African transnational families: Cross‐country and gendered comparisons

Funding information NORFACE Research Programme ‘Migration in Europe ‐ Social, Economic, Cultural and Policy Dynamics’, Grant/Award Number: 315; Netherlands Organization for Scientific Research, WOTRO Science for Development Division, Grant/Award Number: W01.65.316; Seventh Framework Programme, Grant/Award Number: 217206; Spanish Ministry of Science and Technology, Grant/Award Number: (CSO2009‐12816)

investigation of these studies reveals that most of the data come from transnational migrant mothers. Only recently have studies focussed on fathers (Poeze & Mazzucato, 2012;Pribilsky, 2004;Waters, 2009), indicating that fathers also suffer from the separation from their children, albeit in different ways. Finally, these studies are small-scale and thus focus on transnational parents only (Mazzucato & Dito, 2018). They therefore cannot compare their findings with nontransnational parents, in order to identify what may be particular to being in a transnational family versus more general characteristics of a population.
Although gender norms are important to consider, other factors also need to be explored. There are indications from gender and migration studies that women face different contexts abroad, leading to different outcomes, and that they live in different family structures.
For example, a large-scale study investigating how gender, social class, and origin affect migrant health in Spain found that that migrant women of all social classes experienced worse employment conditions, greater material/financial deprivation, and lower health status than migrant men (Malmusi, Borrell, & Benach, 2010). Furthermore, it has been found that African female migrants have different social network characteristics leading to different propensities to migrate and in different ways (Toma & Vause, 2014), Filipino women sometimes migrate to escape problematic marriages (Constable, 2003), and when Ghanaian women migrate on their own, this is associated with higher rates of divorce than when men migrate or when the couple migrates together . None of these studies focus specifically on transnational families, yet they provide important clues that more may be at hand than gender norms and expectations, which have been the predominant explanation for differences between transnational migrant mothers and fathers.
More recently, large-scale transnational family studies emerged that compare transnational mother-and father-away families and the effects that these have on those involved (DeWaard, Nobles & Donato, 2018;Graham & Jordan, 2011;Haagsman, 2015;Liu, Li, & Ge, 2009;Mazzucato et al., 2015;Wen & Lin, 2012). For the most part, these studies are conducted in the origin countries and focus on the effects of maternal or paternal migration on children. Some of these studies indeed confirm that children have more difficulties when mothers migrate than when fathers migrate (DeWaard et al., 2018;Dreby & Adkins, 2010;Graham & Jordan, 2011;Liu et al., 2009;Wen & Lin, 2012). Yet studies in African countries do not show uniform results on this matter. In fact, in Ghana and Nigeria, children in mother-away families did not show any difference in psychological well-being than children living with both parents at origin . Furthermore, as these studies contain only limited information on the parents, there is little knowledge about the disparities between migrant fathers and mothers in transnational families that could potentially cause these different outcomes.
Finally, few studies on transnational families have been comparative in nature. As most studies on transnational families are in-depth ethnographic studies, they have focussed on one particular origin group in one particular host country. Consequently, there is almost no comparative cross-country research on the structure of transnational families (Mazzucato & Dito, 2018). Doing a cross-country comparison allows us to address this gap by investigating whether characteristics of transnational parents are similar for different origin countries and whether we find common patterns in their family trajectories. Until recently large datasets that allow the study of characteristics and family trajectories of transnational parents were unavailable. This paper identifies differences between transnational and nontransnational parents, as well as between transnational mothers and fathers, hereby providing important information to help explain and contextualise differences found in the literature on the effects of maternal and paternal migration on children in the country of origin and on the migrant parents' own well-being. It does so by analysing recent and unique datasets that contain data specific to transnational family life of five sub-Saharan African countries of origin (Angola, Democratic Republic of Congo [DR Congo], Ghana, Senegal, and Nigeria) and eight Western European destination countries (Netherlands, Belgium, United Kingdom, France, Italy, Spain, Portugal, and Ireland). The datasets that are used include information on members of both (male and female) transnational (at least one child in the country of origin) and nontransnational migrant families (living with all the children in the host country) and also include retrospective information suitable for studying family trajectories.

