The Impact of Rainfall on the First Out-Migration: A Multilevel Event-History Analysis in Burkina Faso

Abstract

Using event history analyses, we investigate the impact of rainfall conditions — a crucial environmental factor in the livelihood of Sahelian households— on the risk of the first village departure in Burkina Faso. The distinction of migrations by destination and duration proves critical in studying this relationship. Findings suggest that people from the drier regions are more likely than those from wetter areas to engage in both temporary and permanent migrations to other rural areas. Also, short-term rainfall deficits tend to increase the risk of long-term migration to rural areas and decrease the risk of short-term moves to distant destinations.

Full text

The Impact of Rainfall on the First Out-Migration:
A Multi-level Event-History Analysis in Burkina Faso
Sabine Henry
Bruno Schoumaker
Universite
´
catholique de Louvain
Cris Beauchemin
University of Montreal
Using event history analyses, we investigate the impact of rainfall conditions — a
crucial environmental factor in the livelihood of Sahelian households — on the risk
of the first village departure in Burkina Faso. The distinction of migrations by des-
tination and duration proves critical in studying this relationship. Findings suggest
that people from the drier regions are more likely than those from wetter areas to
engage in both temporary and permanent migrations to other rural areas. Also, short-
term rainfall deficits tend to increase the risk of long-term migration to rural areas
and decrease the risk of short-term moves to distant destinations.
KEY WORDS: migration, Burkina Faso, rainfall, drought, environmental refugees.
INTRODUCTION
The impact of environmental factors on spatial mobility is a recurrent
theme in the literature on migration in Burkina Faso and in other Sahelian
countries. Repeated droughts and low soil productivity have often been
cited as major factors pushing people to leave their villages (Boutillier,
Quesnel, & Vaugelade, 1977; Cordell, Gregory, & Piche
´
, 1996; Marchal,
Please address correspondence to Sabine Henry, Department of Demography, Universite
´
de Montre
´
al, C.P. 6128, Succursale Centre-ville, Montre
´
al, Que
´
bec, Canada H3C 3J7; e-mail:
sabine.henry@umontreal.ca
423
Population and Environment, Vol. 25, No. 5, May 2004 Ó 2004 Human Sciences Press, Inc.

1975; Mathieu, 1994). The concept of environmental refugees is another
illustration of the widespread belief that migration is strongly influenced
by environmental conditions (Loner gan, 1998). Since rain-fed agriculture
is the main source of livelihood in rural Burkina Faso, int uitively it makes
sense that environmental factors (e.g. rainfall and land degradation) will
influence socio-economic conditions and may lead people to e migrate.
For example, in Sahelian and Sudanian regions, crop yields are mostly
controlled by the amount and distribution of rainfall during the growing
season and we may expect that poor harvests tend to drive some house-
holds to migrate further south (Grouzis & Albergel, 1989; Sicot, 1989). A
link between environmental conditions and migration was indeed mea-
sured in several settings in rural Africa (Ethiopia, see Ezra, 2001; Mali, see
Findley, 1994). Yet, despite repeated statements on the effects of envi-
ronmental factors on migration in Burkina Faso, there is little empirical
evidence on this topic. Lack of approp riate data is an important factor in
explaining this situation. Compared to data on mortality and fertility, data
on migration are scarce in most African countries. The most detailed study
of burkinabe migration relies on the 1974–1975 National Migration Sur-
vey (Cordell et al., 1996), almost 30 years old. A national survey on
migration and urbanisation was also conducted in 1992 (Re
´
publique
du Burkina Faso and CERPOD, 1997), but the data are considered unre-
liable and have hardly been used. Finally , censuses also offer some
information on spatial mob ility in Burkina Faso, but they are highly
aggregated and only give information at the time of the census. Environ-
mental data are also in very short supply, and rainfall time-series are in
fact the only reliable data on environmental conditions covering a long
period for the entire country.
The aim of this study is to explore the impact of rainfall conditions on
the first departure from the village in Burkina Faso. This study combines
recent longitudinal multilevel data from a national retrospective migration
survey, a large-scale retrospective community survey and a long time-series
of rainfall data. This study more specifically focuses on the influence of
agro-climatic factors on the risk of ‘‘leaving the village’’, among both men
and women. Individual-level variables an d community-level factors are
combined with rainfall data in discrete-time event-history analyses to model
the determinants of first migration from rural areas over the last 30 years.
Competing risk analyses are performed to explore the differential effects of
rainfall conditions depending on the destination and the duration of
migrations.
The first part of the study summarises the background. Methodology,
data sources and explanatory variables are presented in the second part. In
424
POPULATION AND ENVIRONMENT

the third part we present the descript ive and multivariate results. The paper
ends with the discussion of the results and the conclusion.
BACKGROUND
General Background
Burkina Faso is one of the poorest countries in the world: it ranked
159th of 162 countries in the UNDP’s human development index (UNDP,
2001), and its GDP per capita was approximately 230 US$ at the end of the
1990s (IMF, 2000). The country’s population was estimated at 10.3 million
in the last census (1996), growing at a rat e of about 2.7% per year. With an
urbanisation rate close to 20%, it is one of the least urbanised countries in
the world (PRB, 2001). The country’s economy depends heavily on agri-
culture and cattle-raising: toget her they account for one-third of the coun-
try’s GDP and 90% of the population is engaged in these activities (INSD,
2000). Agriculture in Burkina Faso is to a large extent a rain-fed subsistence
agriculture, and its productivity is mainly determined by environmental
factors (Niemeijer & Mazzucato, 2002). As a consequence, the agro-cli-
matic conditions are critical to rural households, for which agriculture
represents the main source of livelihood.
Environmental Conditions in Burkina Faso
Burkina Faso is a Sahelian country characterised by a strong south–
north decreasing gradient of average annual rainfall (Figure 1). In the
northern part of Burkina Faso, the rainfall is scarce and irregular, with an
average annual precipitation below 500 mm. The agro-climatic conditions
are thus very constraining for agriculture and the main economic activities
are extensive pastoralism and rain-fed agriculture of pearl millet and sor-
ghum (Hampshire & Randall, 1999). The conditions are more suitable for
agriculture in the southern part (average rai nfall above 900 mm), where the
main crops include maize and cotton in addition to millet and sorghum
(Ingram, Roncoli, & Kirshen , 2002).
The country’s rainfall is also characterised by a high degree of seasonal
and annual variability (Roncoli, Ingram, & Kirshen, 2001). Rain falls during
a single wet season lasting three to five months (approximately from May to
September). Large year-to-year variations in tot al precipitation and in the
timing of rainfalls translate into extremely variable crop outcome and
uncertainty at the household level (Reardon, Matlon, & Delgado, 1988;
425
S. HENRY ET AL.

Roncoli et al., 2001). Following the general pattern of Sahelian West Africa,
Burkina Faso has experienced a long-term downward trend of rainfall over
the last 50 years (Nicholson, 2001; Niemeijer & Mazzucato, 2002). The
long period of aridity that began in the late 1960s was accompanied by
several droughts, most notably those in the early 1970s an d in the mid-
1980s (Hampshire and Randall, 1999; Ni cholson, 2001; Roncoli et al.,
2001).
Burkina Faso is also characterised by poor soil fertility and, according
to various authors, land degradation is a serious problem in this part of
Africa (Lindqvist & Tengberg, 1993). The issue of land degradati on in
Burkina Faso is controversial, however. On the basis of an assessment by
FIGURE 1. Map of Burkina Faso showing mean annual rainfall at the
department level, 1960–98.
426
POPULATION AND ENVIRONMENT

regional experts (GLASOD map in Oldeman, Hakkeling & Sombroeck,
1990), a large part of the country (mostly the Mossi Plateau) is considered
degraded. Empirical data on land degradation are sparse however, and a
recent study suggested that neither the agricultural soil productivity nor the
soil chemical fertility had significantly declined over the last few decades in
Burkina Faso (Niemeijer & Mazzucato, 2002). So, while soils are poor in
most parts of the country, there is no convincing evidenc e of a widespread
degradation trend. Other factors seriously constrain agricultural productiv-
ity however. The lack of fertilisers, the low degree of mechanisation of
agriculture, the lack of irrigation, and the spread of parasitic plants all act as
constraints on agriculture (Berner et al., 1997; Ingram et al., 2002).
Migration in Burkina Faso
Burkina Faso has long been characterised by intense mobility, both
within the country and to other co untries such as Co
ˆ
te d’Ivoire and Ghana
(Cordell et al., 1996; Hampshire & Randall, 1999). Migrations to neigh-
bouring countries represent a large proportion of all migrations from rural
areas (the majority of migrations among men), and Co
ˆ
te d’Ivoire has long
been the principal destination for international migrants, attracting
approximately 80% of these in the 1990s.
1
Labour migration to Co
ˆ
te
d’Ivoire can be traced to historical factors such as forced labour policies and
colonial taxation under French colonial rule (Cordell et al., 1996). Co
ˆ
te
d’Ivoire has continued until recently to attract a large number of Burkinabe
migrants, mainly composed of young men moving for economic reasons
(Cordell et al., 1996; Roncoli et al., 2001), alth ough the situation is likely to
change with the recent conflict in Co
ˆ
te d’Ivoire. Migrations from rural areas
have also been directed towards urban centres and have contributed sig-
nificantly to the process of urbanisation since the post-war years (Cordell
et al., 1996). Ouagadougou and Bobo Dioulasso, Burkina Faso’s two largest
urban centres, have attracted the largest share of rural-urban mover s over
the last decades, although smaller towns have also received important
numbers of migrants (INSD, 2000).
Migrations within rural areas are still the dominant type of internal
migration flow and involve both short-distance and long-distance moves.
The former disproportionately concern women migrating for family reasons,
such as marriage and separation, while long-distance moves mainly consist
of migrations for economic reasons , such as agricultural expansion. One
specific type of long-distance move in rural Burkina Faso develope d from
the late 1960s and involves migrations from densely populated areas in the
Mossi Plateau to the less populated areas of Burkina Faso’s southwest
427
S. HENRY ET AL.

