Engaged parenting, gender, and children's time use in transnational families: An assessment spanning three global regions

Abstract Global circuits of migration regularly separate parents from children. How families navigate this separation has changed markedly. The sharp decline in the cost of international communication makes possible new forms of transnational parenting. In many contexts, migrants are now actively engaged parents, involved in decisions, knowledgeable of children's schooling, employment, and activities, and in some cases, even conversant face‐to‐face with children via videoconferencing. These practices, however, are not universal. We use data from surveys in three countries to document the frequency and variability of intensive, engaged transnational parenting in the diverse global regions of Asia, Africa, and the Americas. We then ask whether the organisation of children's lives—specifically, time allocated to school homework, leisure, and household chores—varies by the degree to which migrant parents stay connected to sending homes. The gender of the migrant parent, stay‐behind caregiver, and the gender of the child emerge as explanatory factors for engaged parenting and children's time use. However, and unexpectedly, in the Philippines, migrant mothers are less likely to practice engaged parenting. In sending households, girls in two of the three countries spend more time doing household chores than boys, but parental migration does not mitigate this difference. Although we find some evidence of more traditional gender practices, we also find exceptions that suggest potentially fruitful avenues for future research.

These experiences are far from universal, and whether, when, and why migrant parents stay intensively connected with sending households remains relatively underexamined. Although economic resources likely shape patterns of engagement, well-resourced migrants are not equally engaged in sending community life. Other barriers to communication for migrants exist, including those from employment responsibilities and new domestic responsibilities, and vary widely across origin and destination countries. Further, gendered expectations and norms around how mothers and fathers should be involved in parenting, as well as the gender of the stay-behind child, may influence the practice of transnational parenting.
The implications of a shift towards connected transnational parenting-for union longevity, migrant, and children's welfare alikepromises to be an important avenue of research (Graham, Jordan, & Yeoh, 2015;Nobles, 2011). Decades of scholarship on families divided by other processes (e.g., divorce, separation, and deployment) have emphasised the value of maintained connections for family outcomes, particularly when the relationship between the separated parent and child/children's caregiver is amicable (Amato, 2000;Carlson, 2006).
One of the primary motivations for parental migration is to enhance their children's life chances and schoolwork is a major topic of conversation when parents contact children (Asis & Ruiz-Marave, 2013). Money earned abroad often pays for better, or additional, schooling, and the general expectation is that children in transnational families will apply themselves to their studies both in and after school.
However, parental migration may also require a child to contribute more to the household to cover the tasks previously undertaken by the migrant parent, putting pressure on the time available for study. This reallocation may be gendered, as when daughters are required to take on housework or the care of younger siblings in the absence of their migrant mother. Not only is the way that children's time is distributed across different activities likely to vary by child age and household, but it may also vary by how actively involved migrant parents are in the everyday lives of their children. There is currently very little empirical evidence on the time use of children in transnational families.
This study contributes to addressing this gap by investigating the gendered dimensions of migrants' parenting and children's time use in diverse global regions. First, we use data from sample surveys in three different countries to document the frequency and variability of intensive, engaged transnational parenting, and whether this varies by migrant gender. We then ask whether time allocated to school homework, leisure, and household chores-varies by the degree to which migrant parents stay connected to sending homes. We define engaged migrant parenting as occurring when migrants have a high remittance intensity (contribute financial remittances frequently or, for Mexico, to a degree that covers most of children's expenses) and communicate with children in sending homes at least weekly. We maintain that understanding variability in migrants' opportunities to-and choices to-adopt the role of engaged parent has much to tell us about the implications of transnational family arrangements. By comparing transnational families in Southeast Asia, sub-Saharan Africa, and Latin America, we aim to provide further insight into the diversity of arrangements for "parenting from a distance" (Ambrosini, 2015) and its associations with time use among nonmigrant children.

| TRANSNATIONAL PARENTING AND CHILDREN'S TIME USE: CONCEPTUALISATION AND VARIABILITY
The study of families separated by borders has grown substantially, shifting debates about what constitutes "doing family." Scholars of transnational families have questioned the presumption that physical proximity is necessary for the maintenance of familial ties (Baldassar et al., 2016;Suárez-Orozco & Páez, 2008), capturing diversity in the practices of presence and intimacy within transnational families (Baldassar, 2008;Brownlie, 2011;Diminescu, 2008). We build on these studies in defining parental engagement to reflect two key aspects of transnational parenting: providing material resources through remittances and participating in daily life through contact.