| LITERATURE REVIEW
Literature on gender and migration has shown that women and men have different social network characteristics leading to different propensities to migrate, may have different reasons for migrating, and face different conditions overseas (Constable, 2003;Curran & Rivero-Fuentes, 2003;Eremenko & González-Ferrer, 2018;Liu, Riosmena, & Creighton, 2018;Malmusi et al., 2010;Toma & Vause, 2014). Yet there has been no systematic comparison of transnational families and in particular of migrant mothers and fathers (Mazzucato & Dito, 2018). This literature review summarises findings from different studies investigating the characteristics of female and male migrants and studies on transnational families that took the sex of the migrant parent into account, to inform our analyses.

| Paternal and maternal migration
Although most migration scholarship has historically focused on men, as early as the 1960s, women made up almost half of the share of international migrants and the number of female migrants has been rising since (Dreby & Adkins, 2010). In the 1990s, feminist scholars noted that more women were migrating independently in search of work and that an important share of these women were mothers. One of the main drivers for these changes has been the global division of labour leading to a demand for cheap female labour (domestic and care work) from the Global South in the North (Hondagneu-Sotelo & Avila, 1997;Parreñas, 2000Parreñas, , 2001. Central to these qualitative studies, conducted for the most part in Latin America and South-East Asia, was the hardship and emotional difficulties these mothers faced because they did not conform to the norm of biological motherhood and copresent parenting (Madianou & Miller, 2012;Parreñas, 2000). Moreover, these studies saw mother-child separation as detrimental to the child's well-being in the origin country and for their migrant mothers (Bernhard et al., 2009;Boccagni, 2012;Dreby, 2006;Parreñas, 2001Parreñas, , 2005Suarez-Orozco, Todorova, & Louie, 2002).
Unlike maternal migration, scholars did not see paternal migration as an abrupt break in family life, and therefore, effects for children and fathers were seen as minimal. Hence, transnational families in which fathers migrated received little attention (Haagsman & Mazzucato, 2014;Jordan, Dito, Nobles & Graham, 2018;Poeze & Mazzucato, 2012;Pribilsky, 2004;Waters, 2009). Yet, although mothers are increasingly forming transnational families by independently migrating, it seems that it is still more common for fathers to do so, although exact figures are unknown (Dreby & Adkins, 2010;Suarez-Orozco et al., 2002). It is therefore surprising that there is still limited research on the effects of paternal migration. Even more so because recent scholarship has shown that, in African and Latin American contexts, transnational fathers are engaged in the upbringing of their children, especially when they are still married to the biological mother, and that separation from their children also has negative effects on the emotional well-being of migrant fathers (Carling & Tønnessen, 2013;Poeze & Mazzucato, 2012;Pribilsky, 2004;Waters, 2009).
These studies have been fundamental for highlighting the issue of transnational parenting and the emotional toll that it can bring to parents. Yet, because of their small-scale, they only focus on transnational parents and cannot assess to what degree their findings are attributable to the transnational nature of parenting or to characteristics in a wider population (Mazzucato & Dito, 2018). Furthermore, they have not explicitly compared transnational mothers with transnational fathers. A recent body of literature has engaged large-scale data collection to conduct such comparisons and test the effects of parental migration on the children who stay at origin (DeWaard et al., 2018;Dreby, 2006;Graham & Jordan, 2011;Haagsman, 2015;Liu et al., 2009;Mazzucato et al., 2015;Parreñas, 2005;Wen & Lin, 2012;Wu & Cebotari, 2018). In line with ethnographic studies, some of these studies find that children in South East Asia and China have more difficulties when mothers migrate (Jordan & Graham, 2012;Liu et al., 2009;Wen & Lin, 2012).
Recent work also points to the importance of the caregiver in transnational families Peng & Wong, 2015).
Caregivers are central to the functioning of transnational families as they mediate between the child and the migrant parent. Having a stable caregiving arrangement where children stay with the same caregiver throughout the parent's migration period was an important factor for children's psychological well-being .
Although fathers mostly leave the children in the care of the biological mother, migrant mothers often opt for a female kin member such as their mother or sister (Åkesson, Carling, & Drotbohm, 2012;Banfi & Boccagni, 2011;Haagsman, 2015;Parreñas, 2005).
In sum, the literature has mainly studied the effects of the migration of mothers in transnational families, and the few studies that compared migrant mothers and fathers found that children were affected more when mothers migrated. However, the latter studies surveyed children in the country of origin and not their parents and could therefore not include many parental features as control variables. Consequently, the disparities between the characteristics of migrant fathers and mothers in transnational families remain a gap in the literature, and these are important to explain differences between mother-and father-away families. Moreover, the above literature review drew from cases around the world, but comparative research is limited. In this study, we compare migrants from different origin countries in order to investigate whether findings apply cross-nationally. The next section will draw on literature outside of transnational family studies to sketch the main differences found between the family trajectories of migrant men and women that may inform our investigation of transnational families.