(Cordell et al., 1996; Goldberg & Frongi llo, 2001). Migrations between rural
areas were also encouraged by resettlement programmes, such as the AVV
(Ame
´
nagement des Valle
´
es des Voltas) programme launched in the early
1970s to develop regions freed from onchorcercosis, also called river
blindness disease (Cordell, et al., 1996; Oue
´
draogo, 1986). Spontaneous
population movements also swelled in the villages adjacent to the official
migration areas (Quesnel, 1999).
As in other developing countries, circulation is also a prominent feature
of burkinabe migration (Blion, 1995; Cordell et al., 1996). In other words,
people continue to maintain strong links with their place of origin, and a
large fraction of migrants tend to return to their village at some point. In
Burkina Faso, the probability of returning to the village is especially high
among men: 60% of those leaving their village after age 15 return within
10 years, while the corresponding figure for women is around 15%
(Schoumaker et al., 2002). Those leaving the country also have the highest
probability of returning to their village,
2
but a significant percentage of
those leaving to urban and (to a lesser extent) rural destinations also go back
to their village (Schoum aker et al., 2002). The fact that a considerable
proportion of those leaving their village tend to return home at some point
suggests that one should also look at the factors that influence the risk of
returning to the village. That question will be addressed in further research.
However, by distinguishing short-term and long-term migrations, this study
will partly take into account the fact that a significant fraction of those
leaving actually return to their village of origin.
Environmental Conditions and Migration
The role of environmental factors in explaining migration is a persistent
theme in the literature (Bout illier et al., 1977; Cordell et al., 1996; Marchal,
1975; Mathieu, 1994). As far as climatic conditions are concerned, the
common view is that ‘‘migration rises both immediately and as a long-term
response to the threat of recurrent droughts’’ (Findley, 1994, p. 539). Quite
illustrative in this respect is the anecdote reported in the preface of Hoe and
Wage, a detailed study of migration in Burkina Faso.
In October 1974, in the village of Nanou in Southwestern
Burkina Faso, Siaka Coulibaly, the chief, extended the traditional
welcome. [...] Introductions made and visitors welcomed, he
spoke: ‘‘You want to know why people migrate? Well, make it
pour money and bring rain in abundance. The n you will know
why people leave.’’ (Cordell et al., 1996, p. xii).
428
POPULATION AND ENVIRONMENT

Because rain-fed agriculture is the main source of livelihood in rural
Burkina Faso, it is natural to expect some impact of rainfall on migration.
Migratory responses to climatic constraints can take several forms however:
some people will move for a short duration and return to their village within a
few months (temporary migration). Others will leave their region permanently
to cultivate in better endowed rural areas or to work in urban areas or abroad.
Temporary migration is an important aspect of spatial mobility in the
drought-affected Sahel and is often interpreted as part of a household sur-
vival strategy for coping with drought and high levels of production
uncertainty (Guilmoto, 1998; Hampshire & Randall, 1999; Hill, 1990;
Reardon et al., 1988; Roncoli et al., 2001). Such temporary moves usually
involve only a few members of the household (mainly young men) and
represent a way to diversify incomes. While temporary migration may be
adopted as a strategy even in non-drought years (Ezra, 2001; Hampshire &
Randall, 1999; Reardon et al., 1988), it is likely to vary with fluctuations in
rainfall and harvests (Findley, 1994). Young men may be more likely to
move in periods of economic hardship to earn additional incomes, whether
in rural areas, in urban areas or abroad (Coulibaly & Vaugelade, 1981;
Lallemand, 1975; Roncoli et al., 2001). Migration can also be used to re-
duce the number of consumers by sending children and women to stay with
relatives in a period of drought (Findley, 1994; Roncoli et al., 2001). Finally,
fluctuations in rainfall may also influence the destination of migration.
Since moving abroad or to urban areas usually entails higher costs than
moving to rural areas, such moves may be postponed in periods of eco-
nomic stress (Findley, 1994; Nelson, 1983).
Some people may also leave their village permanently in response to
the risk of repeated droughts (Findley, 1994). Cordell et al. (1996) suggest,
for example, that migrations from the northwestern part of the Mossi Plateau
in the late 1960s–early 1970s were induced by irregular rainfall, an d that
migrants departed for areas with more rainfall (see also Boutillier et al.,
1977). Permanent migrations may also be influenced by rainfall-driven
fluctuations in harvests. For example, migrants may decide to move after
several consecutive years of bad harvests. Female migration could also
increase after poor harvests, as women might be permitted or encouraged to
marry earlier in order to reduce the family’s food demands (Findley, 1994).
On the other hand, long-term migrations to urban areas and to foreign
countries might be deterred in years following bad harvests, as discussed for
the case of temporary migrations.
In summary, long-term rainfall conditions as well as short-term varia-
tions of rainfall are likely to influence both temporary and permanent
migrations. Although poor rainfall conditions are often believed to increase
429
S. HENRY ET AL.

the risk of moving, the effects of rainfall may differ depending on the des-
tination and the duration of migration. Gender differences are also expected
as the motives of migration vary greatly by gender: males generally move for
economic reasons whereas women’s moves depend essentially on family
reasons. As a result, male migration should be more sensitive to rainfall
conditions, even though female migration may also be expected to rise in
some circumstances (Table 1).
The decrease in soil productivity is also generally considered to be a
determinant of migration in Burkina Faso. However, due to the lack of reli-
able data, this question is not addressed in this study. Two land degradation
variables were included in preliminary analyses: a land degradation assess-
ment obtained from the GLASOD map (Oldeman Hakkeling, & Sumbroek,
1990), and a land degradation indicator based on the rain-use efficiency
index obtained by combining satellite data and rainfall data (Prince, Brown
De Colstoun , & Kravitz, 1998). Neither of these indicators was retained in
the models because of their low reliability (GLASOD) or the very small
proportion of areas considered as degraded (rain-use efficiency index).
METHOD AND DATA
This study focuses on the determinants of village de parture after age 15,
using multilevel longitudinal data and event history methods. Analyses are
restricted to the first out-migration from the village after age 15, retained as
the age at which participation in decision making is considered to
TABLE 1
Synthesis of expected effects of rainfall conditions on different types of
migration
Type of migration
Rainfall
conditions Rural–rural Rural–urban Rural–abroad
Long-term
conditions
()) Migrations from
drier regions to
wetter regions
()) Migrations from
drier regions to
urban areas
()) Migrations from
drier regions
to abroad
Short term
fluctuations
(+) Migrations if
rainfall deficit
()) Migrations
postponed if
rainfall deficit
()) Migrations
postponed if
rainfall deficit
430
POPULATION AND ENVIRONMENT

commence (INSTRAW, 1994). The decision to focus on the determinants of
the first departure from the village is justified by the fact that it represents a
significant life event. Moreover, because a significant fraction of migrants
return to their village, taking into account the subsequent migrations would
tend to dilute the effects of environmental factors on migration. Analyses are
performed separately for men and women, as we expect the effect of rainfall
conditions to dif fer by gender. Finally, because the effects of rainfall con-
ditions on the risk of leaving may also vary according to the destination and
the duration of migration, co mpeting risk models are used in addition to
models that treat all the events the same.
As suggested by various authors, it is desirable to include factors at the
community and individual/household levels to investigate the determinants
of migration (Bilsborrow, 1984; Findley, 1987; Lucas, 1997). Migration
behaviours depend strongly on individual and househ old characteristics
such as age, education, economic activity and household socio-economic
status (Lucas, 1997; Hugo, 1998). In addition, the characteristics of the
place of residence are also likely to play a major role in the migration
decision-making process (Gardner, 1981; Hugo, 1985). Place-related fac-
tors provide opportunities and constraints that make an area more or less
attractive to an individual (availability of land, local labour market con-
ditions, etc.), and factors such as its accessibility may also act as facilitators
of migration or influence the awareness of potential destinations (Oberai &
Bilsborrow, 1984). Environmental conditions are considered as one spe-
cific type of contextual characteristics that may influence migration deci-
sions. In this study, both individual factors and contextual factors are
included in the models. Although household-level factors may also be
important determinants of migration (Hugo, 1998), few time-varying
household-level variable s were collected in the survey, and none was
retained in this study. Actually, the relevant variables at the household
level (such as the position of the migrant in the househ old or the agri-
cultural practice at the household level) were not time-varying. We think
that the definition of these household variables is too restricted to be rel-
evant in this migration analysis.
A longitudinal approach was taken to study the determinants of
migration. People are likely to migrate in response to changing individual or
contextual conditions, and longitudinal data and methods are well-suited to
incorporate such factors in migration models. This approach thus not only
requires individual longitudinal data on migration (a migration history), but
also community-level and environmental data about the places of residence
in which the sampled individuals lived in the past, at every point in time.
The data are presented below.
431
S. HENRY ET AL.

