An Origin and Destination Perspective on Family Reunification: The Case of Senegalese Couples

Abstract

European societies are expressing growing concern about the consequences of migrant family reunification on their soil for the management of their borders and the success of the integration process. Many policy makers assume that most migrants intend to bring their relatives to Europe as soon as possible, and argue that it might be difficult for reunified migrants to integrate into their host societies. Our results concerning the process of reunification of Senegalese couples in France, Italy, or Spain strongly challenge this view. Using MAFE (Migration between Africa and Europe) data with a life event history approach, we show that (1) separation is often a long-lasting situation among Senegalese couples; (2) separated couples do not only reunify in Europe but also quite commonly in Senegal; (3) the couples who reunify in Europe tend to be those who adapt most readily to the European culture and economy.

1 Introduction

Family reunification has become a major policy concern across Europe in the last decades, as illustrated by the growing legal restrictions implemented since the late 1970s (OECD 2001; Kraler 2010). This concern is two-sided. First, there was and there still is a fear related to quantity: family reunification is viewed by most policy makers as a powerful immigration multiplier, every migrant being systematically seen as a reunifier who intends to bring his/her whole family to Europe. Secondly, in more recent years, the debate about family reunification policies has shifted from the size of the flows to their composition. The ability of reunified immigrants to successfully integrate into their host societies, both economically and culturally, is now at the forefront of the debate (Grillo and Mazzucato 2008). Migrants admitted on the basis of family reunification are assumed to be mostly dependents, i.e., out of the labor market, who will increase the dependency ratio among the immigrant population. Furthermore, it has become commonplace to blame family reunification for the increasingly inward-looking behavior of immigrant communities and their failure to integrate into the host societies. Yet, most of these preconceptions and the policy measures inspired and justified by them are not grounded in solid empirical evidence. The fact that family reunification has become the leading legal mode of entry into Europe simply reflects the fact that most European countries stopped labor migration in the 1970s. It does not imply that all migrants are reunifiers. In fact, there is a dearth of adequate data to analyze both the potential for family reunification and the integration outcomes of immigrants admitted on family and non-family grounds (Kofman 2004; González-Ferrer et al. 2012).

Using a unique dataset that contains biographical information that is transnational (i.e., based on surveys carried out both in Europe and in Africa) and that includes both documented and undocumented migrants,¹ this paper intends to bring a new perspective on family reunification by conceptualizing family reunification from a dual viewpoint (considering reunification not only in Europe, but also in Africa) and as a dual selection process (not only by the host countries, but also by the migrants themselves). In this perspective, Senegalese migration is taken as an empirical case study to provide new quantitative evidence, in a field where the socio-anthropological approach has so far been predominant. Senegalese migration represents a critical and highly relevant case since Senegalese families, and Sub-Saharan African families in general, have been a clear target of restrictive family reunification policies due to their larger size, their multigenerational structure and their relatively frequent practice of polygamy.²

Building on previous theoretical and ethnographic studies, we argue that not all migrants (and their relatives) wish to reunify as soon as possible; that reunification in Europe is not their only option, some of them preferring to re-join their relatives back in the home country; and that reunification in Europe is—in fact—the choice of the most integrated/integratable ones. Accordingly, we test two hypotheses: (1) Couple separation is likely to work as a long-lasting arrangement for many African migrants, especially if partners can visit each other frequently. (2) Reunification in Europe is more likely when men are economically and culturally integrated, and women have the potential to adapt to the receiving context (more education, higher occupational skills, childlessness, etc.). The article is organized as follows: Sect. 2 provides a literature review that sheds light on the rationales of family reunification by bringing together three separate strands of research (economic theories on migration decision-making, transnational and socio-anthropological approaches to the study of migration and African family dynamics, and previous quantitative research in this field). Next, we present in detail the MAFE survey, its advantages for this kind of study, and the methods used in our analyses. Finally, we summarize and discuss the results in Sect. 4 and conclude in Sect. 5.

2 Literature Review

2.1 Family Reunification as a Dual Selection Process

Receiving countries undoubtedly play a major role in the selection of both migrants who can reunify (the “reunifiers”) and those who join their relative(s) at destination (the “reunified”). However, as Borjas pointed out, they can only select from the available self-selected pool of applicants, i.e., those who already want to migrate (reunify) abroad (1990). Individuals, regardless of whether they are migrants or their “left-behind” relatives, are also actors of selection. Besides, they may decide to ignore government restrictions and reunify de facto, without following the legally established procedure.³ Moreover, whatever the mode of entry they chose, they also decide the timing and the place of reunification.