| Differences in family trajectories of migrant men and women
Few studies have explicitly compared male and female migrants and their family behaviours (Kraus, 2017). Yet studies on migration and gender have shown how men and women have different motivations to migrate (Curran & Rivero-Fuentes, 2003;Kanaiaupuni, 1999;Massey, Fischer, & Capoferro, 2006), migration experiences (Hondagneu-Sotelo, 1999;Wong, 2006), opportunities in destination country contexts (Boyd & Grieco, 2003;Feliciano, 2008), social network characteristics (Curran & Rivero-Fuentes, 2003;Liu et al., 2018;Malmusi et al., 2010;Toma & Vause, 2014), and transnational practices (Goldring, 2001;Itzigsohn & Giorguli-Saucedo, 2005). Goldring (2001) found that Mexican men in the United States were more involved in citizenship practices oriented towards their country of origin. Kanaiaupuni (2000) found that education for Mexican women increases their propensity to migrate to the United States, resulting in Mexican migrant women in the United States being more highly educated than men (Feliciano, 2008). Effects of migration have also been found to differ for men and women. For example, migration generally benefits the career of the husband, whereas the wives' employment status is negatively affected irrespective of employment and income before migration (Boyle, Kulu, Cooke, Gayle, & Mulder, 2008).
Migrants can experience conflicting gender norms in the destination and origin country and within the migrant community. These gender roles can affect men and women differently (Gallo, 2006;Hill, 2004;Jolly & Reeves, 2005) and ultimately affect couples' relationships Fouron & Schiller, 2001;Mahler & Pessar, 2001). Hirsch (2003) found that some Mexican migrant women enjoyed greater freedom in the United States than in Mexico and, as a result, experienced more marriage instability. Migrant women experience a change in gender roles by taking up employment, sometimes becoming the main breadwinner. Their husbands do not always appreciate these new roles and can feel threatened in their masculinity, which can lead to stressed marital relationships (Charsley, 2005;Gallo, 2006;George, 2000;Manuh, 1999;Zontini, 2010).  found that in Ghanaian couples where the wife migrated independently, the risk of divorce was higher than the other way around.
Although these studies do not specifically focus on migrants in transnational family arrangements, they give indications that migration affects women and men differently. It is therefore important to investigate this for transnational families in particular, as it can give us important clues to explain the differences in effects of transnational family arrangements on men and women and on their children who stay at origin. By 2013, approximately 737,217 Ghanaians, 1,117,901 Nigerians, 518,711 Angolans, 1,306,026 Congolese, and 540,363 Senegalese had migrated to search for better opportunities abroad (World Bank, 2016). Although these figures do not include undocumented migrants, they indicate that the phenomenon of migration from our study countries is large. The patterns of movement and the composition of these migration flows differ between the five African origin countries. The economic, political, and environmental situation in many sub-Saharan countries led to more migrants entering European countries, resulting in a significant African presence in Europe. Although initial flows tended to follow colonial ties, throughout the latter half of the 20th century, African migration diversified to more destinations (Grillo & Mazzucato, 2008).

