1 Since the beginning of the AIDS epidemic in Africa, the association between migration and contamination has been an issue. Migrants were initially seen as a medium for the spread of the disease in the host population. This approach was very controversial, and was replaced by a view that treated the migrants above all as sexually and socially vulnerable individuals at risk of becoming infected. This article analyses data from a survey on return migrants in the Senegal River valley. Richard Lalou and Victor Piché consider the issue from a completely different perspective. Individuals actively shape their own behaviour, adapt to the situation and manage risks in given environments. Upon their return to their community of origin, the international migrants tend to remain faithful to their spouses, whereas the internal migrants do not necessarily renounce multiple partnerships, but tend to protect themselves by using condoms. Thus it appears in this part of Senegal that the migratory experience does not result in an actual increase of risky behaviour in the return area. This is doubtless one of the factors, among many others, which explains the persistence of a low prevalence of AIDS in Senegal.

2 Although numerous studies have already shown the existence of an association between geographic mobility and AIDS (Lalou and Piché, 1994; Decosas and Adrien, 1997; Kane et al., 1993), the complex processes underlying this association are still poorly understood (Soskolne and Shtarkshall, 2002). Since the beginning of the epidemic, analyses of the relationship between AIDS and migration have frequently borrowed customary interpretations from public health to account for the health of migrants. In this context, AIDS is either a “pathology of importation” and the migrants are its carriers, or a “pathology of adaptation” and the migrants accumulate vulnerabilities favourable to the infection when subjected to a stressful environment — that of their host area (Gentilini and Du?o, 1986; Gentilini, Brücker and de Montvalon, 1986). In the ?rst case, containing the epidemic implies controlling the migrant population; in the second case it depends on an awareness of the plight of migrants, and on fair access to treatment. The approach shifts from a mode of stigmatization to one of compassion.

3 The ?rst point of view — that of the pathology of importation — is mostly present in research concerned with the epidemiological aspects of AIDS. Like other infections, HIV is transmitted from one person to another and circulates according to the direction and rhythm of human moves. For sub-Saharan Africa, a signi?cant number of studies have con?rmed these clear associations between the epidemic’s spatial dynamics and labour migration (Painter, 1992; Hunt, 1996; Prothero, 1996), forced migration (Prothero, 1994), urbanization (Lydié et al., 2001), and major roads (Marck, 1999). The geography of AIDS, as that of many epidemics of the past, con?rms the existence of such a relationship (Prothero, 1977; Amat-Roze, 1989 and 1993; Lydié et al., 1998).

4 Admittedly, this interpretation has the merit of shedding some light on the epidemic’s dynamics. However, it also addresses the association between migration and AIDS in a mechanical manner, without explicit reference to the modes of transmission. Here, the body of the migrant is an infected and contagious body. It is the vehicle for a virus looking to conquer other bodies and other territories. It is also a body separated from its social and cultural reality. Sexual behaviour, i.e. the practices that connect the intimate with the social and through which the virus is transmitted, is subordinated to the essential characteristic of the migrant, mobility. From this point of view, it can be said that the analysis belongs to the biological rather than to the social plane, and that it is less interested in the association between migration and AIDS than in the migration of AIDS (or the spread of HIV).

5 The second interpretation of the association between migration and AIDS, which is commonly opposed to the epidemiological model, ?ts into a psychosocial and sociological approach. This analysis highlights the social and behavioural mechanisms through which migration increases the risk of HIV infection and places the migrants’ vulnerability at the core of the explanations. For some authors labour migration, initiated during colonization with the introduction of capitalism, contributes in sub-Saharan Africa to a long and thorough process of disorganization of societies, families, and behaviour (Descloîtres, 1972; Doyal and Pennell, 1981; Hunt, 1989; Lalou and Piché, 1994). These studies are generally based on examples of international migration in western and southern Africa that focus on Côte d’Ivoire and South Africa respectively. Here, labour migration fundamentally produces insecurity and vulnerability as a result of the migrants’ legal and social outsider status, their socio-demographic characteristics, and the restrictions imposed by the economic and social organization of the receiving area. These factors in turn create a change in the migrants’ sexual behaviour, resulting in a multiplication of sexual partners and resort to prostitutes (Hunt, 1996; Packard and Epstein, 1992; Anar?, 1993; Lalou and Piché, 1996; Lurie et al., 1997; Campbell, 1997).

6 While this approach has the advantage of focusing on the migrant’s point of view rather than on that of the epidemic, it does not avoid the trap of an essentialist drift that characterizes the epidemiological analysis. The category of migrant is rei?ed and almost promoted as a foundation of identity. Sexuality is imprisoned in an essential determination that is de?ned once and for all. Risk appears inherent to the migrant person. And even when labour migration is not a risk factor in itself, it generates a set of alienating restrictions over which the migrant has no in?uence: poverty, dif?cult living and working conditions (in mines and agriculture), social and emotional isolation, and also xenophobia. Risk is deemed inherent to the migrants’ condition (Delaunay, 1999; Delor and Hubert, 2000).

7 As we shall see, the issue here is not to deny the existence of real and signi?cant determinants of certain forms of migration in certain contexts (Fassin, 2000), but to reject an overly simplistic naturalistic interpretation. Migration de?nes neither a condition which in itself involves risks, nor an environment which necessarily produces risks. It is a social reality that is constructed through life histories and situations, as a function of context and social networks. Within this reality, individuals shape their behaviour, make their choices, and manage their risks.

8 Our analytical perspective reintroduces individual strategies within the boundaries of structural and institutional frameworks (Soskolne and Shtarkshall, 2002). In other words, the perception and management of risk are constructed inside a socio-cultural matrix where several factors are combined and give meaning to the practices. The operating rationality is no longer only that of the structures nor totally that of individuals, but that of the situation within which the behaviour takes place.

9 It is from this theoretical perspective that we have re-examined the association between geographical mobility and the spread of AIDS in the return area. Beyond the classic epidemiological assessment of migration as a vehicle for the transmission of disease (Prothero, 1977), the issue of the spread of HIV encompasses the migrants’ sexual behaviour and the spread of AIDS as sociological facts. The risk of spreading HIV is therefore linked as much to the migrants’ serology as it is to the socio-cultural context of the return migrants’ sexual behaviour. In other words, we assume that the form and level of the spread of HIV in return areas depend at the same time on individual and collective perceptions of the risks and on the social and cultural characteristics of the return area. That area in?uences the migrants’ ability to reproduce or to avoid the sexual conduct that they had in the destination areas and in part determines the strategies used to adapt themselves to the risk of transmitting HIV.