In sum, family reunification is thus a dual act of selection (Bledsoe and Sow 2008) in which at least two types of selection processes overlap: on the one hand, a process derived from the costs and constraints that immigration policies impose on individuals willing to migrate and, on the other, a process related to the families’ decision to reunify. Unfortunately, we still know very little about these two processes and how they interact. Although some studies have examined family migration decisions, they are mostly restricted to the experience of internal migrants and focus on family separation (Sandell 1977; Mincer 1978; Courgeau 1990; Mulder and Wagner 1993; Stark 1988) rather than on the process leading to family reunification after a period of physical separation due to international migration. As Grasmuck and Pessar (1991) brilliantly illustrated in their analysis of Dominican international migration, negotiation among household members concerning who migrates first and who follows—if someone actually does—is far from being straightforward. The living arrangements adopted by families involved in international migration do not necessarily reflect each individual’s preferences but also their bargaining position within the household or family, and hence the normative context that structures their roles, at least loosely, in generational and gendered ways (Grasmuck and Pessar 1991; Hondagneu-Sotelo 1992, 1994, 1999).

2.2 Family Reunification only Partially Addressed in Economic Theories of Migration

There is currently no consistent theoretical model accounting for family-related migration. Family reunification has not been explicitly addressed by either Neoclassical Economics (NE) or New Economics of Labor Migration (NELM). It is, however, possible to derive some predictions about reunification behaviors from these two theoretical strands. The NE approach conceives migrants as “income maximizers” who will stay abroad as long as the expected income differentials between their country of origin and their country of destination persists (Todaro 1976). Accordingly, they are expected to endure relatively long separations from their relatives left behind, until proper arrangements can be made for family reunification in the country of immigration (i.e., obtaining a stable job, being able to pay for a larger apartment, etc., which takes some time). Due to its strongly individualist approach, the reasoning provided by the NE model neglects the possibility that reunifying with the partner at destination may contribute to the couple’s income maximization if this second partner is willing to work as well. Similarly, the possibility of reunification back in the origin country is not considered either, since the economic reasons underlying the migration decision (international wage differentials) are difficult to remove and, consequently, return would not be a rational decision.

In contrast to this, the NELM clearly familized the international migration decision by placing individuals in the larger context of their households and considering the role played by different household members in providing for the family. In addition, Stark (1991) conceived international migration as a household strategy designed to minimize the economic risk deriving from a variety of market failures, rather than as an individual strategy to maximize income. By placing different household members in different countries where employment conditions are weakly or not correlated, families can diversify the sources of risk and better provide for their economic well-being. Thus, the international migrant envisaged by NELM is a “target-earner” migrant who will return to origin as soon as the macro-economic context, and his/her own economic situation allows him/her to successfully surmount the economic risks faced by the household. Accordingly, partners’ reunification at destination makes little sense in the context of NELM, unless the sponsored partner is also willing to work there. In such a case, reunification takes place in order to reduce the number and duration of trips and to increase the probability of return (instead of permanent settlement), by enhancing the household’s ability to meet a given earnings/savings target (Constant and Massey 2002).

Couple reunification at destination can thus be partially explained by both theoretical frameworks, although they interpret reunification as an indication of opposing residential intentions: the couple’s reunification suggests a clear movement toward permanent settlement at destination for NE’s income-maximizing migrants, versus an attempt to accelerate return to the home country for NELM’s target-earners. Moreover, conditional on his/her willingness to work at destination and for the same reasons, the partners’ reunification is expected to be quicker among target-earner migrants than among income-maximizing ones.

2.3 Empirical Evidence on Migrant Couple Formation, Separation and Reunion

NE and NELM theories both suffer from two main limitations that new approaches on transnationalism have repeatedly emphasized (Bryceson and Vuorela 2002). First, both remain largely constrained by a form of “methodological nationalism” (Wimmer and Glick Schiller 2002) and restrict their analyses to either the country of destination or the sending one, and consequently tend to neglect the possibility of repeated movements back and forth by different family members as a strategy to sustain cross-border livelihoods (Sørensen and Olwig 2002). In fact, the best illustrations and analyses are provided by recent socio-anthropological research in the area of transnational families which shows that living apart-together across countries is a rational and even functional strategy for migrant families.⁴ In some particular cases, such as Senegalese separation, transnational arrangements often appear to be a choice rather than a state of affairs forced upon migrants who, while finding themselves in a vulnerable economic and legal position in Europe, are anxiously waiting for the opportunity to reunify in Europe (Bledsoe and Sow 2008; Bledsoe 2008; Riccio 2001; Rodríguez-García 2006).

