Social Position and Migrant Networks in International Migration from Africa to Europe

Abstract

Ample prior research shows that social capital is contingent on a person’s position in society and is, consequently, a significant factor in perpetuating and amplifying social inequalities. In contrast, migration scholarship is relatively quiet about how social position may stratify the role of migrant networks, and instead conceptualizes migrant networks as broadening access to migration. This paper integrates theoretical insights from these two lines of research to offer a novel contribution on the interplay between social stratification and migrant networks. We examine three pathways—network access, network mobilization, and network returns—through which social position may shape migrant networks and, as such, reinforce inequalities in migration. To do so, we employ retrospective data from the multi-sited Migration between Africa and Europe survey that allows us to account for the dynamic nature of migrant networks and to distinguish among these pathways. Our study highlights significant stratification in access, mobilization, and returns to migrant networks among individuals in different social positions, operationalized as educational attainment. In a context of positive educational selectivity such as sub-Saharan African migration to Europe, access to migrant networks increases substantially with higher social position. And while lower-educated individuals rely more financially on migrant networks, which offer them larger relative gains, these networks ultimately exacerbate initial advantages and amplify social inequalities in migration opportunities.

Introduction

Sociologists have long argued that social capital—the resources embedded in social networks—leads to better outcomes across a range of life domains such as education, labor market, and health, among others. They further posit that the role of social capital is contingent on a person’s position in the social structure (Lin 2002; McDonald et al. 2013). Empirical studies consistently reveal that individuals with higher socioeconomic status (SES)—typically operationalized as parental and one’s own level of education—possess broader, more diverse, and more influential social networks (van Tubergen and Volker 2015) and derive greater benefits from them in terms of educational achievement (Zwier and Geven 2023), wages or occupational prestige (Granovetter 1983; Pichler and Wallace 2009; McDonald et al. 2013), or health (Aartsen, Veenstra, and Hansen 2017). An important corollary is that social capital plays a pivotal role in perpetuating and exacerbating socio-economic inequalities within society, enabling advantaged individuals to maintain or enhance their positions (Bourdieu 1986; Lin 2002; DiMaggio and Garip 2012; Jackson 2021).

In research of migration, migrant social capital—the resources embedded in migrant networks—emerged as a critical factor facilitating international migration (Massey and Espinosa 1997; Garip and Asad 2016). Yet the migration scholarship has remained conspicuously silent with respect to how social position stratifies the role of migrant networks, with few exceptions (Garip 2008; Garip and Curran 2010; McKenzie and Rapoport 2010). Furthermore, in contrast to social capital scholars, migration researchers have tended to conceptualize migrant networks as expanding access to migration opportunities (Massey 1990; McKenzie and Rapoport 2010). According to cumulative causation theory, as migrant networks grow within an origin community, the costs and risks of migration decrease, making it more accessible to individuals who hold lower socio-economic status (SES). Instead of exacerbating socio-economic inequalities in access to migration, migrant networks would thus lower them. However, the empirical evidence is so far limited and produced mixed results (Orrenius and Zavodny 2005; Garip and Curran 2010; McKenzie and Rapoport 2010; Neubecker, Smolka, and Steinbacher 2015).

The apparent contrasts between studies of social capital and studies of migrant networks highlight the need for a systematic examination of social stratification and migrant networks. Surprisingly, research on social stratification and migrant networks is sparse, despite significant advancements in social capital scholarship—most often focused on education, labor market, and health outcomes—and a strong body of research that demonstrates the importance of migrant networks in international migration. Indeed, little is known about whether and how social position shapes the role of migrant networks in migration. Also understudied is how migrant networks may, in turn, contribute to socio-economic inequalities in migration propensities.

The major contribution of this paper is to bridge these two, thus far separate, strands of literature: social capital scholarship and migration research on the role of migrant networks. More specifically, we integrate key theoretical insights and findings from research on social capital and social networks into migration scholarship to examine how social position shapes the role of migrant networks. We build upon seminal works by Lin (2000, 2002), Smith (2005), and Pedulla and Pager (2019) and delineate three pathways through which we anticipate migrant networks to be shaped by social position and, in turn, reinforce socio-economic inequalities in migration opportunities: (i) differential access to migrant networks, (ii) differential mobilization of migrant networks, and (iii) differential returns to migrant networks.

This study examines how each of these pathways is conditioned by the social position of prospective migrants, primarily measured by their own and their fathers’ educational attainment. It addresses whether conclusions drawn from social capital studies in education, health, and labor markets apply similarly to high-cost, long-distance international migration. If this is the case, then lower-SES individuals would have less access to and less benefit from migrant networks in terms of their migration chances, thereby exacerbating social inequalities in migration opportunities. Conversely, in line with findings from migration scholarship, migrant networks might partially compensate for limited human and financial capital, resulting in greater reliance on and benefits from such networks among lower-SES individuals.

Second, we contribute to previous research by using detailed time-varying data on personal migrant networks. This approach enables us to better account for the dynamic nature of networks and to measure direct relationships with more distant ties, such as friends and extended kin, than previous surveys (Garip and Curran 2010). Specifically, we employ retrospective, individual-level data from the 2008–2010 Migration between Africa and Europe (MAFE) project which surveyed nonmigrants in the Democratic Republic of Congo (DRC), Ghana, and Senegal, as well as Congolese, Ghanaian, and Senegalese migrants living in Europe. The MAFE data proves most suitable for our research for three reasons. First, by including both migrants and nonmigrants, we can estimate the drivers of international migration, particularly the influence of ties to prior migrants, and assess how prospective migrants’ social positions moderate this influence. Second, the time-varying nature of our migrant network variables allows us to measure premigration access to migrant networks, thus mitigating concerns regarding simultaneity or reverse causality. Lastly, the survey’s comprehensive coverage of social ties facilitates a more thorough evaluation of migrant networks compared to most studies.

To our knowledge, this is the first paper to systematically examine how the social position of prospective migrants shapes their access, returns to, and mobilization of migrant networks. Our findings show all three dimensions to be deeply, though somewhat differently, stratified by social position. They further suggest that migrant networks exacerbate social inequalities in international migration opportunities. The paper proceeds as follows: we review the social capital literature, which emphasizes, across various settings such as education, labor market, and health, how social position conditions the functioning of social networks. Subsequently, we delve into migration scholarship, where migrant networks are recognized as pivotal resources in international migration yet their interaction with social position remains underexplored. After formulating a set of hypotheses, we introduce the study context—international migration between sub-Saharan Africa (SSA) and Europe. The subsequent section outlines the MAFE data and our employed methodologies. After presenting our findings, we discuss the implications, limitations, and suggest avenues for future research.