10 According to the literature on the relationship between migration, risky sexual behaviour, and AIDS, four limitations should be considered. First of all, these studies focus essentially on the migrant’s risk of infection, as the risk of diffusion is believed to result nearly automatically and mechanically from it. Secondly, the analyses concentrate on individual variables and neglect the macro-structural context. Thirdly, the research does not always distinguish between different types of migration. And ?nally, these studies rarely compare migrants and non-migrants. In order to address these shortcomings, our study explicitly examines the link between migration and risky sexual behaviour in the return area (risk of spread) by using a conceptual framework that considers: 1) different types of mobility, 2) different social contexts, and 3) non-migrants. The macro-social level is studied through the choice of two strongly contrasting zones from the point of view of mobility and socio-economic context. Other individual variables that are important as explanatory factors, such as marital status, level of education, and household wealth, are also included in the analyses.

11 The present study is based on a survey carried out in Senegal in the year 2000 in the Senegal River valley on a sample of 1,320 individuals (Piché et al., 2003). This region had already been mentioned as a high mobility zone with a higher level of HIV prevalence than elsewhere in Senegal (Kane et al., 1993). The two zones studied, the town of Richard-Toll and the rural area of Matam, are signi?cantly contrasted both economically and socio-culturally and are characterized by the high mobility of their populations. As migration is not an undifferentiated phenomenon, which would imply uniform risk, we have distinguished three forms of spatial mobility: international migration, internal migration, and temporary short-term moves. The indicator used to de?ne risk in a mobile or sedentary situation is sexual behaviour.

12 Following a brief overview of the economic and socio-cultural contrasts between the two zones, we introduce the concepts, data, and methods used. Next we examine the effect of migratory experience on sexual behaviour in the survey area, which for the migrants is the return area. These analyses are supplemented by a comparison of the migrants’ sexual practices in both the destination and the return areas. Using these results we discuss the in?uence of the cultural and social environment and of perception and knowledge of AIDS on the sexual practices of mobile individuals in the return area.

I – Context of the study: Matam and Richard-Toll, two contrasting zones

13 We chose to carry out the survey Mobility and STD/AIDS in Senegal (MISS – Mobilité et IST/Sida au Sénégal) in the Senegalese part of the Senegal River valley. Situated in the north of the country, the alluvial val-ley of Senegal is a vast border region, with the river constituting a natural separation between Senegal on one side, and Mauritania and Mali on the other. Since the beginning of the 1970s, successive periods of drought have resulted in a significant deterioration of the soil and therefore in a considerable reduction of agricultural yields and the development of internal and international migration (Findley, 1994) [1]. As in the rest of Senegal, internal migration to Dakar is very high. But in this region the rural exodus is also the result of a strong international migration. Migration to Mauritania is the most frequent in the lower valley, while in the middle valley it is more targeted towards west Africa (Côte d’Ivoire), central Africa (Gabon and Congo), and Europe (France). During the 1988-1992 period, 9% of internal migration and 19% of international migration originated in the Saint-Louis region. Among all departures from this region, 38% are to foreign countries. With an international migration rate of 1.65%, which is nearly double the national rate (0.87%), Saint-Louis is the leading region of emigration to foreign countries (Direction de la prévision et de la statistique, 1998).

14 Like some other countries in the sub-region (Mauritania, Mali, the Gambia), Senegal is an African country with a low HIV prevalence: 1.4% of the whole population (Comité national de prévention du sida, 2002). Although information remains fragmentary, this rate almost certainly includes “pockets” of higher HIV prevalence, particularly in the regions of high international mobility [2]. Thus as early as 1990 an exhaustive study of 8 villages (n = 600) in the region of Matam (Senegal River valley) demonstrated signi?cant serological differences between international migrants and other individuals. In the sample studied, 414 people had not travelled out of the country during the preceding ten years and their HIV prevalence was 0.5%. Among the 186 people who had travelled abroad, the prevalence reached 5.4% (3.7% for females and 7.7% for males). The rate for the sample as a whole was 2.0% (Kane et al., 1993). Let us note that the Matam region is one of the regions most affected by international migration. A survey carried out in 1990 in the Ziguinchor region showed a similar association between HIV prevalence and temporary rural-urban migration (Pison et al., 1993).

15 In 1998 the population of the Saint-Louis region represented nearly 10% of the national population. It is characterized by its youth and by an over-representation of females (a sex ratio of 118 females for 100 males). The level of education of the population is generally low, especially among females. The population is composed principally of Haalpoularen (61%) and Wolofs (30%). The Soninké (3%) and the Maures (4%) are other important ethnic groups in the region. Traditionally, the Haalpoular society is a very hierarchical caste-based system. Nearly all the inhabitants of the Senegal valley are Muslim. This region, which was the starting point of the Islamization of Senegal, remains strongly attached to an orthodox version of Islam. Religion is an essential element in the socialization of the individual and in the de?nition of values and norms.

16 The traditional cultural systems based on patrilinearity and virilocality as well as religion weigh heavily on female status and favour strict social control. As we will see below, this control is expressed particularly in nuptial strategies where virginity and faithfulness are valued. In contrast to women, sexual experiences before marriage are sometimes valued for men and male in?delity is better tolerated. As is frequently the case elsewhere, there is a double standard in sexual behaviour.

17 The four sites of this project are situated in two quite contrasting zones, both socio-economically and from the point of view of migration. One of the sites belongs to the delta region (the town of Richard-Toll) while the other three are situated in the middle valley of the river (Matam region).

18 The town of Richard-Toll situated in the Dagana region is home to a large agribusiness, the Compagnie sucrière sénégalaise (a sugar producing company), the majority of whose employees are seasonal workers from the surrounding villages, other regions of the country (particularly Casamance), and Mauritania. As a result of the development of economic activity, the city experienced high population growth until recently. The presence of the factory has reinforced the position of the town as a crossroads and has favoured the development of small trade and prostitution.

