Family Obligation Attitudes, Gender, and Migration

Abstract

This study focuses on attitudes related to fulfilling family obligations and their relationships to migration behavior. We hypothesize that men who highly value fulfilling family obligations will be more likely to migrate in order to fulfill material obligations while women who highly value fulfilling family obligations will be less likely to migrate in order to fulfill care obligations. The empirical analysis examines data from the Chitwan Valley Family Study, located in south-central Nepal. We test whether variation in how much individuals value putting family needs before individual needs and caring for their adult parents matter for whether they migrate at all and if so, to which specific destinations. Our results provide only moderate support for these hypotheses but uncover patterns in how these attitudes toward family obligations are related to migration destinations. Men with strong attitudes toward family obligations are more likely to migrate internationally but especially to nearby India, sacrificing some level of economic returns for proximity. For women, the effect of attitudes is consistent: putting family needs first is negatively related to migration, while caring for adult parents is positively related to migration to India but not domestic or other international destinations. The findings suggest that our conventional typology of gendered labor and gender expectations for masculine breadwinning and feminine care might too strictly dichotomize the reality of how people actually care and provide for their families, obfuscating how they negotiate these competing demands.

1. Introduction

A long history of migration scholarship has identified how family context influences individual migration decisions, framing these moves as either altruistic for the family’s greater good (Stark & Bloom, 1985; Stark & Taylor, 1991) or contested by individuals who feel constrained by rigid gender expectations within family roles (Hondagneu-Sotelo, 2011; Curran 1995). Gender and migration scholars have since theorized how family contexts frequently gender individual family members, imposing a framework that shapes who performs which types of labor, i.e. domestic or otherwise, and where, i.e. within or outside the home. The consequences of these gendered family expectations often translate into gender-specific migration patterns, with men frequently migrating for paid work and women either remaining at home or migrating for domestic or other feminized forms of labor. However, individuals vary in terms of how much they personally feel obligated to fulfil these expectations. Even in cases where family norms, gender expectations, and other theoretically relevant factors are comparable, individuals often make quite different migration decisions pertaining to whether to move at all and, if so, to which destinations. Thus, the present study aims to examine whether individually held attitudes towards fulfilling family obligations produce differences in migration behaviors.

Attitudes—and cognitive factors as a whole—have been shown to influence a number of behaviors, including decisions to cohabit, spousal choice, and educational attainment (Bachrach, 2014; Bachrach & Morgan, 2013; Frye, 2012; Johnson-Hanks et al., 2011; Lizardo, 2017; Miles, 2015; Vaisey, 2009). In alignment with this scholarship that seeks to reintroduce the individual/micro into an understanding of how structural/macro determinants affect behavior, this study looks at the relationship between individually held family attitudes—specifically attitudes towards one’s family obligations—and migration. While gendered norms and expectations involving family obligations might drive overall migration patterns so that, for example, men migrate more than women on the aggregate, we hypothesize that the extent to which individuals agree with and feel the need to adhere to such expectations likely mediate those influences.

For the empirical portion of the study, we use detailed data containing information on migration behavior, sociodemographic characteristics, and attitudes towards fulfilling family obligations from the Chitwan Valley Family Study in Nepal (CVFS). Our analysis leverages event history, multinomial logistic regression models to examine whether variation in how much individuals value prioritizing family obligations affect migration at all, and if so, to specific destinations that allow them to support their attitudes. In settings where norms uphold a gendered division of labor in which men are expected to work and women are expected to remain within the home, we specifically hypothesize that holding attitudes that prioritize fulfilling family obligations will result in men migrating more and women migrating less than their respective counterparts who hold less supportive attitudes towards fulfilling family obligations. In order to examine this relationship between family obligation attitudes and gendered migration behaviors in greater detail, we also present results from an analysis that disaggregates migration by destination types.

We found that men with supportive family obligation attitudes were more likely to migrate internationally than domestically but, surprisingly, only significantly to India rather than to the more remunerative options of the Persian Gulf or wealthy Western and other Asian countries. For women, the effect of attitudes varied: putting family needs first is negatively related to migration, while caring for adult parents is positively related to migration to India but not domestic or other international destinations. Overall, the results detail a complicated yet theoretically important relationship between individual attitudes towards family obligations and migration. Attitudes may either amplify or dull the effects that factors like individual gender and normative social contexts place on prospective migrants. More interestingly, though, the findings suggest that our conventional typology of gendered labor and gender expectations for masculine breadwinning and feminine care might too strictly dichotomize the reality of how individuals actually care for and provide for their families. Gender, after all, is elastic in its mandates and capable of being negotiated. Our interpretation of individual attitudes towards family obligations as aligning with gender norms might in fact elide a far more complex reality of how both men and women themselves understand their family obligations to be both material- and care-oriented, albeit still to varying degrees.

2. Considering Gender and the Family in Theorizations of Migration

Early theorizations in the neoclassical school on the causes of migration largely employed economic reasoning to understand why people move. The rationale was that people overwhelmingly migrated because they wanted to make a better living for themselves through increased wages. Implicit to these explanations was a gendered assumption of who these people were (i.e. working-age men) that ignored the historical record of families who migrated together, women who migrated for feminized types of labor, and others who migrated for reasons unrelated to work. Later scholarship attempted to address these gaps, developing theories that took into account community effects like social networks and cumulative causation as well as the effects of global inequality from capitalist production on labor migration (Massey, 1990; Massey et al., 2010; Massey et al., 1993, 1994; Massey & Espinosa, 1997; Taylor, 1986).

In particular, when the new economics of labor migration (NELM) framework emerged in the migration literature, it sought to address the shortcomings of neoclassical models that only accounted for expected wage differentials as a cause of migration (Stark & Bloom, 1985; Stark & Levhari, 1982; Stark & Taylor, 1989, 1990; G. C. Taylor, 1986; J. E. Taylor, 1987). Instead, the NELM perspective broadened the scope of migration decision making to the household level, arguing that individuals consulted with their families in the decision-making process because the aim of migration was to both diversify household income portfolios and to gain access to capital to finance family consumption. We build on this foundation that migration is frequently a collective decision informed by family dynamics rather than by just the individual. However, we add to the NELM perspective by acknowledging that power differentials among family members to make decisions as well as the varying expectations towards different members regarding their obligations to the family unit must be considered when attempting to understand the relationship between migration, gender, and the family. While all family members may feel obligated in some way to their familial responsibilities, their responsibilities often differ based on gender expectations.

To understand how family members’ attitudes toward their perceived obligations may shape their migration behaviors, it is first necessary to engage with the gendered components of such obligations. Gender constitutes one way in which societies generally organize their social relationships, divisions of labor, and a host of other practices. The family is oftentimes the primary institution where gender makes itself most visible in terms of the specific forms that these social relationships and labor arrangements take. Though there was a dearth of scholarship prior to the 1970s, reviews conducted by Curran et al. (2005), Donato et al. (2006), Hondagneu-Sotelo (2011), Mahler and Pessar (2006), Manalansan (2006), Palmary et al. (2010), and Silvey (2006) chart a rich history of work during the past half century that has moved the field of gender and migration through three main emphases: (1) women’s absolute invisibility in which they were not even considered as potential migrants, (2) an “add women and stir” approach that merely acknowledged the existence of migrant women without developing new theories or models to understand them, and (3) a promising, contemporary theoretical place that investigates not just men and women but gender as fundamentally constitutive to the migration process.