Secondly, studies inspired by transnationalism have been particularly successful in highlighting how mainstream economic explanations of migration decisions remain largely gender and culture-blind. Sociologists and demographers, however, have also increasingly emphasized the gendered nature of migration and family dynamics among migrants in quantitative studies. A good example of this can be found in the growing number of analyses of marriage migration in different European countries, based on the challenging idea of Lievens (1999) that Moroccan and Turkish “women may marry an imported partner in order to satisfy modern goals” (p. 717). While the decision of male migrants to choose a female partner who still lives in the country of origin may well be attributable to the strong sex imbalance that affects the first generation of some ethnic groups in Europe, plus a wish for “unspoiled” traditional wives, the same reasoning does not apply so easily either to female migrants or to second generation young adults. In fact, the empirical evidence available so far tends to support the hypothesis that marriage migration mostly entails a traditional option among immigrant men, as suggested by the negative and significant relationship between men’s educational level and their probability of importing a partner instead of marrying a native or a co-national immigrant already at destination (Lievens 1999; Celikaksoy et al. 2003; González-Ferrer 2006). In contrast, most studies have found a positive, although generally non-significant relationship between the educational level of immigrant women and their propensity to import a partner from their country of origin (Hooghiemstra 2001; Celikaksoy et al. 2003; González-Ferrer 2006: Kalter and Schroedter 2010).

The limited value of education to explain why some immigrant women choose transnational partners instead of co-national immigrants already at destination has led some authors to analyze the potential influence of the social group of reference in some ethnic groups, and also the potential benefits to be derived by the woman from this marital strategy in terms of bargaining power within the couple—because she is the one who already knows the country where the couple will settle—as the two main mechanisms explaining the importation of partners by female immigrants in Europe (Kalmijn 1998; González-Ferrer 2006).

In the case of family reunification of partners who were already in a couple before migration, some studies have also found different effects of education for men and women in explaining the likelihood and length of spouses’ separation due to international migration. Gupta (2003) found that more educated Mexican women are also more likely to migrate with their husbands to the US than being left behind. In addition, for couples who have experienced at least one spell of separation, the duration of separation tends to be shorter if the husband’s educational level is high.⁵ She conjectured that education improves women’s status overall and results in a more egalitarian partners’ relationship, which in turn might make wives more likely to insist in migrating with their husbands or, alternatively, to succeed in persuading them not to migrate at all. González-Ferrer (2007) also found that more educated women are likely to join their husbands in Germany more quickly than less educated ones, whereas having a husband with more years of education substantially increases the odds of joint couple migration but does not significantly affect the pace of the spouses’ reunification at destination.

2.4 Family and Migration Dynamics in Senegal

In Senegal, as in almost all sub-Saharan societies, the basic social unit is some form of extended family. According to the latest Census of 2002, the average size of Senegalese households was 9.1 persons, which is not surprising given that, among other things, polygamy is permitted and relatively common (25 % of all marriages are polygamous, according to the Senegalese Census of 2002) (Vázquez Silva 2010). In addition, after marriage, the wife moves to the house of her husband’s family in the majority of cases, especially in rural areas, where she will share the household chores and caring tasks with other women of the family—maybe other co-spouses if the husband is polygamous, or her new sisters-in-law (Poiret 1996).

Marrying and moving to the husband’s parental home do not necessarily imply a great deal of intimacy between the spouses, at least not in the Western way. According to Findley, in much of sub-Saharan Africa, men and women take their meals separately, rarely socialize together, and have marriages where the level of conjugal interaction is quite low (Findley 1997, p. 121). Indeed, in the Senegalese traditional family model, being in a couple does not necessarily imply living together in the same place. In Africa, quite a high proportion of spouses live far apart for relatively long periods (from 3 to 7 years), frequently as a result of intense internal migration aimed at diversifying sources of income and risk across several locations. Findley estimated that between 43 and 68 % of couples in Senegal experience this situation at some point in their lives (Findley 1997, p. 125). In the urban context and where polygamy is frequent, living in different dwellings even when both spouses reside in the same locality is not rare.