| BACKGROUND
Two countries in our study, the DR Congo and Angola, are postconflict countries, although the eastern regions of the DR Congo still experience regular violence. As a consequence of these wars, large migration and refugee flows emerged, leaving families scattered both within and outside the country (Flahaux & Schoumaker, 2016;Grassi & Vivet, 2014). Nigerian and Ghanaian migration was more specifically caused by the economic crises during the 1980s in both countries, when many migrants left for Europe, sparking chain migration and family reunification later on (Adepoju, 2004;Mazzucato, 2008;Schans, Mazzucato, Schoumaker, & Flauhaux, 2013). Senegalese migration took place initially mainly to France in the 1960s in response to the so-called guest worker schemes (Pison, Hill, Cohen, & Foote, 1997). During the 1980s, Senegal also experienced economic difficulties, substantially increasing migration from Senegal to Europe, and also diversifying European destinations (Sakho, Beauchemin, Schoumaker, & Flahaux, 2013).
The share of female migrants varies greatly by regional and national origin. In particular, the feminization of migration has been much more pronounced in migration flows from Ghana, Nigeria, and DR Congo compared with Senegal and Angola, with women from these former three countries migrating increasingly independently instead of joining or following their husbands' migration (Adepoju, 2004;González-Ferrer et al., 2014;Schoumaker et al., 2015).
These differences in the composition of migration flows can be partly explained by the organisation of family life. Family life in many African countries shares certain characteristics, such as the importance of the extended family, strong division between gender responsibilities, multilocal residence of spouses and children (not living in the same house), and loyalty towards own lineage members over conjugal bonds (Locoh, 1989;. Although exact numbers on child fostering are rare, African Demographic and Health Surveys showed that 9% to 35% of the households foster children that do not live with either parent (Pilon & Vignikin, 2006). There are also differences between these origin countries. In particular, different gender norms and relations shape women's freedom of movement, with countries where gender hierarchy is stricter and women engage in less independent migration (e.g., Senegal), compared with countries where women have independent household budgets or come from matrilineal kinship systems, both of which give women more freedom of movement (e.g., DR Congo, Ghana, and Nigeria; Adepoju, 2004;Schoumaker et al., 2015).
Although polygamy is practiced in all five countries, the prevalence and social acceptance differs: Polygamous marriages are less common in Ghana and DR Congo compared with Nigeria and Angola, and it is most widespread in Senegal (Caarls, Mazzucato, Schans, Quartey, & Tagoe, 2015;Findley, 1997;NDHS, 2008;Nzatuzola, 2006). Even though divorce is prevalent in all five countries, it is most frequent in Ghana and DR Congo (33.2% and 28.8% of first marriages ended in divorce, respectively) and less frequent in Nigeria and Senegal (11.8% and 19.7% of first marriages ended in divorce, respectively; Clark & Brauner-Otto, 2015). In Table 1, we summarise the main differences and similarities between these five origin countries.
As a consequence of different origin contexts, migration patterns, family organisation, and gender roles in countries of origin, family arrangements occurring after migration might also differ-both between origin countries and between men and women. When exploring the different forms, characteristics, and trajectories of transnational families, these differences can help to contextualise our findings on migrant mothers and fathers. were sampled. Table 2 shows the characteristics of the sample.

| Methods and analytical samples
We analyse two types of family arrangements: transnational families (meaning that at least one child lives at origin); and nontransnational families (in which all children live at destination with the mother or father who was interviewed). These categories do not distinguish if parents migrated independently, jointly, or in step-wise fashion. The statistical analysis consists of two parts, each one using a different analytical sample and methodological approach. In the first part, we present descriptive statistics on socio-economic, demographic, and migrationrelated characteristics of transnational and nontransnational parents.
Moreover, we compare transnational and nontransnational parents on grounds of different family and migration-related indicators using statistical significance tests. In the second part, we apply sequence analysis to examine the differences in family formation trajectories from age 18 to 35 of transnational mothers and fathers only. In this second part, we use a reduced sample of transnational parents, as will be explained in more detail below. Our analyses are descriptive with the aim of understanding associations given the lack of knowledge on the basic characteristics of transnational families to date.
We do not intend to test causal relationships. Next, each of the two parts of the analysis will be explained.