Donato et al. (2006) sum up the third orientation by arguing that gender should be defined as relational, meaning constructions of maleness and femaleness must be understood in terms of how they operate as cultural foils to each other, in hierarchies of authority that are constantly being reinscribed and contested across individual, interactional, and institutional domains. Gender, thus, broadly structures masculine and feminine obligations so that they exist in relation to each other. Relatedly, we define family obligations as the “culturally-defined rights and duties that prescribe how family members are expected to care for and provide support to each other” (Diwan et al., 2011). In many contexts around the world, this oftentimes takes the shape of masculine obligations towards family breadwinning and feminine obligations towards family caretaking. With obligations understood in this way, migration within families can frequently be framed as a masculine endeavor as migration becomes the mechanism through which men can participate in the domain to which they are presumed to be entitled, i.e. public spheres of social life that facilitate breadwinning. Women, on the other hand, are not typically seen as rightfully entitled to migration in the same way because their primary duties are rooted within the home. This is of course not to say that women never migrate, but how and when this occurs often requires greater negotiation and is frequently considered behavior that goes against gender norms (Paul 2015).

This conceptualization of gender and its relationship to family obligations thus allows us to understand migration as a gendered process. The identities of the actors involved (e.g. as dutiful sons and daughters or good husbands and wives), the behaviors they are expected to enact (e.g. as competent providers or homemakers), and the possibilities for action available to them due to structural constraints (e.g. job opportunities within gendered labor markets) all influence family-level migration behaviors along somewhat predictable lines, with men tending to migrate more than women in many—though certainly not all—migration corridors. For example, previous research by De Jong (2000) and Paul (2015) demonstrate how family obligations and expectations shape migration behavior. In a study on migration in the Thai context, De Jong found that men and women in families with higher expectations that members should migrate were more likely to migrate themselves. Paul, however, notes that such expectations are often gendered, with men expected to migrate and women expected to stay and take care of the home. Paul further found that women in the Philippines who wanted to migrate but felt constrained in doing so by gender expectations needed to negotiate their migrations more actively with their families, casting their moves as part of their daughterly duties instead of potentially violating gender expectations.

As noted earlier, however, gendered family obligations by themselves fail to account for all variation in migration behavior. One reason why we propose this occurs is because individuals hold varying attitudes towards fulfilling these obligations—with some highly valuing fulfilling these obligations compared to others. We hypothesize that this variation has potential consequences for actual migration behavior. By engaging with the reemergence of ideational factors and behavior in the literature, we investigate this aspect of the gender, family, and migration relationship to bring the realm of attitudes, ideas, and the individual into theorizations of how social structures—including gender—shape action.

3. Attitudes towards Family Obligations, Gender, and Migration Behavior

The contention of the present study is that individually held attitudes—specifically as they concern fulfilling one’s family obligations—likely matter in the stay-move calculus in important ways that can account for some of this remaining variance. We draw upon a long line of work on the importance of cognitive factors for behaviors (Johnson-Hanks et al., 2011; Lizardo, 2017), including theorizing within studies of migration (Carling & Collins, 2017; Carling and Schewel, 2018; Collins, 2018; Creighton, 2013; Koikkalainen and Kyle, 2016; Koikkalainen et al., 2019; Migali and Scipioni, 2019; Schewel 2019; Thornton et al., 2019; Williams et al., 2014). For example, Thornton et al. (2019) discuss that the literature often casts people’s motivation behind migration decisions to be economic in nature, but they actually express varying levels of desire for material goods. It is this attitudinal variation that influences more materially aspirational individuals to migrate to more remunerative destinations compared to their counterparts who value material goods less. This study and other recent work demonstrate that people’s personal attitudes, values, beliefs, or expectations about a range of issues affect an individual’s migration behavior.

We might expect that family members who hold the most supportive personal attitudes towards gendered family obligations will be more or less likely to migrate according to their position within the family. Within social contexts in which men tend to work outside the home and women tend to work within it, men who most strongly value fulfilling family obligations should be significantly more likely to migrate compared to men who place less value on those obligations. Conversely, women who most strongly value fulfilling family obligations should be significantly less likely to migrate compared to women who place less value on those obligations. Thus, we present two gender-specific hypotheses for contexts in which norms expect men to work outside the home and women to work within it:

H1: Men who have very positive attitudes towards fulfilling family obligations will be more likely to migrate compared to men whose attitudes less strongly support those obligations, ceteris paribus.

H2: Women who have very positive attitudes towards fulfilling family obligations will be less likely to migrate compared to women whose attitudes less strongly support those obligations, ceteris paribus.

4. Gendered Family Obligations and Migration in Nepal

Our study focuses on the relationship between family obligation attitudes and migration in the context of Nepal. As a remittance-reliant economy that sends a large number of migrants along several major corridors, Nepal is an important case to study. Located mainly in the Himalayas that border China in the north and India in the south, east, and west, Nepal is home to a diverse geography and landscape. The country also has extensive social and cultural diversity: our generalizations about Nepali society should be taken as analytically necessary, yet partial, descriptions, which do not fully reflect the immense variation in experiences and attitudes across Nepal’s inhabitants.

Our study setting specifically covers the western Chitwan Valley in south-central Nepal. The administrative district of Chitwan borders India and is about 100 miles from Kathmandu. There is one large city, Narayanghat, while the rest of Chitwan’s population—like much of Nepal—lives in small, rural neighborhoods.

Nepal is predominately an agricultural country, with nearly two-thirds of the population relying on labor intensive, subsistence-based farming. In 2008, nearly 74% of women and 53% of men were engaged in agriculture (GoN, 2009). The study setting of Chitwan is no exception. Slightly over 80% of the households were involved in mostly subsistence farming, and a large majority of farmers follow crop-livestock mixed farming (Bhandari & Ghimire, 2013, 2016). Households keep cattle, buffalo, sheep, goats and poultry as well as grow rice, maize, wheat, mustard, and millet. Rural agriculture is transitioning more recently from animal or human powered farming toward mechanized, commercial farming. Many farm households now increasingly rely on farm technologies that use tractors, fertilizers, and pesticides.

The country’s only east-west highway (Mahendra Highway) runs through this study setting. The district also hosts the Chitwan National Park, a UNESCO World Heritage Site and tourist hub. The urban center of Bharatpur/Narayanghat in Chitwan is also linked to Kathmandu, the capital city. In addition, the establishment of the agriculture campus, the Institute of Agriculture and Animal Science (Tribhuvan University) in Chitwan during the 1970s has played an important role in the transformation of the district. More recently in 2010, the establishment of the Agriculture and Forestry University in the district has been instrumental in further transformation of the economy of the valley.

4.1. Caste and Ethnicity in Nepal

As Nightingale (2011) has noted, there is a staggering diversity in Nepal’s ethnic, religious, linguistic, and topographical makeup. According to the 2011 census, 81 percent of Nepalis were Hindu followed by Buddhists (9%), Muslims (4%), Kirati (3%), Christians (1%) and others (Central Bureau of Statistics, 2012). However, the pervasiveness of Hinduism throughout the country explicitly and implicitly plays an important role in organizing social relations. Caste in Nepal is not necessarily religiously specific. Due to a civil code that was established in 1854, every ethnic group—regardless of whether they previously ascribed to a caste system or not—was placed in a hierarchy of variable inclusion according to the Hindu caste system. So, while Nepal as a country houses 126 distinct caste/ethnic groups that do not necessarily identify as Hindu, all groups have been nonetheless subjected to caste differentiation¹. Caste and ethnic differences in our study area, the Chitwan Valley, can be crudely divided as explained below.

Brahmin and Chhetri peoples are generally considered “high caste Hindus,” enjoying access to prestigious work in government service and general political influence. They tend to hail from all regions of Nepal and have had the greatest economic and social resources made available to them (Bista, 1991; Dahal, 2003; Bennet and Govindasamy, 2008).

Newars comprise an ethnic group that has traditionally claimed its origin in the Kathmandu Valley (Gellner, 1986). Since this region operates as the functional seat of the Nepali government and is home to the country’s most populous urban center, Newars have enjoyed some measure of relative advantage in business, trade, and political opportunities (Bennet et al., 2008; DFID/World Bank, 2006). As a central figure in their ethnic identity, Kathmandu Valley’s symbolic importance to the Newars as their historic homeland likely plays a role in their willingness to leave and return for migration that differs from populations with fewer ties to the land (Gellner, 1986).