Significant changes in interpersonal communication technologies
(ICTs) over the past decade (Chib, Wilkin, & Hua, 2013;Madianou & Miller, 2011) make it possible for migrant parents to be actively involved in their children's lives. Contact has thus become an important dimension of "doing transnational family." Financial and social engagements (contact) are often related in transnational family practices (Levitt & Jaworsky, 2007;Mahler, 2001) and, in contrast to the more negative effects of other types of family separation such as divorce, are more likely to be sustained over time (Nobles, 2011).
To date, more migration scholarship is devoted to remittance behaviour than to communication and active parenting, partially in response to the central role of remittances in theoretical models of migration as a family economic project (Lucas & Stark, 1985;Sana & Massey, 2005). The destination context is a primary predictor of remittance behaviour, largely because of its impact on migrant's labour market opportunities, and the possibilities it offers for the regularisation and integration of migrants (Carling, 2008). For example, among Mexican migrants, education, income, and documentation status are all positively correlated with remittance frequency and monetary value (Goldring, 2004;Valentine, Barham, Gitter, & Nobles, 2017). Generally, the remittance amounts sent by migrant mothers are less than the amounts sent by migrant fathers, probably as a result of women's more limited labour market opportunities and lower earnings. Nevertheless, despite being structurally disadvantaged, women often remit a larger share of their income (Abrego, 2009).
Remittance sending is typically defined within the gendered expectations of migrant parenting. Mothers' feelings of guilt may influence their remitting, even at the cost of their own essential needs (Basa, Harcourt, & Zarro, 2011;Schmalzbauer, 2004). Similarly, remittance sending may be considered fundamental by fathers as it is normatively tied to male breadwinning roles (Dreby, 2006). When fathers fail to remit, expectations (of both parents) are unmet and this can lead to transnational family dissolution (Dreby, 2010;Haour-Knipe, 2011).
Characteristics of children and their caregivers may also influence whether and how frequently remittances are received, although Nobles (2011) found no evidence that the age and gender of nonmigrant children in Mexico influenced the financial contributions of migrant fathers. Others have emphasised that gender bias may be more evident in the allocation of remittances (Antman, 2012;Bouoiyour & Miftah, 2016). In some contexts, a bias in favour of girls and younger boys has been found in the spending of remittances on education (Acosta, 2006) and health care (Lopez-Ekra, Aghazarm, Kötter, & Mollard, 2011), whereas other studies have found the opposite effect, with a bias in favour of boys (Hu, 2012;Lu & Treiman, 2007). Some evidence suggests that the relationship between the caregiver and the migrant parent shapes remittances. Divorce, for example, has been found to negatively affect migrant parent-nonmigrant child relationships and also influence the flow and use of remittances (Dreby, 2007).
Knowledge of how nonfinancial aspects of child rearing are accomplished from a distance is mostly based on ethnographic work.
Some studies indicate that the lack of face-to-face contact constrains parent-child intimacy (Boccagni, 2012;Laurie, 2008). Others stress the role ICTs play in enabling a meaningful relationship between parents and children (Cabanes & Acedera, 2012;Haagsman & Mazzucato, 2014;Peng & Wong, 2013). As separation due to migration is the "new normal" in some settings, advancement in ICT has been central to transnational family relationships, leading to a "de-demonisation" of distance (Baldassar et al., 2016). However, these influences vary depending on social class, gender, children's age, notions of family access, and skills to use new technologies (Baldassar, 2007;Madianou, 2016;Madianou & Miller, 2011;Parreñas, 2005Parreñas, , 2008. The current investigation seeks a broader understanding of transnational parenting practices and children's lives. Rather than focusing solely on remittance sending, which captures a single dimension of parental involvement, we combine the provision of financial resources with contact frequency between migrant parents and their nonmigrant children to define engaged parenting. We hypothesise that engaged parents are more likely than other migrant parents to influence the organisation of their children's daily lives. As a co-resident parent plays an important role in managing their child's weekly routine, an engaged migrant parent may well fulfil this role remotely (Fresnoza-Flot, 2009).
Only a few studies suggest a link between parental migration and children's time use, with equivocal results. Some find that children of migrant parents spend more time on school homework (Botezat & Pfeiffer, 2014), whereas others find that girls especially spend less time on education. Girls are more likely to substitute household chores for school homework, whereas boys are more likely to substitute leisure activities for homework (Nguyen, 2016;Pörtner, 2016). In Mexico, children spend less time studying in the short-term after their fathers' departures, but study time is replaced with working for pay and teenage boys experience this shift to a greater extent than younger boys or than girls of any age (Antman, 2012). Engaged parenting may both relax the income constraint that leads families to substitute children's schooling with work and allow migrant parents to encourage children's school performance. We expect that, depending on age and gender, children with engaged migrant parents will spend more time on schoolwork relative to those with less engaged migrant parents.
The effect of parental migration on children's time use will depend in part on the context-specific organisation of children's lives. In general, children in low-income countries spend more time in both market and household work compared to children in wealthier countries (Hsin, 2007;Larson & Verma, 1999;Lloyd, Grant, & Ritchie, 2008). Parental migration may relax the income constraints that drive children's labour, but when mothers migrate from countries with strong gender norms, daughters might be expected to take over their household tasks (Asis, 2002), and this may be reinforced with the presence of an engaged migrant parent.
Other potential determinants of children's time use include household size and the characteristics of the child's caregiver (Maralani, 2008;Steelman, Powell, Werum, & Carter, 2002). Younger children with more siblings may receive less supervision from a co-resident parent or caregiver and thus spend more time on leisure. On the other hand, the absence of a parent may increase the importance of peer friendships for nonmigrant older children and thus lead to an increase in time spent with friends. Moreover, greater material resources (e.g., from remittances) may positively influence time spent in leisure (Larson & Verma, 1999). If remittances allow children in sending homes to substitute other activities for employment, whether this is school homework or leisure is likely to be context-specific.
Migration introduces considerable stress into transnational families, and children's lives are likely to be shaped in important ways by the mental health of nonmigrant caregivers (Graham et al., 2015;Jordan & Graham, 2012;Waldfogel, Craigie, & Brooks-Gunn, 2010).