In any case, and regardless of migration, in the most extended traditional family system, once the newly wed wife moves into “her” new household, she comes under the authority not only of her husband and other older men in his family, but also of the older women, especially her mother-in-law, for whom she becomes a care-giver (see more details in Vázquez Silva 2010). In view of this relationship between daughters-in-law and mothers-in-law, as well as the risk that remittances sent by young male migrants in Europe to their parental household in Senegal will decline if their wife (and children) joins them abroad, it is quite rational for the elders in the family to oppose to any form of “family reunification” in Europe. In addition, the dominant European way of living, as well as the increasingly restrictive immigration policies that have stigmatized large African polygamous families and their living arrangements, has also probably reinforced this view that reunification in Europe is a sub-optimal choice. Thus, transnational couples as a relatively stable living arrangement for Senegalese individuals involved in international migration to Europe may prove to be a totally rational and desired outcome for many of the male migrants currently “living alone” in Italy, France, or Spain.

In Senegal, as in most countries, the dominant type of family organization that we have just described, as well as the family values associated with it, is continuously challenged by groups and individuals who either belong to minority ethnic or religious groups different from the largest one, or who possess larger resources that allow them (at least partially) to deviate from the social norm. In our case, as we said, the previous description corresponds to the most traditional family model among the Wolof, the largest ethnic group in the region of Dakar and among Senegalese migrants in Europe. Yet, the Serer and the Diola groups, for instance, are known to have traditionally followed a lineage system with some matrilineal elements (Dupire 1977; Sy 1991), which probably gives women more bargaining power within their couple and families at large, and will thus modify some of the behaviors described above. For the same reason, men with greater economic stability and women with higher levels of education are clear candidates for deviating from the dominant behavior regarding family reunification, because resources in general provide individuals with additional opportunities to successfully circumvent social pressures from their peers and, in addition, will make them more “integratable” from the point of view of Western countries.⁶

3 Data and Methods

The objective of this article is to study the timing and the determinants of reunification among Senegalese couples who at some point lived geographically apart (in different countries) because one of the partners had migrated to Europe. We call these couples “transnational couples”: they are “living apart together” across borders because they are physically “separated” but may maintain their emotional or legal ties (marriage), economic exchanges, frequent visits, family obligations, common children, etc.⁷ “Reunification” occurs when partners start living in the same country (not necessarily in the same dwelling) after a period of transnational partnership. Thus, reunification is not restricted to the legal procedure; we analyze “de facto” reunification, which takes a variety of forms: reunification in Senegal or in Europe, in accordance with the legal procedure specifically established for that goal or via any other available channel, including irregular migration.

A proper analysis of this kind requires information on individuals who are both at origin (in Senegal) and destination (in Europe). In addition, multi-level time-varying information is also needed in order to characterize the couples—and their members—not only at the time of the survey but also at the time of reunification (or just before), and the same information is required for those couples who have not reunified yet. There is thus a need for a transnational sample that includes partners who formed transnational couples at some point during their lives, some of whom have reunified (either in Europe or in Senegal), while others are still separated at the time of the survey.

3.1 The MAFE-Senegal Survey

Few datasets present the features that are needed to study the determinants of family reunification. The MAFE-Senegal survey is, for two reasons, one of the rare quantitative sources that allows such analyses. First, it comprises a transnational dataset containing individual data collected both in European countries and in Senegal using identical questionnaires. A total of 603 Senegalese migrants were surveyed in France, Italy, and Spain⁸ (about 200 in each country), and 1,067 persons were interviewed in the region of Dakar (including 197 returnees and 101 migrant’s partners at the time of the survey in 2008).⁹ Second, the data are time-varying by nature, since they result from individual life-histories recorded in biographical questionnaires. The questionnaire was designed to collect retrospective biographical information on a yearly basis from birth until the time of survey (2008), for each sampled individual, whatever his/her country of residence at the time of the survey. The data collected include a wide range of information on respondents’ migration and occupational histories, as well as on their family history (children, partnerships). Interestingly, the questionnaire includes an entire module on the international migration history of the respondent’s relatives (including his/her current and past partners), international migration being defined as a stay of at least 12 months outside Senegal. This 12-month threshold also applies to couple separation and reunification: a separation or a reunion lasting less than 12 month is not considered in our analyses.

