Verlag Breitenbach.

Gotanegre, J.F., Pierre Sirven and Christian Prioul. 1974. gangrannjg du
gnanda. Brussels: Editions a. de Boeck.

Gould. HTS. 1976. WWW
Innnjna1_nfnida. African Population Mobility Project Working Paper
No. 29. Liverpool, U. K: University of Liverpool Department of
Geography.

. 1985. Migration and Development in Western Kenya, 1971-82: A
Retrospective Analysis of Primary School Leavers. Afniga 55 no. 3:
262-85.

Gourou, Pierre. 1953. La Dans jgd da la EnnglaLinn an Bgan da-Urnndj;

Eagnjsaa d’una Et ndg Gédgranni 19nd. (Mém. no. 8, XXI- -6) Bruxelles:
Institut Royal Colonial Belge.

. 1971 WWW
gnlle ga dd Eranggd a 124 Z a 1222. Paris: Ecole Pratique des Hautes

Etudes- Sorbonne.

Grigg, David. 1979. Ester Boserup’s Theory of Agrarian Change. Endgnaas
1n flnman Gadgranny 3, no. 1: 64-84.

. 1980a. Migration and Overpopulation. In Ing_§adgnann1ga1
o , ed. Paul White and Robert Woods, 60-83.
London: Longman Group Limited.

. 1980b. 0 at' n ro r i : i
Ednanddnlyg. Cambridge, U.K.: Cambridge University Press.

Haeringer, Philippe. 1968. L’Observation Rétrospective Appliquée é

l’Etude des Migrations Africaines. 9an1an§_QB§IQML_2dn1a_§d1anga§
flnnainaa 5 no 2: 3-10.

Haggett, Peter, Andrew 0. Cliff and Andrew Frey. 1977. Ldgat_dna1
i um 1 G 1cr 11 -_.n. . 111 1 .0; 1 -l
Mndala. New York: John Wiley and Sons.

Hance, William A. 1970. PonulaLjdn, Mignagjdn, and Urbanizagjnn 1n
Afnida. New York: Columbia University Press.

Harris, J. R. and M. P. Todaro. 1970. Migration, Unemployment and

Development: A Two Sector Analysis. Ananidan_£dnndnid_flayian LX:
139-149.

161

Haynes, Kinglsey E. and A. Stewart Fotheringham. 1984.
lngara91129_flgda1§. Beverly Hills, CA: Sage Publications.

Jones, Richard C. 1980. Behavioral Causes and Consequences of Rural-
Urban Migration: Special Reference to Venezuela. In .Int arga al
I 1r t on m he I- v l11 n1 .1 l1 .i 1
1g_La11n_Amgg1ga, ed. Robert N. Thomas and John M. Hunter, 26- SD.
Cambridge, Massachusetts. Schenkman Publishing Co.

Jones, William I. and Roberto Egli. 1984. ' °
W. World Bank
Technical Paper Number 27. Washington, D.C.: The World Bank.

Kabera, John B. 1985. Populating Uganda’s Dry Lands. In
ngglopmgnt Erojacta in Afrjga, ed. John 1. Clarke, Mustafa
Khogali and Leszek A. Kosinski, 112-122. Cambridge, U.K.:
Cambridge University Press.

de Lacger, Louis. 1939. Buanda. Namur, Belgique: Grands Lacs.

Lappé, Frances Moore and Joseph Collins. 1978. Eggg_£1r§11_flgygng_1ng
My1h_gf_§§arg111. New York: Ballantine Books.

Lawson, Victoria A. and Lawrence A. Brown. 1987. Structural Tension,
Migration, and Development: A Case Study of Venezuela.

W 39 no. 2: 179-188.
Lee, Everett 5. 1966. A Theory of Migration. Qamggraphy 3 no. 1:47-57.

Lemarchand, Réné. 1970. Rwanda and Burundi. London: Pall Mall Press.
. 1982 WWW
Bloomington,

Indiana: African Studies Program, Indiana University.

Lewis, W. Arthur. 1954. Economic Development with Unlimited Supplies of

Labor. WM 22 (May):
139-134.

Lipton. Michael 1977 W
W9:1d_flgyg1gpmgn1. Cambridge, Massachusetts: Harvard University
Press.

Loveridge, Scott. 1988. Uses of Farm and Market Survey Data to Inform
Food Security Policy in Rwanda. Ph.D. diss., Michigan State
University.

Lugan, Bernard. 1976. Deux Raisons Historiques de la Prédominance de
l’Habitat Rural au Rwanda. L11nfgrma1gur IX no. 3 (Juin): 56-61.

162

Mabogunje, A. L. 1970. A Systems Approach to a 8Theory of Rural -Urban
Migration WW2: - 8.

Marijsse, Stefaan. 1981. Income Distribution and Political Economy of

Rwanda. In Emapgjpg Development Pattarps; Eprppean Cpngribgtipna,
ed. Istvan Dobozi and Peter Mandi, 196-232. Budapest: European

Association of Development Research and Training Institutes.