Data
This study uses data from three sources. The individual life history data
come from the nationally-representative retrospective survey on migration,
3
conducted in 2000 by the UERD at the University of Ouagadougou, the
Demography Department of the University of Montreal and the CERPOD in
Bamako (Poirier et al., 2001). The full sample comprises 8644 individuals
(men and women) aged 15–64 at the time of the survey. The questionnaire
covered several topics such as migration, employment, marital and fertility
histories. Fo r each sampled individual, a complete migration history from
the age of six was collected. Information on every spell of residence that
lasted more than 3 months was recorded: the date of arrival, the name of
the village (rural areas) or neighbourhood (urba n areas), the motive for
leaving the place of residence, etc. A similar approach was adopted for the
employment history. This study uses data from both the migration history
and the employment history.
The community-level data come from a ret rospective community sur-
vey conducted of 600 settlements in early 2002. The survey was designed to
be linked with the individual migration survey. It comprises all the villages
in which people lived at the time of the survey and a large sample of villages
in which they lived in the past. Each village for which at least three spells of
residence were reported in the migration histories was covered by the
survey (600 villages of a total of 1800 villages in which at least one spell of
residence was reported). Retrospective data was collected from groups of
community informants, consisting of ‘‘de
´
le
´
gue
´
s de village’’ (administrative
representatives), village chiefs and other knowledgeable informants. It
covered a broad range of topics, including land availability, transportation,
agriculture, and employment opportunities. Efforts were made to obtain
retrospective information since 1960 on most village characteristics. For
example, informants were asked to recall all the years in which harvests
were ‘‘particularly bad’’, the year since which uncleared land was no longer
available, etc. A village calendar, including national events and events di-
rectly linked to the life in the village (such as the construction of a school)
was used to improve the quality of dates. The interview of several infor-
mants together, as it was done in this survey, is also believed to improve the
quality of the data (Axinn, Barber, & Ghimire, 1997; Frankenberg, 2000). In
addition, a detailed study of the quality of data was performed by com-
paring the statements of interviewees to external databases (Schoumaker
et al., 2004).
As mentioned above, community-level data was only collected for a
sample of settlements. Community-leve l variables are missing for 12.2% of
432
POPULATION AND ENVIRONMENT

the residence spells among the 3911 individuals in the analysis sample
(5.6% of the 145,000 person-periods). Random hot-deck imputation sup-
plied the missing values at the village level. The imputation is based on two
classification variables: the province (45 provinces) in which the village is
located and the size of the village (less than 5000 or between 5000 and
10,000 inhabitants). Adjustment cells were formed by combining these two
variables, and the missing values for a village were replaced by the values
from a randomly selected village in the same cell, that is a village of the
same size located in the same province. Multiple imputation was performed
to incorporate uncertainty due to missing community-level data (Allison,
2002).
Rainfall data coveri ng the 1960–1998 period were obtained from the
global monthly precipitation data set produced by the Climatic Research
Unit at the University of East Anglia (New, Hulme, & Jones, 2000). These
data were interpolated from a network of stations at a spatial resolution of
0.5°latitude and longitude. Monthly rainfall data were extracted from this
database and two rainfall variables were constructed at the department level
using geographical information systems (GIS).
4
Analysis Sample and Statistical Model
The analysis sample is restricted to people living in rural areas at age
15,
5
and covers the 1970–1998 period. Migration is defined here as a
change of residence involving a departure from the village for a duration of
at least three months. Each individual is ‘‘followed’’ from age 15 until his or
her first migration or unti l the time of censoring.
6
The data are organised as a
person-period data file in which each line represents a 3-month period, and
the dependant variable indicates if a migration occurs during each three-
month interval.
7
Overall, the sample consists of 3911 individuals (1800 men
and 2111 women), and a total of approximately 145,000 person-periods.
Binary and multinomial logistic regression methods are used to esti-
mate discrete-time event history models (Allison, 1995). Models that do not
distinguish among the event types are fitted with binary logist ic regression.
The statistical model is specified as follows:
log
p
ti
1  p
ti

¼ a
t
þ b
0
:X
ti
ð1Þ
where p
ti
is the conditional probability that individual i experiences the
event (first migration) at age t, given that the event has not already occurred.
433
S. HENRY ET AL.

a
rt
represents the baseline hazard function, and X
ti
is a vector of individual,
contextual and environmental covariates. Both time-constant and time-
varying covariates are included in the models.
Multinomial logistic regression is used for competing risk analyses that
distinguish among the destinations and/or types of migration. The discrete-
time competing risk model assumes that the log-odds of experiencing an
event of type r rather than an event of type s (the reference category) at time
t are given by
log
p
rti
p
sti

¼ a
rt
þ b
0
r
:X
rti
ð2Þ
where p
rti
is the conditional probability of an event of type r occurring at
time t for individual i, given that no event has occurred prior to time t. a
rt
represents the baseline hazard function for an event of type r,andX
rti
is a
vector of covariates. Censored cases (no migration) are treated as the ref-
erence category, and the destinations (rural, urban and abroad) and/or types
of migration (temporary or permanent) are distinguished as separate events.
All of the models take into account the fact that the data are clustered, and
the standard errors of the regression coefficients are adjusted accordingly
using Huber-White standard errors (Hox, 2002). Multiple imputation was
also performed to correct for the underestimation of standard errors due to
missing community-level data (Allison, 2002).
EXPLANATORY VARIABLES
Table 2 presents the explanatory variables as well as the proportion of
respondents in each category at age 15. The individual-level, community-
level and environmental variables included in the models are discussed
below. The two migration models, among men and women, include the
same explanatory variables (with identical categories). Although the num-
ber of cases may be small for some categories of the explanatory variables,
particularly for women, this approach has been preferred to ensure the
comparability of the models across gender.
Individual Variables
Numerous studies from both developed and developing cou ntries have
shown that age is strongly linked to the risk of migration. In this sample, the
434
POPULATION AND ENVIRONMENT

TABLE 2
Descriptive statistics of factors affecting migration conside red in the
analyses
% of the sample at age 15
Explanatory variables Male Female Total
Individual-level variables
Education
No education 84.8 93.0 89.3
Primary 12.2 5.9 8.7
Secondary and over 3.0 1.1 2.0
Ethnic group
Mossi 43.3 46.1 44.9
Fulani 10.6 9.2 9.8
Other 46.1 44.7 45.3
Activity
Agriculture 83.4 76.1 79.3
Cattle-raising 8.0 0.9 4.1
Other 8.6 23.0 16.6
Community-level variables
Uncleared land available
Yes 30.3 26.9 28.4
No 69.7 73.1 71.6
All-season road
Yes 41.4 44.9 43.3
No 58.6 55.1 56.7
Water conservation techniques
Yes 28.6 27.1 27.8
No 71.4 72.9 72.2
Rainfall variables
Average rainfall (mm)
200–499 5.7 5.0 5.3
500–699 27.0 26.6 26.8
700–899 47.6 47.9 47.7
900 and over 19.7 20.5 20.1
Rainfall variability
a
< 85% 5.8 6.0 5.9
85 – 94% 30.2 31.9 31.2
95% and over 64.0 62.1 62.9
Sample size 1800 2111 3911
a
Percentages computed on the total of person-periods.
Source: Migration Dynamics, Urban Integration and Environment Survey of Burkina Faso
(EMIUB), 2000.
435
S. HENRY ET AL.

risk of first migration among men increases with age for a few years after age
15 and then declines (Figure 2). Among females, the risk of first migration
decreases sharply with age. The non-linear relationships between age and
the risk of migration are model led here by a function of age and its loga-
rithm. These two measures form the baseline hazard of first migration.
8
The
positive relationship between education and migration is also one of the
most consistent findings of migration studies, especially for migrations to
urban areas (Lututala, 1995; Todaro, 1997). Education is measured by a
time-constant variable indicating the level attained by the individual at the
age of 15
9
.
Differential propensities to migrate among ethnic groups have also
been shown in various studies in Burkina Faso; in particular the Mossi and
the Fulani are distinguished from other ethnic groups. The Mossi, the
majority ethnic group in Burkina Faso, constitute 45% of the analysis
sample. Mossi people live mainly on the densely populated Central Plateau
(the Mossi Plateau), but are also known for their migrations to southwestern
regions (Marchal, 1975; Mathieu, 1994). The Fulani (10% of the sample)
live essentially in the northern and eastern parts of the country, and are also
a highly mobile population (Hampshire and Randall, 1999). Although the
Fulani used to be nomadic people specialised in cattle-raising, many of
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
15 20 25 30 35 40
Age
Probability of leaving the village
Males
Females
Males (parametric)
Females (paramtric)
FIGURE 2. Probability of leaving the village by age and sex.
436
POPULATION AND ENVIRONMENT

them settled under the influence of the colonial power and the independent
State (Copans, 1975). They are still dominant in cattle-raising activities
however: 57% of the men engaged in cattle -raising at age 15 are Fulani. The
third category assembles 10 ethnic groups, among which the Senoufo
(14%), the Gourmantche (8%) and the Gourounsi (7%) are the most rep-
resented.
Finally, a time-vary ing variable indicates the principal activity per-
formed by the individual at each point in time. Three categories are com-
pared in the models. Cultivators represent the large majority of the sampl e
(79% of the population at age 15). A little more than 4% of the respondents
declared that they were cattle-raisers, and 17% of the sample is engaged in
other types of activities, mainly craft and petty trade of food (students
account for 3% of the total).
Rainfall Variables
Two variables measured at the department level are used to capture
two dimensions of the potential impacts of rainfall on migration: the mean
annual precipitation over the 1960–1998 period and the percent of normal
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
15 20 25 30 35 40
Age
Probability of still being in the village
Males
Females
FIGURE 3. Probabi lity of still being in the village by age and sex.
437
S. HENRY ET AL.

precipitation over the three preceding years.
10
Besides their simplicity, the
two indicators used in this study were selected on an empirical basis.
Several rainfall indicators were tested to predict the occurrence of poor
harvests reported in the community survey, and the two indi cators retained
were the only significant predictors of the poor harvests.
People living in regions with a higher vulnerability to drought are
expected to be more likely to leave their villages — temporarily or per-
manently — than those living in regions with more rainfall. The first variable
(mean annual precipitation) is considered a good indicator of agricultural
productivity and of vulnerability to drought. Four categories are compared:
less than 500 mm per year, 500 to 699 mm per year, 700 to 899 mm per
year, and more than 900 mm per year (Figure 6). These categories corre-
spond to areas where crops with similar yield responses to water are cul-
tivated (Doorenbos & Kassam, 1987). In addition, we tried to have a
sufficient number of observations in each category.
Overall, it is also expected that the risk of first migration will increase if
the preceding years were unfavourable, although unfavourable conditions
could also have an opposite effect if people postpone migrations in periods
of economic stress (Findley, 1992; Nelson, 1983). The second variable is a
time-varying variable indicating the extent to which rainfall in the depart-
ment over the three preceding years differed from the long-term rainfall
conditions in the department. The measure is the ratio of the mean rainfall
over the three preceding years to the mean rainfall over the 1960–1998
period, and three categories
11
are compared in the models (less than 85%,
85 – 94%, and more than 95%). Several time windows (1, 2, and 3 years)
were tested and the three-year window was selected for several reasons.
First, rural househo lds may have sufficient stocks of cereals or enough
money, livestock and assets to purchase cereals follow ing a poor harvest
(Lallemand, 1975; Reardon et al., 1988). In other words, people may be
able to cope with one poor harvest without resorting to migration. They are
more likely to have depleted their assets after two or three consecutive bad
harvests however, a situation that could force them into mi gration. More-
over, several consecutive years of poor precipitation may also increase
people’s perception of worsening rainfall conditions and affect their deci-
sion to move to better watered regions. A longer time window would dilute
the relevance of this time-varying factor.
Community-level Variables
Three community-level variables are included to capture the effects of
opportunities and constraints at the village level that may influence
438
POPULATION AND ENVIRONMENT