Social Capital, Social Networks, and Social Stratification

The concept of social capital, rooted in the notion that our lives are shaped not only by what we know but also by whom we know, was initially theorized by Pierre Bourdieu (1980) as the social resources embedded within social networks that individuals can leverage. We follow Bourdieu (1980), Lin (2002), and Portes (1998) and consider social capital an individual attribute. Social capital is typically assessed through the size, structure, and composition of social networks and was found to lead to better outcomes, such as educational achievement (Sacerdote 2011; Burke and Sass 2013), occupational status, prestige, and earnings (Granovetter 1974; Marsden and Gorman 2001; McDonald et al. 2013; Chen and Volker 2016; Vacchiano, Lazega, and Spini 2022), healthy behaviors (Smith and Christakis 2008), and improved physical and mental health (Ehsan et al. 2019), although studies may suffer from bias due to nonrandom selection into networks (Mouw 2003, 2006).

Social capital is also a significant factor in perpetuating and amplifying social inequality. Bourdieu (1980) identified social capital as pivotal in the reproduction of inequalities, alongside economic and cultural capital, utilized by privileged classes to uphold their social dominance. Lin (2002) explicitly linked his social capital theory to socio-economic inequality, proposing that these resources are not uniformly accessible. Subsequent research argues that social networks can reinforce social inequality when network-based resources are unevenly distributed or confer unequal benefits across groups (DiMaggio and Garip 2011, 2012; Pedulla and Pager 2019; Zhao and Garip 2021). As such, networks can be understood as one of the pathways leading to “cumulative advantage,” wherein greater benefits accrue to individuals who already possess an initial advantage (DiPrete and Eirich 2006). Expanding on this, we delineate three pathways through which social capital may contribute to amplify inequality, closely aligned with Lin’s framework: (1) differential access¹ to social capital (in terms of quantity and quality); (2) differential mobilization of social capital; and (3) differential returns to social capital.

A first source of inequality arises from differential access to social capital, based on both quantity and quality. High-status individuals typically belong to expansive, heterogenous networks, with more weak ties, resulting in favorable outcomes (Granovetter 1974; Burt 2009). Conversely, low-status individuals often possess smaller, more homogeneous networks, with more strong ties and disposable connections (Pichler and Wallace 2009; Desmond 2012). Moreover, high-status individuals enjoy greater access to influential contacts, yielding resource-rich networks (Erickson 1996; Verhaeghe, Li, and Van de Putte 2013; van Tubergen and Volker 2015). Findings support Lin’s “strength of position” (Lin 2002, 65), wherein the structure and composition of personal networks are shaped by both inherited and attained positions, most frequently measured in the empirical literature via parental and one’s own educational attainment. Inequalities in access are largely driven by homophily (Jackson 2021)—individuals’ tendency to associate with similar others (McPherson, Smith-Lovin, and Cook 2001).

Differential mobilization of social capital, which refers to whether ties give tangible support that may influence the outcome, was briefly mentioned by Lin and emphasized by Smith (2005), as another source of inequality. Smith’s study on low-income African Americans underscores the significance of factors such as the job seeker’s reputation, the SES of both the seeker and the job contact, and contextual elements in determining whether contacts provide assistance in job search. Other scholars also note that lower-status individuals receive less support from their networks (Miller-Cribbs and Farber 2008; Desmond 2012). Conversely, quantitative studies suggest high-status individuals are less inclined to utilize social ties for job search (McDonald and Elder 2006). Yet, networks of high-SES individuals may still hold utility, since networks with diverse resources often routinely circulate useful information without the need for social capital activation, a process called “the invisible hand of social capital” (Lin and Ao 2008).

Differential returns to social capital constitute a third source of inequality between groups, referring to the varying impact of similar social capital (in terms of quality and quantity) on the respective outcome. While fewer studies have investigated this aspect, existing evidence indicates that high-status individuals reap greater benefits from comparable social resources with respect to educational achievement, employment, or health outcomes (Lin 2000; DiMaggio and Garip 2012). Lin (2000) identifies three potential explanations for these lower returns: failure of low-SES individuals to utilize appropriate social capital; networks’ reluctance to assist the less advantaged; and differential responses from labor market structures such as bias and discrimination.

Overall, previous research suggests that high-status individuals possess greater access to superior social capital and tend to experience higher returns in terms of labor market outcomes compared to low-status individuals. While high-status individuals may be less inclined to activate their networks for job search purposes, when they do, these networks are more likely to offer the requested assistance.

Migrant Networks and Social Stratification

In migration studies, migrant social capital emerged as a pivotal driver of international migration (Garip and Asad 2016). It has been defined as the “resources of information or assistance that individuals obtain through their social ties to prior migrants” (Garip 2008, 591). These resources are accessed through migrant networks, which consist of interpersonal connections based on kinship, friendship, or shared origin community, linking both migrants and nonmigrants (Massey et al. 1993). Across various contexts, ties to prior migrants have been found to increase the likelihood of migration (Massey and Espinosa 1997; Garip 2008; Stecklov et al. 2010; Kalter 2011; Liu 2013). Migrant networks appear to be particularly important where migration is costlier, as in international versus internal migration (Davis, Stecklov, and Winters 2002) and in nonbrokered corridors (Williams et al. 2020) and there is growing understanding of their gendered role, as men and women have access to and mobilize different type of ties for migrating (Creighton and Riosmena 2013; Toma and Vause 2014; Liu, Riosmena, and Creighton 2018; Anastasiadou et al. 2023). This scholarship mainly measured migrant networks and much less so the actual resources embedded in them (but see Garip [2008] for an exception). Hence, in discussing migration literature findings, we primarily refer to migrant networks rather than migrant social capital.

Despite its strengths and breadth, migration literature has largely neglected the interplay between social stratification in origin societies and the functioning of migrant networks. In her seminal work, Garip (2008) distinguishes three dimensions of migrant social capital—(1) resources (information or assistance with migration); (2) sources (prior migrants); and (3) recipients (potential migrants)—and finds that, in Thailand, the greater the amount and the more even the distribution of one’s networks’ resources, the more likely is an individual to migrate internally. In research of Senegalese migration to Europe, Liu (2013) finds that the impact of network resources on migration differs for men and women and depends on the strength of ties. Both sets of findings echo some elements of Lin’s typology, particularly those relating to access and mobilization of migrant networks. Garip also acknowledges that “potential migrants may differ in their access to resources or in the benefits they extract from them” (Garip 2008, 592). However, these studies do not examine how prospective migrants’ socio-economic standing might influence their access and returns to migrant networks.

Based on the general social capital literature discussed above, we may expect higher-SES individuals to be better connected to migrant networks, more likely to mobilize them and to enjoy higher returns from them (in terms of likelihood to migrate). Should this be the case, migrant networks would increase social disparities in access to migration. In contrast, migration scholars theorize that as migrant networks grow within an origin community, migration becomes less selective in terms of education or skill (Massey et al. 1993). As such, migrant networks are expected to make access to migration more rather than less equitable. The scholarship linking social stratification and migrant networks is so far relatively scant; below we review the main findings.