| DATA AND METHODS
Conducting a comparative study across different contexts presents a number of challenges (Mazzucato & Dito, 2018). One critical challenge is operationalizing the conceptual domains across unique data sets to allow cross-country comparison. We use the same measures when possible, with a few exceptions due to data uniqueness (see Appendix S2 for bivariate distributions: "engaged parenting" by each measurement domain for each study country). Further, although each dataset contains the most detailed information available for the study of transnational families, they have different sampling strategies and coverage that impose constraints on comparative analysis. The implications for interpretation and relevant sensitivity tests are considered below.   Table 1 summarises selected characteristics of these surveys for the subsamples included in the current study.

| Data sources
Each survey employed a different sampling strategy. In CHAMPSEA, eligible households were either (a) transnational (one or both parents working overseas) or (b) nonmigrant (both parents usually resident at the same address as the index child) for at least 6 months prior to interview. Sampling followed a three-stage design, with flexible quotas defined by household migration status, child gender, and child age for two groups of children aged 3, 4, and 5, and 9, 10, and 11. The samples include approximately 1,000 households in each country and one index child per household. They are not nationally representative, but, due to tightly specified protocols, they are replicable (see Graham & Yeoh, 2013, for further details).
For the current subsample, 65% of children had migrant fathers, 24% had migrant mothers, and 11% had both parents migrant. In the Nigerian TCRAf-EU data, the school-based sample from 25 selected schools was stratified by school quality (public/private and junior/ senior secondary). One classroom from different grades was included and purposive sampling ensured a sufficient number of children with migrant parents (see Mazzucato et al., 2015, for further details). The dataset, although not nationally representative, contains information on a total of 2,168 children. For the subsample used here, children with migrant fathers (58%) are more common than those with migrant mothers (12%) and those with both parents migrating (29%).

| Method
The analysis proceeds in two stages. For the study sites in which multiple children from a household are included, parenting inputs from migrants are recorded for each child individually.
We begin by examining the prevalence of engaged parenting among migrants whose children reside in each study country. We define engaged parents as those who call children at least weekly (for all three studies) and who practice high remittance intensity, either sending remittances "frequently/regularly" 1 (for CHAMPSEA and TCRAf-EU or in amounts that "cover most of children's expenses" (for S3). In each of the three study sites, the sending household reports this information. Among 9-to 11-year-olds (CHAMPSEA) and 9-to 10-year-olds (MxFLS), the child's primary caregiver reports this information. Among 10-to 14-year-olds (TCRAf-EU) and 11-to 14-year-olds (MxFLS), the child reports the information. We recognise that the difference in who reports may influence the results (see Jordan & Graham, 2012). Single-year age controls are used to adjust for systematic variation in reporting.
To investigate the main determinants of engaged parenting (Stage 1), we assess within-and across-population variation in engaged parenting among mother and father migrants. We regress the dichotomous engaged parenting indicator on a set of child, migration, caregiver, and sending household-specific characteristics ( Figure 1). The measures were selected based on prior literature and common availability across the three surveys for the key measurement domains (see Appendix S1). school homework and leisure are measured in categorical units (less than an hour per day, 1-2 hr, 2-3 hr, 3-4 hr, 4-5 hr, and more than 5 hr). To enhance comparability and retain the greatest possible level of detail, we assign midpoint minutes to the outcomes in (TCRAf-EU).
Doing so assumes that the distribution of minutes within the category is either uniform, normal, or otherwise has a mean of the midpoint. We have no reason to believe time use follows alternative distributions.
We regress time use on the indicator of engaged parenting adjusting for the controls described above. In study sites capturing families with mothers or fathers (or both) absent, we introduce interactions between the engaged parenting indicator and whether or not the migrant is the child's mother or father. We include each of the  We now turn to our two research questions. The characteristics of the caregiver (most likely to be the mother), including age, mental health, and education, are important   Table 2ac, Model 1).