migration behaviour. The first community-level variable is a time-varying
binary variable indicating, for each year, if uncleared land was still avail-
able in the village.
12
The common hypothesis is that out-migration tends to
be more pronounced from areas where land is no longer available or no
longer sufficient to maintain people’s levels of livelihood (Hugo, 1985;
Tallet, 1985). In Burkina Faso, the Mossi Plateau is very densely populated
compared to the eastern and southern parts of the country, and some
authors have suggested that migrations from the Mossi Plateau may be
induced by the lack of land (Tallet, 1985).
13
The second time-varying community-level variable indicates if the
village is connected by an all-season road. Transportation networks can
have complex impacts on the risks of migration (Bilsborrow, 1984). The
usual assumption is that higher accessibility increases the risk of migra-
tion, especially short-term migrat ion (Findley, 1987). However, access to
roads may also contribute to reducing out-mi gration by creating oppor-
tunities for income diversification (Reardon et al., 1988) or by facilitatin g
food aid distribution in periods of drought (Coulibaly & Vaugelade,
1981).
The third time-varying community-level variable indicates whether or
not contour stone walls (diguettes) are used in the village.
14
This soil and
water conservation technique not only reduces erosion but also improves
the availability of water for cultivation by encouraging infiltration (Bandre &
Batta, 1998). Such techniques have been shown to significantly improve
cereal yields in Burkina Faso (Bandre & Batta, 1998), and are expected to
reduce out-migration.
RESULTS
Descriptive Results
Figure 2 compares the risk of leaving the village for the first time after
the age of 15 for men and women, and Figure 3 shows the corresponding
survival curves, that is the probabilities of still being in the village by age.
Before the age of 18, women’s risks of leaving the village are significantly
higher than men’s risks, but are two to three times lower after that age. The
survival curves cross at around 21, and the probability of remaining in the
village is eventually much higher for women (50%) than for men (35%).
These gender differences are probably largely explained by marriage
migrations. Women’s marriages are concentrated in a narrow age range and
are often acco mpanied by a migration to the husband’s place of residence,
439
S. HENRY ET AL.

which means that a large proportion of women will migrate for the fi rst time
after age 15 at the time of their marriage (see below). Women’s migrations
for marriages before the age of 15 may also explain why, overall, a lower
percentage of women leave their village after the age of 15.
Table 3 gives some characteristics of first migrations. Gender differ-
ences are even more patent for these characteristics than for the timing of
first migration. While almost two-thirds of males’s first migrations are
directed to a foreign country, only 15% of their female counterparts en-
gage in such migrations. On the other hand, 70% of women’s first
migrations are directed to an other village, whereas this concerns only 23%
of men’s first migrations. Urban areas repre sent 14% of the destinations of
first migration among both men and women. As seen in Table 3, motives
for first migration also vary strongly according to gender. Almost 80% of
TABLE 3
Characteristics of first migrations from rural areas
Characteristics of migrations
(% of migrations) Males Females Total
Destination
Rural 23.0 70.4 48.2
Urban 13.6 14.3 14.0
Abroad 63.4 15.3 37.8
Type of migration
All destinations
Short-term (62 years) 31.3 5.2 17.4
Long-term (>2 years) 68.7 94.8 82.6
Rural areas
Short-term 12.7 1.5 4.0
Long-term 87.3 98.5 96.0
Urban areas
Short-term 19.3 9.1 13.8
Long-term 80.7 90.9 86.2
Abroad
Short-term 40.6 18.3 35.8
Long-term 59.4 81.7 64.2
Migration motives
Economic motives 68.1 14.0 39.7
Family 6.2 78.1 44.0
Schooling 6.4 1.3 3.7
Other 19.3 6.6 12.6
Sample size 1010 960 1970
440
POPULATION AND ENVIRONMENT

women move for family reasons (65% for marriage), and only a low 14%
for economic motives. A completely opposite pattern emerges among
men, where economic reasons are reported for two-thirds of migrations.
Finally, Table 3 also shows that 83% of those leaving do not return to
their village within 2 years, indicating that short-term migrations are not
the dominant type of move. Strong gender differenc es are observed
however: while only 5% of women go back to their village within two
years of departure, almost one-third of their male counterparts are
involved in short-term migrations. Migrants to foreign countries are also
much more likely to return to their villag e within 2 years than those
moving to urban and rural areas.
Multivariate Results: All Destinations Combined
The main hypothesis is that individuals who live in areas with unfa-
vorable rainfall conditions — including long-term conditions and short-term
fluctuations — are expected to have a higher risk of leaving their villag e
than those who live in areas with more favorable rainfall conditions. We
first look at the factors influencing the risk of the first departure from the
village, treating all migration types the same. All of the models are fitted
separately for males and females (Table 4). The first model includes the
individual-level variables and the rainfall variables (models 1a and 1b); the
community-level variables are added in the second series of models
(models 2a and 2b).
Individual-level variables. Models 1a and 1b (Table 4) show that, as
expected, the risk of leaving the village for the first time is higher among
educated people (2.41
***
). Among men, the risk of moving is significantly
higher only for those who have had some secondary schooling (2.04
***
).
The effect of primary education is more pronounced among women, while
there is no difference between those who reached primary school and the
few women that went to secondary school. The type of activity (time-
varying variable) is also strongly related to the risk of moving. The odds of
leaving the village are twice higher among people engaged in non-agri-
cultural activities (2.06
***
). The risk of leaving the village is higher among
women engaged in cattle-raising (2.78
***
, only a few women). On the other
hand, males engaged in cattle-raising (0.92) are slightly (but not signifi-
cantly) less likely to move than those involved in agriculture (reference).
The higher propensity to migrate among the Mossi (reference) is in line with
the literature on the specificity of Mossi migration.
441
S. HENRY ET AL.

TABLE 4
Event history models of individual, contextual and environmental effects
on the risk of leaving the village, 1970– 1998a
Males Females
Explanatory variables Model 1a Model 2a Model 1b Model 2b
Baseline hazard
Age 0.87
**
0.86
***
0.83
**
0.83
***
Log age 2.03 2.07
***
1.30 1.35
Education
No education (R) 1.00 1.00 1.00 1.00
Primary 1.12 1.08 1.66 1.46+
Secondary and over 2.72 2.41
***
2.46 2.04
***
Ethnic group
Mossi (R) 1.00 1.00 1.00 1.00
Fulani 0.68 0.64+ 0.42+ 0.39
***
Other 0.62
*
0.62
***
0.60+ 0.59
***
Activity (time-varying)
Agriculture (R) 1.00 1.00 1.00 1.00
Cattle-raising 0.91 0.92 2.49 2.78
***
Other 2.08 2.06
***
1.51+ 1.58
***
Average rainfall (mm)
200–499 0.92 0.91 1.18 1.04
500–699 0.87 0.91 1.33 1.43
700–899 0.80+ 0.72 1.10 0.97
900 and over (R) 1.00 1.00 1.00 1.00
Rainfall variability
(time-varying)
< 85% 0.88+ 0.89 0.78+ 0.78+
85–95% 0.83+ 0.83 1.01+ 1.00
95 and over (R) 1.00 1.00 1.00 1.00
Water conservation
techniques (time-varying)
No (R) 1.00 1.00
Yes 1.11 1.00
Uncleared land available
(time-varying)
No (R) 1.00 1.00
Yes 0.96 1.05
All-season road
(time-varying)
No (R) 1.00 1.00
Yes 1.50
***
1.88
**
R – Reference category;
a
Results expressed as odds-ratios;
***
p < 0.01;
**
p < 0.05;
*
p < 0.10;
+: p < 0.20 (two-tailed tests).
442
POPULATION AND ENVIRONMENT

Rainfall and community-level variables. Models 1a and 1b (Ta-
ble 4) also include the two rainfall variables. Contrary to our expectation,
results show that men living in poor agro-climatic regions tend to have
slightly lower risks (0.92) of moving than those in areas with greater rainfall
(reference), but the relationship is not statistically significant. Among wo-
men, there is no relationship between the agro-climatic region and the risk
of leaving the village. Unexpectedly, results for the second rainfall variable
suggest that, for both men and women, the risk of moving is in fact slightly
higher if the previous years were favourable (reference) than if they were
unfavourable (0.88). How ever, these results are not statistically significant
either. The effects of the individual-level and rainfall variables do not
change notably with the incl usion of community-level variables (models 2a
and 2b, Table 4). As expected, the presence of an all-season road increases
the overall risk of moving (1.50
***
), and is the only statistically significant
community-level variable. However, people living in villages where un-
cleared land is still available do not have a lower risk of leaving the village.
Finally, there is no significant effect of the use of a water conservation
technique on the risk of moving.
Multivariate Results: Distinction Between Destinations
As discussed before, there are reasons to expect that the effects of
explanatory variables —notably environmental variables — may differ
across destinations. Three types of destinations (rural areas, urban areas and
abroad) are distinguished with compet ing risk models. Results are reported
in Table 5 (models 3a and 3b). The results are presented in three separate
columns, each column pertaining to the contrast between a single desti-
nation and no migration (reference).
Individual level variables. The effects of the individual-level vari-
ables vary strongly depending on the destination. The relationship between
education and the risk of moving for the first time to urban areas is very
strong and positive (5.82
***
), but is weak and non-significant towards rural
areas (1.88) and rather negative for migrations to a foreign country (0.56).
The fact that educated people are more likely to move to urban areas than
the uneducated is thus clearly confirmed in Burkina Faso. The link between
ethnic group and the risk of leaving the village for the first time also varies
strongly by destination. While the Mossi are overall more likely to leave
their village when no distinction is made between destinations, Fulani
males have the highest propensity to leave their village for another village
(rural-rural migration, 1.67). This is not observed among females however,
443
S. HENRY ET AL.