Quantitative research on social position and migrant networks primarily focuses on returns to migrant networks, adopting a macro-level approach and using aggregate data on bilateral migration flows, to explore network influence across skill groups. Social position is typically measured by educational attainment. While some studies found no significant differences in network effects by education level (Orrenius and Zavodny 2005), others observed networks to be more beneficial for the lower-educated (Beine, Docquier, and Özden 2010; Neubecker, Smolka, and Steinbacher 2015). However, these macro-level studies fail to elucidate individual-level mechanisms.

Micro-level studies are better suited for this purpose but are limited. For example, McKenzie and Rapoport (2010) in Mexico and Bertoli (2010) in Ecuador found that in communities with higher migration prevalence (which the authors use as a proxy for migrant networks), the lower-educated are more likely to migrate to the United States. Conversely, communities with low migration prevalence exhibit no or slightly positive educational selectivity. The authors interpret these findings as evidence of the cost-reducing function of community-based networks. As networks expand, migration costs and risks decrease for all community members, particularly benefiting the lower-educated due to their limited financial capital. Moreover, ethnic enclaves at destination mainly cater to migrants with low skills and limited host-language proficiency (Bauer, Epstein, and Gang 2005; McKenzie and Rapoport 2007). Furthermore, Garip (2008) found the influence of community migration networks to be negatively correlated with prospective migrants’ own migration experience, which can be interpreted as destination-specific human capital. Taken together, the above findings suggest that human capital (skills and education) and migrant social capital (migrant networks) may partially substitute in enabling migration.

Limited attention has been given to how social position shapes access to migrant networks in origin contexts. The abovementioned studies use migration prevalence to measure migrant networks, which is problematic as it assumes social relationships exist among (all) members of the community. In other words, it presumes equal access to the social resources embedded in migration networks. This overlooks how internal socio-ethnic differentiation and kinship structures may condition access to migrant networks in most contexts (De Haas 2003; de Haas 2010b). Garip (2008) finds that migration is more likely in communities where migrant social capital is more evenly distributed, and hence more accessible to everyone. Furthermore, Garip and Curran (2010) show that migration selectivity also depends on how accessible migrant network resources are. If migration experience is concentrated among fewer members of the community, migration remains a highly selective process in terms of education and skill. Different levels of accessibility of migrant networks thus produce divergent community outcomes in terms of migration. These authors call for more research on (differential) access to migrant networks, a gap this paper addresses.

Finally, possessing migrant connections does not automatically entail their (successful) activation, a distinction that is not often made in the literature (but see Park, Lai, and Waldinger [2022] for an exception). Prior research suggests that migrant network mobilization is socially stratified. Ethnographic studies indicate that established migrants consider the social status of potential migrants before offering assistance (Bashi 2007). Moreover, tightening immigration regulations may turn established migrants from “bridgeheads” to “gatekeepers,” reducing their willingness to assist newcomers, especially those with lower skills (Böcker 1994; Snel, Engbersen, and Faber 2016). Additionally, settled migrants’ own social position influences their likelihood of assisting contacts in migration, with those in secure job positions more inclined to help (Paul 2013). Findings further suggest that social position also determines the type of migrant ties prospective migrants utilize. For example, higher-status individuals, despite having access to kin-based migrant networks, rely more on extensive networks of weak ties, such as classmates, friends, or former colleagues, for migration support (Wong and Salaff 1998). Conversely, working-class prospective migrants heavily depend on stronger kin networks.

In summary, existing evidence, primarily derived from aggregate or ethnographic data, suggests that the lower-educated derive greater benefits from migrant networks in terms of migration chances than the higher-educated. By decreasing the costs of migration, migrant networks are argued to make migration more accessible, especially for those lacking other forms of capital (human, financial). Consequently, lower-educated prospective migrants may need to rely more on migrant networks for migration assistance. However, established migrants may be more inclined to refuse aiding the low-educated migrate, particularly in restrictive economic and political destination contexts. Lastly, existing research provides limited insights into how social position may condition access to migrant networks.

Migration from DRC, Ghana, and Senegal to Europe: a Highly Selective Practice

International migration from the DRC, Ghana, and Senegal to Europe, the focus of this paper, provides a particularly provocative case study. Overall, young sub-Saharan Africans are under great pressure to consider intercontinental out-migration due to demographic and environmental pressures, lack of adequate job opportunities, and slow wage growth, as well as violent conflict (Hatton and Williamson 2003; Naudé 2010). Across the continent, youth experience increasing difficulties in their efforts to gain employment, status, and autonomy and find themselves suspended between childhood and adulthood: scholars have named this a “social moratorium of youth” (Vigh 2010) or simply “waithood” (Honwana 2012). In these contexts, international migration has become a key survival strategy and a highly desired life prospect, with 32 percent SSA youth desiring to emigrate (OECD 2015). Although still making up a majority of the moves, intra-African migrations have been declining, while extracontinental mobilities have been on the rise (Flahaux and De Haas 2016). Europe is by far the major destination of sub-Saharan migrants leaving Africa (Schoumaker et al. 2018). Yet, SSA migration to Europe is very costly and may involve migration brokers (Alpes 2011) and complex migration itineraries (Baldwin-Edwards 2006; Schapendonk 2008; Castagnone 2011).

As a result of the high costs and risks, migration to Europe is highly selective, both in terms of human and social capital. Unlike in the case of Mexican–US migration, scholars of SSA migration to Europe have found consistent evidence for positive educational selectivity (Docquier 2006; Shaw 2007; Gonzalez-Ferrer et al. 2013). Migrants from DRC, Senegal, and Ghana to Europe tend to be higher-educated and better-off than nonmigrants. Furthermore, family and friendship migrant networks play important roles in the migration process (González-Ferrer et al. 2018; Schoumaker et al. 2018). These are precisely the conditions that DiMaggio and Garip (2012) identify for network effects to exacerbate socio-economic inequalities in access to a social practice—in this case, international migration.

The three countries—DRC, Senegal, and Ghana—exhibit also important differences in their migration patterns to Europe shaped by their unique characteristics. DRC, rich in natural resources and with a relatively well-educated population, faces economic challenges due to civil war and political instability, leading to increased emigration rates. However, the great majority of moves occur still within Africa, and flows toward Europe have stagnated since 1990. Relatively stable politically, Senegal experiences economic crises and remains among the least educated countries in SSA. Since the 1990s, emigration flows intensified, mainly to Europe and North America (Ndione and Broekhuis 2006; Fall 2010). Since 1990, Ghana has achieved political and economic success, transitioning to a middle-income country, with decreased emigration within Africa but increased migration to Europe and North America. The three countries also differ in the gender composition of migration flows: women outnumber men among Congolese migrants to Europe, while men predominate among the Senegalese, with Ghanaian flows somewhere in the middle (Schoumaker et al. 2018). Similarly, prior work found no gender differences in the functioning of migrant networks among the Congolese, but pronounced variations in the type of migrant ties, as well as their role, in Senegalese men’s and women’s migrations (Toma and Vause 2014).

