Migration and Urbanisation in Francophone West Africa: An Overview of the Recent Empirical Evidence

How does migration contribute to the urbanisation process in Africa? Migrants have traditionally been viewed as responsible for excessive urban growth, for the uncontrolled expansion of urban areas and for urban surplus labour. In light of recent research, this article proposes to re-examine these views in Francophone west Africa. The first part of the article presents the context of urbanisation in this region, the theoretical framework and the recent surveys on which most of the analyses are based. Next, adopting a demographic perspective, the paper shows that migration in fact contributes moderately to urban growth and that new trends in urban to rural migration flows are emerging. The focus then shifts to examine how migrants influence the shape of cities. The final section of this article re-examines the role of migrants in the urban economy.


Introduction
In a recent UN report, almost all developing countries declared their dissatisfaction with the spatial distribution of their populations and approximately 80 per cent indicated that they had initiated policies to slow or reverse the trends of rural-to-urban migration (United Nations, 1998). This near-unanimous attitude towards population movements is the result of a common view according to which migration in the developing world is seen as a major factor contributing to urban surplus labour and to the uncontrolled expansion of urban areas. One imagines the downtown streets of cities overflowing with homeless and unemployed new arrivals from the surrounding countryside. Is this sad portrait of the rural-to-urban migrant really accurate?
In light of recent research, this article pro-poses to re-examine the contribution of migration to urbanisation in the developing world. It is concentrated on Francophone west Africa where several surveys have been conducted on the topic since the mid 1980s. These surveys are of particular interest both because they fill a gap-migration is much less studied in Africa than fertility or mortality (Bilsborrow, 1998)-and because they allow comparisons between countries. Due to language barriers, these data and results remain largely unknown in the Anglophone literature. We synthesise here their principal findings. Even though political and cultural aspects are also relevant, urbanisation is here conceived as a process of accumulation: of people, buildings and capital. Thus, the contribution of migration to urbanisation will be examined from different points of view: demographic, geographical and economic. We first present the context of urbanisation in Francophone west Africa and the recent surveys from which most of our analyses are inferred. Next, we adopt a demographic point of view to study how migration contributes to urban growth and to describe new trends in migration flows between urban and rural areas. Finally, our focus moves from migration to the migrants themselves: how do they influence the shape of cities and how do they fit into the urban economy?
The overview of the recent empirical evidence in Francophone west Africa indicates that migrants into urban areas are quite well adapted in economic and residential terms and that their contribution to urban demographic growth is less and less important. In contradiction with conventional wisdom, migration trends in this region are responsible for a clear slowdown of urban growth. Most of the population projections give a picture of a highly homogeneous urban world, as if each country should go through a transition (although at a different pace) from a predominantly rural population to a predominantly urban population. This vision of an urban future contrasts with the evidence of deepening economic inequalities both within and between countries. The evidence presented in this paper for west Africa is an attempt to reconcile the economic and demographic visions.

A Paradoxical Urbanisation?
Although historically west Africa nurtured great empires and developed large urban centres for capital cities, commercial and religious centres (Chandler, 1994), urbanisation in this region of the world is mostly a recent phenomenon. 1 As in other sub-Saharan regions, this urbanisation was linked with colonisation that developed new centres to control and administer the colonised population and to exploit and ex-port natural resources. Labour migration was an important factor of urban development in colonial times (Coquery-Vidrovitch, 1992) when workers were needed for public infrastructure (runways, railways, ports), administration (soldiers, clerks) or various private services (maids, etc.). The migration flows towards urban centres continued after independence. However, this movement could appear counter-intuitive. Although urbanisation is commonly associated with a gradual reallocation of labour from agriculture to industry, as experienced in the First World during the 19th century (Bairoch, 1985), urban in-migration in sub-Saharan Africa occurred (and continues to occur) without industrial development (Coquery-Vidrovitch, 1992;Arnaud, 1998). In addition, African countries witnessed a massive migration of the rural population into urban areas despite rising levels of urban unemployment and underemployment (Todaro, 1997).
These apparent contradictions of migration and urbanisation were first attributed to irrational behaviours. A better explanation was found in the notion of economic dualism between the formal and informal urban sectors, first introduced by Lewis (1954) and extended by Todaro (1976). In contrast with the formal sector (geared towards capital-intensive and large-scale modern production), the informal sector is a traditional, unregistered, subsistence sector geared towards labour-intensive and small production. The existence of the informal sector explains why rural-to-urban migrations persisted-leading to rapid urban growth-despite limited absorbing capacity in the urban formal employment sector. Although it is sometimes discouraged by public authorities because of its irregularity and suspected tax evasion, most researchers now recognise that the informal sector has economic advantages: it absorbs surplus labour and provides a safetynet in the face of high unemployment and poverty (Becker et al., 1994;Snrech, 1994). In fact, this sector employs the majority of urban workers (65-80 per cent in Africa, according to the country) and serves the purpose of receiving and integrating migrants and providing them with the minimum means of subsistence (Arnaud, 1998). However, this pattern has evolved in the past two decades because of the economic crisis.
Since the late 1970s, west Africa, still mostly dependent on the export of raw materials (coffee, cocoa, bananas, etc.), has suffered from decreasing prices in international markets. West African states first delayed the crisis by maintaining public expenditures. As the debt became intolerable, structural adjustment programmes (SAPs) imposed by international organisations and donors focused on reducing public expenditures by privatisation and staff cuts. The formal sector, of which parastatal enterprises were an important part, was reduced and the workers turned to self-employment or informal employment. The standard of living deteriorated and urban unemployment reached levels never before attained, in particular among educated people in search of jobs in the formal sector (Lachaud, 1994;Charmes, 1996;Bocquier and LeGrand, 1998). As a result, the crisis has favoured the informalisation of the urban economy. Due to the reduction of consumption (linked to the deterioration of the standard of living) aggravated by the irruption of those left out by the formal economy (inducing higher competition in the informal sector), the informal sector can no longer adequately integrate migrants (Arnaud 1998). This new context justifies a new investigation of the relationships between migration and urbanisation.