3.2 Survey Sample

The MAFE survey sample is made of individuals aged between 25 and 75 at the time of the survey, born in Senegal and of present or past Senegalese nationality. Varied sampling methods were used to select the individuals. In Senegal, a stratified probabilistic sample was drawn. In Europe, several sampling methods were used. In Spain, the municipal register (Padrón) offered a national sampling frame from which documented and undocumented migrants could be randomly sampled. Respondents in France and Italy were sampled through varied non-probabilistic methods (e.g., snowballing, intercept points, contacts obtained from migrant associations) in order to fill pre-established quotas by sex and age (Beauchemin and Gonzalez-Ferrer 2011; Schoumaker et al. 2013).

Even though the MAFE survey offers a unique opportunity to study family reunification, the survey was not specifically designed for this purpose. For this reason, the analyses carried out in this article rely on a sub-sample of only 459 individuals that are or were previously in a transnational couple.¹⁰ How were these individuals and transnational couples selected? First, the respondents (male or female) had to have been in a transnational union for a period of at least one year, married, or otherwise,¹¹ at some point in time (i.e., at the time of the survey and/or in the past). Second, the couple had to comprise a woman living in Senegal and a man living in France, Spain, or Italy during the period of separation¹² (see Table 1, especially cells with bold characters). This last restriction unfortunately prevents us from analyzing emerging couple arrangements in which the female is the pioneer partner (42 cases, see Table 1), and the male the one left behind in Senegal (25 cases). But numbers were too small to allow for specific analyses, and priority was given to the constitution of a homogeneous sample to facilitate interpretation of results. For the same reason, cases that involve destination countries other than France, Italy, or Spain were also eliminated from the analysis sample (Table 1). Finally, we used a sample of 546 transnational couples (restricted to their first spell of separation),¹³ for whom the data were obtained either from men interviewed in Europe (347 cases) or from women surveyed in Senegal (199 cases).

Table 1 Number of transnational couples in the MAFE-Senegal survey by respondent’s sex and country of residence (non-weighted)

To take the changing characteristics of the couples (and of the partners themselves) into account, the data were arranged as a couple-year dataset in which each couple appears when it becomes transnational for the first time (i.e., when the man migrates out of Senegal, leaving behind his wife, or when the partners start their relationship while living in separate countries) and disappears when it stops being transnational, either because the couple reunifies (in Europe or Senegal), or because it breaks up (separation, divorce, widowhood), or—by default (censoring)—at the time of the survey (see Table 2 for a detailed account of these outcomes). Each year of life of a transnational couple is thus a line in the dataset and is considered as an observation in the analyses. The analytical sample is thus extended from 546 couples to 3,742 couple-years.

Table 2 Outputs of the first period of separation of transnational couples (analysis sample, weighted data)

Thanks to the longitudinal nature of the MAFE data, the variables describing the partners in the dataset may change every year. However, a major constraint of our analysis sample is that it contains asymmetrical information on the partners: the dataset contains a wealth of variables describing the interviewee at any point in time (his/her entire history of family formation, education and occupation, migration experience, etc.), but much less information describing his/her partner (only six variables: age, country of birth, nationality, couple status, educational level, and socio-economic status, and only at the time when the couple first formed). Additional variables are available to describe the couple itself: whether it started as a transnational partnership (i.e., the partners started their relationship while living in different countries, which is the case for 50 % of the total sample, see Table 3); whether the couple is part of a polygamous family at any moment (i.e., whether the male has several partners, or whether the female has co-wives), and the number of children at each point in time.

Table 3 Description of the sample of transnational life of each couple (weighted data)

3.3 Methods

Using the longitudinal nature of the MAFE data, we computed discrete-time survival functions of the time between the start of the couples’ transnational life until their reunification. Couples still separated at the time of the survey, who had divorced or been widowed were treated as censored when the first of these events took place. “Pseudo-survival” functions¹⁴ (Blossfeld and Rohwer 2002) were computed to account for two possible outcomes (Fig. 1): couple reunification in Senegal (when the male partner returns to Senegal) or couple reunification in Europe (when the female partner migrates to join her husband in Europe).

Fig. 1

Probability that the couple is separated, according to the duration of separation and the place of (re)unification

Second, we performed multivariate discrete-time event history analyses to study the determinants of couple reunification. Since our objective is to analyze the factors explaining reunification either in Europe or in Senegal, we applied multinomial logistic regressions in a competing risk analysis that distinguishes between both destinations, including a set of explanatory variables describing both the individuals and the couples they form. Models are specified as follows (Yamaguchi 1991):

$$ \log \left[ {P_{riy} /\left( {P_{siy} } \right)} \right] \, = \alpha_{r} + \, \beta 'X_{riy} , $$

where P _riy is the conditional probability that transnational couple i experiences a first reunification either in Europe or in Senegal (the place being denoted by the subscript r) versus remaining separated (denoted by the subscript s) at year y, given that reunification has not already occurred. α is a constant term, and X _riy a vector of individual, couple, and contextual covariates (including the baseline hazard function), with β denoting the value of the estimated coefficients of the model for each variable.¹⁵