Next on the caste hierarchy sit Hill Janajati.² They were known for their bravery and, hence, comprised approximately 99% of the British Gurkha regiments in the United Kingdom and India since the 1800s (Pariyar, 2016). This specific migration of the Hill ethnicities to India for military service continues to influence their migration to that region in the present through network and historical legacy effects (Levine, 1987; Gellner, 2007).

Descended from groups that have traditionally resided in the lowland areas that border India, Terai Janajati people (such as Tharu, Darai and Kumal) are considered somewhat different from Nepal’s other ethnic groups given their tendency to live in villages comprised exclusively of their own group and that they are heavily disadvantaged in terms of social, economic, and political opportunities and resources. These groups of people are indigenous to the terai region of Nepal. More importantly, they are very well connected to the land and practice mostly farming and are relatively less integrated into the market economy as compared to other ethnic groups. Thus, the Tharu people are specifically important to consider given that Chitwan is their claimed homeland. This likely plays a role in influencing their willingness and ability to migrate out of the area given that Chitwan is their claimed homeland. In addition, Tharus are among the most marginalized indigenous group of people. The Tharu identity movement has become one of the strongest ethnic political movements in the country, calling specific attention to the systemic marginalization of Tharu people (Sapkota, 2014). Part of this movement has mobilized claims regarding how gender operates quite differently for Tharus compared to other Nepalis, in which Tharu families tend to participate in distinctive practices like polygyny and Tharu women are more likely to jointly decide on the allocation of family labor with men compared to other ethnic groups (Devkota et al., 1999; Maslak, 2003). One of the goals of the movement, which is still ongoing, is to establish a separate federal state of Tharus called Tharuhut.

Lastly, those who occupy the lowest position in Nepal’s caste hierarchy are called Dalits (Bennet et al. 2008). Due to the symbolic practice of considering Dalits as religiously impure, a number of restrictions and forms of discrimination result in their pervasive marginalization along the dimensions of employment, schooling, housing, and social interactions. Social norms advise higher caste groups from eating with or sharing cooked food or water touched by Dalits. It is also socially discouraged for Dalits to even enter the homes of higher caste people. Consequently, this social ostracization exacerbates the more material forms of exclusion that Dalits experience.

4.2. Gendered Family Obligations in Nepal

Within this context of ethnic and caste variation, it is still nevertheless the case that Hinduism’s immense cultural influence holds sway among Nepal’s population, including its expectations regarding gender. According to the Hindu Dharmashastras (Manusmriti, the Hindu civil code), “…women should be under the strict control and supervision of their fathers until marriage, under the control of the husband after marriage, and that of a son after the death of the husband” (Bista, 1991:63). Patriarchal control thus envelopes women in Nepali society throughout the life course—albeit for some more than others—by exerting power over their bodies, labor, income, mobility, sexuality, ideology, and identities (Acharya & Bennet, 1981; Luitel, 2001; Paudel, 2011).

Specifically, household activities in Nepal are highly gendered. Men are expected to be breadwinners, and as a result of this expectation, are cast as the ones held most accountable for the material successes and failures of their families and households. Government agencies, politics, the market economy, and employment are largely controlled by men, and men generally perform the most physically demanding agricultural work and other forms of labor outside of the home. In contrast, women are primarily confined within the home to perform domestic tasks like housework and childcare as well as some agricultural work perceived as compatible with domestic tasks (Acharya & Bennet, 1981; Bista, 1991). In the absence of government social security, elderly parents commonly live with their married sons and daughters-in-law. As caretakers within the home, women are also often held responsible for the day-to-day care of elderly parents in need of assistance.

Expectations for the fulfillment of family obligations within the home are hence tied to men and women’s mobility. Men appear to have far fewer restrictions placed on where they can go and work, while women’s obligations can heavily restrict their mobility outside the home. As Dyson & Moore (1983) have documented in India, the long-held tradition that women move to their husband’s house immediately following marriage to tend to his family is one such example that is also found in Nepal. In general, patriarchal norms often restrict the mobility of women by requiring them to receive permission from the household head to even travel locally, let alone for more distant travel. Thus, this gendered power structure within the family likely holds important implications for men’s and women’s migration decisions and opportunities.

It is important to note, however, that the gender context in Nepal has changed somewhat in recent years. For example, Nepali women have had increased access to education, to land and property ownership, and to participation in the public sphere (ADB, 2010; Pudasaini, 2015; Williams, 2009). Likewise, Ghimire et al. (2006) found that generational processes of cultural change may also be taking place in regard to family and gender expectations. They found that non-family experiences, in general, and youth club participation and media exposure, in particular, influence spouse selection among youth who live in Nepal’s arranged marriage society. It appears, then, that while gender expectations of male breadwinning and female homemaking prevail, there is some variation and change taking place within Nepal’s gender structure.

4.3. Migration in Nepal

As for the phenomenon of interest, i.e. migration, Nepal has a long history of emigration that has only expanded over time. The culture of out-migration has especially increased after the 1990s with the political change in the country. More recently, migration has become a rite of passage and a matter of social status and prestige for individuals, especially among the young (Thieme & Wyss, 2005). In terms of who migrates, migration has always been heavily influenced by gender and age, as well as economic, educational, and other social predictors that closely align with migration theory and empirical patterns from other parts of the world (Bhandari, 2004; Bhandari & Ghimire, 2016; Bohra & Massey, 2009; Donato & Gabaccia, 2015; Williams, 2009).

Young adults, particularly young adult men, overwhelmingly comprise Nepal’s migrant population. This gender gap in migration is indicated by the number of labor permits provided to both men and women by the Government of Nepal. In 2008–2009, of the more than two hundred thousand total labor permits issued, only 4% of those were for women. This proportion of women only slightly increased to 5% in the 2016–17 year (Government of Nepal 2018). However, that these statistics may elide informal ways through which women migrate that are not captured in official estimates. In 2012, for example, the Government of Nepal announced a new ban on women under the age of 30 who wanted to migrate to the Arab states for domestic work. Despite the ban and well-publicized abuses of Nepali domestic workers, migrant women continued to pursue more vulnerable work abroad in informal markets using unauthorized travel (ILO, 2015; Kern & Müller-Böker, 2015). However, this ban only came into effect after the end of the data collection period for the CVFS survey (2008–2012) so it matters less for the empirical purposes of this study. Additionally, the data we use relies on household and individual reports of monthly migration information and destination choice instead of these less reliable official statistics of registered migrations, lending more credibility to the veracity of our empirical results.

Overall, the 2011 population census reported about 2 million individuals as migrants (Central Bureau of Statistics 2012). This same census indicated that one in every four households (25%) reported that at least one member of their household was absent or was living outside of Nepal. However, in the Chitwan study setting this proportion was higher; nearly two-thirds of households had at least one migrant in 2013 (Bhandari & Chaudhary, 2017). Nepali migrants are distributed worldwide and are working in as many as 131 countries (Government of Nepal, 2014). India has been the most popular international destination due to its open border, socio-cultural and linguistic similarities, and well-established social networks. The 2009 Nepal Migration Survey estimated that of the total 2.1 million Nepali migrants, 41 percent were in India, 38 percent were in the Middle Eastern Gulf countries, 12 percent were in Malaysia, and 8.7 percent were in other countries (World Bank, 2011). However, more recently, countries in the Middle East, Southeast Asia, the West (Northern Europe and North America), and Australia have become more popular destinations (Ozaki, 2012).

Evidence shows that about 75 percent of Nepali international migrants are unskilled and employed mainly in entry-level jobs such as cleaning and construction (Kern & Müller-Böker, 2015). Men are primarily employed in construction and driving, yet it has become increasingly common for women to migrate to places like the Gulf and Israel to provide domestic and care work (Adhikari et al., 2006; Bhadra, 2007). These migration trends still reflect an overall gender division of labor in which migrant women perform domestic labor while migrant men perform non-domestic work.