Migration, Urbanisation and Development
These apparent contradictions of urbanisation in Africa, largely linked with economic problems, raise the issue of relationships between migration, urbanisation and development. In simple terms, are rural-to-urban migrations (and the resulting urbanisation) favourable for economic development? This is one of the main topics of concern for policy-makers of Southern countries (United Nations 1998) and is also a very contro-versial subject among scholars, especially in the sub-Saharan African context.
An anti-migration point of view supports the idea that rural-to-urban migration is excessive and that it should be curtailed because it leads to a "a less than optimal allocation of labour between the rural and the urban sectors" (Gugler, 1982, p. 181). On one hand, it causes an exacerbation of unemployment and underemployment in African cities (Bairoch, 1985;Todaro, 1997). On the other hand, it entails a loss of potential agricultural output since the rural population is deprived of its more innovative and stronger members. Furthermore, rural-to-urban migration is condemned as the primary contributor to the uncontrolled expansion of urban areas (Bairoch, 1985). Finally, rural-to-urban migration increases the cost of providing for the country's population in two ways. First, new infrastructure is required by the migrants and amenities are more expensive in cities or towns than in rural areas or simply not needed there (housing, transport, garbage and sewage disposal, etc.) (Gugler, 1982). Secondly, urban job creation is generally more costly than rural job creation because most jobs in the industrial sector require substantial complementary resource inputs (Todaro, 1997). For those reasons among others, rural-to-urban migration is seen "both as a symptom of and a contributor to African underdevelopment" (Todaro, 1997, p. 25). This view is largely shared by policy-makers as suggested in the above-mentioned UN report (United Nations, 1998).
From the opposite point of view, pro-migration authors argue that it is completely inappropriate to regard migration as an undesirable force to be suppressed. From an urban viewpoint, authors acknowledge that substantial unemployment and severe underemployment characterise most large African cities today, but they underline that, on average, migrants must be better off in urban places, or rural out-migration flows would slow (Becker et al., 1994). From a rural viewpoint, they argue that there is no workforce deficit since emigration has not prevented the rural population from increasing far more quickly than it has ever increased in the Northern countries (except for the US). From 1950 to 1990, the rural population of west Africa (excluding Nigeria) increased from 27 to 60 million, according to the GEOPOLIS database. 2 On the contrary, rural out-migration is seen as a means of relieving human pressure on natural resources (Arnaud, 1998). In addition, they contend that urbanisation stimulates agricultural activities since it creates a market for specialised food production with a high added value (market gardening or poultry farming).
Beyond rural areas, migration and urbanisation are conceived as key parts of the development process. At the global level, although the causal relation is far from clear, Davis and Henderson (2003) show evidence of high correlation between urbanisation and GDP per capita, after controlling for purchasing power parity, this correlation being true for Africa as well as for any other part of the world. There is also correlation between urbanisation and the UNDP human development index, a composite measure of health, education and standard of living (Njoh, 2003). Becker et al. (1994) argue that, for several reasons, increased urbanisation causes increases in per capita income and other measures of economic welfare. First, urbanisation, when driven by rural-to-urban migration, tends to move workers from agriculture to higher-productivity occupations (such as urban services, commerce and manufacturing). Secondly, urbanisation offers the cost-reducing advantages of agglomeration economies and economies of scale and proximity as well as numerous economic and social externalities (skilled workers, cheap transport, amenities, etc.). Finally, urbanisation also contributes to development in more subtle ways: for example, it may lead to lower levels of fertility and mortality, as individuals benefit from improved access to health care. Theory apart, Becker et al. (1994) underline, from a macroeconomic point of view, that while less than a third of Africa's population lives in cities or towns, these centres generate over half of the continent's gross domestic product (GDP).
The purpose of this paper is not to close the debate about the impact of population movements on development, the perception of which is highly dependent on the conception of what development should involve and what level of analysis is preferred. However, the data described below provide some insights into the complex relationships between migration, urbanisation and development.

Data and Sources
Until the mid 1980s, most data on west African migration originated from sources whose primary purpose was to collect information about something else. For instance, the first study of migration for west Africa as a whole was based on the analysis of mid 1970s census data (Zachariah and Condé, 1980). Few national surveys have been carried out on migration, with the exceptions of Burkina Faso  and Senegal (1979). Furthermore, the concept of migration differs considerably from one source to another, making comparisons difficult or impossible.
The Network of Surveys on Migration and Urbanisation in West Africa (NESMUWA) 3 was the first international collaborative project to study migration in Africa. Comparisons are now possible since surveys based on the same methodology were carried out simultaneously (1993) in eight west African countries (Burkina Faso, Côte d'Ivoire, Guinea, Mali, Mauritania, Niger, Senegal and Nigeria; see Figure 1). The methodology for data collection and analysis was inspired by that of the 1974-75 survey in Burkina Faso (Cordell et al., 1996). A retrospective questionnaire was used to record migration history from birth to the time of interview and another questionnaire provided an indirect record of out-migrants who had been part of the household in the five years preceding the survey. In this way, and in contrast to most other migration studies, the NESMUWA project was able to produce data not only on domestic migrations but also on international migrations. The surveys were carried out on nationally representative samples, measuring all types of flows be- tween urban and rural areas (Bocquier and Traoré 2000). In this article, data distinguish between rural areas, secondary and principal towns (defined in the Appendix), and the capital city of each country.
NESMUWA data paint a picture of urban dynamics at the national level. As a complement at the local level, several surveys of urban integration in capital cities were conducted in Francophone west African countries (Dakar, Senegal, 1989 and2001;Bamako, Mali, 1992;Yaoundé, Cameroon, 1996;Ouagadougou, Burkina Faso, 2000;Lomé, Togo, 2001). Each of these surveys followed the same methodology, using samples representative of the cities' populations. In these local surveys, the concept of urban integration is broader than the idea of migrants' adjustment, based on "the assumption that the native born are 'well adjusted' to their place of residence since they have lived there all of their lives" (Goldscheider, 1983). In fact, urban integration is relevant to both migrants and non-migrants (city-dwellers from birth), both of whom may experience difficulties accessing urban resources such as housing and employment (Antoine et al., 1998). To compare the integration of migrants and non-migrants, a life history from birth was collected from each individual. Three aspects of integration were covered by the questionnaire: activity (education and employment), family formation and dissolution (births and deaths of children, marriages, divorces, widow(er)hood) and residential history (mobility, access to housing, etc.). Since methods are similar, comparisons of the process of integration between capital cities are possible. By its use of a retrospective questionnaire, this type of analysis offers a social and demographic diagnosis of the city over the 30 or 40 years preceding the survey. The retrospective data compensate for the lack of reliable and continuous cross-sectional data. As far as the integration of migrants is concerned, one study explicitly compared the cases of Bamako and Dakar (Antoine, Ouédraogo et al. 1998) but few results are published on that topic regarding other capital cities.