Research Hypotheses

As discussed above, existing theories of social capital on one hand, and of migrant networks on the other, lead us to derive contrasting predictions for the interplay between social stratification and migrant networks. This could be partly due to differences between domestic job searches, the focus of influential social capital studies, and high-cost international migration. International migration requires aligning an individual’s aspiration and capability to move abroad (Carling 2002; de Haas 2010a). Young sub-Saharan Africans with higher human, cultural, and financial capitals would have the necessary resources to undertake a costly relocation but may have less incentive to move abroad, preferring to compete for limited high-level positions at origin. Migrant networks can influence both aspiration and capability by providing the necessary motivation, information, and funds for the journey. They may partly compensate for the lack of other capital and thus be particularly helpful for the less-resourced. Such a process has been called “resource substitution” (Ross and Mirowsky 2006) or “compensatory leveling” (Schafer, Wilkinson, and Ferraro 2013) with evidence that students from disadvantaged backgrounds experience larger returns to higher education. In contrast, social capital can boost one’s cultural and human capital in the job search but can rarely substitute for it. Integrating labor market and international migration scholarship, we anticipate varying dynamics of social stratification in access, mobilization, and returns to migrant networks.

Given the positive educational selectivity of SSA migration to Europe and due to the action of homophily mechanisms, this model predicts that individuals in higher social positions will have greater access to migrant networks (H1). In line with Granovetter and Burt, we expect this to be the case particularly with respect to weaker migrant ties to extended kin and friends (H1a).

Nevertheless, we expect those in lower social positions to rely more on their migrant networks for assistance with migration (H2). We expect migrant networks to be more involved in the decision-making and financing of migration trips of lower-status individuals, who have fewer other resources to draw on.

Moreover, and connected to the above, we expect that at similar levels of access to migrant networks, individuals in lower social positions will derive larger relative benefits from their migrant networks in terms of migration chances (H3). In line with prior findings (Bertoli 2010; McKenzie and Rapoport 2010), we thus envisage migrant networks as partially compensating for the more limited human and material capital of the lower-status individuals, and potentially also for their less resourced local social networks.

Last, and somewhat in contrast to the above, we anticipate that migrant networks may actually amplify group-level social inequalities in access to migration (H4). Because of larger migration propensity among higher-status SSA individuals, their absolute gains from accessing networks in terms of migration chances may actually be bigger. Analyzing both relative and absolute interaction effects is increasingly recommended in the literature, as the two may have different substantive interpretations (Aradhya, Grotti, and Härkönen 2023).

Data and Methods

We use retrospective data from the Migration between Africa and Europe (MAFE) survey (2008–2009),² an example of the multi-sited data collection on migration and life course (Willekens et al. 2016; Beauchemin 2018). MAFE interviewed nonmigrants and return migrants at origin (the Democratic Republic of Congo - DRC-, Ghana, Senegal) and current migrants in European destinations (Congolese in Belgium and the United Kingdom; Ghanaians in the Netherlands and the United Kingdom; and Senegalese in France, Italy, and Spain). In total, 3,943 individuals were randomly sampled in selected urban areas of DRC (Kinshasa), Ghana (Accra and Kumasi), and Senegal (greater Dakar). In Europe, except for Spain, where a random sampling frame was used, MAFE employed quota sampling and a variety of recruitment channels, including snowballing, intercept points, contacts from origin households, or through migrant associations and public places, to interview 1,456 Congolese, Ghanaian, and Senegalese migrants in Europe (Beauchemin 2018). Our study population comprises 5,399 individuals, including both individuals based in their origin countries and in Europe at the time of the survey. We rely here on individual retrospective survey data. Nearly identical individual questionnaires in each location documented individuals’ detailed housing, family formation, work, and education trajectories, among others. Most important for this study, MAFE collected respondents’ detailed migration trajectories since birth, as well as the migration trajectories of members of their personal network.

Operationalizing migrant networks:

MAFE records respondents’ social ties with current or past migration experience. As such, it provides an ego-centric measure of respondents’ personal migrant network.³ MAFE remains one of the few datasets collecting a dynamic measure of migrant networks, despite recent calls for more studies into the dynamism of transnational ties (Lubbers, Verdery, and Molina 2020). Qualitative and theoretical work has emphasized the changing nature of transnational ties over time and space (Hagan 1998; Ryan and D’Angelo 2018; Wissink and Mazzucato 2018), but surveys on these aspects remain rare, especially in origin countries. In this study, we utilize dynamic, transnational, ego-centric measures of migrant networks, derived from full retrospective (year-by-year) migration histories of respondents’ parents, siblings, spouses, extended kin, and friends. For each network member, years abroad and countries lived in are recorded, as are sex, the year they met the respondent, and year of death (if applicable). The composition and location of each respondent’s migrant networks can thus change over time as family or friends migrate abroad, move onward, or return to origin. Given our interest in migration to Europe and considering prior findings on the importance of distinguishing between region of destination-specific networks (Liu 2013; Bertoli and Ruyssen 2018), we define respondents’ migrant networks as their kin or friends living in Europe in a given year, allowing us to capture yearly variation in (Europe-based) migrant networks.⁴ Hereafter for simplicity’s sake, migrant networks and ties refer to Europe-based migrant networks and ties.

We further distinguish between ties to close kin (parents or siblings) and ties to extended kin or friends. For each of these, we construct both a dichotomous variable and a count variable of the number of ties (both time-varying). Unfortunately, MAFE does not provide direct measures of the strength of the relationship. Our distinction, by type of relationship (i.e., close kin versus extended kin/friendship tie), therefore does not directly map onto the weak versus strong ties distinction pioneered by Granovetter (1983). We expect it to be nonetheless consequential with respect to the amount of assistance that can be expected among SSA youth, since the level of (material and social) obligations underlying a social tie is strictly codified by the type of (kin) relationship (Buggenhagen 2012). Throughout our analysis, we examine migrant spouses separately due to their strong connection with legal family reunification. Following Liu (2013) and Toma and Vause (2013), we argue that this represents a distinct influence channel, separate from social network effects. We control for a time-varying variable taking the value “1” if the respondent has their partner in Europe that year, and “0” otherwise.

We measure migrant network mobilization through two binary variables: whether respondents’ migrant connections were involved in (i) the decision to migrate (for the first time, to Europe) and/or (ii) the funding of their first migration trip to Europe.

Social Position is a key independent variable in our analysis. Prior research found access to social capital to be stratified both by attained and inherited social position; we also measure both here. Educational attainment is the most widely used proxy in both cases (Lin and Erickson 2008; van Tubergen and Volker 2015). In line with this literature, we use father’s highest level of education to proxy for respondents’ inherited social position. Since educational distributions vary quite substantially across the three countries (with DRC and Ghana the most educated and Senegal the least), we construct relative measures of education—low-, medium-, and high-educated—which differ between Senegal, on the one hand, and Ghana and DRC, on the other.⁵ MAFE also contains some information on father’s level of occupation,⁶ which we use as an alternative for father’s level of education in robustness checks, with substantively equivalent results (results available upon request).