19 The small town of Ourossogui is one of the survey sites in the Matam area. It is situated 7 km from the capital of the Matam département and has developed because of its geographic position. Easily accessible, even during the rainy season, Ourossogui is at the crossroads of two major roads: the axis Saint-Louis – Ourossogui – Bakel – Kidira and the axis Ourossogui – Linguère – Dakar. This situation has enabled the town to grow rapidly, notably as a result of the arrival of migrants from the groundnut producing zone and the settlement of returning international migrants from the region. Furthermore, the town is home to a military camp.

20 The other two sites are rural areas in the Matam département. Like many of the other rural communities of the département, these two villages have been much affected by international migration, particularly to Côte d’Ivoire. The population of these villages belongs nearly exclusively to the Haalpoularen and Soninké ethnic groups.

21 The differences between the Richard-Toll and Matam zones are ?rst of all socio-economic. As a result of urbanization, the population of Richard-Toll is characterized by great ethnic diversity. All the large ethnic groups of Senegal (Haalpoular, Soninké, Wolof, Sereer, Diola, etc.) are represented in Richard-Toll. In contrast, the largely rural Matam zone is essentially inhabited by Haalpoularen (77% of the population) [3]. Moreover, the population of Richard-Toll is much better educated than that of Matam. The proportion of the population who have never attended school is 1.45 times higher in Matam than in Richard-Toll (p = 0.001) (Table 1). Furthermore, a far higher proportion of inhabitants of the Matam area have little knowledge of AIDS than in Richard-Toll (RR = 3.70; p< 0.000) [4]. From an economic perspective, the sugar industry and the derived commercial activities employ the majority of Richard-Toll’s working population. According to our data, nearly a third (31.5%) of Richard-Toll’s active males were employed by the Compagnie sucrière du Sénégal during the twelve months preceding the survey. In Matam, on the other hand, the major part of economic activity is centred on agriculture and livestock farming.

Table 1

Characteristics of the surveyed population

Richard-Toll Matam Total Migrants’ country of destination (most recent migration) Mauritania 86.0% 20.0% 65.1% Other neighbouring countries 11.6% 15.0% 12.7% Other African countries (with >5% HIV prevalence) 2.3% 65.0% 22.2% Individuals with no schooling 43.8% 63.4% 48.5% Attitudes Virginity: high value of women’s virginity 92.3% 95.4% 93.0% Faithfulness: high value of faithfulness 95.1% 93.7% 94.8% Individuals with poor knowledge of AIDS (score below 8 on a scale of 0 to 13) 11.3% 42.2% 18.7% Average age at first sexual intercourse Males 18.2 years 18.1 years 18.2 years Females 17.1 years 16.5 years 16.9 years Sexual intercourse before first marriage Males 73.9% 67.0% 72.2% Females 13.9% 6.9% 12.5% Sexual intercourse outside of marriage Males 9.3% 13.9% 10.3% Condom use Never used a condom 63.1% 73.2% 65.3% Non-systematic use in occasional sexual intercourse 56.8% 40.5% 54.1% Unweighted numbers 925 395 1,320 Source: MISS Survey, 2000.

Characteristics of the surveyed population

22 Moreover, the two zones are characterized by very different migratory dynamics. The development of industrial and commercial activity transformed Richard-Toll into an important regional attraction pole. As the activity of the Compagnie sucrière du Sénégal is to a great extent seasonal, internal migration to Richard-Toll also partly presents this characteristic. The population of the middle valley (Matam zone) is traditionally characterized by high mobility, particularly to foreign countries. According to our survey, nearly 22% of males in Matam (aged 15-49) declared having lived abroad at some point since 1985. Finally, when we examine international migration dynamics in the two zones it appears that in Richard-Toll migration is oriented particularly towards the neighbouring countries (86% to Mauritania), whereas in Matam it is for the majority (65%) directed towards Côte d’Ivoire and central Africa (Central African Republic, Congo, Burundi, etc.), i.e. towards countries with a high HIV prevalence (Table 1). In Mauritania, HIV prevalence is slightly lower than in Senegal.

23 We have stressed the differences between the two zones because this distinction is essential to our analysis. Our hypotheses are partly based on this distinction. Urbanization, schooling, and the development of wage employment, as observed in Richard-Toll, are probably factors that in?uence the loosening of social ties and the individualization of practices. Furthermore, sexual services are readily available. In this context, we can assume that sexuality is little controlled by society. On the other hand, in the Matam zone the socio-cultural conditions suggest strong social control of sexuality. This control, which is traditionally exercised over young girls, could also affect international migrants in view of the risks that the medical and social discourse assigns to them. On the whole, the Richard-Toll and Matam zones constitute two contrasting contexts, within which sexual practices and strategies for risk prevention should be manifested in very different ways.

II – Data and method of analysis

2 – Measurement of risky sexual behaviour (dependent variable)

33 Like several publications before this (Brockerhoff and Biddlecom, 1999; UNAIDS, 1998; Adrien and Cayemittes, 1991; Caraël, 1995; Spira et al., 1993), we de?ne risky sexual behaviour as a function of the number and type of sexual partners (regular, occasional, sex worker) and condom use. In order to distinguish regular partners (other than married individuals) from occasional partners, we consider the length of the relationship, the nature of the relations (intercourse in exchange for gifts or money), and the sexual network of the partner (whether he or she has partners other than the respondent during the relationship). Therefore, a sexually active person is demonstrating risky sexual behaviour if he or she meets at least one of the following conditions:

• At least one case of intercourse with a sex worker during the last twelve months;
• At least one occasional sexual partner during the last twelve months. An occasional partner is defined as a partner with whom the indi-vidual has been having sexual relations for less than a year, or with whom intercourse is obtained in exchange for money or gifts and who, according to the respondent, has other sexual partners;
• For married individuals, at least one extra-marital sexual partner, regular or occasional, during the last twelve months;
• For unmarried individuals, at least two regular sexual partners. A regular partner is either a fiancé(e) or a partner that is not occasional.

Furthermore, the respondent must declare that condom use was not systematic during these relationships.