The Contribution of Migration to Urban Growth
Even if it remains one of the least urbanised regions of the world (30 per cent urban, according to GEOPOLIS), Francophone west Africa experienced one of the highest urban growth rates during the second half of the 20th century: the urban population grew 7.8 per cent per year between 1950 and 1990. Having reached a peak in the 1960s (11.5 per cent per year), urban growth in Francophone west Africa has since slowed to a low of 4.7 per cent , which remains one of the highest in the world. This section examines to what extent migration, particularly rural emigration, is responsible for this extraordinary urban growth. We will first consider the different components of urban growth. Next, we will present new trends in migratory flows and take a closer look at urban-to-rural migration.

Migration and Urban Growth
Urban growth can be caused by any of three mechanisms: natural growth, reclassification and migration. Natural growth in west Africa remains one of the highest in the world: 2.7 per cent as compared with 1.9 per cent for the developing world as a whole (PRB, 2001). This partially explains why urban growth remains so high in this region compared with the rest of the world. A UN study showed that the contribution of natural growth to urban growth is much more substantial in Africa than in other parts of the world: in this continent natural growth represents 75 per cent of urban growth while this share was only 50 per cent in Asia (without China) in the 1980s (Chen et al., 1998).
Another particularity of urban growth in Africa is the importance of reclassification. Reclassified settlements are those which cross a population threshold (10 000 inhabi-tants in the GEOPOLIS database), beyond which they are considered urban instead of rural. These newly urban settlements contribute to increasing the urban population although their migratory growth is not necessarily positive. In Africa between 1950 and 1980, the share of reclassification in urban population growth was 26.4 per cent. This means that, in 1980, more than one new African urban dweller in four lived in an agglomeration that was classified as rural in the preceding 30 years. This rate was 13.4 per cent globally, excluding China and Korea for which reliable data are not available (Moriconi-Ebrard, 1993). Thus, urban growth in Africa is due, much more than in the other continents, to the proliferation of new small urban centres which were previously considered rural (Bertrand and Dubresson, 1997).
Considering those results, the role of migration in urban growth seems to be less important than is commonly assumed. 4 The process of urbanisation, from a demographic point of view, is more complex than a simple rush of migrants from rural to urban areas. Furthermore, the contributions of migration and reclassification to urban growth appear to be decreasing. Indeed, according to the GEOPOLIS database, two-thirds of urban growth in west Africa was due to migration and reclassification in the 1960s and only one-third in the 1990s (Bocquier and Traoré, 2000). This regional result is consistent with comparative studies in other parts of the continent (Makannah 1990;Chen et al., 1998). How can we explain this evolution?
A first explanation is mathematical. Urban growth due to migration is necessarily reduced as the urban population gradually increases compared with the rural population, simply due to the reduction of the rural population who could potentially leave their villages. When the majority of people live in villages, net out-migration (i.e. a negative migratory growth) from a rural environment means, for the same volume of migrants, a higher net in-migration rate towards urban areas. As the respective weights of the rural and urban populations balance out, the num-ber of migrants who feed the urban growth decreases, even though the probability of out-migration from the rural areas may remain the same. This is what happens for example in developed countries: when more than 60 per cent of the population lives in urban areas, urban growth by immigration from rural areas is necessarily low.
But is the mathematical explanation sufficient to explain the decrease in urban growth in the 1980s and 1990s in Africa, where the level of urbanisation remains approximately 30 per cent? Or could low economic growth also be a factor? Given the relationship between urban growth, rural outmigration and per capita GNP (Ledent, 1982;Moriconi-Ebrard, 1993), the context of crisis depicted above may be at least partly responsible for the decrease in rural out-migration and also the increase of urban out-migration. The declining contribution of migration to urban growth could be interpreted as an adjustment to a depressed economy. This hypothesis will be discussed in the following sections.
Whatever the cause of the decreasing contribution of migration to the urbanisation process, the influence of migrants on urban demographic dynamics is double: they contribute directly to urban growth through their arrival, but they also have an indirect role through their reproductive behaviours. It is generally assumed, due to their origin in high-fertility rural areas and their youth (most are of reproductive age), that migrants from rural areas experience higher fertility rates than the rest of the urban population. However, there is little specific evidence on this subject in Francophone west Africa. On a broader scale, one of the pioneer studies on this subject (Brockerhoff, 1998), based on the analysis of 14 Demographic and Health Surveys (DHS) conducted in west Africa, east Africa and southern Africa, showed that high levels of female in-migration have lowered fertility rates in sub-Saharan African cities significantly in recent years [i.e. 1980s and early 1990s], by over one birth per woman, and have reduced na-tional rates by an average of one-third birth per woman (Brockerhoff, 1998, p. 386). This is mainly due to the "disruptive effects of migration" and to the "migrants' partial adaptation to urban fertility regimes". Along the same lines, a recent study in Ghana confirmed that migrants are "more receptive to fertility control" (Stiff and White, 2002). The same conclusions on the disruption and adaptation effects were reached by a more recent and broader study using all DHS with information on migration in developing countries (National Research Council, 2003). Therefore, it seems that the indirect contribution of migration to urban growth is not as straightforward as is commonly assumed.