TABLE 5
Competing risk models of individual, contextual and environmental effects on the risk of leaving the village for
different destinations, 1970–1998
a
Male (Model 3a) Female (Model 3b)
Rural
vs. no
migration
Urban
vs. no
migration
Abroad
vs. no
migration
Rural
vs. no
migration
Urban
vs. no
migration
Abroad
vs. no
migration
Baseline hazard
Age 0.98 0.87
***
0.80
***
0.84
***
0.73
*
0.84
**
Log age 0.98 1.80
**
3.26
***
1.24 2.69 1.25
Education
No education (R) 1.00 1.00 1.00 1.00 1.00 1.00
Primary 0.61+ 1.34 1.18 1.36 2.71
***
0.92
Secondary
and over
1.88 5.82
***
0.56 0.61 6.82
***
0.16
*
Ethnic group
Mossi (R) 1.00 1.00 1.00 1.00 1.00 1.00
Fulani 1.67 0.35
**
0.30
***
0.49
***
0.06
**
0.48+
Other 0.77 0.68 0.57
***
0.53
***
0.36
**
1.88
**
Activity
Agriculture (R) 1.00 1.00 1.00 1.00 1.00 1.00
Cattle-raising 1.50 0.36 0.71 2.39
**
0.53 8.99
***
Other 4.25
***
4.40
***
1.00 1.34
**
3.68
***
1.51
Average rainfall (mm)
200–499 3.16
**
1.10 0.61 1.69+ 0.58 0.46+
500–699 2.73
***
1.57 0.67 2.52
***
1.22 0.38
**
700–899 1.81
**
1.62 0.51
**
1.43 1.39 0.48
**
900 and over (R) 1.00 1.00 1.00 1.00 1.00 1.00
444
POPULATION AND ENVIRONMENT

TABLE 5 (Continued )
Male (Model 3a) Female (Model 3b)
Rural
vs. no
migration
Urban
vs. no
migration
Abroad
vs. no
migration
Rural
vs. no
migration
Urban
vs. no
migration
Abroad
vs. no
migration
Rainfall variability
< 85% 1.58
*
0.62 0.70+ 0.78+ 0.37+ 1.27
85 – 95% 1.42
**
0.76 0.69
**
0.99 0.60
**
1.48+
95 and over (R) 1.00 1.00 1.00 1.00 1.00 1.00
Water cons. Techniques
No (R) 1.00 1.00 1.00 1.00 1.00 1.00
Yes 0.69+ 0.78 1.38
*
1.13 0.42
**
1.13
Uncleared land available
No (R) 1.00 1.00 1.00 1.00 1.00 1.00
Yes 0.97 0.68 1.01 1.21 0.47+ 0.95
All-season road
No (R) 1.00 1.00 1.00 1.00 1.00 1.00
Yes 1.57+ 3.30
***
1.31+ 1.97
**
3.30
***
1.16
***
: p < 0.01;
**
: p < 0.05 ;
*
: p < 0.10; +: p < 0.20 (two-tailed tests)
a
Results are expressed in odds-ratio
445
S. HENRY ET AL.

where the risk of moving to another village among the Fulani (0.49
***
)is
only half the risk of the Mossi.
15
The Mossi, both men and women, have a
much higher propensity to move to urban areas than the other ethnic
groups, and Mossi males are also significantly more likely to leave for Co
ˆ
te
d’Ivoire, confirming a known character istic of foreign migration from
Burkina Faso (Blion, 1995; Cordell et al., 1996; Deniel, 1967).
The effect of the type of activity on the risk of moving for the first time
also differs across destinations and gender. Regardless of gender, individuals
involved in cattle-raising are more likely to move to rural areas than those
involved in agriculture. On the other hand, males involved in cattle-raising
are less likely to move to urban areas. As far as temporary moves are
concerned, this could be explained by the year-round commitment
required: cattle-raisers cannot easily leave their herds just for a few months
(Hampshire & Randall, 1999) . No explanation was found for the higher risk
of moving abroad among females involved in cattle -raising, but only a few
women are involved in such activities. Finally, people involved in activities
other than agriculture or cattle-raising are overall much more likely to leave
their village.
Rainfall variables. The relationships between rainfall variables and
the risk of leaving the village for the first time also vary significantly across
destinations. The link between average rainfall conditions and migration is
especially clear in rural areas. As expected, the odds of leaving the village
for another village are three times higher for men living in the poorest agro-
climatic region (3.16
**
) than for those living in areas with an average rainfall
over 900 mm (reference) (Table 5). Even though the relationship is less
pronounced for women, those living in wetter areas (reference) are also less
likely to move to another rural destination than women living in the drier
regions (2.52
***
). The relationship between the average rainfall and the risk
of migrating to urban areas is much less consistent. There is however a
significant relation ship between the agro-climatic conditions and the risk of
leaving for a foreign country, among both males and females. Results
indicate that those living in the wetter areas (reference) are in fact much
more likely to leave their village for a foreign country (0.61). Although it
may seem surprising, this result is not completely une xpected. First, this
could partly be interpreted as a proximity effect, since the wetter regio ns are
located near Co
ˆ
te d’Ivoire. However, migrations to Co
ˆ
te d’Ivoire are gen-
erally directed to the more remote areas in the forest zone, and this suggests
that factors other than distance are at play. One tentative explanation is that
people from the wetter regions of Burkina Faso must necessarily go further
south (out of Burkina Faso) to find better watered areas.
446
POPULATION AND ENVIRONMENT

Rainfall conditions in the three preceding years are also significantly
related to the propensity for leaving the village for the first time. As ex-
pected, the risk of moving to rural areas is significantly higher among males
if there was a rainfall deficit over the three preceding years (1.58
*
). The odds
of leaving are 60% higher if the conditions were unfavourable (less than
85% of the long-term average) than if they were normal (more than 95%).
This result thus suggests that not only average rainfall but also short-term
unfavourable rainfall conditions tend to push men to leave for other rural
areas. Somewhat surprisingly, there is no significant effect of recent rainfall
conditions on the risk of leaving among females, and they seem rather less
likely to leave thei r village in years following poor rainfall conditions. An
interesting result is the opposite pattern that emerges for migrations to urban
areas and foreign countries. Overall, people are more likely to migrate to
these destinations if the recent rainfall conditions were favourable. Both
men and women seem to be more likely to move to urban areas in normal
conditions than in unfavourable periods (0.62 for men, 0.37
+
for women),
although results are not significant. Among men, the odds of moving to a
foreign country are also approximately 50% higher in normal periods than
in unfavorable periods (0.69
**
). One tentative explanation might be that
rural people are waiting for good economic conditions in the preceding
years before moving to urban areas and to foreign countries (Findley, 1994;
Nelson, 1983), as they may need a production surplus to finance their
migration.
Community-level effects. Two of the three community-level vari-
ables are significantly related to the risk of leaving the village for the first
time (Table 5). First, as expected, people living in villages connected by an
all-season road are more likely to migrate to rural and urban areas than
those living in places without roads. The effect is particularly strong on the
risk of moving to urban areas, with odds of moving more than three times
higher in places accessible by road. This agrees with results from studies in
other settings (Findley, 1987). Contrary to our expectation however, there is
no significant relationship between the availability of uncleared land and
the risk of moving. This somewhat surprising result thus suggests that the
link between land availability and migration is weak, or that a more refined
analysis is needed. One tentative explanation might that agricultural
intensification in more densely populated places may offset the expected
effect of land availability on migration. Finally , men living in places where
soil and water conservation techniques are used are somewhat less likely to
migrate to rural and urban areas (not significant), while they are significantly
more likely to move abroad. As discussed before, the slight negative effect
447
S. HENRY ET AL.

in rural and urban areas could be explained by the fact that these techniques
improve cereal yields and as a result may decrease the need for migration.
16
However, no explanation was found at this stage for the fact that people
living in places where water conservation techniques are used are more
likely to move to foreign countries.
Multivariate Results: Distinction Between Short-term and Long-term
Migrations
While the previous results identify some effects of environmental
conditions on the risk of leaving the village, they do not indicate whether
people leave their village permanently or for a short period. In this section,
we further refine the dependant variable to distinguish between short-term
and long-term migrations. Six types of migrations — classified according to
destination and duration — are now considered. A short-t erm migration is
defined here as a change of residence involving a departure from the village
for a duration of between 3 months and 2 years.
17
As discussed before,
short-term migrations represent 17% of first migrations, but that proportion
varies str ongly by gender and destination (Table 3). Males are much more
likely to engage in short-term migration (31%), and international migrant s
are also much more likely to return to their village within 2 years (36%).
Multivariate results are presented in Tables 6 (males) and 7 (females).
Interesting patterns emerge for individual an d community-level vari-
ables. For example, more educated men are significantly more likely to
engage in long-term moves than uneducated people, and are much less
likely to move for a short duration. Among males, the presence of an
all-season road in the village has a stronger effect on the risk of undertaking
a short-term rural-urban migration than on the risk of a long-term move. For
the purpose of this paper, however, we will concentrate on the effects of the
rainfall variables.
In regard to average rainfall conditions, additional insight is provided
by the distinction between short-term an d long-term migrations. Results
from the previous section indicated that men from the drier regions were
much more likely to migrat e to rural areas than those from wetter regions.
Table 6 shows that this is to a large extent the result of their much higher
propensity to engage in short-term moves (21.43
***
), although permanent
migrations to rural areas are also more likely among people from the drier
regions (2.18). The same conclusion applies to women. The risk of short-
term migration to urban areas is also higher (although not significantly)
among males (3.09) and females (2.17) living in the drier regions. Finally,
the odds of a short-term migration to a foreign country is slightly higher
448
POPULATION AND ENVIRONMENT