Second, we measure achieved status with respondents’ own educational attainment. We use a time-constant variable measuring highest level of educational attainment at the time of the survey, recoded in three groups: low-, medium-, and high-educated, which again differ between Senegal, on the one hand, and Ghana and DRC, on the other.⁷ Since there may be some bias induced by not measuring level of education prior to the primary outcome (migration), we run the same analyses using a time-varying education variable counting the number of years each respondent spent on education. The drawback is that this computed variable does not offer information on the specific grade level attained each year, which is why in our preferred specification we retain the time-constant variable. Notwithstanding, results are almost identical (available upon request).

Covariates:

We account for several time-invariant covariates—gender, number of siblings, birth rank (respondent is a first-born child or not—reference), country of origin (DRC—reference, Ghana, Senegal)—as well as for a series of time-varying variables: age, partnership status, and partner location (single—reference, spouse outside Europe, spouse in Europe), number of children under 6, asset ownership (whether the respondents owns any land, real-estate, or business), activity status (employed—reference, studying, not working—which includes the inactive and the unemployed); period⁸ (before 1990—reference, 1990–2002, 2003–2008). We run pooled models and only control for country of origin, as we lack both theoretical reasons to expect different interactions between social position and migrant networks across the three countries of origin, as well as sufficiently large sample sizes to test such differences. Table 1 presents descriptive statistics for our analytic sample at the time of the survey, by migration status.

Table 1.

Descriptive Statistics by Migration Status presented as Weighted Percentages (%) and Means [95% confidence intervals], Migration between Africa and Europe (MAFE) individual questionnaires (2008–2010).

	Never migrated to Europe		Has already migrated to Europe		Total
Individual factors
Female	57.8	[55.3,60.3]	35.7	[31.5,40.0]	56.1	[53.8,58.5]	***
Age	41.4	[40.8,42.1]	43.7	[42.7,44.7]	41.6	[41.0, 42.2]	***
Level of education (%)
Lower educated	30.9	[28.6,33.3]	13.3	[10.9,16.1]	29.6	[27.5,31.8]	***
Medium educated	49.1	[46.5,51.6]	37.4	[33.5,41.5]	48.2	[45.8,50.6]
Higher educated	20	[18.0,22.1]	49.3	[45.0,53.6]	22.2	[20.3,24.2]
Current Activity status (%)
Employed	69.8	[67.4,72.1]	78.5	[75.0,81.7]	70.5	[68.3,72.6]	***
Student	5	[4.0,6.3]	5.9	[4.0,8.7]	5.1	[4.1,6.3]
Not working	25.2	[23.0,27.4]	15.6	[13.1,18.4]	24.4	[22.4,26.6]
Asset Ownership status (land, real estate, business) (%)	36.1	[33.8,38.6]	54.0	[49.7,58.3]	37.5	[35.3,39.8]	***
Country of origin (%)
RDC	40.2	[37.8,42.8]	14.2	[1 1.2,17.9]	38.3	[36.0,40.6]	***
Ghana	31	[28.6,33.4]	29.4	[25.4,33.7]	30.8	[28.7,33.1]
Senegal	28.8	[26.6,31.1]	56.4	[52.0,60.7]	30.9	[28.8,33.0]
Family background
First born	23.9	[21.8,26.1]	25.9	[22.6,29.6]	24.0	[22.1,26.1]
Number of Siblings	6.9	[6.67,7.09]	6.48	[6.14,6.82]	6.86	[6.6,7.0]
Father’s level of education (%)
Lower educated	50.8	[49.8,51.7]	42.6	[41.8,43.3]	50.1	[49.2,51.0]	***
Medium educated	36.2	[35.3,37.3]	29.9	[29.1, 30.7]	35.7	[34.8,36.7]
Higher educated	12.9	[12.6,13.2]	27.5	[26.8,28.2]	14.1	[13.8,14.4]
Current family status
Have children under 6 (%)	0.57	[0.53,0.61]	0.41	[0.34,0.48]	0.55	[0.5,0.6]	***
Partnership status and partner location (%)
Single	33.3	[30.9,35.7]	23.7	[20.7,27.1]	32.6	[30.4,34.8]	***
Partner not in Europe	64.3	[61.8,66.7]	43.1	[38.7,47.7]	62.7	[60.4,64.9]
Partner Europe	2.4	[1.8,3.2]	33.1	[29.5,36.9]	4.7	[4.1,5.5]
Current Europe-based ties (%)
Has any Europe-based tie (besides partner)	33	[30.7,35.4]	67.9	[63.5,72.0]	35.7	[33.5,37.9]	***
Has close kin in Europe	16.3	[14.5,18.2]	46.7	[42.5,5 1.0]	18.6	[16.9,20.4]	***
Has extended kin or friends in Europe	21.2	[19.3,23.4]	41.0	[37.0,45.1]	22.7	[20.9,24.7]	***
Migration-related factors
Age at first migration to Europe			28.5	[27.9,28.9]
Primary reason for first migration to Europe (%)
Family			19.4	[16.4,22.9]
Economic			43.7	[39.6,48.0]
Study			24.6	[20.3,29.4]
Other			12.3	[10.2,14.7]
Timing of first migration to Europe (%)
Before 1990s			33.1	[29.1,37.4]
1990s			26.9	[23.4,30.8]
2000s			39.9	[35.8,44.2]
Number of observations (unweighted)	3927		1472		5399
Weighted % of study sample	92.4		7.6		100.0

Note: Legend 1 MAFE data. Weighted statistics. Imputed values for Father education level.

Empirical Strategy

We conduct three sets of analyses to test our four main hypotheses.

First, to assess if individuals in higher social positions have greater Access to migrant networks (H1), we conduct cross-sectional ordinary least-squares regression with “Number of Europe-based migrant ties” as the dependent variable. Access to migrant networks is measured in the year before individuals’ first migration to Europe to avoid capturing ties developed after migration, a bias common in studies lacking a time-varying network measure. For nonmigrants, access is measured at average age of first migration to Europe (29 years) or at time of survey (if younger than 29), to enhance comparability. We further test whether social position has a stronger effect on weaker ties (extended family and friends) compared to closer ties (close kin) (H1a), in two additional models. While linear regression models are reported, logistic regression models with binary access measures yield similar findings.

Second, we estimate cross-sectional logistic regressions to examine whether social position influences migrant network Mobilization. We examine whether migrant networks were involved in (i) decision making and/or (ii) financing of respondents’ first migration trip to Europe. This cross-sectional analysis is restricted to Europe-bound migrants who reported having at least one migrant tie in the year before migration.

Third, we examine Returns to migrant networks, specifically whether respondents’ social position moderates their networks’ effect on migration likelihood. Our dependent variable is first adult migration to Europe. We exclude childhood migrations since networks dynamics are likely to be different. We employ logistic regression models within a discrete-time event history framework, following individuals from age 18 until their first migration to Europe or the survey time. We censor individuals who migrate elsewhere first, by excluding them from the risk set once they migrate. We use time-varying measures for migrant networks and most other control variables. Measuring access to migrant networks prior to migration, albeit retrospectively, promises to reduce potential endogeneity bias due to reverse causality. Other sources of endogeneity are due to selection into (friendship) networks (Palloni et al. 2001): individuals tend to associate with similar others, which may bias network effects in the absence of proper controls. Given networks are particularly homophilous with respect to gender, education, and age, controlling for these variables should partially mitigate such confounding effects (Boujija et al. 2022). The discrete-time survival models specified here thanks to our longitudinal data further alleviate these concerns.