34 From this de?nition we created an exclusive variable with three modalities. This dependent variable takes the value of 0 if there is no multiple sexual partnership and no resort to prostitution. This reduced risk of meeting an HIV-positive person should re?ect a strategy of risk avoidance through faithfulness to the regular partner or spouse. The variable takes the value of 1 if the respondent reports several sexual partners and/or visits to sex workers while declaring having always used a condom during intercourse with these occasional partners. Systematic condom use in a risky situation should in this case indicate a strategy of avoiding the risk of infection through protection. Finally, the variable takes the value of 2 if the individual reports sexual relations with several partners and/or sex workers while not using condoms or using them occasionally. This behaviour could indicate a logic of risk-taking [8].

35 Multiple sexual partnership and the absence of systematic condom use are signi?cant factors in risky sexual behaviour. We present some descriptive elements concerning these indicators. First, regarding sexual behaviour at the time of the survey in the Senegal River valley, we observe that multiple sexual partnerships are not very frequent among the sexually active respondents (Table 2). Only 8.1% reported having had two or more sexual partners over the last twelve months (for married individuals, spouses are counted in the number of partners). Multiple sexual partnerships are almost non-existent among females but involve nearly one out of six males (16.5% in Richard-Toll, 14.4% in Matam). In migration, multiple sexual partnership is always more frequent: depending on the type of mobility considered, the practice is 2 to 3 times higher than what is observed in the general population of the survey sites. It is always dif?cult to evaluate the reliability of responses to such intimate and sensitive questions. Nevertheless, in comparing our results with those of the Senegalese survey on health indicators in 1999 (ESIS), we ?nd that the levels of multiple sexual partnership are quite similar. For the Saint-Louis region, the proportions were 1.6% for females and 29.6% for males (Sénégal, 2000, pp. 106-107) [9]. It may be that the levels have been underestimated in both surveys, but as we have already mentioned the under-reporting should not be too different according to migratory status, which is the key variable in our study.

Table 2

Frequency of multiple sexual partnerships in the survey areas and during the most recent migration or temporary move, by sex and residential area

Area of reference for sexual behaviour Richard-Toll Matam Total Males Females Total Males Females Total In the survey area(1) % 16.5 1.1 8.8 14.4 0.0 5.8 8.1 Number 380 373 753 131 183 314 1 067 International migration(2) % 28.9 8.0 21.4 25.0 7.1 18.4 20.6 Number 48 41 89 41 19 60 149 Internal migration(2) % 32.5 0.8 14.1 36.8 3.1 21.4 16.4 Number 73 114 187 50 42 92 279 Temporary moves % 11.6 8.7 10.3 27.9 11.3 19.5 12.4 Number 192 147 339 80 67 147 486 (1) More than two sexual partners during the last 12 months. (2) Sexual intercourse with occasional partner or with prostitutes (for males) during the most recent migration or temporary move. Source: MISS Survey, 2000.

Frequency of multiple sexual partnerships in the survey areas and during the most recent migration or temporary move, by sex and residential area

36 As for systematic condom use (with reported partners in the survey zone, or during occasional sexual intercourse or relations with prostitutes — for males — during the most recent migration or temporary move), the numbers are too small to present detailed data. We will therefore give general indications using raw (unweighted) data. In the survey zones, approximately 44% (76/173) of those who have had several partners declared systematic condom use. The proportion is higher in Matam (57%) than in Richard-Toll (39%). In the migration areas, the frequency of condom use is generally greater at 51% (49/96), and is higher for international migrants (64%) than for internal migrants (31%). The reliability of such information is again at issue. The comparison with ESIS data indicates that the frequency of systematic condom use is quite similar for males: 55.9% for the Saint-Louis region against 50.3% in the survey zones (Sénégal, 2000, p.110) [10]. At any rate, it is possible that these proportions are overestimates. With awareness campaigns and the frequent allusions to the association between migration and HIV/AIDS, we may assume that the respondents are unwilling to declare multiple sexual partnerships and the non-use of condoms at the same time, and that this is particularly true for international migrants. But we remind the reader that more returning international migrants report sexual relations with occasional partners and sex workers and the frequent use of condoms during their last stay abroad.

III – The influence of migratory experience on sexual conduct

41 Before beginning the analysis of the relationship between migratory experience and sexual behaviour, let us look again at Table 1 in order to present some elements that characterize the survey zones from the point of view of norms and sexual practices. The contrast between the two zones, mentioned earlier at the economic and migratory levels, seems much less marked when considering sexual attitudes and behaviour. Female virginity at marriage is a value shared by 90% of the population in Richard-Toll and Matam (Table 1). However, the number of women who do not accord such importance to virginity is 4.1 times higher in Richard-Toll than in Matam (p = 0.006). Moreover, in both Richard-Toll and Matam, more than 9 out of 10 respondents declared unfaithfulness unacceptable within marriage, both for men and women. Regarding sexual practices, 20.8% of respondents aged 15-49 have never had sexual intercourse. This proportion is slightly higher for females (23.2% against 18.0% for males; p = 0.023). The differences between the two zones are not signi?cant. The average age at ?rst intercourse is 17.5, and females begin their sexual life 1.3 years earlier than males (16.9 years against 18.2 years; Student’s p < 0.001). Once more the differences between Richard-Toll and Matam are not signi?cant. 40% of the total population had sexual intercourse before their ?rst marriage. This proportion is 72% for males and does not vary signi?cantly between Matam and Richard-Toll. However, among females this proportion is only 12%, but it is twice as high in Richard-Toll (14%) as in Matam (7%) (p = 0.024). In the whole sample only males reported having extra-marital relations during the 12 months preceding the survey. They represent slightly over 10% of married males. The practice of the levirate is at the origin of 3% of marriages contracted by the survey population; there is no difference between Matam and Richard-Toll in this respect.

42 Finally, we note that two thirds of sexually active persons have never used a condom. Systematic condom use is more likely to occur during occasional sexual relations: approximately 45% reported using condoms systematically in such circumstances (Table 1). This proportion does not vary signi?cantly between the two zones and is quite similar to that reported in the Saint-Louis region and at the national level.