Capital Cities and Secondary Towns in Migration Routes
Although migration is no longer the principal reason for urban growth in west Africa, it remains true that migratory flows reveal the differential dynamics of rural and urban areas and also distinct dynamics within these two categories of space. African cities have always been notable for their large share in urban population and their spectacular growth, especially during the 1970s. For example, the growth rate of Abidjan, Côte d'Ivoire, was ϩ 10.3 per cent per year between 1965 and 1975. However, in west Africa as well as in the rest of the continent, most cities now have a slower pace of growth (Dubresson, 2003). In fact, the NES-MUWA results show that capital cities are still attractive in the early 1990s, but the rates are not so spectacular (Table 1). The annual migratory growth for the population aged 15 and over varies from 0.5 per cent in Niamey, Niger, and Dakar, Senegal, to 3.1 per cent in Ouagadougou, Burkina Faso, where international immigration is responsible for almost 40 per cent of the growth. International immigration is also responsible for threequarters of the annual growth of 1.7 per cent of Abidjan, Côte d'Ivoire. The recent turmoil in this country has probably changed the  migratory pattern in the region. Abidjan is likely to be much less attractive for international migrants and may have lost a part of its share of foreign inhabitants. Furthermore, the return of those migrants originating from Mali or Burkina Faso may have enhanced the growth of Ouagadougou and Bamako, which were already preferential destinations for return migrants (Bocquier and Traoré, 2000). Finally, Dakar, the second urban pole of attraction in the Francophone region, might have been boosted by Abidjan's crisis. Un-fortunately, no data are available to verify these assumptions. One of NESMUWA's striking results is that secondary towns in all countries have negative migration growth rates (Table 1). In Senegal and Niger, secondary towns continue to receive rural migrants but this does not compensate for departures towards bigger cities (Appendix , Tables A4 and A5). In Guinea, Burkina Faso and Côte d'Ivoire, migrants leave secondary towns to move both to bigger cities and to rural areas (Appendix ,  Tables A1-A3). Fur-thermore, in Guinea and Côte d'Ivoire, principal towns also have a migratory deficit in domestic migrations (Table 1). Actually, in Côte d'Ivoire, principal towns lose migrants to Abidjan and to rural areas in equal numbers. Between 1988 and 1992, principal towns in the country lost almost 20 000 people per year to villages. These departures were not entirely counteracted by international migration, although it is still quite significant ( ϩ 10 000 per year towards principal towns). The same situation was found in Guinea (Appendix , Tables A1  and A3).
The examples of Niger and Mauritania suggest that migration routes may be more complex than the step-migration model, according to which rural migrants leave their villages for the nearest small town before moving to a larger agglomeration and then, eventually, to the capital city (Adepoju, 1983). The role of small or medium-sized towns in demographic dynamics has become a controversial subject in the Francophone migration and urbanisation literature: since the mid 1980s, some researchers have assigned them a more complex role. For instance, a comparison of case studies from Côte d'Ivoire and Togo (Dupont and Dureau, 1988) demonstrated that immigrants to small and medium-sized towns originated from capital cities as well as from rural areas. Furthermore, it showed that migrants from rural villages do not originate, in most cases, from the surrounding areas. Therefore, small and medium-sized towns serve more as redistribution places than as steps on the way to capital cities.
Finally, it appears that the urban migratory dynamic of Francophone west Africa is not universally thriving: capital cities now attract fewer migrants than in the 1970s; negative net migration has been observed in some principal and secondary towns; and some rural areas now gain from rural-urban flows.