We present interactions between social position and migrant networks on both relative (multiplicative) and absolute (additive) scales, measured respectively in odds ratios (ORs) and percentage point differences. While sociology lacks theoretical reflection on the choice of scale (relative versus absolute) for interactions (Aradhya, Grotti, and Härkönen 2023), an established debate in epidemiology argues for the necessity of presenting both scales for comprehensive understanding (VanderWeele and Knol 2014; Mehta and Preston 2016). Relative scale interactions reveal mechanistic differences: in our case, lower-status individuals may benefit more from migrant networks by utilizing them more for assistance, such as funding. Conversely, absolute scale interactions elucidate heterogeneity in population-level consequences. These could inform us about migrant networks’ role in social inequalities regarding migration likelihood. Since absolute scale interactions depend on baseline rates, differences in migration probability (due to social status) may result in larger absolute effects of network access. For example, if the higher educated have higher baseline rates (of migration probability), ties to migrants may translate into larger absolute gains for them than for their lower-educated counterparts, even if they benefit similarly (on a relative scale). Therefore, considering both scales provides a comprehensive view of how migrant networks shape social inequalities in migration opportunities.

Results

A Clear Social Stratification in Access to Migrant Networks

We begin our analysis by examining whether social position conditions access to migrant networks. Using linear regression models, we assess the size of Europe-based migrant networks the year before migration (for migrants) or at age 29 or at survey time (for nonmigrants). We estimate total number of migrant ties excluding partners (Model 1), as well as number of ties to migrant close kin (parents and siblings, Model 2) and to extended kin and friends (Model 3) separately. Only premigration ties are considered. In addition to our variables of interest—father’s and respondents’ educational level—we control for various individual and contextual attributes, including age, gender, sibship size, birth rank, activity status, asset ownership, partnership status and partner location, country of origin, and period. Figure 1 displays coefficients for our two variables of interest across the three specifications (panels A, B, and C), with the full table available in the Supplementary file (Table S1). Interestingly, men and women have a similar sized migrant network (Table S1). Owing to the comparatively longer history and larger size of Senegalese migration to Europe, the Senegalese report the largest migrant networks, followed by the Congolese and the Ghanaians.

Figure 1. Access to Migrant Networks by Level of Education (95% and 90% Confidence Intervals).

Source: Migration between Africa and Europe (MAFE) Individual Surveys, 2008–2009.

Notes: Results are presented in coefficients from weighted ordinary least-squares regressions. Models control for age, gender, sibship size, birth rank, activity status, asset ownership, partnership status and partner location, country of origin, and period.

We observe a clear and significant social gradient in migrant network access, with the number of migrant ties increasing linearly with both father’s and individual’s education levels, supporting H1. Additionally, these variables have cumulative effects, each maintaining independent effects net of the other. The total number of migrant ties increases by about 0.4 among those with the highest-educated fathers compared to those with the lowest-educated fathers, and by an additional 0.24 for those who are highest-educated themselves (compared to the lowest-educated). Having a highly educated father displays a significantly larger correlation with the number of weaker ties to extended kin or friends (Model 3, panel 1C) than to stronger ties to close kin (Model 2, panel 1B), as confirmed by t-tests. However, this distinction is not evident based on personal education, yielding only partial support for Hypothesis 1A.

Our findings remain robust across various checks. First, we conducted separate analyses by country of origin (Table S2 in Supplementary file), revealing consistent patterns of social stratification in migrant network access. Second, we dichotomized outcomes (access to migrant ties or not) and conducted logistic regression models. Third, we employed alternative specifications for the individual education variable—utilizing a time-varying measure—and for father’s social position—using father’s occupation—with highly similar results.

Limited Evidence of Social Stratification in the Mobilization of Migrant Networks

Accessing social networks is not the same as using such networks. We transition to examining how prospective migrants’ social position influences their mobilization of migrant networks. We estimate, through logistic regression, whether the respondents’ migrant networks were involved in their migration decision making (Model 1) or in funding their migration trip (Model 2). Our analysis is confined to those who migrated to Europe for their first adult migration and had at least one Europe-based migrant tie (besides their partner) at the time of migration (N = 699). Alongside our variables of interest—father’s and respondents’ educational level—we control for individuals’ age, gender, activity status, asset ownership, partnership status and partner location, country of origin, period, and the type of migrant ties possessed. Women are more likely than men to rely on their migrant ties (besides their partner) in their migration decision, but similarly likely to have migrant ties contribute to funding their trip. ORs for our two variables of interest across both models (adjusted for all mentioned controls) are presented in Figure 2, with the complete table available in the Supplementary file (Table S3).

Figure 2. Mobilization of Migrant Network in Migration Decision-Making and Funding by Level of Education (Odds Ratios, 95% and 90% Confidence Intervals).

Source: Migration between Africa and Europe (MAFE) Individual Surveys, 2008–2009.

Notes: Weighted Logistic Regressions. Models control for age, gender, activity status, asset ownership, partnership status and partner location, country of origin, period, and type of migrant ties.

Overall, we find no or little evidence between level of education and mobilization of migrant networks for either deciding or financing migration. There is no significant relationship between level of education (either personal or paternal) and networks’ involvement in migration decision. In terms of funding the migration trip, highly educated individuals appear less reliant (p < .05) on migrant ties for financial assistance than their low-educated counterparts. Robustness checks, including alternative measures of father’s status (occupational status) and individual education (time-varying measure), yield similar results.⁹ Hence, there is only marginal and partial evidence in support of Hypothesis 2, concerning how an individual’s education is associated with their reliance on migrant ties for funding their migration.

Differential Returns to Migrant Networks

Lastly, we investigate whether individuals from diverse social backgrounds experience varied benefits from their migrant networks regarding migration opportunities. Using a discrete-time event history framework, we assess the likelihood of a first migration to Europe in a given year, considering social position (measured by both individual and father’s education), access to Europe-based migrant networks (distinguishing between close kin and extended kin or friends), and their interaction. We include various control variables known to influence migration likelihood, both constant—gender, sibship, birth rank, country—and time-varying—age, age square, activity status, asset ownership, number of children under 6, partnership status and location, period. Instead of aiming to fully explain European-bound SSA migration, our analyses concentrate on how migrant networks and social stratification intersect to influence migration access, incorporating other factors only as needed for model specification. Our results confirm well-established gender (women less likely to migrate to Europe then men) and country of origin (the Senegalese, most likely to move to Europe, followed by the Ghanaian and the Congolese) patterns. Robustness checks using random-effects logistic regressions refute the hypothesis of significant unobserved individual-level heterogeneity¹⁰ and provide consistent results. Therefore, we present ORs from simple logistic regression models.