Lastly, this context of heavy migration that now dominates Nepal’s cultural and economic landscape has also come to bear on gender relations within the country. Alongside the changes to education, land ownership, political participation, and spousal choice mentioned above, Yamanaka (2005) has shown how migration has specifically disrupted the ideal of the patrilocal family. Using a case of Nepalese labor migrants in Japan, they document how both husbands and wives have begun to migrate abroad for work, leaving children behind to be cared for by other relatives and disrupting normative arrangements of gender and labor within and outside the home. Additionally, studies by Adhikari and Hobley (2015), Lokshin and Glinskaya (2009), Maharjan et al. (2012), and Ghimire, Zhang, and Williams (forthcoming) have shown that male outmigration has differential effects on women’s status and responsibilities at home, with some demonstrating women’s lower labor force participation and others documenting women’s increased agricultural responsibilities and household decision-making power. Despite these changes, however, gender hierarchies that privilege men over women still characterize normative gender relations, particularly when matters concern work and the family.

5. Data

We use panel data from the Chitwan Valley Family Study (CVFS) between 2008 and 2012 (Axinn et al., 2018). Neighborhoods were originally selected in a stratified sampling approach in 1996 from a different component of the CVFS. The same neighborhoods were used again in 2008 when a new sample was drawn. The survey consists of sampled individuals living in the western Chitwan Valley in 2008. Residents aged 15–59 in 2008 were administered baseline interviews during that year. Residents 12–14 in 2008 were administered baseline interviews once they turned 15 years old. The CVFS was collected by the Institute for Social and Environmental Research (ISER) in Nepal with assistance from the University of Michigan. Because the initial sample of residents did not include enough households with minority ethnic group members, the CVFS also oversampled for them. The final sample comprises 4.415 individuals with an overall response rate of 97 percent for those contacted for an interview.

If baseline respondents continued to reside in Nepal—either in Chitwan or elsewhere within the country—they were re-contacted and interviewed three times per year from the beginning of the survey collection period in 2008 through the end of 2012. 93 percent of people in this group were retained through the end of survey data collection. Those who moved internationally were re-interviewed after their migration if they had moved prior to June 2011, with a completion rate of 95 percent. All respondents who migrated, either domestically or internationally, were also administered an extra migration experiences survey in which the timing and destinations of their moves were recorded.

In addition to the individual-level surveys, the CVFS also contains household-level interviews collected three times per year, which provide basic demographic information on all household members for every month throughout the data collection period. This provides an average of 49 months of observation, depending on the date of first and last interview. Sample loss is lower compared to other similar prospective surveys, with very few households lost due to migration, attrition, or refusal. The household reports of individual residential information thus cover 49 months and 98 percent of our original sample.

Consequently, the survey contains information on migration destinations and months of migration from both migrants themselves and from their households. We use a combination of information provided by the migrants for whom it was available and migration information from a household survey to supplement when it was not. Given the internal consistency between individual and household reports on migration histories at 98%, this is a justifiable approach. This approach provided us with information on the migration destinations and migration months for 97% of our original sample.

6. Measures

6.1. Defining Migration

The people at risk of migration in our analyses are Chitwan residents aged 15–59 when first interviewed in the baseline survey. Because attitudes can change over time as a result of migration, we limit our analysis to the first migration event after the baseline interview. In our first analysis, we defined our dependent variable as any migration out of Chitwan, with no differentiation among destinations. In our second analysis, we defined our dependent variable as migration out of Chitwan, with the destinations divided into four categories: inside Nepal, India, the Persian Gulf, and the relatively wealthy Western and Asian (WWA) countries. The destinations categorized as the Persian Gulf were Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, and the United Arab Emirates. We also included in this category a few destinations in the same general region as the Persian Gulf—Afghanistan, Iraq, Israel, and Lebanon. Migrants to these four countries comprised 7 percent of our migrants labeled as Persian Gulf migrants.

The primary international destinations in the WWA countries were Australia, the United Kingdom, the United States, Japan, Malaysia, South Korea, and Thailand, with these seven countries accounting for 86 percent of migrants categorized as going to WWA countries. The remaining 14 percent of the migrants to WWA countries were divided fairly evenly among Belgium, Belize, Canada, China, Finland, Germany, Iceland, the Maldives, Poland, Portugal, and Spain. Although these countries categorized as WWA represent a wide range of geographic locations and cultures, all are quite culturally dissimilar from Nepal and all are classified by the World Bank as upper middle- or high-income countries, justifying their grouping together for this analysis. We do not decompose international destinations further because of limitations of sample sizes within these regions. Our regional groupings are both fine-tuned enough to provide distinct and meaningful migration destination types and have large enough sample sizes to support the analyses.

We treat as a migrant anyone who left Chitwan and resided in another geographic area for six months or more after the 2008 baseline interview. People who left Chitwan but returned prior to the six-month cut-off were not considered to be migrants. As shown in Table 1, out of the original baseline sample, approximately 16 percent subsequently migrated out of Chitwan for six months or more—distributed roughly equally between domestic and international migration. Dividing the migrant destinations more finely, we find that 8.5 percent went within Nepal but outside of Chitwan, 2.3 percent went to India, 3.5 percent to the Persian Gulf, and 2.0 percent to the WWA destinations. Table 1 also provides descriptive statistics on all the specified covariates used in the analysis.

Table 1.

Descriptive Statistics for Respondents in Chitwan at the Baseline Interview (N=4415)

Variable	Mean/% (s.d.)
Dependent Variable
First Migration Destination
No Migration	83.6
Within Nepal	8.5
India	2.3
Persian Gulf	3.5
Wealthy Western & Asian (WWA)	2.0
Respondent Family Obligation Attitude Predictor Variables
Put Family Needs First	0.86 (.34)
Adult Children Taking Care of their Parents	0.88 (.33)
Respondent Demographic Characteristics
Gender:
Female	60.8
Male	39.2
Age at Time of Baseline (Time Varying)ab:
15–19	27.7
20–24	11.6
25–29	10.7
30–34	9.8
35–39	9.5
40+	30.7
Caste Status:
Brahmin-Chhetri	44.5
Newar	6.6
Hill Janajati	15.5
Dalit	11.0
Terai Janajati	22.4
Family Characteristics
Respondent Ever Married (Time Varying)a	65.0
Number of Household Members (Time Varying)a	3.4 (1.5)
Respondent Non-Family Work Experience:
Ever Worked	45.9
Wage Work Only	34.9
Any Salary Work	19.2
Socioeconomic Characteristics
Distance to Narayanghat (Miles)	8.6 (3.9)
Respondent School Attainment (Time Varying)a:
None	25.4
1–5 Years	15.5
6–8 Years	26.7
9–10 Years	14.6
11+ Years	17.9
Household Resources	.45 (2.48)
Relative Household Resources:
Lower Third	24.4
Middle Third	37.3
Upper Third	38.3
Migration Specific Capital
Respondent Migration Experience from Age 15 to Baseline
No Migration	62.8
Domestic Migration Only	23.2
Any International Migration	13.9
Logged % of Household Members Migrating (Time Varying)a	3.2 (1.1)
Logged % of Neighbors Migrating (Time Varying)a	3.7 (.30)

^aMean or percent distribution in Table 1 is calculated from the value of the variable at each respondents’ baseline interview month.

^bThere were a few contradictory reports of age for some young people in that they were recorded as ages 14 or 15 in different reports. We categorized them here as age 15.

6.2. Explanatory Measures - Family Attitudes Measures

We identified two measures to capture individually held attitudes towards family obligation in order to test our hypotheses. We describe each below.