The Growing Importance of Urban-to-rural Migration
Among NESMUWA countries, Côte d'Ivoire alone has a positive rural migratory growth rate: almost ϩ 2 per cent per year, of which 1 per cent derives from internal migrations (Table 1). In this country, when only internal flows are considered, urban areas lose more people than they gain from migration. This is true for secondary and principal towns, and surprisingly also for Abidjan that lost more than 12 000 people a year to rural areas in the period from 1988 to 1992 (Appendix ,  Table A3). This result caused much surprise when published because it ran counter to the conventional wisdom of a rural exodus in Africa. The first reaction was doubt concerning the reliability of Côte d'Ivoire's survey. However, any problems of definition or sampling were insufficient to disqualify the results (Beauchemin, 2001(Beauchemin, , 2002b. In fact, the NESMUWA survey simply confirmed Côte d'Ivoire's 1988 census, which showed that the rural-urban migratory flows benefited the villages. And this finding was again confirmed by the 1998 census (Zanou, 2001). Since then, due to civil war, domestic movements have probably evolved, but it is impossible to draw a clear picture. On the one hand, the departure out of the southern region of non-native planters (originating from the northern part of the country or from abroad) may have encouraged some urban dwellers to return to their villages to serve as manpower or to take advantage of the abandoned plantations. On the other hand, insecurity may have encouraged mobility according to the conflict situation, sometimes from villages to towns and cities, sometimes in the reverse direction. Due to lack of data, all the comments below on urban-rural migration are relative to the prewar period.
Urban out-migration in Côte d'Ivoire is not only a circular migration in which migrants engage in constant comings and goings. For one thing, primary migrants born in urban areas form a significant proportion of urban emigrants (25 per cent in 1993): for them, it is clear that urban out-migration is not a return migration. Furthermore, qualitative research on migrants' integration in rural areas shows that some migrants choose to stay in villages where they feel they have better employment opportunities and a lower cost of living (Beauchemin, 2002a). However, the economic crisis may increase circular migration between towns and villages, especially for young people seeking employment and for whom integration difficulties exist in towns and also in villages (Faussey-Domalain and Vimard, 1991). Finally, even if there is an intensification of circular migration in Côte d'Ivoire, rural areas are experiencing a net migratory gain.
In all other Francophone countries, the NESMUWA results show negative migration rates in rural areas (Table 1). However, internal migration rates are quite low for rural areas (from Ϫ 0.2 per cent to 0 per cent per year). In fact, the negative balance of the rural areas is essentially due to international emigration. In every country where rates were calculated for both villages and small towns, the international rates of net rural migration (from Ϫ 1.4 per cent to Ϫ 0.4 per cent per year) are much higher than the internal rates. Thus, migrations towards urban areas within the countries are not responsible for a large demographic deficit in rural areas.
Various case studies have shown that return migrations are very important in some west African countries. For instance, in the Senegalese part of the Senegal River valley, the villages' migration balance is positive for people older than 40 (Guilmoto, 1997). In Cameroon, migration flows have not been precisely measured, but several authors have remarked that urban-to-rural migrations have increased (Franqueville, 1987;Gubry et al., 1996). Urban out-migration is not a new phenomenon but it seems to be increasing in importance, even outside Côte d'Ivoire. In addition to the traditional return flows of migrants, a new kind of urban-to-rural migration, linked to the economic crisis, has appeared in west African countries since the early 1980s. Evidence of urban-rural migration as a consequence of SAPs has not only been shown in Francophone west Africa, but also in Zambia, Uganda, Tanzania (Potts, 1995) and Zimbabwe (Potts, 2000). Finally, in Nigeria, a study showed that the desire of urbanites to return to their villages was greater in 1987 than in 1961; and this difference has been attributed to the deteriorating economic fortunes of many urban dwellers (Gugler, 1991).
In most African countries, migrants still maintain close relations with their birth village even from a distance: they return to visit; they invest in housing, social activities, education and health amenities; they send money and sometimes receive goods or host visiting relatives (Vidal, 1991). Traditionally, the birth village is the preferred place for eventual retirement: Franqueville in Cameroon, Caldwell in Ghana and Gibbal in Côte d'Ivoire reported that most migrants living in towns wanted to return to their village upon retirement (Caldwell, 1969;Gibbal, 1973;Franqueville, 1987). Another traditional motive for return is inheritance: returning to take care of the family, the house or the farm is often an obligation for sons. This type of return migration has been documented since the beginning of the 1960s. Gender differences exist: Margaret Peil's research shows that women are more likely than men to prefer remaining in town (Peil, 1995). Peil reports that women do not want to participate in village chores and prefer to remain in town where they can fulfil their economic and social roles by trading or looking after their grandchildren and be supported by their children. However, with the economic recession, circumstances of urban out-migration have become more varied.
Job market degradation and deterioration in the standard of living have created new relationships between migration, employment and education. In the past, people moved to town to attend school or to find a job. Today, the opposite is often the case. A large number of people who have lost their formal-sector jobs return to villages. Most first try to find a new job in cities or towns, sometimes using their 'golden handshake' to launch a new career, a strategy observed in Francophone (Côte d'Ivoire, Togo, Niger) and also Anglophone areas (Ghana, Tanzania) (Dureau, 1989;Giraut, 1994;Satterth-waite and Tacoli, 2003). If they fail (as many do), the village is their last resort. In addition, some urban residents with jobs, comparing their incomes with the urban cost of living, choose to return to rural areas where incomes are lower but where food and housing are almost free. Another explanation of urban out-migration is the adjustment of domestic arrangements of some urban households trying to expand their incomes or to reduce their expenses. For example, a husband and his wife (or wives) may dissociate their residences with the woman (or women) returning to the village while the man remains in town. Incomes are thus diversified and the family can better face economic hazards. In another example observed in Côte d'Ivoire (Guillaume and Vimard, 1997), Cameroon (Eloundou-Enyege, 1992), Benin (Pilon and Vignikin, 1996) and Niger (Gado and Guitart, 1996), young foster children and even the children of the household head are sent back to villages to attend school. Similarly, young people who failed school and graduates unable to find jobs may return to villages by their own choice or be sent home by their hosts, as shown in Cameroon (Gubry et al., 1996) and Côte d'Ivoire (Chauveau, 1997). Those situations show that education is no longer a determining factor of urbanisation: children can be sent to villages to attend school and young educated people can leave cities to find better employment opportunities and living conditions in rural areas. This shows a reversal of the traditional direction of migration of the west African population.
These new types of urban-to-rural migration explain the evolution of the return migrant profile. In the past, the return migrant was most often an old retired person (Zachariah, 1978). Today, urban emigrants to rural areas are mostly young people. In Cameroon in 1992, three-quarters were between 15 and 45 years old (Gubry et al., 1996). In Côte d'Ivoire in 1993, three-quarters were less than 30 years old and, in the late 1990s, the chances of a 20-24 year old man leaving a city for a village were twice those of a man of 50 or older (Beauchemin, 2001). The traditional return migration at the age of retirement continues, but youth is now the major component of urban out-migration.
In west Africa, as well as in other African regions, decreasing urban growth through the upsurge of urban out-migration and the reduction of the rural exodus is largely interpreted as a consequence of the economic crisis, which has been having unequal impacts in rural and urban areas (Potts, 1995;Gubry et al., 1996;Chen et al., 1998). In a large number of African countries, the fall in real urban incomes has reduced, and sometimes even reversed, the rural-urban income gap (Jamal and Weeks, 1988). Structural adjustment plans (SAPs), by reducing the income differential between rural and urban areas and increasing urban unemployment in the parastatal sector, may be at least partly responsible for this (Becker et al., 1994;Potts, 1995Potts, , 2000. Some authors even hypothesise that SAPs could be disguised migration policies due to their differential effects on urban and rural areas (Antoine, 1991;Guillaumont and Lefort, 1993). Regardless of the possible indirect effects of SAPs, we must underline that those urban-torural migrations are not due to any direct migration policy even if some countries have attempted to reduce the rural exodus. In Côte d'Ivoire, for instance, the failure of the return migration policy of the 1970s and 1980s is unanimously recognised-people who were sent from cities to rural areas came back to town (Affou-Yapi, 1985). As a consequence, there is in this country a paradox in the migratory policy: while many beneficiaries of the return-to-land programmes (who received subsidies to settle in rural areas) came back to town, at the same time young people spontaneously left urban areas for villages without any public support.