| Part 1: Comparing transnational and nontransnational parents
In the first part of the analysis, we explore characteristics of transnational mothers and fathers and use nontransnational parents, that is,  (Feliciano, 2008).
Activity (working, unemployed, other): Labour market opportunities and economic integration depend on the destination country context and have been found to be different for men and women (Boyd & Grieco, 2003;Feliciano, 2008).
Duration of residence at destination (1 to 2 years, 3 to 5 years,

| Part 2: Sequence analysis of transnational fathers and mothers
In the second part of the analysis, we focus only on transnational parents and conduct a sequence analysis to study differences in family  (Abbott, 1995;Billari, Fürnkranz, & Prskawetz, 2006).
We analyse family formation histories (i.e., the formation and dissolution of relationships and having children) between 18 and 35 years of age of those who were transnational parents at time of interview.
The studied age bracket between 18 and 35 captures the respondents in their young adulthood, which is the age range most people start relationships. Sequence analysis is best suited for sequences that are complete and of equal length (Robette & Thibault, 2008); therefore, we had to restrict our sample to respondents for whom we had information about the entire 18-year period. Respondents who were younger than 35 at the time of the survey were dropped. Furthermore, the sample was restricted to individuals who already had a child at age 35 (see "reduced sample" in Table 2). To create the alphabet to perform sequence analysis, we combined two variables: relationship status and whether the respondent has children. 4 The final alphabet contains  Significance levels: *p < .1;**p < .05;***p < .01. six different states: (i) Single, no child; (ii) single, with child; (iii) in a relationship, no child; (iv) in a relationship, with child; (v) single, divorced/separated, no child; and (vi) single, divorced/separated, with child. The differences in family formation trajectories between mothers and fathers are calculated using a dissimilarity matrix (Studer, Ritschard, Gabadinho, & Müller, 2011). Table 3     with previous studies, men seem to migrate more often alone than women, leaving their wife and children in the country of origin (Dreby, 2006;Hondagneu-Sotelo & Avila, 1997;Kraler, Kofman, Kohli, & Schmoll, 2011). Women more often join their husbands when they migrate, and when they migrate alone, they are more often single mothers and in a transnational family arrangement.

| FINDINGS
In line with previous findings, the mother takes up the care of the children when the father migrates, whereas female kin members provide care when mothers migrate (Åkesson et al., 2012;Banfi & Boccagni, 2011;Haagsman, 2015;Parreñas, 2005). This is true for most of the origin countries, with the exception of the DR Congo.
Transnational parents more often have children from multiple relationships, with the exception of Senegal, where there is no difference between transnational and nontransnational fathers. Although formation, and they have the first child earlier, which is not specific for migrants, but corresponding to general demographic behaviours (Fussell & Furstenberg, 2005). Other cross-country differences that are present in the general population in Sub-Saharan Africa can be observed also in the trajectories of transnational mothers and fathers (e.g., age differences between men and women at first marriage, or the share of divorced; Clark & Brauner-Otto, 2015).
Furthermore, single parenthood after divorce is relatively frequent among Angolan, Ghanaian, and Nigerian parents and among Senegalese and Congolese mothers in transnational families. Only Senegalese and Congolese fathers seem different in this respect, as there are only very few cases of divorced fathers with a transnational family arrangement. Although rather rare or non-existent in most of the countries, singles with children, who have not been in a relationship before, seem relatively common in the DR Congo. Again, scattered families among migrants from Angola and the DR Congo are very likely to be the result of the conflict situations in these countries (Flahaux & Schoumaker, 2016;Grassi & Vivet, 2014).