First, results presented in Table 2 (full details in Supplementary Table S4) reaffirm prior findings in the literature: migrant network access strongly correlates with migration likelihood (OR = 4.8, Model 1), with cumulative and similar effects for ties to close kin and to extended kin/friends (OR = 3.7 and 3.1 respectively, Model 2). Additionally, higher personal education significantly increases the likelihood of migration to Europe (OR = 5.5 for highest-educated compared to lowest-educated; Model 1), with no notable impact from father’s education when individual education is controlled. Models with interactions (Models 3 and 4) expressed in relative (multiplicative) terms (ORs) reveal negative and significant effects, indicating that higher-educated individuals derive lower benefits from migrant networks relative to lowest-educated individuals. Access to networks leads to a seven-fold increase in migration hazard for the lowest-educated, compared to only a 3.5-fold increase for the highest-educated (OR = 6.8 × 0.52 = 3.5). This holds for both personal education and father’s education. Table S4 shows similar interaction effects for close kin (Models 4 and 7) and extended kin and friends (Models 5 and 8). These findings support the hypothesis (H3) that migrant networks yield smaller relative gains for higher-educated individuals and those with highly educated fathers, irrespective of their type.

Table 2.

Likelihood of First Migration to Europe by Own Education, Father’s Education and Networks (Odds Ratios, Standard Errors).

	M1	M2	M3	M4
Personal education (ref: lower educated)
Ego medium educated	1.79*** (0.27)	1.7*** (0.25)	1.8*** (0.38)	1.72*** (0.26)
Ego higher educated	5.49*** (1.01)	5.46*** (0.98)	7.61*** (1.89)	5.26*** (0.97)
Father’s level of education (ref: lower educated)
Father medium educated	0.97 (0.12)	0.99 (0.13)	0.95 (0.12)	0.96 (0.21)
Father higher educated	1.1 (0.15)	1.13 (0.15)	1.1 (0.15)	1.67** (0.33)
Access to migrant networks (ref: no EUR-based migrant networks)
Has EUR-based migrant ties	4.78*** (0.6)		6.83*** (1.89)	5.88*** (1)
Has EUR-based close migrant kin (ref: no close kin)		3.7*** (0.43)
Has EUR-based extended kin or friends (ref: no ext. kin)		3.06*** (0.35)
Any migrant ties × Ego medium educated			0.95 (0.29)
Any migrant ties × Ego higher educated			0.52** (0.17)
Any migrant ties × Father medium educated				1.01 (0.27)
Any migrant ties × Father higher educated				0.51*** (0.12)
Observations	88,313	88,313	88,313	88,313

Source: Migration between Africa and Europe (MAFE) individual surveys, 2008–2009.

Notes: Weighted logistic regressions. Models control for gender, sibship, birth rank, country of origin, age, age squared, activity status, asset ownership, number of children under 6, partnership status and location, and period. EUR = Europe.

^***p < .01

^**p < .05

^*p < .l.

However, when considering absolute gains, a different story emerges. Figure 3 illustrates predicted probabilities (panel 3A) and average marginal effects (AME) of migrant networks (panel 3B) depending on personal education levels, based on the same models. Despite the smaller relative gains observed for higher-educated individuals, migrant networks lead to higher absolute gains for them due to their higher baseline migration probability compared to the lower-educated. For instance, AME = 0.015 for highest-educated compared to 0.005 for lowest-educated, translating into a predicted migration probability of 0.017 for highest-educated with access to migrant networks compared to 0.002 for lowest-educated with networks. These findings support Hypothesis 4.

Figure 3. Predicted Probabilities (A) and Average Marginal Effects (B) of Access to Migrant Networks by Personal Level of Education (95% Confidence Intervals).

Source: Migration between Africa and Europe (MAFE) Individual Surveys, 2008–2009.

Notes: MN = migrant networks, AME = average marginal effects.

Robustness checks using alternative education variables (i.e., father’s occupation and time-varying personal education) produce consistent results. Additionally, similar trends emerge when analyzing data by country of origin: interaction terms exhibit consistent directional trends, though statistical significance is only observed for Senegalese migration to Europe (in relative interaction terms, see Table S5 in Supplementary file) and for Senegalese and Ghanaian flows (in absolute terms, see Figure S1).

Discussion and Conclusion

This paper explores how the social position of prospective migrants influences the role of migrant networks in international migration. We are motivated by existing social capital research which suggests that social networks often perpetuate socioeconomic inequalities in status attainment, and migration research which notes, contrastingly, that migrant networks may democratize access to migration for lower-SES individuals. Our findings reveal significant social stratification in the access, utilization, and benefits of migrant networks in migration from Africa to Europe. Individuals in higher social positions, operationalized through their own level of education and their father’s educational or occupational status, appear better connected to Europe-based migrant networks. However, they are less reliant on these networks for financial support and derive lower relative returns than their lower-SES counterparts. Nonetheless, the absolute gains from migrant networks are higher for the higher-SES individuals. We discuss each of these findings in turn.

First, our findings align with Lin’s (2002) strength-of-position proposition, indicating a positive correlation between both parental and personal education levels and the accessibility and size of Europe-based migrant networks. This proposition, extensively validated in sociological research on social capital, suggests that higher inherited and attained status lead to larger and more diverse social networks. Our study extends this validation to migrant social networks in the context of SSA migration to Europe. Except for qualitative studies (De Haas 2003) or community-level analyses (Garip and Curran 2010; DiMaggio and Garip 2011), the migration literature has largely overlooked how internal stratification processes shape migrant network access. Indeed, our findings reveal significant social inequalities in accessing Europe-based migrant ties. Higher-SES individuals’ greater access to Europe-based migrant networks is likely related to the positive educational selectivity of Congolese, Senegalese, and Ghanaian migrants to Europe we observe, alongside the educational homophily of social networks (Jackson 2021).

Second, greater access to migrant networks does not necessarily translate into increased mobilization. Indeed, despite their differential network access, we detect little to no difference between how lower and higher status migrants use their networks. It does appear that higher-educated migrants, despite being more connected to prior migrants, may be less likely to rely on them for initial migration funding to Europe compared to their lowest-educated counterparts (p < .05). However, in one of the robustness checks, this finding is only statistically significant at the 90 percent confidence level and should be interpreted with an abundance of caution given this and the small sample size for this part of the analysis. The still scant empirical research within and outside migration studies on network mobilization has shown mixed results. Prior studies mainly adopt the perspective of settled migrants, and find these to be more inclined to assist migration candidates with higher human capital (Paul 2013; Snel, Engbersen, and Faber 2016), in contrast to our findings. However, our perspective differs, focusing on potential recipients rather than providers of support. Moreover, we only observe network mobilization among successful migrants and overlook failed attempts where ties may have been unwilling or unable to help. Among successful Europe-bound migrants, we find that lower-educated individuals rely more on network funding compared to their higher-educated counterparts. This suggests a greater dependence on network support among the less-educated, while the higher-educated may access alternative resources. However, our analysis overlooks other network influences, particularly information sharing. Proponents of the “invisible hand of social capital” argue that higher-SES individuals do not need to actively mobilize their networks, as they benefit from valuable information routinely circulating through them. Further research, ideally qualitative, is necessary to elucidate how social position shapes individuals’ (attempts at) mobilization of migrant connections.