Migrants and Cities: In and Out
The contribution of migrants to urbanisation is not exclusively a demographic matter. Migrants are also active participants in urbanisation, when this is conceived as a process of accumulation in terms of economy or in terms of geographical space. This section deals with how migrants influence the shape of urban areas and how they contribute to the urban economy.

Migrants Building the Cities
Except in the case of reverse flows between villages and towns (as observed in Côte d'Ivoire), migration contributes, at least demographically, to the growth of urban areas in west Africa. Here, we examine the role of migration in urban growth in terms of the occupation of space. From this point of view, it appears that migrants' contribution to the spatial extension of towns is not as simple as commonly imagined.
Settling in the city. It is commonly supposed that migrants move from rural areas to the outskirts of cities, where they build slums. In fact, their trajectories are generally more complex. In Dakar (Vernière, 1973), Yaoundé (Franqueville, 1987), Bamako (Antoine et al., 1998), and Abidjan (Antoine et al., 1987), studies have shown that migrants first settle in peri-central areas, where they are housed by friends or family, or where they find rented accommodation. 5 If they stay in town, and especially if they are responsible for a family, migrants may eventually move to the outskirts where they can buy a plot and build their own house. As a result, the population of the outlying districts mainly comes from the city (even if they are not city-born) and not directly from villages or other towns. For instance, in Dakar, Senegal, the population of Pikine, a huge and mostly informal suburb, consists primarily of large families, most of whom (84 per cent) come from central areas of the capital city (Antoine et al., 1995). It appears therefore that migrants do not immediately contribute to the growth of outlying districts. However, points of arrival are changing: as time goes on, peri-central areas are becoming too dense to receive new arrivals; former outskirts are progressively integrated into the city and become new points of arrival for new migrants; and new outskirts are created by urbanites from more central areas (from peri-central districts or former outskirts) who want to buy a plot to build their own house. The direct cause of urban expansion is therefore intraurban mobility rather than direct in-migration.
The primary reason that new migrants are not directly responsible for the spatial extension of urban areas is that they are not, for the most part, autonomous settlers. More often than not, migrants are housed in existing urban households, whether they are related to a household member or not (Locoh 1988). For instance, in Abidjan at the beginning of the 1970s, 62 per cent of migrants were housed by relatives (Gibbal, 1973). This percentage rose with the economic crisis. Studies of migrants' integration into capital cities show that migrants increasingly depend on others when arriving in the city: migrants in a dependent status are more and more numerous and they stay dependent longer. For instance, in Dakar, 33 per cent of migrants who arrived during the 1960s stayed with friends or family, as compared with 60 per cent of migrants in the 1980s. Furthermore, among those who arrived in the 1960s, 40 per cent owned a house within 6 months of their arrival, as opposed to only 15 per cent of those who arrived in the 1980s (Antoine et al., 1995).
Migrants do contribute to a change in the shape of cities, if not directly to urban expansion. Real estate is adapted to their requirements. Rental accommodation is made available for those who are not housed. For instance, in Abidjan, migrants often reside in small terraced accommodation (a room with an entrance) inside a common compound. This type of housing is widespread (70 per cent of the population of Abidjan) and mainly rental (80 per cent). Although it is not exclusive to migrants, its development is interpreted as a response to the migrants' arrival (Antoine et al., 1987). Migrants also contribute to the densification of the pericentral areas, where dwellings become more and more crowded. As a consequence, owners adapt their rental accommodations either vertically (adding new floors, as in the Grand Yoff or Grand Dakar districts in Dakar) (Tall, 1994) or horizontally (filling compounds with new rooms).
International migrants and real estate investments. Despite their absence, international emigrants also play a role in the structural transformation of west African cities. A few recent studies have pointed out the importance of international emigrants' real estate investments and their influence on the transformation of cities. Even when west African migrants live in distant countries, they maintain strong links with their birth countries. Their foreign wages give them high purchasing power; and, for those now living outside west Africa, the devaluation of the CFA franc in 1994 has doubled their investment potential. It is well known that emigrants invest in their villages-they build houses, health centres, schools and mosques; they contribute to social activities, etc. They also invest in towns and cities where it is more profitable.
In Dakar (Tall, 1994), international emigrants have played a role in urbanisation since the 1970s. Peri-central areas (Grand Dakar, Grand Yoff) attracted the investments of the first generation of Senegalese emigrants to France. The rental housing market in peri-central areas is very attractive due to its proximity to working areas. Emigrants invest in housing units and progressively add floors in order to rent more rooms. In doing this, they become the main actors of the densification of those areas. In the outskirts of Dakar, new emigrants also contribute to the extension of squatter settlements. They contribute directly when they buy a plot of land and build a house, and indirectly by adding value to those informal areas. By building cement houses and paying to connect electricity or water networks, they make the area more permanent. As a result, their investments consolidate squatter settlements and make resettlement more difficult.
Emigrants have two motivations for investing in real estate: either purely economic (investing in rental housing) or familial (building accommodation to house the family). In both cases, it is part of the emigrants' retirement plan. But their interest in real or landed estate often results in speculation because their financial capabilities are higher than those of non-migrants. Landowners give them priority over non-migrants. For instance in Touba, capital city of the Murids and the second-largest town in Senegal, some areas can no longer be afforded by non-migrants because international emigrants caused a huge increase in real estate prices in the 1990s (Gueye, 2002). In Sikasso, the fourth-largest town in Mali, prices offered to emigrants are twice as high as to non-migrants, and emigrants pass on this cost by increasing their rents (Bertrand, 1994).
No study has evaluated the exact percentage of the contribution of international emigrants to real estate transactions. Nevertheless, authors agree on the growing influence of emigrants and on the essential role they already play in urbanisation. For instance, Tall (1994) considers that, in Dakar, international migrants form a replacement for the state since its withdrawal from the provision of public housing and social amenities.