Following the same technical specifications, two multinomial models were run to test the effects of various characteristics of the couples and of each of their members. As mentioned above, several variables were only available (or relevant) for one member of the couple. Accordingly, in Model 1, we investigate whether the fact that a woman lives with in-laws has any effect on the odds of reunification, while in Model 2, we control for several variables pertaining to the man (whether he holds a work or residence permit, whether he reports a bad financial situation, and whether his father decided or paid for his migration). Table 3 presents some descriptive information on the explanatory variables included in the analyses and specifies whether they are time-varying or time-constant and for whom they are available (the respondent and/or his/her partner). For instance, the “ethnic group” variable refers to the respondent’s first answer on his/her ethnic group. Based on ethnographic information (Dupire 1977), we classified the main ethnic groups according to the degree of patrilineality. “Weakly patrilineal” groups include mainly the Serer and Diola; “strongly patrilineal” groups include the Wolof, Mandingue, Soninké, and Bambara. All results presented are weighted to account for the different sampling probabilities in each of the countries in which the survey took place. Details on the sampling strategy can be found in Beauchemin (2011).

4 Results

As Fig. 1 clearly illustrates, couple reunification appears to be relatively uncommon among Senegalese migrants, be it at destination, i.e., in Europe, or in the country of origin, i.e., Senegal. Approximately 60 % of migrants had not reunified with their partners after 10 years of separation. According to our multivariate models, the probability of reunification seems to be somewhat higher in the 3 initial years of separation and to decline afterward; nevertheless, in most models, the probability does not vary significantly with the duration of separation (see Table 4).¹⁶ Although for a small group of migrant couples, reunification occurs relatively quickly, most couples seem to endure rather long separations. This result is consistent not only with theoretical perspectives that emphasize family dispersion as a functional strategy in order to diversify the household’s sources of income and risk (see above), but also with the anthropological literature that highlights the importance of complex family structures in Sub-Saharan Africa and especially in Senegalese culture. Furthermore, this pattern of long separations as an apparently stable family arrangement contrasts with the behavior of other migrant groups in countries like Germany or Spain who join their partners at destination much more quickly (González-Ferrer 2007, 2010).

Table 4 Model results

Figure 1 reveals a second important finding: couple reunification in Senegal is almost as likely as reunification in Spain, Italy, or France (these “pseudo survival functions” for each destination are not significantly different). This result challenges the widely held belief that family reunification is very intense and only possible at destination and highlights the importance of return migration. We will see below, however, that the determinants of reunification differ across locations. Overall, these results suggest that couple reunification is an important component of the migration dynamics between Senegal and Europe, rather than of migration from Senegal to Europe only. It is, therefore, crucial to understand the choices of both members of these couples, including whether and where they reunify.

The duration of separation may depend on the role of visits that partners make to each other. These visits seem to be quite frequent, as they involve 20 % of the sample’s person-years. In fact, it is the partner living in Europe who is almost always the visitor: 27 % of the men living in Europe visited their partner at home at least once in the first year of separation, while only 2 % of the women left behind in Senegal visited their partner in Europe (Table 3). Obviously, visits are facilitated by holding a work or residence permit in the relevant European country, which is the case for two-thirds of the men living in France, Italy, or Spain in our sample. Even though frequent visits might suggest a stronger wish to be together, our results indicate that having visited the partner during the previous year substantially reduces the odds of reunification. There appears to be a strongly negative link between reunification in Senegal and visits, since the odds are significantly reduced (0.07) in Model 1 (Table 4). These results are consistent with stable transnational arrangements, for which visits, together with remittances, “(…) help to oil the functioning of the split families” (Grillo and Mazzucato 2008).

Another striking finding is that as many as half of the couples started their union while living in different countries (duration of union 0 years at start of separation; see Table 3). This is related to the fact that most of our sample involves young adult years, in which both migration and partnership formation take place. Furthermore, many Senegalese migrants living abroad seek their partners in Senegal, not in their country of residence. Marriage migration may be connected to the strong gender imbalance in the Senegalese migrant population living in Europe, and the fact that few women migrate independently (Vause and Toma 2012). It also suggests that having migrated facilitates union formation with someone living in Senegal, as already mentioned in qualitative studies (Mondain 2009). Our multivariate results show that starting a union at a distance has no significant impact on the chances of reunifying compared to couples who had previously lived together.