TABLE 6
Competing risk models of individual, contextual and environmental effects on the risk of short-term
( < =2 years) and long-term (>2 years) migrations for different destinations, males, 1970–1998
a
To rural areas To urban areas Abroad
Long-term vs.
no migration
Short-term vs.
no migration
Long-term vs.
no migration
Short-term vs.
no migration
Long-term vs.
no migration
Short-term vs.
no migration
Baseline hazard
Age 0.98 0.96 0.87
***
0.87
**
0.77
***
0.84
***
Log age 0.95 1.47 2.08
***
0.98 3.76
***
2.92
***
Education
No education (R) 1.00 1.00 1.00 1.00 1.00 1.00
Primary 0.63 0.59 1.74+ 0.51 0.95 1.50
*
Secondary and over 2.28
+
(n.a.)
1
10.58
***
0.07
*
0.47 0.76
Ethnic group
Mossi (R) 1.00 1.00 1.00 1.00 1.00 1.00
Fulani 2.05
+
0.55 0.29
**
8.69
**
0.39
**
0.22
***
Other 0.75 1.10 0.38
**
35.89
***
0.46
***
0.78
Activity
Agriculture (R) 1.00 1.00 1.00 1.00 1.00 1.00
Cattle-raising 1.10 5.62
***
0.27+ 0.59 0.60 0.85
Other 3.83
***
8.12
***
3.28
**
7.64
***
1.02 0.93
Average
rainfall (mm)
200–499 2.18 21.43
***
0.52 3.09 0.27
**
1.61
500–699 2.37
**
7.65
*
1.08 3.55 0.46
*
1.12
700–899 1.87
**
1.43 1.37 2.27 0.39
***
0.75
900 and over (R) 1.00 1.00 1.00 1.00 1.00 1.00
Rainfall variability
449
S. HENRY ET AL.

TABLE 6 (Continued )
To rural areas To urban areas Abroad
Long-term vs.
no migration
Short-term vs.
no migration
Long-term vs.
no migration
Short-term vs.
no migration
Long-term vs.
no migration
Short-term vs.
no migration
< 85% 1.64
*
1.44 0.68 0.73 0.95 0.37
**
85 – 95% 1.31
+
2.25
*
0.68 1.29 0.66
*
0.73
*
95% and over (R) 1.00 1.00 1.00 1.00 1.00 1.00
Water cons.
techniques
No (R) 1.00 1.00 1.00 1.00 1.00 1.00
Yes 0.51
*
3.70
***
0.59
+
2.45
*
1.43
*
1.32
Uncleared
land available
No (R) 1.00 1.00 1.00 1.00 1.00 1.00
Yes 0.94 1.30 0.65 1.30 1.27 0.69
+
All-season road
No (R) 1.00 1.00 1.00 1.00 1.00 1.00
Yes 1.66
+
0.75 2.69
**
6.23
***
1.55
**
1.01
***
: p < 0.01;
**
: p < 0.05 ;
*
: p < 0.10; +: p < 0.20 (two-tailed tests).
1
No estimate available because no event occurred among the observations in this category.
a
Results are expressed in odds-ratio.
450
POPULATION AND ENVIRONMENT

TABLE 7
Competing risk models of individual, contextual and environmental effects on the risk of short-term
( < = 2 years) and long-term (>2 years) migrations for different destinations, females, 1970–1998
a
To rural areas To urban areas Abroad
Long-term vs.
no migration
Short-term vs.
no migration
Long-term vs.
no migration
Short-term vs.
no migration
Long-term vs.
no migration
Short-term vs.
no migration
Baseline hazard
Age 0.83
***
1.06 0.69
*
0.97 0.71
***
1.05
Log age 1.28 0.44 3.51
+
0.60 2.37
*
0.49
**
Education
No education (R) 1.00 1.00 1.00 1.00 1.00 1.00
Primary 1.36 1.15 2.96
***
1.29 0.80 1.35
Secondary and over 0.60 (n.a.)
1
6.80
***
4.66 0.19
+
(n.a.)
1
Ethnic group
Mossi (R) 1.00 1.00 1.00 1.00 1.00 1.00
Fulani 0.46
***
4.39
+
0.06
**
0.24 0.51 0.24
Other 0.52
***
2.21 0.29
**
2.39 1.65
*
3.00
Activity
Agriculture (R) 1.00 1.00 1.00 1.00 1.00 1.00
Cattle-raising 2.32
**
4.42
+
0.57 (n.a.)
1
10.26
***
(n.a.)
1
Other 1.36
***
0.37
*
3.83
***
2.76 1.58 1.29
Average rainfall (mm)
200–499 1.56 34.41
***
0.51 2.17 0.40
*
0.73
500–699 2.46
***
20.45
**
1.14 2.55 0.34
***
0.58
700–899 1.41 8.33
*
1.23 3.98 0.37
***
1.28
900 and over (R) 1.00 1.00 1.00 1.00 1.00 1.00
Rainfall variability
< 85% 0.78
+
0.83 0.41 (n.a.)
1
1.31 1.12
451
S. HENRY ET AL.

TABLE 7 (Continued )
To rural areas To urban areas Abroad
Long-term vs.
no migration
Short-term vs.
no migration
Long-term vs.
no migration
Short-term vs.
no migration
Long-term vs.
no migration
Short-term vs.
no migration
85 – 95% 0.99 1.14 0.60
*
0.60 1.79
*
0.56
95% and over (R) 1.00 1.00 1.00 1.00 1.00 1.00
Water cons. techniques
No (R) 1.00 1.00 1.00 1.00 1.00 1.00
Yes 1.13 0.86 0.38
**
0.87 1.32 0.49
Uncleared land available
No (R) 1.00 1.00 1.00 1.00 1.00 1.00
Yes 1.19 2.76
+
0.51 0.00 0.95 1.00
All-season road
No (R) 1.00 1.00 1.00 1.00 1.00 1.00
Yes 1.97
**
1.68 3.83
***
0.99 1.20 1.06
***
: p < 0.01;
**
: p < 0.05 ;
*
: p < 0.10;
+
: p < 0.20 (two-tailed tests).
1
No estimate available because no event occurred in this category.
a
Results are expressed in odds-ratio.
452
POPULATION AND ENVIRONMENT

among males living in regions with scarce rainfall (1.61 ), and lower among
females (0.73). These results thus indicate that men from regions with scarce
and irregular rainfall are much more likely to engage in temporary migra-
tion, mainly to other rural areas and to a lesser extent to urban and foreign
destinations. This agrees with results from two small -scale studies in Burkina
Faso (Homewood, 1999; Reardon et al., 1988), and supports the idea that
short-term migration is a way to diversify income sources in a context of
severe production variability (Reardon et al., 1988). Interestingly, people
living in agro-climatically poor regions are also more likely to leave their
village permanently for another place in rural areas (2.37
**
for men, 2.46
***
for women). On the other hand, they seem rather less likely to engage in
long-term migration foreign destinations than thei r counterparts in wetter
areas.
The distinction between short-term and long-term moves also offers
new insight into the link between rainfall variability and the risk of leaving
the village. Overall, it seems that poor rainfall conditions in the previous
years do not increase the risk of undertaking a short-term move
18
. While
there is a slight increase of temporary migrations to rural areas (2.25
*
),
males are in fact significantly less likely to move abroad for a short durat ion
(0.37
**
), and there is no relationship for migrations to urban areas. The
lower risk of moving abroad following poor rainfall conditions might be
explained by the fact that such moves entail higher costs than moving to less
distant areas. On the other hand, the odds of long-term male migration to
rural areas do significantly incr ease in bad years (1.64
*
). This is consistent
with the hypothesis that people might decide to move permanently after
severe rainfall deficits. The opposite effect is observed among females
though: women are more likely to leave their village permanently for an-
other village following years of relatively high rainfall. Given that most
long-term female migrations to rural areas are marriage migrations, one
tentative explanation might be that marriages tend to be delayed in periods
of climatic stress. This runs counter to the hypothesis of increased marriages
in drought periods as suggested by Findley (1994).
DISCUSSION AND CONCLUSION
The influence of environment al factors on migration has been the ob-
ject of increased interest over the last few years (Lonergan, 1998). Because
of the large environmental disparities within the country and the significant
proportion of the rural population involved in agriculture, Burkina Faso is
an interesting setting in which to investigate the links between environ-
453
S. HENRY ET AL.

mental conditions and migration. Using recent longitudinal multilevel data,
event history models were used to test the effects of rainfall conditions and
community-level characteristics on the first mi gration, in addition to clas-
sical socio-demographic individual factors.
The main hypothesis was that individuals who live in areas with
unfavorable rainfall conditions — including long-term conditions and short-
term fluctuations — are expected to have a higher risk of leaving their
village than those who live in areas with more favorable rainfall conditions.
The findings for the environmental variables, of special interest in this pa-
per, show the following points. First, there is no evidence of an effect of
rainfall conditions on the risk of first migration from rural areas when no
distinction by destination or duration is made. This suggests that environ-
mental conditions either have no effect on migration, or that opposing ef-
fects are at play depending on the type of migration. Overall, these results
suggest that migration behaviour is not very responsive to community-level
variables and to environmental factors as measured by rainfall variables, but
rather depends on individual characteristic s such as the educational level,
the type of activity or the ethnic group to which the individual belongs.
Although these results are somewhat surprising, the absence of a relation-
ship between rainfall variability and migration agrees with results of the
1974–1975 Nationa l Migration Survey (Coulibaly & Vaugelade, 1981).
Opinion data from that survey showed that the vast majority of people
declared that their migration behaviour had not been affected by the
drought conditions, and only a low 4% of migrants declared that they had
moved because of the drought . This is all the more surprising given that the
early 1970s were characterised by a severe drought
19
. One explanation
suggested by the authors was that the effect of the drought on migration was
weakened by massive food distribution. The same explanation was offered
by Findley to account for the fact that the volume of migration had not risen
during the 1983–1985 drought in Mali (Findley, 1994). Coulibaly and Va-
ugelade (1981) also advanced the idea that the effect of poor rainfall con-
ditions on migration was not more apparent was because all migrations
were considered together, with no distinction by sex, destination or
migration type.
The effect of rainfall conditions was then hypothesized to be different
depending on the destination and the duration of migrations. In the case of
Burkina Faso, the distinctions by destination and duration of migration
prove critical in measuring a relationship between rainfall conditions and
the risk of leaving the village. Overall, results indicate that environmental
conditions as measured by rainfall variables are indeed linked to migratory
behaviour, but in rather intricate ways depending on the varying types of
454
POPULATION AND ENVIRONMENT