Masser, I. and W.T.S. Gould. 1975. Intpp-Bagjppal MigraLipn jn lrppjpal
Afpipa. London: Institute of British Geographers.

Mbithi, Philip M. and Carolyn Barnes. 1975. Ina_§ppp1apapp§_§p111amap1
Eggplam 1n Kgpya. Nairobi: East African Literature Bureau.

Meach, Sam Ell. 1986. Analyse de la Productivité Agricole au Rwanda.
WM 19. no 3 (Juillet): 143-151 (part
one). 19, no. 4 (octobre): 215-222 (part two).

Mellor, John W. and Uma Lele. 1973. Growth Linkages of the New Foodgrain

Technologies WWW 28. no 1
(January- March): 35- 55.

Ministere de l’Agriculture, Rép. Rwandaise. 1985. BappprLs Annuelp ppgr
lg; Annaaa 1269-1285. Kigali, Rwanda: Min. de l’Agriculture.

Ministere du Plan. 1970. Epgpé1p_Q§mpgpaph1gpa1_flwapga. Kigali, Rwanda:
Ministere du Plan, Rép. Rwandaise.
1981 MW. Kigali. Rwanda:
Ministere du Plan, Rép. Rwandaise. (unpublished data).
. 1987. EQMWMMMI Kigali.

Rwanda: Ministere du Plan.

Monmonier, M.S. 1983. Computer A§§j§tag Cartpgpaphy. Englewood Cliffs,

New Jersey: Prentice Hall.

Monsted, Mette and Parveen Walji. I978.
° . Uddevalla, Sweden:
Bohuslaningens AB.

Mukamanzi, Monique. 1985. La Situation Socio- -Démographique au Rwanda. In
BMW
Cpmmpp1pa11pp_§p_1a1a. Kigali: Office National de la Population.

Nwafor, James C. 1979. Agricultural Land Use and Associated Problems.

Wu VII (4) no 28: 624- 636

ONAPO (Office National de la Population). 1982. Ap1g§_pp_§pllpggg
on '. Kigali, Rwanda: Office
National de la Population.

163

1985.

Cpmmpp1pa11pp_§pp1a1g. Kigali, Rwanda: Office National de la
Population.

Olsson, G. 1965. Distance and Human Interaction: A Migration Study.
Geogra£1§Ba_Annaler 47: 3-43.

Préfol, B. and G. Delepierre. 1973.

Ierres_au_B_andaa_S1tuat1on_Actuelle_et_flersnecti_es Butare.
Rwanda: Institut des Sciences Agronomiques du Rwanda.

. I975. Disponibilité et Utilisation des Terres au Rwanda
(S ituation Actuelle et Perspectives). Bp11a11n_Agp1ppla_dp_Bwapda
8 no’ 5 I, 2, 3, 4:9 no.1.

Prioul, Christian. 1976. Pour une Problématique de l’Aménagement de

l’Espace Rural au Rwanda. 111pfppma1gpp 9 no. 3 (Juin): 71-109.
1981. Les Densités de Population au Rwanda. In Bpphgzphg§_§pp
Hautes Terre d’ fri ume d

fiapgraphja Tropjcalp No. 42, ed. Centre d’Etudes de Géographie
Tropicale, Ill-126. Talence, France: Domaine Universitaire de
Bordeaux.

Prioul, Christian and Pierre Sirven. I981. A11a§_pp_fluanda. Kigali,
Rwanda: Imprimerie Moderne Nantasie Coueron.

Pryor, R. J. 1975. Migration and the Process of Modernization. In Egpplp

on_tbe_Boxea_5tud1es_o_.1nternational_nigration. ed Leszek A
Kosinski and R. Mansell Prothero, 23-38. London: Methuen & Co
Ltd.

Ranis, Gustav and J. C. H. Fei. 1961. A Theory of Economic Development.
American_£conomic_Bexiew 51 (September): 533- 565

Ravenstein, E.G. 1889. The Laws of Migration. Jppppa1_pf_1pa_flpya1
fitatiétisal_§osietx 52 (June): 241-301-

Reintsma, M. 1981. Lapp_1gpprp_1p_flgappa. Washington, D.C.: Agency for
International Development.

Rempel, Henry and Michael Todaro. 1972. Rural-to-Urban Migration in
Kenya. In

Booulat1on_Brow1n_and_Economic_nexelooment_in_A£rica
eds. S.H. Ominde and C.N. Ejiogu, 214-233. London: Heinemann.

Rép. Rwandaise. 1977. in '
upf1p111fp. Kigali, Rwanda: Rép. Rwandaise.

- 1982. II1emg_Elan_du_Dexelnnnemen1._1282;12§§- Kigali.
Rwanda: Rép. Rwandaise.

164

Riddell, J. Barry and Milton E. Harvey. 1972. The Urban System in the
Migration Process: An EvaluatiOn of Step- -Wise Migration in Sierra

Leone. Eppnpnip_§ppgrapny’48, no. 3 (July): 270- 283.