In all the models computed, the educational level provides strong and statistically significant coefficients. Thus, the odds of reunification strongly increase with the women’s educational level (Models 1 and 2, Table 4): for women with at least a secondary level of education, the odds ratio is 2 to 4 times higher—depending on the place of reunification—than for women with primary level or less. The results for men in Model 2 show somewhat weaker effects of education; compared to less educated men, the odds ratios for men with tertiary education are 1.5 and 1.7 for reunification in Senegal and Europe, respectively. This result is highly consistent with our general hypothesis that individuals more economically or culturally integrated into European society will be more likely to reunify, in particular in a European country. Education is an indicator of availability of personal resources as well as a sign of potential cultural assimilation to Western culture and European languages. In fact, it is very difficult for individuals with no or little formal education (some of whom may be illiterate) to successfully integrate into European labor markets and find legal employment, so many of them work in the underground economy (Reyneri 2006). The jobs available to them may not only be restricted to just a few occupations (such as agricultural laborers or street vendors), but also very precarious, with few opportunities to enter the formal labor market or to obtain a residence and work permit. For women, formal education may imply more gender equality in the couple and, perhaps more importantly, greater chances of finding employment that may help to cover the high cost of living in Europe. According to MAFE survey data, in 2008, nearly two-thirds of interviewed Senegalese partnered women living in Italy, Spain, or France held a job; about one-fourth reported “homemaker” as their main occupation; and 4 % reported being unemployed. By contrast, 95 % of partnered men living in these countries reported being employed and 5 % unemployed.

The results concerning educational level are consistent and complementary to those of the “socio-economic status” variable. In particular for men, the higher the occupational status, the higher the probability of reunification. This result holds both for reunification in Senegal and in Europe. Men holding jobs in skilled occupations and in professional/employer occupations show reunification odds that are 1.5 times higher than men in unskilled occupations or who are self-employed without employees (Model 1, Table 4). This variable indicates a greater ability to afford the costs of the partner’s migration and expenses while in Europe, as well as better economic prospects and economic integration. It is interesting to see that for men living in France, Italy, or Spain who are unemployed or inactive (mainly students), the odds of reunifying are 3.3 times higher than for men in unskilled jobs; however, these higher odds are entirely due to the much higher probabilities of returning to Senegal and reunifying with their partner there (Model 1). The importance of economic resources is confirmed by the results of the variable “bad financial situation”¹⁷ that yields strongly negative and significant results for reunification in Senegal (odds of 0.1 with respect to individuals with a good financial situation; Model 2) while also negative but insignificant results for reunification in France, Italy, and Spain. Taken together, these results clearly demonstrate that men’s availability of economic resources is crucial for couple reunification, both in Senegal and particularly in Europe. The results of the socio-economic status variable for women provide non-significant coefficients for reunification in Senegal, but in contrast, holding a skilled or professional occupation doubles the odds of reunification in Europe. For women with a good job in Senegal, the opportunity costs of leaving their employment are high, although they probably have a higher chance of obtaining a job in Europe, thus reducing the costs of remaining there (Model 2). These results can be related to the generally subordinate economic position of women in the couples; they also highlight the importance of economic and social integration in Europe for reunification.