Migrants in the Urban Economy
Employment is central to all rural-urban migration theories because the search for a job is seen as the primary, if not the only, motivation for migration (Lututala, 1995). Establishing a relation between limited access to the job market and persistent rural out-migration, one is tempted to conclude that urban unemployment is a consequence of migration. Here, we examine this assumption asking two questions. Do migrants increase unemployment in urban areas? And, are migrant workers confined to precarious activities, particularly in the informal sector?

Migrants and unemployment.
In the literature about the economic integration of migrants, there are two contradictory hypotheses. One emphasises migrants' difficulties in the urban labour market due to their supposed disadvantages in comparison with non-migrants in terms of, among other things, education, social networks and family support in town. The other affirms that migrants fit easily into the labour market because of migratory selectivity: migrants are selected from among the more dynamic and educated rural people. What conclusions can we draw on this topic from recent research in Francophone west Africa?
NESMUWA surveys measured unemployment in the two weeks preceding the interviews (Table 2). These measures of declared urban unemployment do not respect ILO's criteria, but are sufficient to observe any differences between migrants and non-migrants. The most striking result is that, in all Francophone countries and regardless of sex or the type of town, migrants have much lower unemployment rates than non-migrants. More specifically, in most countries, differences between migrants and non-migrants are higher in capital cities than in other towns. For instance, in Abidjan and Ouagadougou, males' employment rates are more than twice as low for migrants (Table 2), while the differential according to migratory status is much less important in other towns in Côte d'Ivoire and Burkina Faso. This suggests that migrants have fewer difficulties than non-migrants finding work in highly competitive contexts (unemployment rates are generally higher in capital cities than in other towns). These results counter the belief that migrants are handicapped in the urban labour market. How can we explain the lower unemployment rate of migrants?
A first hypothesis is based on a methodological issue. Surveys on urban integration are for the most part based on urban samples. Thus, those who have been constrained to leave towns or cities because they could not find a job are excluded from analysis. This may result in an optimistic bias in urban integration studies: migrant unemployment rates may be underestimated when considering only those who have remained in urban centres. This hypothesis is consistent with the surge of urban out-migration (Potts, 2000), but it has recently been challenged in Burkina Faso by using a national survey of both rural and urban areas, designed on the same model as the urban integration surveys in capital cities (Poirier et al., 2001). Integrating in the analysis those who have left the cities, the authors found that migrants still have better access to urban jobs (Zurkaleini and Piché, 2003).
A second hypothesis is that migrants are selected: only those villagers who have information or even job offers from the urban labour market move to town. Non-migrant urban residents, on the other hand, live in town with or without any particular job information. Little evidence confirming this hypothesis exists.
A third explanation is that migrants have a lower unemployment rate because, being forced to find a job quickly when they arrive, they accept to work for lower wages and in more precarious conditions than non-migrants. Non-migrants therefore have a higher probability of being unemployed, having been priced out of the labour market by the new arrivals. It is true that migrants quickly find jobs when they arrive in urban areas: Piché and Gingras (1998) have shown that most migrants have found a job within a year in Bamako as well as in Dakar. However, that study did not compare migrants with non-migrants and it is not known if their jobs are less lucrative. 6 We will discuss below if migrants have more precarious working conditions than non-migrants.
A final explanation takes into account the worsening economic situation: since the 1980s, access to formal employment has been reduced to almost nothing while, at the same time, the level of education has risen. As a result, unemployment among educated youth is now higher: in Dakar and Bamako, for instance, unemployment among educated youth (35 and under) is twice as high (26 per cent) as among the non-educated (13 per cent) (Bocquier, 1996;Bocquier and LeGrand, 1998). An extension of this comparative study on similar retrospective data showed that in Yaoundé, Cameroon, the unemployment rate of young people with an upper secondary education and above was twice that of those with less education and that in Antananarivo, Madagascar, the difference is four-fold (Antoine et al., 2001). Educated youth often refuse to work in petty jobs  Source: Piché and Gingras (1998).
or, when they accept an informal activity, they consider it equivalent to unemployment and thus declare it as such on the surveys. This could explain why non-migrants have higher unemployment rates than migrants, who are typically less educated. 7 Are migrants prisoners of the informal sector?. The sector of activity (informal or formal) can serve as a proxy for job quality since work in the formal sector is much more desirable in terms of wage, social advantages, job security, etc. (Becker et al., 1994). The informal sector is commonly viewed as a sector of integration for migrants arriving in the urban labour market. As a result, migrants are mainly reputed to work in more precarious conditions than non-migrants. But are migrants confined to this sector? We do not have information for all of Francophone west Africa. However, surveys on urban integration in capital cities give some insight on this topic. From the descriptive results, it is clear that migrants in Dakar and Bamako were more often engaged in the informal sector than non-migrants when only their first job in the city was considered (Table 3). This is probably due to the fact that migrants are less educated than non-migrants. However, this employment differential according to migratory status tends to disappear with time. The differential is high for the older men (aged 45-54 at the time of the survey): for instance, among migrants to Bamako, 66 per cent were engaged in the informal sector compared with only 26.4 per cent of non-migrants. However, for the younger generation, the differential according to migratory status is very weak: in Dakar, for instance, migrants and non-mi-grants are equally employed in the informal sector. In the long run, the convergence of migrants' and non-migrants' profiles can be explained by the fact that the urban labour force increased much more quickly than the pace of employment creation in the formal sector. This was essentially due to the absence of industrial take-off and, in the short term, the implementation of SAPs that drastically reduced access to the formal economy through downsizing in the public and parastatal sectors.
Regression analysis of longitudinal data from Bamako and Dakar further explains the relationship between migratory status and type of activity. Education level appears to be a more discriminating factor than migratory status in explaining entry in one or the other sector of the urban economy. Piché and Gingras (1998) showed that, when controlling for education, the chance of a migrant entering the informal sector is lower than for a non-migrant in Dakar, whereas the migratory status has no effect on the probability of working in the informal sector in Bamako. Considering entry in the formal sector in the same cities, Bocquier and LeGrand (1998) showed that, education level being equal, new migrants stand a better chance of finding formal employment than non-migrants. Again, as was suggested above about unemployment, migrants seem to be more competitive than non-migrants in the formal labour market. It is also worth noting that the place of birth (rural or urban) and the social origin (ethnic group and caste) of the migrant have no effect on access to employment in the two capital cities studied. A similar study using retrospective data collected in Abidjan (Kouamé and Gueye, 2000) showed that the chances of a male migrant finding wage employment were 1.4 times higher than for a non-migrant, whereas for women the chances were equal.
Overall, these studies show that, for equivalent levels of education, migrants are more likely to begin their career in the formal sector (and less likely to begin it in the informal sector) than non-migrants. As a result, two ideas emerge. First, it seems that education gives migrants more incentive to look for better jobs in the cities. And, secondly, the key factor that leads migrants to the urban informal sector is (lack of) education, not migration itself. Adding to this the fact that migrants are less affected by unemployment than non-migrants, we conclude that recent research in Francophone west Africa gives a totally different picture from the one generally describing migrants as ill-adapted to city life and engaged in lower-level economic activities. Migration could be seen as a qualifier rather than a hindrance in the urban job market.