Third, we observe greater relative returns to migrant networks for less-educated prospective migrants compared to their more-educated counterparts. In essence, those with lower education levels benefit more from networks of similar size and composition than those with higher education levels. This contrasts with studies indicating that high-SES individuals reap larger benefits from accessed social capital in areas such as employment or health outcomes (Behrman, Kohler, and Watkins 2008). Our findings align more closely with McKenzie and Rapoport (2010), who found that lower-skilled Mexicans experience larger returns to community-level migrant networks in terms of migration likelihood to the United States. This is also consistent with our observation that less-educated individuals rely more on their migrant networks for migration funding, a mechanism likely explaining heterogeneity in relative gains. In summary, at the individual level, migrant networks function as an “equalizer,” substituting for other key resources (Andersson 2018) in the migration process, such as human, financial, or local social capital.

In contrast, we find larger absolute gains from networks for the higher educated: access to migrant ties increases to a greater extent migration probability among the higher educated compared to their low-educated counterparts. This is because higher-educated individuals have higher baseline migration probabilities than the lower-educated. The presence of migrant networks, despite yielding larger relative returns for the lower-educated, exacerbates initial educational inequalities in migration opportunities. Thus, as anticipated by H4, in a context of positive educational selectivity into migration to Europe, migrant networks intensify population-level educational inequalities. Additionally, the greater access of higher-status individuals to Europe-based migrant networks, crucial for European-bound migration, further amplifies social inequities in migration likelihood.

In summary, we conclude that due to initial social (i.e., educational) disparities both in access to migrant networks and in migration likelihood, migrant networks exacerbate social inequalities in migration opportunities. Our empirical findings illustrate DiMaggio and Garip’s (2012, 94) formal model, fulfilling the three conditions they identify for networks to aggravate inequality: (1) “a behavior, if adopted, is likely to improve adopters’ well-being”: in our context, Europe-bound migration is seen as highly desirable in SSA; (2) “the probability of adoption is a function both of individual endowments and of one’s social ties with earlier adopters”: there is clear positive educational selectivity in Europe-bound migration, also influenced by ties to prior migrants; and (3) “likely adopters tend to associate with other likely adopters,” due to homophily: in our case, higher-educated individuals are more likely to associate with prior migrants, arguably since migrant networks are homophilous with respect to human capital, itself positively associated with migration. Under these conditions, DiMaggio and Garip argue, network influences compound advantages individuals gain from initial endowments, exacerbating intergroup inequality in the adoption of rewarding practices, beyond what could be expected based on individual differences alone (2012, 94). This effect should be particularly strong in the case of migration, a risky and high-cost practice requiring what the authors call a “complex contagion,” or a sustained input from social networks. Our study of migration between Senegal, DRC, Ghana, and Europe aligns with these formal intuitions.

In the end, our findings somewhat challenge previous studies that link migrant network expansion to negative migration selectivity (McKenzie and Rapoport 2007, 2010; Bertoli 2010), which tended to overlook differential access to networks. These studies typically measure community-based networks, assuming equal accessibility regardless of skill level or class background. Our research suggests otherwise, highlighting the importance of measuring fine-grained, ego-centric migrant networks. Our analysis complements prior studies situated at the village level (Garip 2008; DiMaggio and Garip 2011), which indicate how uneven migrant network accessibility within communities affects migration flows and exacerbates inequality between villages in migration outcomes. Furthermore, it extends these findings with evidence from a less studied context: that of migration between SSA and Europe.

Several study limitations warrant consideration in interpreting our results. First, pooling the probabilistic origin-country samples with nonprobabilistic European samples requires a relatively strong assumption that factors determining sample inclusion are uncorrelated with variables of interest or adequately adjusted by quotas and poststratification weights (Schoumaker and Mezger 2013). Second, although our survival models provide relatively strong support for a causal interpretation of network effects, while random-effects specifications refute the hypothesis of significant individual-level unobserved heterogeneity, correlations between unobserved individual effects and other variables may bias estimates. Fixed-effects models—a possible fix—were not an option since key education-related variables were time-invariant. Nevertheless, literature suggests minimal bias with random-effects approaches (Frees 2004). Third, data age (collected in 2008–2010) may limit relevance to current migration dynamics. However, we contend that enduring processes such as the social stratification shaping migrant networks remain relevant across diverse contexts and timeframes. Fourth, although MAFE improves several network measurement issues (dynamic measures, broader set of ties, direct versus supposed relationship), it lacks information on the resources embedded in networks, such as those related to network members’ status. With such data, we could examine network homophily based on education and occupation and would likely observe stronger stratification in social capital resources. Future migration surveys should gather more comprehensive information on resources within migrant networks. Fifth, prior work has shown that migration patterns and the functioning of networks are gendered (Anastasiadou et al. 2023), but these vary across contexts of origin (Toma and Vause 2014). While our study, due to limited sample sizes, only controls for gender and country of origin, a promising avenue for future research would be a comparative analysis of the intersection of gender, social stratification, and migrant networks across contexts.

Despite these limitations, our findings have at least two important implications for theoretical frameworks on network mechanisms in shaping migration opportunities. First, we highlight migrant networks as amplifiers of social (primarily educational) inequalities in accessing Europe-bound migration for sub-Saharan Africans. Social position, operationalized here primarily as own and parental level of education, strongly stratifies access to Europe-based migrant ties and migration opportunities. Consequently, even though lower-educated individuals rely financially more on migrant networks and derive larger relative benefits from migrant ties access, these networks exacerbate educational gaps in migration probability. Second, this mechanism’s importance is likely greater in contexts with stronger positive educational selectivity and network homophily. Although the MAFE project provided scope for comparison by collecting data from three origin countries, all three display positive educational selectivity in European-bound migration. Results disaggregated by country suggest such selectivity is more important in the Congolese flow, but our sample sizes prevent us from testing whether the degree of educational selectivity shapes findings. Further comparative research on this topic would be highly valuable. A compelling comparison could be made with the Mexican–US migration flow, likely characterized by negative educational selectivity (Rendall and Parker 2014).

This paper bridges two expanding research lines: social capital scholarship focusing on inequality in status attainment and the migration literature examining the role of migrant networks in international mobility. Its primary contribution to the former extends findings on social networks exacerbating social inequalities across a variety of outcomes, to international migration, particularly from SSA to Europe. Its main contributions to the latter are to highlight social stratification dynamics shaping migrant networks and to challenge assumptions of their equal accessibility and benefits across social positions. Neglecting these processes obscures migrant networks’ role in amplifying social inequalities in migration.

Supplementary Material

Supplemental material for this article is available online.