| Engaged parenting and children's time use
Stage two considers the relationship between engaged parenting and children's time allocation. There are a variety of ways in which engaged parenting might influence the amount of time children spend  There is also an indirect migration effect in the Philippines, whereby children whose parents are working overseas in less common destinations are significantly more likely to spend time doing school homework compared to children whose parents are working in the Middle East, Asia, or seafaring (

| Leisure
Model 3 (Cabanes & Acedera, 2012;Chib et al., 2013). We observe this result first in our bivariate analyses ( Figure 2), and then in the multivariate analyses for the determinants of engaged parenting (Table 2aa-c, Model 1). In our sample, Filipina migrant mothers are taking a less active role in parenting from a distance than some scholarship suggests (Madianou & Miller, 2011). This might be due to structural barriers reducing the opportunities for migrant mothers to contact their families back home, but further examination of occupational type and documentation status failed to support this suggestion (results not shown). As Eremenko and Gonzalez (2018) demonstrate, structural factors of destination and origin contexts are influential determinants of transnational family dynamics.
Given current dataset limitations, we are not able to conduct a more detailed analysis of the possible meaning and influence of financial constraints on migrant mothers' contact patterns in this study.
It is also possible that the relationship between the migrant parent and the stay-behind caregiver influences engaged parenting.
Perhaps when the child is in the primary care of the co-resident father, the migrant mother feels more confident about the child's well-being and therefore feels less need for frequent contact, although Mazzucato et al. (2015) found that Angolan migrant parents in Europe experienced lower emotional well-being when the caregiver of the child in Angola is the other biological parent. Alternatively, the reverse causal pathway may operate, with qualitative aspects of the relationship between the migrant mother and stay-behind father inhibiting contact. Maternal migration is sometimes referred to as "Filipino divorce" (Timmerman, Martiniello, Rea, & Wets, 2015) and the marriages of some couples in the sample may be under strain.
The available data do not include information on the quality of relationships. Qualitative findings by Manuh (1999) and Schmalzbauer (2004) indicate the tensions and distrust that couples experience due to separation following international migration, and such tensions could limit engaged parenting. Formal divorce in Mexico has been rare (Frank & Wildsmith, 2005) until recent increases in the last decade (Arias, 2013 (Hsin, 2007;Larson & Verma, 1999) and the few studies that have examined time use and parental migration (Nguyen, 2016;Pörtner, 2016). In societies with gendered norms about the division of labour within families, time spent doing household chores may be greater among girls (Nguyen, 2016 and/or that children respond to the "sacrifice" of family life made by their mother and father by studying harder to please absent parents. Additionally, in the Philippines, children whose migrant parents are in less common destinations are more likely to spend more time doing school homework. These destinations account for about 18% of the total sample and include places in Europe and North America.
Such destinations could be associated with increased financial security for migrants because of higher earning potential. These settings also offer higher returns to human capital that make educational qual- Overall, our study offers a range of insights into the practices of transnational families within three global regions of significant international out-migration. It is not, however, without limitations. We have already noted the limitations imposed by the lack of comparable financial data that could allow greater specification of household wealth.
An added challenge to comparability comes from the different sampling designs and content of the three surveys, as well as child versus adult reporting on key measures, which could influence the findings. In particular, the school-based Nigerian survey raises issues of children's knowledge about their parents' remitting behaviour. Another notable limitation is the lack of precise measures for comparison (e.g., on migration and caregiver characteristics including caregiver mental health). Finally, the data used in this study were collected between 2008 and 2010. Since then, the cost of communication technology has decreased further whereas simultaneously the methods of ICT have increased significantly. How this has shaped the entry into and maintenance of transnational parenting will be an important avenue for future research. Our assessment of existing scholarship, along with the results presented here, suggests that some barriers to staying connected to children in sending households will not be easily overcome by reductions in communication costs. Nevertheless, it is possible that the prevalence rates of "engaged parenting" presented in this study underestimate rates observed today.