43 The study of the association between migratory experience and sexual conduct is based on multinomial regressions produced separately for the zones of Richard-Toll and Matam and only on the male population (Table 3) [14]. Generally, we observe that the increase in likelihood obtained by the derivation of the function varies between 30% and 45% (pseudo R²) depending on the model, and that the models are signi?cantly different from the zero model (i.e. with all the coef?cients equal to zero) [15]. We must nevertheless add that the relatively small sample size (n = 105 in Matam and n = 323 in Richard-Toll) causes a certain imprecision in the estimates: the con?dence intervals of the coef?cients are sometimes large. This situation does not affect the statistical signi?cance (con?dence intervals do not include the value of 1) or the direction of the association, but it does not make it possible to really qualify its intensity [16].

Table 3

Factors associated with behaviour of faithfulness and protection with regard to risky sexual behaviour in Matam and Richard-Toll (male population)

Explanatory factors Dependent variables Matam Richard-Toll Relative risk ratio (RRR) p Relative risk ratio (RRR) p Age at survey (Ref. 15-29 years) Protection vs. Risk 1.06 0.952 0.62 0.466 30-45 years Faithfulness vs. Risk 0.20 0.107 0.94 0.919 Marital status (Ref. Unmarried) Protection vs. Risk 0.92 0.935 0.62 0.525 Married Faithfulness vs. Risk 14.32 0.007 22.1 0.000 Level of education (Ref. No schooling) Primary Protection vs. Risk 0.62 0.557 0.69 0.471 or above Faithfulness vs. Risk 0.46 0.307 0.94 0.881 Migratory status (Ref. Non-migrant) International migrant Protection vs. Risk 2.91 0.350 0.39 0.174 Faithfulness vs. Risk 10.30 0.042 0.37 0.154 Protection vs. Risk 8.58 0.024 3.66 0.067 Internal migrant Faithfulness vs. Risk 2.72 0.293 2.68 0.120 Temporary Protection vs. Risk 4.35 0.216 0.90 0.879 move Faithfulness vs. Risk 3.35 0.229 0.53 0.160 Household’s wealth (Ref. Poor household) 0.64 0.720 0.77 0.755 Rich households Protection vs. Risk Faithfulness vs. Risk 2.57 0.309 1.16 0.856 Number of observations 105 323 Pseudo R2 0.4553 0.3087 Source: MISS Survey, 2000.

Factors associated with behaviour of faithfulness and protection with regard to risky sexual behaviour in Matam and Richard-Toll (male population)

44 The models applied to Matam and Richard-Toll evaluate the statistical associations between independent variables and risky sexual behaviour, when these interact with the survey zone. Looking at Table 3 the only factor associated with risky sexual behaviour common to both zones, and which is therefore not inherent to the context, is marital status. Married males, more so than unmarried males, avoid risky situations (i.e. occasional or regular multiple sexual partnerships without condom use) by being faithful (i.e. absence of multiple sexual partnerships and of visits to sex workers). On the other hand, marital status does not discriminate between protection behaviour (i.e. multiple sexual partnerships or visits to sex workers with systematic condom use) and risk-taking. Marriage and the disapproval of unfaithfulness by the community seems for males to constitute a framework which limits multiple sexual partnerships, even though this compliance with the social norm may also be the result of a reporting effect, with married males having no interest in admitting extramarital relations. By the same token, we observe once more that males, even when they are married, are more likely to report sexual relations with occasional partners during migration or a move.

45 On the other hand, risky sexual behaviour does not have any signi?cant statistical association with age, level of education, or household wealth for males resident in Matam or Richard-Toll. These results are quite unusual in the literature on sexual behaviour in Africa, as education in particular is sometimes described as a behavioural risk predictor for males. Better educated men often report more occasional sexual relations than others (Caraël, 1994; Cleland and Ferry, 1995). According to some authors, this link could be explained by educated individuals’ greater capacity to evade the control of their circle of family and friends (Spira, 1998). Regarding the absence of an effect of age on risky sexual behaviour, this result could be linked to the fact that only two age groups were taken into consideration. However, when three age groups are used (1524 years, 25-34 years, and 35-49 years) in a multinomial model which includes all sexually active males and females in both zones, we do not ?nd any more statistical association with sexual behaviour [17].

46 The association of migration experience with sexual behaviour varies according to the type of migration (national or international), and the survey area. We observe that the inhabitants of Matam who were involved in international migration are more likely to be faithful than to have unprotected casual sexual intercourse, compared to non-movers (RRR  [18] = 10.3; p = 0.042). On the other hand, no statistical association is observed between international migration and protection behaviour. These observations seem to conform with other results obtained elsewhere (Piché et al., 2003). The international migrants of Matam report more frequently that they have modi?ed their sexual behaviour (mainly by being faithful) since they became aware of AIDS (OR = 7.9; Fischer’s p = 0.005). Furthermore, the same migrants in Matam expressed the most unfavourable opinion of condom use (OR=4.5; Fischer’s p = 0.0005) [19]. We must nevertheless note that this lack of intercourse with occasional partners and these reservations towards the condom are expressed essentially in the context of the area of origin. When international migrants are interviewed on their sexual behaviour while abroad, they report having sexual relations with occasional partners or sex workers and nearly always using a condom. In comparison with the sedentary inhabitants, potentially risky sexual behaviour (i.e. with occasional partners or sex workers, with or without protection) are, for the migrants of Matam, much more frequent in the migration areas (OR _adjusted  [20] =18.9; p = 0.001) (Piché et al., 2003). Similarly, we observe that international migrants are the only moving individuals to signi?cantly increase condom use when they are away compared to when they are in their zone of origin (OR _adjusted =6.5; p =0.002) (Piché et al., 2003). Multivariate analysis indicates that the international migrants of Matam do not consider themselves to be more at risk from HIV contamination than the sedentary population (Lalou et al., 2004).