| DISCUSSION AND CONCLUSION
The recent and unique datasets on African migrant families living between Africa and Europe analysed in this paper shed new light on transnational families and offer possible avenues to explore explanations for why maternal migration seems to lead to poorer well-being effects for mothers and children. Until now, the predominant explanation for such findings centres on the strong affective ties between mothers and children and the gendered parenting norms that affect family functioning (Mazzucato & Dito, 2018;Parreñas, 2005). Our work shows that there are structural differences, not only between transnational and nontransnational families but also between male and female transnational families. These structural differences may offer important clues to dynamics that may be involved in affecting mothers and children negatively other than gendered parenting norms.
Our findings point to several important structural differences between transnational families and nontransnational families and between mother's compared with father's transnational families that correspond closely to what family sociologists find make "fragile families" among nonmigrant families (Boynton-Jarrett, Hair, & Zuckerman, 2013;Bzostek & Beck, 2011). Transnational mothers are more often single or in an unmarried union, compared with nontransnational mothers and to transnational fathers. In line with previous studies (Nzatuzola, 2006), single mothers are most frequent among Angolans.
This may be related to the aftermath of the Angolan civil war leading to scattered families (Grassi & Vivet, 2014). Irrespective of their marital status, when women migrate independently, they are more often in a transnational family arrangement.
There are several possible explanations. Mothers might migrate alone because they are single and need to search for economic means to raise their children. High-income requirements for family reunification in many of the European countries studied are difficult for single parents to meet when employed in low-paying sectors Caarls & Mazzucato, 2016;Eremenko & González-Ferrer, 2018;Kraler, 2010). Being single also makes it difficult to raise children while working full time, so women have been found to leave children at origin in the care of their own mothers or sisters as a preferred solution (Åkesson et al., 2012;Banfi & Boccagni, 2011;Parreñas, 2005;Poeze & Mazzucato, 2016). On the contrary, transnational fathers mostly have their child's mother living with the child at origin, which is in line with previous studies stating that mothers take care of the children if the father migrates.
Our findings further showed that transnational parents are more likely to have children from multiple relationships (except for Senegalese parents). Several factors may influence this. The fact that they are more often single means they can start new relationships and new families or because transnational relationships are more vulnerable than relationships where the partners live in the same location Pribilsky, 2004). Senegalese and Ghanaian transnational parents have lower educational attainment than their nontransnational counterparts, which might be related to the importance of socio-economic status as a precondition to successfully reunify with one's children or to form a family at destination, at least for these two origin countries (Beauchemin, Nappa, et al., 2015;Caarls & Mazzucato, 2016;Eremenko & González-Ferrer, 2018). Finally, nontransnational parents have lived in Europe longer than transnational parents. Being in Europe for a longer period allows migrants the time to meet the prerequisites and go through the lengthy procedures for reunifying with children at destination and for reunifying with the partner in order to form a family Kraus & González-Ferrer, 2016).
Overall, this study points to three main conclusions. First, transnational families differ from nontransnational families on important sociodemographic and socio-economic characteristics: single parenthood, children from multiple relationships, and lower educational attainment. Second, these findings pertain particular to transnational mothers, as they more frequently face single parenthood and more often are divorced than transnational fathers. In line with more "mainstream" families, transnational mothers also start family life earlier than transnational fathers. Third, this study demonstrates the importance of taking migrants' origin context into account (DeWaard et al., 2018;Mazzucato & Dito, 2018). Family and gender norms in the country of origin and the political context (civil war) not only shape the composition of the migration flow but also seem related to differences in transnational family life and family formation trajectories across groups from different countries of origin.
Our study identifies more issues related to transnational family life and family formation and dissolution trajectories that should be addressed in future studies. Our aim was to understand some of the main structural differences between transnational and nontransnational families, and the differences between father-and mother-away families. Yet the sample sizes have limited the possibility of carrying out multivariate analyses. Future studies can investigate how individual characteristics as well as migration-related characteristics, such as the timing and duration of migration and the specific country of destination, influence these different family formation trajectories of father-and mother-away families.
Notwithstanding these limitations, this study is among the few, together with DeWaard et al. also in this issue, that examined transnational family life across five different migration flows, taking into account the gendered differences within these families. Importantly, it draws questions around the prevailing explanations in policy discourse and academic studies that point to the particularly strong bond between children and mothers, and gendered parenting norms as the reason for the worse well-being and emotional toll of migrant mothers and their "stay-behind" children. Rather, this study indicates that there are important structural differences in the form and characteristics of mother-away versus father-away families that put women at a disadvantage when it comes to conducting their family life at a distance. The findings reveal more fragile circum-

ENDNOTES
1 For more information, refer to www.tcra.nl 2 For more information, refer to http://mafeproject.site.ined.fr/en/ 3 Additional analyses showed that respondents in the MAFE and TCRA Ghanaian flow had similar characteristics. We therefore do not anticipate problems with pooling these datasets.