Although having a residence or work permit does not guarantee entitlement to legal family reunification with immediate relatives from the home country, it is obviously a first step toward it. Accordingly, it is not surprising that the results of the “permit” variable show that having a permit multiplies by seven the chances of reunification in Europe, while it has no significant effect on reunification in Senegal. Consistent with the cultural and legal restrictions on polygamy in European societies, the odds of reunification for this type of family arrangement are especially low in Europe (odds ratio 0.6), but it has no significant effect for reunification in Senegal. This result suggests that polygamy tends to facilitate long periods of couple separation, since it may be negatively correlated to the emotional and social closeness of partners. Similarly, the presence of children, usually left behind in Senegal with their mother, greatly speeds reunification in Senegal (odds of 2.9), but reduces the odds of reunification in Europe (0.7). In both instances, the presence of a large family in Europe would be at odds with traditional family arrangements and would increase the costs of staying there, in contradiction with a “target earner” type of migration. Women living with in-laws in Senegal have a much lower probability of reunification (odds of 0.03 for reunification in Europe, with respect to women who do not live with in-laws), which again provides evidence that an arrangement that fits the traditional organization of family roles and is at odds with a conjugal type of family delays family reunification.¹⁸ Living with in-laws is a feature of extended families characterized by patrilocality, a strongly hierarchical system in which elders and men hold the power, and where the roles of women are bounded to the domestic sphere. In particular, qualitative studies have shown that women are expected to care for their in-laws, a requirement that restricts migration and more generally their freedom to pursue individual goals (Vázquez Silva 2010). We included a variable on the ethnic group which shows that where patrilineal practices are less strong, e.g., among the Serer and Diola, the probability of reunification in Europe increases significantly (odds ratio 1.35) with respect to the clearly patrilineal groups, such as the Wolof. Additionally, the variable “men whose father decided or paid for migration” could also indicate a strong vertical hierarchy in the family and subordination to the interests of the extended family. Consistent with this interpretation, this variable does have a negative effect on reunification, albeit with no significant coefficients. Finally, we included the “religion” variable which shows that members of the Mouride brotherhood have lower probabilities of reunification than other Muslims, irrespective of place of reunification. Odds of reunification in Europe about twice as high for Christians as for “other Muslims” (Model 2).¹⁹ Overall, these results are consistent with our hypothesis that individuals displaying behavior more distant from European culture will favor transnational living apart-together arrangements.

5 Conclusions

The MAFE data offer a unique opportunity to present new evidence on the process of family reunification among African migrants in Europe. Thanks to its transnational nature, this dataset allowed us to overcome the usual “methodological nationalism” often criticized in the recent literature on migrants’ families. Furthermore, its longitudinal nature allowed us to study the determinants of couple reunification via a life course approach.

As expected, couple reunification appears to be relatively uncommon among Senegalese migrants, be it at destination or in the country of origin. 10 years after separation (due to migration), approximately 60 % of migrants had not (re)unified with their spouses. Although for a small group of migrants, couple reunification occurs relatively quickly, most couples seem to endure rather long separations. This result is not only consistent with theoretical perspectives that emphasize family dispersion in order to diversify the sources of income and risk, but also with the anthropological literature that highlights the importance of complex family structures in sub-Saharan Africa, and especially in Senegalese culture. Our results also reveal a second important finding: couple reunification in Senegal is almost as likely as reunification in Spain, Italy, or France. This result challenges the widely held belief that family reunification is very intense and only possible at destination and highlights the importance of return migration.

In addition, the model results generally support the hypothesis of increasing likelihood of reunification in destination countries with increasing economic and cultural integration and/or potential adaptability of both partners in Europe. For instance, men with secondary or tertiary education, i.e., those with a closer affinity to western culture, are more likely to reunify in Europe than in Senegal. On the contrary, less educated women (which may mean illiterate and with a lesser command of European languages) are more likely to reunify in Senegal, with the return of their partner, than in Europe through their own migration. In line with our integration hypothesis is the result that individuals holding a residence or work permit are also more likely to reunify in Europe. The variables indicating the “traditional” character of the couple (in contrast to a conjugal family type) is also consistent with our hypothesis. For polygamous couples, the odds of reunification of are much lower in Europe than in Senegal, which are consistent with the cultural and legal restrictions on this type of family arrangement in European societies. Moreover, couples who are enmeshed in strong “traditional” networks at origin (e.g., with the left-behind woman living with her in-laws) have a much lower probability of reunifying at destination.

In short, contrary to common wisdom, family reunification is generally more likely to occur in Europe among people able to integrate easily into the host society. To what extent is this due to state polices or to a self-selection of migrants and their families into reunification? Even though our data do not provide any clues about the selection process in government administrations, our results tend to suggest that both levels are at play. Receiving country selection criteria typically include living conditions (earnings, housing) and legal requirements (having a residence permit, being married, not being polygamous). Our models include these variables (or some proxies) and show that they play an expected role. Interestingly, other variables that are also related to integration prospects in Europe also have significant results (education, “traditional” family). Their effects, net of the country selection-related variables, suggest that migrants and their families also tend to self-select according to their integration potential. Their knowledge of the European economic situation and social context (hostility toward Sub-Saharan migrants) seems to negatively influence their probability of reunifying in Europe. Indeed, family reunification appears to be a dual selection process. Further research is needed, however, to better identify the impact of policy orientations on migration behaviors, to identify how decisions are taken within families and to understand the bargaining process that leads to the choice of reunification here or there or the choice of maintaining a transnational way of life.