47 In Matam, protective behaviour is statistically associated with hardly any of the respondents’ other characteristics. Only migratory experience within Senegal has any in?uence on systematic condom use with multiple or occasional partners. Thus we observe that internal migrants, all other things being equal, are more likely to systematically use a condom upon their return to Matam than non-movers (RRR = 8.58; p = 0.024). This protective behaviour is con?rmed by some declarations of the internal migrants of Matam that we have regrouped in Table 5. More than half assert having changed their behaviour since they became aware of AIDS (63.9%), and of those, 37.6% have changed to using condoms (Table 5). Of the sedentary population interviewed in Matam, only 10.6% reported the same choice of condom use (OR =4.7; Fischer’s p = 0.048) [21]. Finally, multivariate analysis (not presented here) has shown that, in contrast to the internal migrants of Richard-Toll, the internal migrants of Matam perceive themselves to be more at risk of HIV contamination than sedentary individuals of the same area (OR _adjusted  [22] = 2.21; p = 0.026) (Lalou et al., 2004). Finally, we note that, according to Table 5, internal migrants are not more likely than others to report having modi?ed their sexual behaviour towards faithfulness.

Table 5

Sexual practices, perception of risk, and knowledge of AIDS by survey area ( frequency in %)

Characteristics Richard-Toll Non-migrates International migrates Internal migrates Temporary moves Total Total number Potentially risky sexual behaviour (Males) 29.5 41.5 35.4 29.8 32.2 322 Little knowledge of AIDS (score< 8) for males 9.1 5.8 8.4 7.9 7.9 322 Males acquainted with a person living with HIV 10.6 15.4 9.7 16.0 12.3 481 Males with more than one partner 12.9 22.0 27.3 19.1 18.6 322 Do not perceive themselves to be at risk of infection from HIV (Males) 58.0 69.2 67.6 47.9 58.8 441 AIDS is a serious threat to the community 43.9 59.7 49.0 48.9 47.7 816 Migrants are a group at risk of HIV infection 18.5 8.0 17.9 16.8 17.0 914 Males who modified their sexual behaviour since the epidemic 50.0 64.0 60.7 59.6 56.7 380 Males who chose faithfulness as behavioural change 43.6 46.9 34.0 52.2 44.5 210 Males who chose condom use as behavioural change 23.9 36.6 53.7 25.5 32.7 210 Characteristics Matam Non-migrates International migrates Internal migrates Temporary moves Total Total number Potentially risky sexual behaviour (Males) 53.8 13.0 52.0 46.2 39.2 105 Little knowledge of AIDS (score< 8) for males 40.5 34.6 31.0 50.0 37.6 105 Males acquainted with a person living with HIV 30.6 53.8 43.9 26.3 39.3 122 Males with more than one partner 23.1 9.1 28.0 15.4 19.2 105 Do not perceive themselves to be at risk of infection from HIV (Males) 61.1 42.3 31.7 68.4 48.4 157 AIDS is a serious threat to the community 63.4 70.0 66.2 65.5 65.4 260 Migrants are a group at risk of HIV infection 22.6 19.5 26.6 32.8 25.3 387 Males who modified their sexual behaviour since the epidemic 51.9 60.0 63.9 52.9 58.1 138 Males who chose faithfulness as behavioural change 14.3 60.0 34.8 25.0 35.0 81 Males who chose condom use as behavioural change 10.6 15.6 37.6 27.5 24.3 81 Source: MISS Survey, 2000.

Sexual practices, perception of risk, and knowledge of AIDS by survey area ( frequency in %)

In Richard-Toll, being faithful or using protection are not associated with the experience of international or internal migration. In this city, no explanatory variable introduced in the model explains risky behaviour, apart from marital status whose effect is not speci?c to this context.

IV – Putting sexual behaviour during migration into perspective

48 In order to better understand the effect of the migratory experience on different risk avoidance strategies (being faithful and using protection), we must return to sexual practices and protection during the most recent internal or international migration, and compare them to those reported for the survey zone. As we already mentioned in a previous section, the data collected on sexual behaviour in the migration areas are not as detailed as those collected on sexuality in the survey areas. We have therefore created a comparable risk indicator: having had sexual intercourse with a sex worker or an occasional partner [23] during the most recent migration and during the twelve months preceding the survey. Condom use is not taken into account at this stage and so the risk measured is potential only. Nonuse of condoms, which indicates the real risk, will be considered at a second stage of the analysis.

49 By thus comparing behaviour in the migration and survey areas, a signi?cant reduction in relations with casual partners or sex workers is observed in the survey area (Table 4). The prevalence of this sexual type of behaviour for internal and international migrants together falls from 31.4% to 11.0% for the zones under consideration.

Table 4

Sexual relations with casual partners or sex workers among migrant males by area of reference (area of most recent migration/returning area) and survey zone (in %)

Survey zone International migrants Relative risk Fischer’s p In the destination area In the return area Richard-Toll 28.7 4.9 0.17 0.001 Matam 25.1 3.2 0.13 0.003 Total 27.4 5.3 0.19 0.000 Internal migrants Relative risk Fischer’s p In the destination area In the return area Richard-Toll 33.9 15.5 0.46 0.000 Matam 34.9 14.0 0.40 0.055 Total 34.2 15.1 0.44 0.003 All migrants Relative risk Fischer’s p In the destination area In the return area Richard-Toll 31.7 12.1 0.38 0.001 Matam 30.7 8.7 0.28 0.001 Total 31.4 11.0 0.35 0.000 Source: MISS Survey, 2000.

Sexual relations with casual partners or sex workers among migrant males by area of reference (area of most recent migration/returning area) and survey zone (in %)

This tendency is found again when international and internal migrants are considered separately, but at markedly different levels. Upon their return, international migrants reduce potentially risky sexual practices more than internal migrants. The reduction is 80% for international migrants against 56% for internal migrants and the difference is signi?cant.

50 The results obtained for each zone are not globally different from the preceding observations. In both Matam and Richard-Toll, migrants report less sexual intercourse with occasional partners in the survey area than during their most recent migration (Table 4). Furthermore, the international migrants interviewed in Matam and Richard-Toll tend to reduce risky behaviour more than internal migrants. In Richard-Toll, the relative risk for internal migrants is 0.46 against 0.17 for international migrants. In Matam, it is 0.40 and 0.13 respectively (Table 4).