Conclusion
From the conclusions of the above-mentioned analyses of Francophone west Africa, drawn from cross-sectional and longitudinal data, it appears that migrants are not disadvantaged when compared with non-migrants. From the point of view of housing and employment, migrants adapt quite well to the city. Urban integration problems do not concern exclusively migrants but all citydwellers. The opposition between the 'poor, uneducated, informally employed or unemployed migrant' and the 'better-off, educated, formally employed non-migrant' is, in the end, not supported by the facts. Certainly, economic inequalities are important, but the divide is not along the migration line: the economic crisis struck the urban population equally, regardless of origin-as noted by other authors for Africa as a whole as early as the 1980s (Kannappan, 1988;Jamal and Weeks, 1993).
All urban integration surveys agree in observing that the economic crisis has increased the difficulties faced by both migrants and non-migrants in urban areas in terms of employment, housing and family constitution. Young people in particular appear to have more difficulties becoming independent in all respects. The crisis and the implementation of SAPs have made cities and towns less desirable for migrants but also for urban natives: job opportunities are becoming scarcer; urban facilities are deteriorating, etc. In this context, urban growth registers signs of a slowdown largely because of the redistribution of rural-urban migration flows. Although the urbanisation process might differ in Anglophone Africacircular migration of single males is more frequent there-results for later periods show the same trend as in Francophone Africa (Potts, 1995;Bigsten, 1996;Mazumdar and Mazaheri, 2002).
To date, the effects of the recent Ivorian turmoil are quite difficult to assess in terms of urbanisation processes within the country and abroad in West Africa. However, the current economic trends in the region might be favourable to the continuation of the diminishing contribution of migration to urban growth. Furthermore, in the coming years, the slowdown in urban growth should be enhanced by a reduction in the natural growth in urban areas, where, in contrast to rural areas, the demographic transition towards lower fertility is already starting. In the end, if the slowdown in urbanisation and urban-to-rural migration trends were confirmed for Africa as a whole, United Nations projections predicting that the level of urbanisation in the continent would reach 50 per cent by 2025 and would continue to grow thereafter (United Nations, 2002) should seriously be questioned (Bocquier, 2003b;Cohen, 2004). The observed trends suggest, rather, that Francophone west Africa may remain for long predominantly rural. Would it be a pause in the urban transition? In any case, it would reflect the marginal position of Africa in a mostly urbanised world economy.

Notes
1. Bairoch (1985) estimates that the urbanisa-tion level of Africa as a whole was 5 per cent in 1900, 12 per cent in 1950 and 28 per cent in 1980. 2. This database uses a constant threshold to define urban and rural areas: settlements with population of over 10 000 are considered as urban (Moriconi-Ebrard, 1994). In this paper, we prefer to use the GEOPOLIS database for comparative purposes, preferably with the figures published by the UN. The UN use national, non-homogeneous definitions of urban areas, generally with a much lower threshold (2000 or 5000 inhabitants), sometimes combined with nonconsistent administrative criteria. The national urban definitions lead to an overestimation of urbanisation levels (Bocquier, 2003a governments of developing countries believed that internal rural-to-urban migration was the dominant factor contributing to city growth (United Nations, 1998). 5. We use the expression 'peri-central areas' to designate the districts that are close to the centres of cities. In west Africa, the centres were generally built and first occupied by the colonial power. Indigenous people historically occupied the surrounding peri-central areas. At first outlying, these districts are today quite central because of urban expansion. However, their principal function remains residential, whereas the centres are mainly occupied by administrative and commercial functions. 6. However, empirical studies conducted in other parts of the developing world have shown that migrants do not seem to face particular difficulties when entering the urban job market: regardless of their education, they obtain employment faster than non-mi-grants (Goldscheider, 1983;Banerjee and Bucci, 1995). 7. According to the NESMUWA surveys, in all Francophone countries, between 40 and 68 per cent of migrants in urban areas were educated as compared with 44-82 per cent of non-migrants (Bocquier and Traoré, 2000). Secondary towns are all other administrative centres.
b Secondary towns include all remaining administrative centres, settlements of over 10 000 and settlements of over 4000 with more than 50 per cent of households involved in agricultural activities. Source: République de Côte d'Ivoire/CERPOD (1996). Principal towns are regional capital cities.
b Secondary towns include all remaining settlements of over 10 000.
c National data do not allow the computing of international flows. The total results were computed with the data of the other NESMUWA countries. Source: Bocquier and Traoré (2000).