6.2.1. Attitude towards Putting Family Needs before Individual Needs

The first measure, which addresses attitudes about “putting family needs before individual needs” is taken from a survey item which, when translated into English, asks the respondent “Overall, which do you think is better for most people in Nepal today—to put individual needs first or to put family needs first?” The answers were coded so that 1 equals family needs first and 0 equals individual needs first or about the same. We interpret this measure as referring broadly to a family’s needs, meaning, whether the respondent thinks it is important for Nepali people to put their family’s well-being and interests, broadly defined, before their own personal preferences. As stated earlier, how individuals understand their own responsibility towards fulfilling certain aspects of those needs will likely vary depending on their gender, where men may feel more obligated to take care of their families’ economic, material needs, while women may feel more obligated to meet their families’ domestic, emotional, and care needs. Eighty six percent of the individuals reported that “it is better to put family needs before individual needs.”

6.2.2. Attitude towards Adult Children Caring for Their Parents

The second measure we use is taken from a survey item which, when translated into English, asks the respondent “Overall, which do you think is better for most people in Nepal today—adult children taking care of their parents and in-laws or parents and in-laws taking care of themselves?” Answers of adult children taking care of their parents and in-laws were coded 1 and answers of parents and in-laws taking care of themselves or about the same were coded 0. In practice, Nepali marriage customs expect that married women are thereafter most obligated to care for their in-laws while men remain most obligated to care for their own parents. Therefore, this attitude measure should be realistically interpreted as asking how much respondents believe Nepali women should care for their in-laws while asking how much they believe Nepali men should care for their biological parents. However, it should be noted that this is not explicitly clarified in the survey to primarily refer to the husband’s parents and not both the husband’s and wife’s parents. We infer from general Nepali norms and behavioral patterns that this was how this question was likely interpreted by respondents.

Generally, we draw upon the literature to forward the idea that what constitutes care is also broadly defined and likely varies by gender. While men likely feel as though they need to care for parents in primarily monetary ways, women likely feel obligated to care for parents in more everyday ways that require their physical presence. Eighty eight percent of the sample reported that “it is better for adult children to care for their parents.”

6.3. Covariates

We used a number of theoretically-motivated individual-, household- and community-level controls to account for factors that the literature has identified as enabling, constraining, or moderating the likelihood of migration (Massey & España, 1987; Stark & Taylor, 1991; VanWey, 2005). In the context of Nepal, work by Thornton et al. (2019), Williams (2009), Massey et al. (2010), Bohra and Massey (2009), Thornton et al. (2019), Williams et al. (forthcoming), and others have shown that gender, age, marital status, ethnicity, education and work experience, economic resources, social services and rurality, and social networks influence migration. Individual-level controls include gender (= 1 if female), a time-varying measure of age in years (measured in six categories: 15–19, 20–24, 25–29, 30–34, 35–39 and 40+), a time-varying measure of marital status (= 1 if ever married), caste/ethnicity (Brahmin-Chhetri, Newar, Hill Janajati, Terai Janajati, and Dalit), non-family work experience by individuals prior to the baseline interview (never worked, wage work only, and any salary work), and a time-varying measure of school attainment (grades 1–5, 6–8, 9–10, 11+, and no schooling). The controls for gender and caste/ethnicity are particularly included because of our previous discussion on the importance of social categories for determining placement within Nepal’s social hierarchy, which differentially allocates opportunities and resources that likely shape whether someone migrates, who undertakes a migration, and where they decide to go.

In addition, we also used the migration experience of individuals prior to the baseline interview to control for the effect of migration-specific capital. Subsequent migrations are hypothesized to be less costly than initial migrations since initial migration events are undertaken with less knowledge, experience, and migration-related capital. Accounting for past migrations divorces the dependent variable of prospective migration from past migration experiences. This migration-specific capital is measured with the following categories: respondent has no migration experience from age 15 to baseline interview, has had domestic migration experience only, and has had any international migration experience.

At the community level, we included neighborhood proximity in miles to the urban center of Narayanghat. Past migration literature has shown that the extent of rurality matters for migration, though the direction of this effect has varied depending on the context (Adams, 1969; Brown & Wardwell, 2013; Ploch, 1977). As the community-level measure of migration specific capital, we used a time-varying measure of logged percent of community members who have migrated. The migration literature on cumulative causation and migrant social capital have identified that the propensity to migrate is influenced by an individual’s access to migrant networks, which lower the informational, social, and logistical costs of moving (Massey & Espinosa, 1997; Williams et al. (forthcoming). By increasing the prospective migrant’s access to potential employment opportunities, housing, and community resources at destination, larger migrant networks should increase the likelihood of migration.

Like community migration networks, household-specific networks might have similar but slightly different effects. Though household-level capital can also assist prospective migrants via resources, information, and logistical support, it might also be the case that having other household members who have migrated in the past actually decreases the likelihood that someone in that same household migrates. For this reason, a separate measure for migration networks at the household level is included specified as a time-varying measure of logged percent of household members who have migrated.

The household-level controls include a time-varying measure for the number of residents and household resources in 2006—a variable that is a composite of land ownership, livestock ownership, housing quality, and income measured in 2006. Each of these indicators provides partial measures of economic status and together provide a more comprehensive record of household resources. For each of these four measures, we logged the indicators to correct for skewness, calculated a z-score for each of the logged variables, and then added the z-scores. In addition, we also controlled relative household wealth of the respondent’s household. For this purpose, we compared the wealth of the respondent’s household with the wealth of the households in the respondent’s neighborhood. The relative household wealth measure was grouped into thirds (lower, middle, and upper third of relative household wealth).

Because the 2008 baseline data survey did not collect household resource information, we relied on a 2006 household survey that collected this information. Thus, household wealth and relative household wealth variables were unfortunately not available for 10 percent of the initial 2008 sample living in households that were not surveyed in 2006. We excluded this 10 percent from our main analyses. With the exclusion of individuals without economic resource data, our analysis sample consists of a maximum of 4415 individuals. However, the caring for parent measure was asked in the first follow-up interview occurring after the baseline interview. Only respondents who had not migrated by the first follow-up interview were included in the analysis using this measure, while the monthly hazard file only begins the month following their first follow-up interview. Consequently, 4233 respondents were eligible to be included for the analysis using this measure.

7. Methods

We analyzed both of our dependent migration variables as monthly hazards of migrating. Each person at risk of the transition is followed from the baseline interview until they experience migration or are censored because the observation period terminated. We estimate the hazard as a function of the predictors using logit regression and multinomial logit regression. We analyze the rates of migration using discrete-time multivariate event history models with person-months as the unit of analysis (Axinn & Thornton, 1993; Massey et al., 2010; Thornton et al., 1995; Thornton & Rodgers, 1987; Williams, 2009; Williams et al., 2012). Because the odds of migrating are so small within each one-month interval, the estimates from discrete-time methods are very similar to those that would be obtained using continuous time models (Petersen, 1991). While using person-months of exposure to risk as the unit of analysis in discrete-time models substantially increases the sample size, it does not deflate standard errors, and thus provides appropriate tests of statistical significance (Allison, 1984; Petersen, 1986, 1991).

As noted earlier, we specified two separate outcomes, both as the rate of making a transition in residence to either (1) any location outside of the study area, and (2) specific regional destinations divided into domestic migration inside Nepal but outside Chitwan, migration to India, to the Persian Gulf, and to the WWA. The first specification directly tests our hypotheses, which postulate that greater support for fulfilling family obligations will increase high-support men’s odds of migrating and decrease high-support women’s odds of migrating as a whole.

We use the second specification to see if granularity in destination type reveals a clearer picture of how attitudes influence migration behavior since where migrants decide to ultimately go, if they decide to go at all, will enable them to meet these family obligations in different and substantively important ways. For example, migrating within Nepal might yield the least monetary gain compared to migrating to India and especially compared to the Persian Gulf and WWA, where remittance income tends to be highest. Conversely, migrating within Nepal increases the ability of migrants to return home more easily and more often due to cost of travel and proximity. Due to Nepal’s open border with India, ease of return from there is also relatively informal, particularly compared to returning from the Persian Gulf and WWA, which might require breaking the terms of a labor contract, acquiring all the necessary travel documents and paperwork, and paying for the trip back itself.