migrations. A significant result from our models is that men and women
living in areas where rainfall is scarce are much more likely to leave their
village for another village (rural–rural migration) than are those living in
areas with greater rainfall. This greater propensity to leave is to a large
extent the consequence of their greater tendency to engage in short-term
moves, which supports the theor y that short-term migrations are part of a
strategy to diversify income sources in a risky environment (Hampshire &
Randall, 1999; Reardon et al., 1988). Interestingly, overall short-term
migrations do not rise following a severe rainfall deficit: international
temporary migrations are in fact less common among males in such periods.
The former result agrees with a study on migrations in northern Burkina Faso
that suggested that there was no increase in migrations in periods of drought
(Homewood, 1999). The second result suggests that, rather than encour-
aging migration, rainfall deficits and bad harvests tend to limit people’s
ability to invest in long-distance moves (Findley, 1994) . Overall, long-term
migrations seem to be less related to environmental conditions than short-
term moves, and the effects also differ across gender. In short, men are more
likely to move permanently to another village if they live in a region where
rainfall is scarce and in years following poor rainfall conditions. This is
consistent with the hypothesis that migration rises immediately and as a
long-term response to the threat of recurrent droughts (Findley, 1994).
While women are also more likely to leave their village for another village if
they live in the drier regions, they are less likely to move after bad rainfall
conditions. Permanent migrations to urban areas seem to be less related to
rainfall conditions. Finally, men and women from better watered areas are
more likely to engage in a long- term migration to a foreign country than
those living in the drier regions.
The results of this paper should not be considered as definitive and
further research is needed on sever al topics. This study focused on the first
migration, which is only one part of the story. Further analyses could
investigate the effect of environmental conditions on subsequent migrations
from rural areas. Perhaps more importantly, it could be fruitful to examine
the effect of environmental conditions of the village of origin on the risk of
returning to that village. One might expect that those leaving better-off
places would be more likely to return to their home-village than those
leaving places where rainfall is scarce. In fact, our results that distinguish
migrations by duration tend to suggest that this is not the case, but more
work is needed on this topic. Looking at exactly where migrants move and
what they do in their new place of residence could also inform on their
underlying motives for leaving their village. Many additional individual,
household, community and environmental variables also remain to be
455
S. HENRY ET AL.

investigated. Individual and household economic conditions, secon dary
activities, the spread of parasitic plants and the availability of off-farm
activities in the community are some of the possible candidates for the
formulation of more complex mod els. Other environmental factors could
also be taken into account in migration models. The effects of land degra-
dation on the risk of migration deserve special attention, although the lack
of validated data at the national level constitutes a serious obstacle to the
exploration of this issue. Finally, the characteristics of potential destinations
could also be taken into account in further research (Baydar et al., 1990),
although this would lead to more complex models and would require data
that are not readily available. Such research will further our efforts to ex-
plain the complex interactions of the factors which influence migration in
Burkina Faso.
ACKNOWLEDGMENT
This research was supported by the Canadian International Develop-
ment Agency and by the Andrew Mellon foundation. We are grateful to the
Unite
´
d’Enseignement et de Recherche en De
´
mographie (Burkina Faso), the
Universite
´
de Montre
´
al (Can ada) and the Program Majeur en Population et
De
´
veloppement (Mali) for granting us access to the demographic and
community data. We express our gratitude to the Climatic Research Unit
(UK) for the rainfall data. We thank Eric Lambin for his comments on a
earlier draft.
ENDNOTES
1. Results from the survey used in this study, ‘‘Migration Dynamics, Urban Integration and
Environment Survey of Burkina Faso’’.
2. One should note at this point that only those international migrants who had returned to
Burkina Faso were included in the survey. As a result, the proportion of international
migrants who return to their village is overestimated. However, analyses using informa-
tion collected from other household members on international migrants who had not
returned by the time of the survey confirm that these have a much higher probability of
returning to their village.
3. Migration Dynamics, Urban Integration and Environment Survey of Burkina Faso (EMIUB
for Enquete ‘‘Dynamique migratoire, Insertion urbaine et environnement au Burkina
Faso’’).
4. Burkina Faso is composed of approximately 350 departments.
5. Rural areas include all communities of less than 10 000 inhabitants (Beauchemin et al.,
2002).
456
POPULATION AND ENVIRONMENT

6. The individuals entering the risk set in 1970 may be more than 15 years old at the time of
entry. This is a situation of late entry into the risk set that is easily handled in discrete-time
event history models (Allison, 1995).
7. A 3-month migration definition was used to include temporary migrations in the dry
season and migration related to short-term activities in urban areas (Poirier et al., 2001).
8. Models with dichotomous indicators of each age were also estimated. The estimates of the
other coefficients in these models did not change. We thus chose the more parsimonious
parametric specification of the baseline hazard.
9. The educational level does not vary very much after the age of 15.
10. The timing of the rains (delays in the start of the rainy season, occurrence of rains in
relation to crop growth stages) could also be an important factor. The intra-annual vari-
ability of rainfall was also tested but not used in this study because this variable was not
correlated with bad harvests.
11. This categorization was arbitrarily defined but several tests were performed to test the
robustness of the results.
12. This information was collected in two stages. A first question was asked about the
availability of uncleared land in the village at the time of the survey. In the communities in
which uncleared land was no longer available, respondents were then asked to recall the
(approximate) year when uncleared land was last available.
13. Preliminary analyses indicate a strong relationship between the availability of uncleared
land in the village and population density measured at the province level.
14. Due to the absence of data at the household level, the presence of contour walls in a
community is assumed to equally affect all community members.
15. This could be explained by the lower age at marriage among Fulani females. Since a
larger proportion of Fulani females are already married by age 15, they will probably be
less likely to move after that age.
16. One should note that this variable may be endogenous and that its measured effects
should be interpreted with caution. For example, given that the implementation of such
techniques requires a significant involvement of the workforce, the observed result for
migrations to rural and urban areas might be partly explained by a greater probability of
implementing soil and water conservation techniques in places where male migration is
low. Further research is thus needed to fully understand the measured effects.
17. This definition is similar to that used by De Jong, Chamratrithirong, & Tran, (2002).
18. Models that only distinguish short-term and long-term moves (results not shown) in fact
indicate that short-term migrations are significantly less likely following severe rainfall
deficits.
19. Similar results were reported in Ethiopia where less than 2% of people mentioned drought
as the cause for leaving their parent’s home (Ezra, 2000).
REFERENCES
Allison, P. D. (1995). Survival analysis using the SAS system. A practical guide. Cary: SAS
Institute.
Allison, P. D. (2002). Missing data. Thousand Oaks: Sage.
Axinn, W., Barber, J., & Ghimire, D. (1997). The neighborhood history Calendar: A data
collection method designed for dynamic multilevel modelling. Sociological Methodology,
27, 355–392.
Bandre, P. & Batta, F. (1998). Soil and water conservation in Burkina Faso. London: Overseas
Development Institute.
Baydar, N., White, M., Simkins, C., & Babakol, O. (1990). Effects of agricultural development
policies on migration in Peninsular Malaysia. Demography, 27(1), 97–109.
457
S. HENRY ET AL.

Beauchemin, C., Beauchemin, E., & Le Jeune, G. (2002). TABVILLES BF: rapport de pre
´
sen-
tation. Technical report n°. 2002-1. Department of Demography, University of Montreal.
Berner, D., Winslow, M., Awad, A., & Cardwell, K. (Eds.). (1997). Striga research methods.
A manual. Ibadan: International Institute of Tropical Agriculture.
Bilsborrow, R. (1984). The need for and design of a community-level questionnaire. In Bilsborrow,
A. Oberai, & G. Standing (Eds.), Migration surveys in low income countries: Guidelines for
survey and questionnaire design (pp. 407–486). London and Sidney: Croom Helm.
Blion, R. (1995). Les Burkinabe
´
sdeCo
ˆ
te d’Ivoire Entre ’Inte
´
gration’ et Circulation Migratoire.
Mondes en De
´
veloppement, 23(91), 81–93.
Boutillier, J.-L., Quesnel, A., & Vaugelade, J. (1977). Syste
`
mes Socio-Economiques Mossi et
Migrations. Cahiers ORSTOM Sciences Humaines, 14(4), 561–581.
Copans, J., (Ed.) (1975). Se
´
cheresses et Famines au Sahel. Dossiers Africains. Paris: Franc¸ois
Maspe
´
ro.
Cordell, D., Gregory, J., & Piche
´
, V. (1996). Hoe and wage. A social history of a circular
migration system in West Africa. Boulder: Westview Press.
Coulibaly, S., & Vaugelade, J. (1981). Impact de la Se
´
cheresse 1969–1974 sur les Migrations en
Haute-Volta. Paper presented at the International Population Conference, Manila 1981
(unpublished).
De Jong, G. F., Chamratrithirong, A., & Tran, Q.-G. (2002). For better, for worse: life satis-
faction consequences of migration. International Migration Review, 36(3), 838–863.
Deniel, R. (1967). De la Savane a
`
la Ville: Essai sur la Migration des Mossi vers Abidjan et sa
Re
´
gion. Aix en Provence: Centre africain des Sciences Humaines Applique
´
es.
Doorenbos, J., & Kassam, A. H. (1987). Re
´
ponse des rendements a
`
l’eau. Bulletin FAO d’irri-
gation et de drainage. Rome: FAO, 235 pp.
Ezra, M. (2000). Leaving-home of young adults under conditions of ecological stress in the
drought prone communities of northern Ethiopia. Genus, 56(3–4), 121–144.
Ezra. M. (2001). Ecological degradation, rural poverty, and migration in Ethiopia: A contextual
analysis. Policy Research Division Working Papers 149. New York: Population Council.
Findley, S. (1987). Rural development and migration. A study of family choices in the Philip-
pines. Boulder: Westview Press.
Findley, S. (1992). Circulation as a drought coping strategy in rural Mali. In C. Goldscheider
(Ed.), Migration, population structure, and redistribution policies (pp. 61–89). Boulder:
Westview Press.
Findley, S. (1994). Does drought increase migration? A study of migration from rural Mali
during the 1983–1985 Drought. International Migration Review, 28(3), 539–553.
Frankenberg, E. (2000). Community and price data in the Living Standards Measurement
Surveys. In Grosh, M., & Glewwe P., (Eds.), Designing Household Survey Questionnaires
for Developing Countries: Lessons from Fifteen Years of LSMS Experience. New York: The
World Bank.
Gardner, R. (1981). Macrolevel influences on the migration decision process. In G. De Jong &
R. Gardner (Eds.), Migration decision making: Multidisciplinary approaches to microlevel
studies in developed and developing countries (pp. 59–87). New York: Pergamon Press.
Goldberg, A. & E., Frongillo. (2001). Cultural perspectives for understanding food security
among the Mossi: A background paper on food security in Burkina Faso. Washington,
D.C.: Food and Nutrition Technical Assistance Project.
Grouzis, M., & Albergel, J. (1989). Du risque climatique a
`
la conduite e
´
cologique. In M. Eldin &
P. Milleville (Eds.), Le risque en agriculture (pp. 243–253). Paris : Editions de l’ORSTOM,
collection a
`
travers champs.
Guilmoto, C. (1998). Institutions and migrations. Short-term versus long-term moves in rural
West Africa. Population Studies, 52(1), 85–103.
Hampshire, K. & Randall S. (1999). Seasonal labour migration strategies in the Sahel: Coping
with poverty or optimising security? International Journal of Population Geography, 5(5),
367–385.
458
POPULATION AND ENVIRONMENT