Rossi, Georges. 1984. Evolutions des Versants et Mise en Valeur Agricole

au Rwanda. Anna1p§_pg_§gpgrapn1a 515 (Jan-Fev): 23-43.

Rwamasirabo, Serge. 1988. Personal communication.

SESA (Service des Enquétes et des Statistiques Agricoles). 1984.
1 ti mm 1 d Pri e a t i d
W. Kigali. Rwanda: Min. de
l’Agriculture, de l’Elevage et des Foréts, Rép. Rwandaise.

SESA and the ASPAP (Service des Enquétes et des Statistiques Agricoles
and the Agricultural Surveys and Policy Analysis Project). 1988.
MW. Kigali. Rwanda: Min. de
1’ Agriculture, de l’ Elevage, et des Foréts, Rép. Rwandaise.

Sibomana, J.M.V. 1983. Le Probleme Démographique au Rwanda; Ses

Consequences et Approches de Solutions. BgIIaLjn Agnjcplp pg
Buanda (Octobre): 244-250.

Silvestre, Victor. 1974. Différenciations Socio- -Economiques dans une
Société a Vocation Egalitaire: Masaka dans le Paysannat de
l’ Icyanya WW XIV (53): 104- 169

Sirven, Pierre. 1975. Transports et Urbanisation au Rwanda.

L_1nfppna1ppp VIII no. 3 (join): 25- 36.
1981. Etude de la Croissance Urbaine. In Bppnprpnpp_ppp_1p§

e ’Afr v
No. 42, ed. Centre d’ Etudes den Géographie
Tropicale, 111-126. Talence, France: Domaine Universitaire de
Bordeaux.

. 1983. La Sous-Urbanisation et les Villes du Rwanda et du
Burundi. Ph.D. diss., Université de Bordeaux III.

Shaw. R Paul 1975 W
W Bibliography Series Number 5
Philadelphia: Regional Science Research Institute.

Shrestha, Nanda R. 1988. A Structural Perspective on Labour Migration in

Underdeveloped Countries. Eppgna§p_1n_flpman_§ppgpapny 12 no. 2
(June): 179-207.

Simmons, Ozzie G. 1988. Eppppp
fir9uth_1n_1hg_ln1pd_flppld. New York: Plenum Press.

 

165

Skinner, Elliott P. 1985. Labor Migration and National Development in
Africa In Women. ed Beverly
Lindsay, 18- 39. University Park: the Pennsylvania State
University Press.

Statistical Package for the Social Sciences, PC Plus Version 3.0.
Chicago.

Tabah, Léon. I970. Mesure de la Migration Interne au Moyen des
Recensements. £ppp1a11pn 25 no. 2 (Mars/Avril): 309-46.

Timmer, C. Peter. 1988. The Agricultural Transformation. In fianpppp5_pf
prplppman1_£ppnpm1p§ Volume 1, ed. Hollis Cherery and T. N.
Srinivasan, 275-331. New York: Elsevier Science Publishers B.V.

Thomas, Robert N. and C.W. Minkel. 1984. Population, Land Tenure, and
Development in Latin America: The Case of Guatemala. In La11n
Ampp1pan_ga§a_§1pd1pp, eds. Richard G. Boehm and Sent Visser, 109-
120. Dubuque, Iowa: Kendall/Hunt Publishing Co.

Todaro, Michael P. 1969. A Model of Labor Migration and Urban
Unemployment. Ampr1pan_£ppnpm1p_3gy1pu 69: 183-193.

. 1981. International Migration in Developing Countries: A
urvey In 22W.
ed. Richard Easterlin, 361- 402. New York: National Bureau of
Economic Research.

Twagiramutara, Pancrace. 1976. Rapport a l’Espace, Rapport Social,
Contribution a l’Etude de Leurs Liaisons a Partir du Cas du
Rwanda. Ph.D. diss., U.C.L.

United Nations. Department of Economic and Social Affairs. 1970.
111.: on Me hods 0 u. o ’oo_ . A. W‘ ‘11 1
' .Population Studies, No. 47. New
York: United Nations.

United Nations. Economic Commission for Africa. Population Division.
1983 '00! . a. I 1_ 0‘ .‘o 0.. . 11 ; I “0
African Population Studies Series Number 7.

White, Paul and Robert Woods. 1980.
M1gra11pn. London: Longman Group Limited.

Woods, Robert. 1979. Eppp1a11pn_Analy§1§_1n_§ppgraphy. London: Longman
Group Limited.

Zachariah, K.C., and Julien Condé. 1981. M1gpa11pn_1n_flp§1_Afp1pa1
. Washington, D.C.: Oxford University Press for
the World Bank.

166

Zelinsky, Wilbur. 1971. The Hypothesis of the Mobility Transition.
' 61, no. 2:'219-49.

Ill-1|“l‘ll‘ iiilvi“ ‘4‘.

‘ll ‘l../ i (ll) ‘I JI.||) I III .i ll:

I‘ll!!! lllllllllll ‘ll‘lfll I‘-

"mmmllmm