The analysis primarily inspects the influence of the attitudinal measures on both any migration and migration to specific destinations. Because we hypothesize that the effects of our attitudes toward family obligations are different for men and women, we use an interactive model where we enter family obligation attitudes, gender, and gender multiplied by attitudes into the equation.

8. Results

8.1. Attitudes towards Putting Family Needs before Individual Needs

The first family attitudes analysis examines how the degree to which the respondent agrees with the sentiment that it is better for Nepali people to put the needs of their family before their own influences migration. Table 2 displays the results for the analysis using this measure for both outcomes of any migration versus no migration and migration to the disaggregated regional destinations. In general, the estimates for the attitudinal measure in the first row can be interpreted as the relative odds of migration for men who support putting family needs first compared to men who do not since men are coded as 0 and high support respondents are coded as 1in the data. The second row indicates the effect of the gender-times-attitudes coefficient—the extent to which being a woman modifies the coefficient for family attitudes observed for men. Multiplying the estimates from the first row with the estimates for the interaction term in the second row can be interpreted as the odds of migration for women who support putting family needs first compared to women who do not since women are coded as 1 and, again, high support respondents are coded as 1. Therefore, the estimates for the attitudinal covariate in Table 2 indicate that men who express support for putting family needs before individual needs have a 0.99 odds ratio of migrating at all compared to men who do not express support—a coefficient that is not statistically significant. This result does not provide evidence to support H1.

Table 2.

Odds Ratios and T-Statistics for the Put Family Needs First Attitudinal Predictor and Other Control Variables from Multilevel Logistic and Multinomial Regression Models Predicting Migration of Six Months or More

		Analysis 1		Analysis 2
		Any Migration	Within Nepal	India	Persian Gulf	WWA
Family Needs First		0.99 (.03)	0.54***(3.28)	1.87+(1.42)	1.43 (1.23)	1.40 (.91)
Family Needs First x Female		0.86 (.77)	1.75*(2.22)	0.33*(1.80)	0.76 (.38)	0.39*(1.65)
Female		0.51***(3.66)	0.72+(1.48)	0.70 (.62)	0.08***(3.78)	0.34*(2.06)
Age: 15–19		Ref	Ref	Ref	Ref	Ref
20–24		1.76***(4.99)	1.27(1.62)	1.56 (1.52)	5.92***(5.71)	4.06***(3.40)
25–29		0.87 (.83)	0.56*(2.46)	0.42+ (1.75)	5.02***(4.08)	1.82 (1.16)
30–34		0.58**(2.90)	0.37***(3.52)	0.19**(2.88)	3.20**(2.71)	1.12 (.20)
35–39		0.35***(4.78)	0.20***(4.48)	0.16**(3.07)	1.84 (1.34)	0.78 (.39)
40+		0.13***(9.56)	0.20***(5.54)	0.04***(5.17)	0.21**(3.07)	0.20*(2.56)
Caste Status: Brahmin-Chhetri		Ref	Ref	Ref	Ref	Ref
Newar		1.20 (1.17)	1.55*(2.27)	1.66 (1.13)	0.56 (1.20)	.57 (1.10)
Hill Janajati		1.14 (1.16)	1.10 (.60)	2.08*(2.52)	1.03 (.12)	.73 (.85)
Dalit		1.05 (.36)	0.82 (.98)	1.17 (.45)	1.72+ (1.94)	0.84 (.35)
Terai Janajati		0.74*(2.22)	0.73+(1.66)	0.69 (.99)	0.62 (1.60)	.89 (.26)
Respondent Ever Married		0.75*(2.20)	0.49***(3.84)	0.73 (.90)	1.16 (.54)	2.21*(2.32)
Number of Household Members		0.92**(2.59)	0.84***(3.81)	0.86 (1.61)	1.10 (1.56)	0.89 (1.27)
Respondent Non-Family Work Experience:
Never Worked		Ref	Ref	Ref	Ref	Ref
Wage Work Only		1.16(1.47)	1.11 (.77)	1.77*(2.26)	1.66*(2.06)	0.42*(2.16)
Any Salary Work		1.22+ (1.69)	1.26 (1.32)	1.00 (.01)	1.54+ (1.74)	0.45*(2.58)
Distance to Narayanghat		1.01 (1.32)	0.99 (.43)	1.09**(2.75)	1.07*(2.48)	0.58 (1.51)
Respondent School Attainment:None		Ref	Ref	Ref	Ref	Ref
1–5 Years		1.88***(3.13)	1.95*(2.40)	2.08 (1.26)	1.43 (.90)	1.73 (.64)
6–8 Years		1.62*(2.40)	1.43 (1.29)	1.69 (.90)	1.63 (1.21)	2.75 (1.27)
9–10 Years		1.92**(3.17)	2.09**(2.63)	1.75 (.93)	1.34 (.67)	2.73 (1.22)
11+ Years		2.16***(3.75)	2.23**(2.85)	1.12 (.18)	1.26 (.52)	6.10*(2.28)
Household Resources		0.99 (.12)	1.00 (.07)	0.94 (.88)	0.99 (.01)	1.08 (.94)
Relative HH Resources: Lower Third		Ref	Ref	Ref	Ref	Ref
Middle Third		1.00 (.05)	0.94 (.40)	1.03 (.10)	1.18 (.68)	0.91 (.25)
Upper Third		1.01 (.08)	0.90 (.53)	1.29 (.68)	0.99 (.01)	1.03 (.07)
Respondent Migration Experience
No Migration		Ref	Ref	Ref	Ref	Ref
Domestic Migration Only		1.12***(3.76)	1.93***(4.18)	0.57 (1.01)	0.90 (.35)	2.01*(2.30)
Any International Migration		2.61***(7.07)	1.35 (1.20)	7.05***(5.41)	1.88**(2.50)	3.78***(3.88)
Logged % of HH Members Migrating		1.04 (.96)	0.99 (.09)	1.02 (.20)	0.99 (.12)	1.40*(2.57)
Logged % of Neighbors Migrating		1.04 (.26)	1.11 (.53)	1.18 (.39)	0.68 (1.22)	0.69 (.87)
Number of Person Periods		192598	192598	192598	192598	192598
Number of Moves		721	375	103	155	88
Fit Statistics
AIC		8800.12	5032.07	1576.39	2175.60	1371.22
BIC		8889.83	5181.79	1666.11	2265.31	1460.93

Significance:

^+.10

^*.05

^**.01

^***.001 (one-tailed for predictor variables, two-tailed for controls).

Note: The reference category for both Analysis 1 and Analysis 2 is “No Migration.”

Results from the multinomial analysis tell a more complicated story. Men who express support for putting family needs before individual needs have a statistically significant 0.54 (p < 0.001) odds ratio of migrating within Nepal, a marginally significant 1.87 (p < 0.10) odds ratio of migrating to India, a 1.43 not significant odds ratio of migrating to the Persian Gulf, and a 1.40 not significant odds ratio of migrating to WWA. These results suggest that men who express support for putting family needs before individual needs are less likely to migrate domestically and more likely to migrate internationally to at least India. While the estimates for the Persian Gulf and WWA are not statistically significant, the direction of the estimates corresponds to this domestic versus international interpretation. These estimates suggest that attitudes that prioritize family obligations influence men’s migration to relatively more remunerative destinations, though the significance of the India result compared to the farther and generally more remunerative destinations of the Persian Gulf and WWA might speak to a desire to balance monetary gain with proximity.