51 These analyses suggest that upon their return, migrants largely abandon sexual relations with occasional partners. Overall, this tendency is stronger for international migrants than for internal migrants and is slightly more important in Matam than in Richard-Toll. Without undertaking a detailed interpretation of the results we can see the — probably large — effect of the social context. Social systems and religion promote strict social control, particularly in the rural area of the middle valley of the Senegal River. This control is manifested, particularly regarding the sexuality of young women, by the practice of early and sometimes arranged marriages, and by the increased value given to virginity and faithfulness. Under these circumstances, it is clear that migrants have a reduced capacity to reproduce the sexual behaviour of migration, namely, occasional relations in the areas where they live and get socialized.

52 When considering systematic condom use during intercourse with occasional partners and sex workers, the comparison between practices during and after migration indicates a signi?cant reduction in condom use by international migrants in the return area (Figure 1). The prevalence of protection behaviour thus falls from 79.6% to 36.5% (Fischer’s p = 0.001). A similar tendency is not observed among internal migrants and those who have made a temporary move. On the contrary, exclusive condom use appears to be a bit more frequent in the survey areas than during internal migration or temporary moves. However, these differences are not statistically signi?cant.

Figure 1

Systematic condom use during migration or a temporary move and in the return area, by migratory status (in %)

Systematic condom use during migration or a temporary move and in the return area, by migratory status (in %)

53 Behavioural changes regarding condom use appear therefore to be determined more by crossing the national border than by internal migration or temporary moves. As is often the case, the danger is perceived as being greater beyond national borders. On the other hand, as we will see in the following section, internal migrants are less aware of the risk in relation to a different area than they are in relation to their own sexual practices. This no doubt also explains the fact that they do not change their protection practices in the survey area.

V – Discussion

54 The study carried out in Senegal has made it possible to shed light on the complex dynamics of the association between migration and risky sexual behaviour in the return zones. On the one hand, we observe that return migrants abandon to a large extent their risky behaviour. On the other hand, the analyses suggest different AIDS risk avoidance strategies according to migratory choices (internal and international migrants), and according to return area (Richard-Toll or Matam). Movers returning to Richard-Toll do not seem to adopt different sexual behaviour from that of the sedentaries. In Matam, on the contrary, international migrants report, more so than the sedentaries, a sexuality which is limited to their regular partner; on their part the internal migrants report a tendency to use condoms in occasional relations.

55 In order to understand behaviour in the return area, we must interpret it not only as a function of the value that the migrants give to avoiding the risk of infection, but also as a function of the value (or the satisfaction) attributed to the strategy which produces this result. In Matam, the choice of faithfulness by international migrants doubtless makes it possible to avoid the risk of infection [24]. But it is probably also the expression of a greater desire (particularly in comparison with internal migrants) to respect the social norms of the community. In the middle valley of the Senegal River international migration is an unavoidable phenomenon that affects the whole society. Whether the issue is funding irrigated farming, promoting the private sector, satisfying the ?nancial needs of households, modernizing housing, or providing social and religious infrastructure in the villages, everything is related to international migrants and the money that comes from migration (Daum, 1999). Return migrants are therefore in the centre of the community rather than on its periphery. Beyond their central economic role, they are integrated again in the society of origin through a logic that is not one of discontinuity or change (particularly from the social point of view), but of conformity. This conformity is also expressed through respecting the sexual codes prescribed and sanctioned by the community. Finally, the international migrant is implicitly suspected of “importing diseases”. The inhabitants of Matam consider international migrants more as a “risk group” (25.3%) than the inhabitants of Richard-Toll (17.0%) (OR = 1.62; p = 0.002) (Piché et al., 2003), but this attitude does not produce reactions which elsewhere could be a threat to groups discriminated against, such as homosexuals or immigrants in the northern countries. International migrants returning to Matam are neither excluded from the community nor marginalized.

56 It remains to be seen whether the management of risk, as it is practiced by the international migrants of Matam, is an individual choice resulting from a better understanding and a greater awareness of the threat of infection or merely a reaction to the society’s indictment of a threatening sexuality—because it does not conform to collective norms or because it carries a threat to health. It is obviously very dif?cult to answer this question. In order to shed more light on this discussion, we will present some data on sexual practices, risk perception, and knowledge of AIDS (Table 5) on the one hand, and on the other hand, call on other analyses that we have presented elsewhere which reinforce the elements of interpretation that we are suggesting here.

57 On the one hand, international migration, and more particularly migration to countries with high HIV prevalence, does not promote better knowledge of AIDS. According to Table 5, the international migrants of Matam are no better informed about the disease than the rest of the male population of Matam and there are proportionally six times more of them with a low awareness of AIDS compared to the international migrants of Richard-Toll (RR = 6.0; p = 0.002) [25]. Analyses based on ordered logistical models con?rm that migration does not have a positive effect on knowledge about AIDS (Lalou et al., 2004). In Matam the results even show that international migrants have on the whole less knowledge of HIV/ AIDS than the sedentary population. The only signi?cant advantage that the migrants possess is that they know that a person infected by HIV can appear to be in good health (OR = 1.66; p = 0.015). However, this advantage is restricted to the international migrants of Richard-Toll, i.e. individuals who have migrated mainly to Mauritania.

58 But on the other hand, this relative lack of scienti?c knowledge concerning AIDS in Matam is not exacerbated by a misunderstanding of the reality of the disease (Lalou et al., 2004). The inhabitants of the Matam zone, and in particular those males who migrated abroad, are clearly aware of the threat of AIDS. Indeed, in the Matam zone, AIDS is considered as a serious threat to the community by 65.4% of the population, compared to 47.7% in Richard-Toll (RR = 1.37; p < 0.001). Similarly, nearly 40% of males in Matam report knowing one person living with HIV. This proportion is only 12% in Richard-Toll (Table 5). More than one out of two returning international migrants in Matam know one person living with HIV (53.8%) as against 15.4% in Richard-Toll.

59 Finally, the international migrants of Matam are less knowledgeable about the infection but also closer to the sick. They do not perceive themselves to be more at risk of becoming infected than the sedentary males of the middle valley of the Senegal River [26] (Lalou et al., 2004). As is the case for much of the native population of the Senegal River valley, AIDS is considered the disease of others and from elsewhere. Consequently, the native community is perceived as a space protected by Islam and culture, a space untouched by the “transgressions” which favour HIV transmission (multiple sexual partnerships and extra-marital relations) [27].