Hill, A. (1990). Demographic responses to food shortages in the Sahel. In G. McNicoll & M.
Cain (Eds.), Rural development and population: Institutions and policy (pp. 169–192). New
York/Oxford: Population Council/Oxford University Press.
Homewood, K. (1999). Land use, household viability and migration in the Sahel. In CTA (Ed.),
Summary Reports of European commission supported STD-3 projects (1992–1995) (pp. 10–13).
Wageningen: Technical Centre for Agricultural and Rural Cooperation ACP-EU.
Hox, J. (2002). Multilevel analysis: Techniques and applications. Mahwah, NJ: Lawrence Erl-
baum Associates.
Hugo, G. (1985). Investigating community-level effects on population movement. In J. Cas-
terline (Ed.), The collection and analysis of community data (pp. 157–175). Voorburg:
International Statistical Institute.
Hugo, G. (1998). Migration as a survival strategy: the family dimension of migration. In U.N.P.
Division (Ed.), Population distribution and migration (pp. 139–149). New York: United
Nations.
IMF. (2000). Burkina Faso: Recent economic developments. Washington, D. C .: International
Monetary Fund, Staff Country Reports from International Monetary Fund 00/103.
Ingram, K., Roncoli, M., & Kirshen, P. (2002). Opportunities and constraints for farmers of West
Africa to use seasonal precipitation forecasts with Burkina Faso as a case study. Agricul-
tural Systems, 74, 331–349.
INSD. (2000). Analyse des Re
´
sultats du Recensement Ge
´
ne
´
ral de la Population et de
l’Habitation de 1996. Ouagadougou: Institut National de la Statistique et de la
De
´
mographie.
INSTRAW. (1994). The migration of women: Methodological issues in the measurement and
analysis of internal and international migration. Santo Domingo: INSTRAW.
Lallemand, S. (1975). La Se
´
cheresse Dans un Village Mossi de Haute-Volta. In J. Copans (Ed.),
Se
´
cheresses et Famines au Sahel: Paysans et Nomades (pp. 44–61). Paris: Franc¸ois Mas-
pe
´
ro.
Lindqvist, S., & Tengberg, A. (1993). New evidence of desertification from case studies in
Northern Burkina Faso. Geografiska Annaler. Series A, Physical Geography, 75(3), 127–
135.
Lonergan, S. (1998). The role of environmental degradation in population displacement. Global
Environmental Change and Human Security. Report 1. Victoria: University of Victoria.
Lucas, R. (1997). Internal migration in developing countries. In M. R. Rosenzweig, & O. Stark
(Eds.), Handbook of population and family economics (pp. 721–798). Amsterdam: Elsevier
Science Publishers.
Lututala, M. (1995). Les Migrations Africaines Dans le Contexte Socio-Economique Actuel. Une
Revue Critique des Mode
`
les Explicatifs. In H. Ge
´
rard, & V. Piche
´
(Eds.), La Sociologie des
Populations (pp. 391–416). Montre
´
al: Presses de l’Universite
´
de Montre
´
al/AUPELF-UREF.
Marchal, J.–Y. (1975). Ge
´
ographie des Aires d’E
´
migration en Pays Mossi. In ORSTOM (Ed.),
Enque
ˆ
tes sur les Mouvements de Population a
`
Partir du Pays Mossi (Haute-Volta): les
Migrations de Travail Mossi (pp. 29–71). Paris: Orstom.
Mathieu, P. (1994). Mouvements de Population et Transformations Agricoles: le Cas du Burkina
Faso. In P.-J. Laurent, P. Mathieu, & M. Totte (Eds.), Migrations et Acce
`
sa
`
la Terre au
Burkina Faso (pp. 17–40). Louvain-la-Neuve/Paris: Academia/L’Harmattan.
Nelson, J. (1983). Population Redistribution Policies and Migrants’ Choices. In P. Morrisson
(Ed.), Population movements: Their forms and functions in urbanization and development
(pp. 281–312). Lie
`
ge: Ordina.
New, M., Hulme, M., & Jones, P. (2000). Representing 20th century space-time climate vari-
ability. II: Development of 1901–1996 monthly terrestrial climate fields. Journal of Cli-
mate, 13, 2217–2238.
Nicholson, S. (2001). Climatic and environmental change in Africa during the last two cen-
turies. Climate Research, 17, 123–144.
Niemeijer, D. & Mazzucato, V. (2002). Soil degradation in the West African Sahel. How serious
is it? Environment, 44(2), 20–31.
459
S. HENRY ET AL.

Oberai, A. & Bilsborrow, R. (1984). Theoretical perspectives on migration. In Bilsborrow, R.
Oberai, A., & Standing G., (Eds.), Migration surveys in low income countries: Guidelines
for survey and questionnaire design (pp. 14–30). London & Sidney: Croom Helm.
Oldeman, R., Hakkeling, R., & Sombroeck, W. (1990). World map of the status of human
induced soil degradation: An EXPLANATORY Note. Wageningen/Nairobi: International
Soil Reference and Information Centre/UNEP.
Oue
´
draogo, D.O., (1986). Ame
´
nagements hydro-agricoles, operations ‘‘terres neuves’’ et
de
´
placements de population au Burkina de 1900 a
`
nos jours. The
`
se de doctorat d’E
´
tat,
ge
´
ographie, Universite
´
de Bordeaux III.
Poirier, J., Piche
´
, V., Le Jeune, G., & Dabire, B. (2001). Projet d’E
´
tude des Strate
´
gies de
Reproduction des Populations Sahe
´
liennes a
`
Partir de l’Enque
ˆ
te ’Dynamique Migratoire,
Insertion Urbaine et Environnement au Burkina Faso’. Cahiers Que
´
be
´
cois de de
´
mographie,
30(2), 289–309.
PRB. (2001). Fiches de Donne
´
es sur la Population Mondiale: Donne
´
es et Estimations De
´
mo-
graphiques des Pays et Re
´
gions du Monde. Washington, D. C.: Population Reference
Bureau.
Prince, S., Brown De Colstoun, E., & Kravitz, L. (1998). Evidence from rain-use efficiencies
does not indicate extensive Sahelian desertification. Global Change Biology, 4(4),
359–374.
Quesnel, A., (Ed.) (1999). Population dynamics in rural areas freed from onchocercosis in West
Africa: Synthesis of national monographs. Paris: CICRED.
Reardon, T., Matlon, P., & Delgado, C. (1988). Coping with household level food security in
drought effected areas of Burkina Faso. World Development, 16(9), 1065–1074.
Re
´
publique du Burkina Faso & CERPOD. (1997). Enque
ˆ
te sur les Migrations et l’Urbanisation
au Burkina Faso. Rapport Descriptif. Bamako: CERPOD.
Roncoli, C., Ingram, K., & Kirshen, P. (2001). The costs and risks of coping with drought:
Livelihood impacts and farmers’ responses in Burkina Faso. Climate Research, 19(2), 119–
132.
Schoumaker, B., Dabire, B., Piche
´
, V., & Kouaouci, A. (2002). Enque
ˆ
te sur les Migrations,
l’Insertion Urbaine et l’Environnement au Burkina Faso. Unpublished document. Atelier de
Ouagadougou. Ouagadougou: Unite
´
d’Enseignement et de Recherche en De
´
mographie.
Schoumaker, B., Dabire B. et Bilampoa G.T. (2004). La collecte de donne
´
es communautaires
pour l’analyse des de
´
terminants contextuels des comportements de
´
mographiques. L’ex-
pe
´
rience d’une enque
ˆ
te au Burkina Faso. Unpublished manuscript. Universite de Oua-
gadougou.
Sicot, M. (1989). Contraintes et risques hydriques encourus par l’activite
´
agro-pastorale au
Sahel. In M. Eldin, & P. Milleville (Eds.), Le risque en agriculture (pp.1131–141) Paris :
Editions de l’ORSTOM, collection a
`
travers champs.
Tallet, B. (1985). Espaces Ethniques et Migrations: Comment Ge
´
rer le Mouvement? Politique
Africaine, 20, 65–77.
Todaro, M. (1997). Urbanization, unemployment and migration in Africa: Theory and policy.
Policy Research Division Working Paper 104. New York: Population Council.
UNDP. (2001). Human development report 2001, UNDP. New York : Oxford University Press.
460
POPULATION AND ENVIRONMENT