Turning now to the gender-times-attitudes interaction effects in row 2 of Table 2, we see that the interaction coefficient is 0.86 and statistically insignificant for any migration. This means that the family attitude effect for women is not significantly different from the effect for men. In examining the results for the interaction between gender and the attitudinal measure multiplied with the estimates for the attitudinal measure itself, we find that women who express support for putting family needs before individual needs have a 0.99 × 0.86 = 0.85 odds ratio of migrating anywhere compared to women who have weaker attitudes supporting family obligations. Thus, the results for the any migration analysis do not lend statistically significant support for H2 either, though the direction of the effect for women (0.85) corresponds to what we predicted.

Again, however, the multi-destination analysis paints a more complex picture. The estimates indicate that the interaction term for gender-times-attitudes is statistically significant and positive for migration within Nepal and negative for all international destinations (and statistically significant for India and WWA). This suggests that the large negative effects of family attitudes on migration within Nepal for men is much less negative for women. It also means that the general positive effects of family attitudes on international migration for men is much less positive for women.

Multiplying the coefficients for family first attitudes in row 1 times the interaction coefficients in row two provides estimates of the effects of family first attitudes on migration for women with high support for family obligations. Such calculations indicate that if a woman expresses support for putting family needs before individual needs, she still has a 0.54 × 1.75 = 0.95 odds ratio of migrating within Nepal, a 1.87 × 0.33 = 0.62 odds ratio of migrating to India, a 1.43 × 0.76 = 1.09 odds ratio of migrating to the Persian Gulf, and a 1.40 × 0.39 = 0.55 odds ratio of migrating to WWA compared to a woman who expresses less support for putting family needs before individual needs. These results suggest that the effects of family attitudes for women migrating domestically or migrating to the Persian Gulf are relatively small compared to remaining at home, but the effects of family attitudes on women’s migration to India and to WWA are substantial and negative.

Overall, the estimates concerning the influence of family obligation attitudes on any migration were not significant and thus do not lend support for H1 or H2. However, the multi-destination analysis demonstrates that attitudes matter in a more complex way. Men who express greater support for putting family needs first are less likely to migrate domestically and more likely to migrate internationally. While India was the only international destination with a significant coefficient, the direction of the estimates for the Persian Gulf and WWA correspond to this interpretation. This suggests that support for putting family needs first for men translates into migrating to generally more remunerative destinations, though perhaps to ones that are more proximate. For women, the estimates indicate that having supportive family first attitudes results in less migration, especially to the farther destinations of India and WWA.

8.2. Attitudes towards Adult Children Caring for Their Parents

The second family attitudes analysis examines how the degree to which the respondent agrees with the sentiment that it is better for Nepali people if adult children care for their parents influences migration. Unlike the small and statistically insignificant effect of the family needs variable on overall migration for men, the estimate for the caring for the parents attitudinal covariate on overall migration for men in Table 3 indicates that men who express support for adult children caring for their elderly parents have a marginally significant 1.28 odds ratio of migrating at all compared to men who do not express support. This result, therefore, provides moderate support for H1.

Table 3.

Odds Ratios and T-Statistics for the Adult Children Taking Care of their Parents Attitudinal Predictor and Other Control Variables from Multilevel Logistic and Multinomial Regression Models Predicting Migration of Six Months or More. Predicting migration of 6 months or more from Better for Adult Children to Care for Parents^a with gender interaction & control variables for Respondents at risk for Migration

	Analysis 1		Analysis 2
	Any Migration	Within Nepal	India	Persian Gulf	WWA
Children Taking Care of their Parentsa	1.28+(1.44)	0.96 (.17)	3.62*(2.14)	1.44 (1.10)	1.09 (.21)
Children Taking Care of their Parentsa x Female	1.02 (.08)	1.50 (1.16)	0.30+ (1.40)	0.88 (.11)	1.09 (.10)
Female	0.46**(3.01)	0.76 (.83)	0.87 (.16)	0.05**(2.64)	0.12**(2.57)
Age: 15–19	Ref	Ref	Ref	Ref	Ref
20–24	1.85***(4.83)	1.39*(1.97)	1.50 (1.23)	5.71***(5.02)	4.03**(3.04)
25–29	0.91 (.48)	0.58*(2.07)	0.72 (.61)	4.57***(3.29)	1.68 (.88)
30–34	0.62*(2.23)	0.36***(3.29)	0.36 (1.59)	3.08*(2.23)	1.28 (.40)
35–39	0.42***(3.55)	0.21***(4.11)	0.36 (1.54)	2.22 (1.50)	0.88 (.18)
40+	0.14***(8.11)	0.16***(5.54)	0.10***(3.43)	0.17**(2.83)	0.21*(2.26)
Caste Status: Brahmin-Chhetri	Ref	Ref	Ref	Ref	Ref
Newar	1.26 (1.33)	1.59*(2.18)	0.98 (.04)	0.85 (.32)	0.60 (.90)
Hill Janajati	1.19 (1.37)	1.11 (.58)	2.31**(2.65)	1.10 (.29)	0.58 (1.26)
Dalit	1.07 (.45)	0.83 (.83)	1.43(.95)	1.69 (1.69)	0.89 (.24)
Terai Janajati	0.77+(1.73)	0.70+(1.67)	0.89 (.29)	0.67 (1.19)	.81 (.45)
Respondent Ever Married	0.73*(2.15)	0.56**(2.92)	0.56 (1.35)	1.06 (.17)	2.24*(2.01)
Number of Household Members	0.88***(3.31)	0.81***(3.97)	0.79*(2.16)	1.07 (.88)	0.88 (1.20)
Respondent Non-Family Work Experience:
Never Worked	Ref	Ref	Ref	Ref	Ref
Wage Work Only	1.09 (.77)	1.12 (.78)	1.57 (1.62)	1.38 (1.12)	0.42*(1.99)
Any Salary Work	1.18 (1.20)	1.28 (1.23)	0.79 (.58)	1.34 (.98)	0.46*(2.26)
Distance to Narayanghat	1.03*(2.37)	1.01 (.41)	1.09**(2.65)	1.09**(2.84)	0.98 (.51)
Respondent School Attainment: None	Ref	Ref	Ref	Ref	Ref
1–5 Years	1.75*(2.41)	1.75+(1.78)	2.76 (1.52)	0.95 (.09)	1.69 (.61)
6–8 Years	1.66*(2.22)	1.17 (.49)	2.20 (1.17)	1.74 (1.12)	2.96 (1.36)
9–10 Years	2.07**(3.13)	2.07*(2.34)	2.45 (1.30)	1.14 (.25)	2.52 (1.11)
11+ Years	2.33***(3.63)	2.22**(2.56)	2.11 (1.05)	1.19 (.33)	4.33+(1.81)
Household Resources	0.99 (.12)	1.03 (.77)	0.91 (1.32)	0.98 (.27)	0.99 (.02)
Relative HH Resources: Lower Third	Ref	Ref	Ref	Ref	Ref
Middle Third	0.99 (.08)	0.85 (.94)	1.36 (.90)	1.11 (.36)	1.00 (.00)
Upper Third	0.98 (.12)	.75 (1.26)	1.69 (1.21)	0.86 (.39)	1.53 (.83)
Respondent Migration Experience
No Migration	Ref	Ref	Ref	Ref	Ref
Domestic Migration Only	1.41*(2.51)	1.76**(3.22)	0.34 (1.40)	0.67 (1.00)	2.02*(1.97)
Any International Migration	2.51***(5.89)	1.23 (.72)	6.00***(4.40)	1.99*(2.21)	3.53***(3.30)
Logged % of HH Members Migrating	1.04 (.95)	0.97 (.53)	1.12 (.90)	1.09 (.84)	1.38*(2.23)
Logged % of Neighbors Migrating	1.17 (.95)	1.28 (1.05)	1.27 (.49)	0.72 (.86)	0.76 (.56)
Number of Person Periods	164699	164699	164699	164699	164699
Number of Moves	569	309	82	109	69
Fit Statistics
AIC:	7030.42	4211.12	1288.88	1569.84	1113.76
BIC:	7120.14	4300.83	1378.59	1659.55	1203.47

^aAdult Children Taking Care of their Parents was collected at first re-interview after baseline and this analysis uses months subsequent to that time point.