60 The return of internal migrants does not have the same social implications as does that of international migrants. By returning to their community of origin, the internal migrants’ social status does not necessarily change; they are reintegrated into the community, but not in a central position. Moreover, and contrary to migrants returning from Côte d’Ivoire or central Africa, internal migrants are not suspected of carrying the disease. Under these circumstances the community, and the migrants themselves, do not give a different social signi?cance to their sexuality than before migration. In other words, internal migrants, because they are neither a reference group nor a risk group for the community, are not a threat to the social body through their sexual behaviour (transgression of sexual codes), nor to the human body (contamination of the spouse by HIV). For the internal migrant, while the risk is not associated with a territory or with otherness, it is probably more closely associated with sexual conduct with multiple and occasional partners. The fact that signi?cantly more of the returning internal migrants of Matam perceive themselves to be at risk from infection than do the sedentary inhabitants (OR = 2.21; p = 0.026) doubtless con?rms this interpretation (Lalou et al., 2004). Finally, while internal migration has favoured multiple sexual partnerships and condom use, return to the community does not seem to impose different sexual and protection practices or greater conformity with the social norms of the area.

61 In conclusion, we propose the hypothesis that while internal migrants make greater use of the condom to avoid health risks, the strategy adopted by international migrants from the Matam region seems to tend more towards avoiding the risk of stigma.

Conclusion

62 The literature on AIDS in Africa places much emphasis on migration as a factor associated with both the risk of infection and the risk of spread of the disease. In this study we have chosen to focus our attention on the risk of spread associated with return migrants. Two basic hypotheses have guided our analysis. First, we have assumed that the links between migration and risky sexual behaviour are differentiated according to the type of migration. We have therefore distinguished between three forms of mobility, which apart from international and internal migration also include temporary moves. The second hypothesis stipulates that the links between migration and risky sexual behaviour are differentiated according to the social context in which the migrants’ return occurs. This is why we have chosen to study zones with strongly contrasting social contexts. Richard-Toll, the ?rst site, corresponds to an urban context marked by high international migration to a country of low HIV prevalence (Mauritania), while the second site, the Matam region, is essentially rural with emigration mainly directed towards countries of high HIV prevalence (Côte d’Ivoire and central Africa).

63 The results presented here lead us to re-examine the in?uence of migration on the prevalence of risky sexual behaviour and on the dynamics of the AIDS epidemic. Generally, the analyses show that the net effect of the migratory experience is not signi?cant, with the exception of migration in the Matam region. In that instance, the relationship is in the opposite direction of what is generally expected: all else being equal, returning internal and international migrants in Matam have less pronounced risky sexual behaviour than sedentary individuals. These results indicate that the social context and macro-structural factors play an important mediating role (Soskolne and Shtarkshall, 2002) in the de?nition of a behavioural strategy — that of the international migrants — but also that individuals adopt behaviour according to an itinerary (internal migration) and an experience (the multiple sexual partnership).

64 Moreover, the inhibiting in?uence of international migration on risky sexual practices in the return zone may explain why the HIV prevalence rate (2%) has remained stable in the Matam region between 1990 and 2002. More generally, these results make it possible to expand on explanations already put forward for the favourable situation of Senegal (Meda et al., 1999). Since 1989 the AIDS epidemic has been moderate and relatively stable, in comparison to the majority of countries of the region [28]. The reasons for this are attributable to health factors (low prevalence of ulcerative STDs), social conditions (social control of sexuality and the role of religion), and policy (early and signi?cant involvement of the Senegalese government in AIDS prevention). Our results con?rm some of these explanations, but they also demonstrate the relatively small role of international migration in the spread of HIV [29].

65 In summary, the migratory experience does not seem to be a major factor in the spread of AIDS at the community level. Casual sexual relations are relatively rare between the general population and international migrants and risky behaviour is not widespread in the general population of the Senegal River valley. Faced with the threat of AIDS, both at an individual and at a collective level, the international migrants’ response shows the strength of social control.

66 From the point of view of public health, the conclusions of the study invite an examination of sexual practices both in terms of the risk to individuals and of the threat to the community. In the past, the stress placed on ideas of contagion and danger was inherent to an approach based on protecting the community through the identi?cation of risk groups, and their stigmatization and possible exclusion. Later the concept of risk occupied central stage in public health debates. This progression from the notions of danger and contagion to those of risk and infection has made it possible to raise issues concerning the disease and its prevention at the individual level (Douglas and Calvez, 1990; Calvez, 2001). The notions of risky sexual behaviour and individual responsibility bear witness to this development. Yet, without contesting the ethical usefulness of such an approach, it can present the risk of not fully considering the community context in which individual behaviour is occurring.

67 The case of return migrants in the Matam region is particularly instructive. The individualist approach is certainly partially at work at the level of the migrants themselves, and particularly of the internal migrants. Even a reduced exposure to health messages undoubtedly increases the returning internal migrants’ capacity to adjust their behaviour according to indicated and perceived health risks. But faced with the threat of AIDS, the response of Matam’s international migrants obviously involves a signi?cant social dimension. Faithfulness to the spouse is therefore the most adequate response, in relation to the social context, to minimize the risk of infection. However, this strategy does not eliminate all the risks of transmission (through sex, and from mother to child) particularly within the family and the kinship group (through the practice of the levirate). In such circumstances the families of the migrants obviously appear to be most vulnerable, as women are unable to ask their spouse to use a condom or to take a test. In summary, it can be said that the returning international migrant manages the risks (health and social in particular), but that the migrant’s family bears them.

68 At the end of these analyses and considerations, it is essential to develop policies to inform and educate, not only to close a particularly glaring gap in this region of Senegal, but also to promote the emergence of responsible behaviour at the individual level as well as communal responses geared towards reciprocal protection. The success of prevention programmes depends on this articulation between individual and communal responses. Furthermore, in the context of the Senegal River valley, solidarity between members of the community must be expressed particularly with regard to migrants’ spouses, who are vulnerable and sometimes victims of exclusion at the time of widowhood. The community must therefore facilitate and relay information to improve communication within couples and negotiated risk management. Desired and shared knowledge of spouses’ serological status could, in this perspective, enable women to initiate a process of negotiation necessary for the establishment of preventive behaviour within the couple.