Significance:

^+.10

^*.05

^**.01

^***.001 (one-tailed for predictor variables, two-tailed for controls).

Note: The reference category for both Analysis 1 and Analysis 2 is “No Migration.”

Results from the destination-specific, multinomial analysis again tell a more intricate story. Men who express support for adult children caring for elderly parents have a not significant 0.96 lower odds of migrating within Nepal, a significant and much higher 3.62 (p < 0.05) odds ratio of migrating to India, a not significant 1.44 odds ratio of migrating to the Persian Gulf, and a not significant 1.10 odds ratio of migrating to WWA. These results suggest, akin to the other analysis, that men who express support for caring for parents are less likely to migrate domestically and more likely to migrate internationally, especially to India. These estimates again imply that attitudes that prioritize family obligations encourage men to migrate to relatively more remunerative destinations. But the significance of the India result compared to the farther and generally more remunerative destinations of the Persian Gulf and WWA might evince a desire to balance monetary forms of care with other forms of everyday care or their desire to be relatively close by should the need to return home arise.

Turning now to the gender-times-attitudes interaction effects in row 2 of Table 3, we see that the interaction coefficient is a statistically insignificant 1.02 for any migration. This means that the caring for elderly parents attitude effect for women is not statistically different from the effect for men. In examining the results for the interaction between gender and the attitudinal measure multiplied with the estimates for the attitudinal measure itself, we find that women who express support for adult children caring for their elderly parents have a 1.27 × 1.05 = 1.33 odds ratio of migrating anywhere compared to women who have weaker attitudes supporting children caring for parents. Given that this is in the opposite direction as hypothesized, the results for the any migration analysis for women do not lend support for H2 either.

Turning to the multinomial results in which destinations are disaggregated, the estimates indicate that the interaction terms for gender-times-attitudes are not statistically significant for any of the possible destinations except India, for which it is marginally significant at the p < 0.10 level. This can be interpreted to mean that the family attitude effects for women are not statistically different from those for men with the exception of India.

Multiplying the coefficients for the care for parents indicator in row 1 times the interaction coefficients in row two provides estimates of the effects of the caring for parents attitude on migration for women. Such multiplications indicate that if a woman expresses support for caring for elderly parents, she has a 0.96 × 1.50 = 1.44 odds ratio of migrating within Nepal, a 3.62 × 0.30 = 1.09 odds ratio of migrating to India, a 1.44 × 0.88 = 1.26 odds ratio of migrating to the Persian Gulf, and a 1.09 × 1.09 = 1.18 odds ratio of migrating to WWA compared to a woman who expresses less support for caring for parents. However, again, the interaction term coefficient estimates were not statistically significant for any of the possible outcomes except marginally for India.

Overall, the results from this portion of the analysis provide some support for H1 but somewhat contradicts H2. Men who express greater support for caring for parents are somewhat more likely to migrate overall compared to men who express less support. However, these findings come with a caveat. While migrating to WWA and the Persian Gulf would be most remunerative, there is no statistically significant difference in migration risk between men who support or do not support this attitude for these destinations. The estimate for migration to India, on the other hand, was substantially higher in comparison to any other possible destination for men who expressed greater support for caring for their parents. It is unclear why this would be the case, but we propose that one possible explanation may be that men who value caring for adult parents may themselves define care more capaciously. Rather than limiting care to just masculinized, material and monetary definitions related to breadwinning, men who highly value caring for parents may also seek to balance the material obligations to which they feel beholden with other forms of care like physical presence and everyday support. Migration to India would facilitate such a balance since the border is permeable, work is oftentimes more short-term, and the distance is shorter while the remittance payoff is still notable. Lastly, the estimates provide no support for H2, where women who express support for adult children to care for their parents were shown to not be different from low support women except perhaps in the case of India. The direction of this estimates is opposite from what was hypothesized, suggesting that women’s definitions of care may also be more capacious to extend beyond gendered definitions of care to encompass financial contributions.

9. Conclusion

This study examines the influences of individual attitudes towards fulfilling family obligations on migration. Specifically, we investigate the effect of individual attitudes towards family obligations on overall migration and destination-specific migration net of other important factors influencing who migrates, such as caste and ethnicity, household resources, social networks, and many others. Because there are differential gender expectations for men and women within and outside of the household, we also focus on whether the effects of such attitudes vary by gender. The character of Nepali migration remains gendered and our empirical results provide some support for the claim that such gendered patterns are somewhat mediated by the extent to which individuals support prioritizing family obligations. Gender and migration scholarship helps to make sense of this pattern, suggesting that migration is largely understood as masculine. Particularly in the Nepali context, migration is often undertaken to find employment or pursue higher education, both of which also constitute masculine domains³.

Given this, we expected that the extent to which women and men value family obligations would affect their overall migration experience—increasing men’s overall migration and decreasing women’s overall migration. However, as discussed above, our results concerning this overall hypothesis were mixed, with the strongest support being for the adult care for parents indicator where the effect for men was positive, as hypothesized, but was only marginally statistically significant.

However, our destination-specific analyses were more complex and powerful. One substantial observation was that the effects of high valuation of family commitments vary dramatically between domestic and international destinations. Although the coefficients were not consistently statistically significant, the general pattern for men was for positive attitudes toward family obligations to negatively affect migration within Nepal and to positively affect migration outside of Nepal to destinations that are further but also more remunerative. Men who expressed support for caring for their parents were more likely to migrate to India compared to not migrating at all (a result that was statistically significant), but they were not more likely to migrate to the faraway and more remunerative destinations in the Persian Gulf and WWA compared to not migrating at all. These findings compel us to question the strict and dichotomous ways we oftentimes conceptualize gendered family obligations and the individuals we think are held to those standards. For the men who were more likely to migrate to India but not to the Persian Gulf or WWA compared to remaining at home, perhaps rigid definitions of what constitutes care for men do not fully encapsulate the scope of their obligations or their own desires to remain close to home while still being able to provide monetarily for their families.

A second substantial observation is that there were gender interactions in the effects of family obligations on migration destinations. For the family needs indicator, the interaction coefficient was large and positive for migration within Nepal. This means that while the effect of family obligation attitudes for men’s migration within Nepal tends to be small or negative, the effect for women’s migration within Nepal tends to be small or positive.

The effects of the interactions for domestic migration were positive and large, but all of the interaction effects on international destinations, with one exception, were negative. This suggests that the effects of strong attitudes favoring family obligations for international destinations are generally weaker among women than among men. These results indicate that while strong family obligation attitudes tend to lead men to migrate internationally, the effect of strong family obligation attitudes either have little effect on women or influence them to avoid international destinations.

This study contributes to family, gender, and migration scholarship by bringing the role of individual attitudes towards family obligations into the literature. The incorporation of attitudes toward family obligations into our consideration of the gendered institution of the family contributes to an expanding recent literature assessing the role of cognitive factors in migration (Carling & Collins, 2017; Carling and Schewel, 2018; Collins, 2018; Creighton, 2013; Koikkalainen and Kyle, 2016; Koikkalainen et al., 2019; Migali and Scipioni, 2019; Schewel 2019; Thornton et al., 2019; Williams et al., 2014). Our analysis helps explain why some individuals might migrate more than others and to which destinations despite the overall gendered social structure in which they all collectively live. The unexpected yet noteworthy findings from our disaggregated, destination-specific analyses additionally reveal how gender and family expectations are not rigid mandates for who migrates and where they migrate. People may act in ways that reconcile competing demands by migrating to closer destinations and rearticulate what it means to fulfil family obligations depending on their circumstances.

We close with a call for more research to ascertain why these migration patterns arise. Of importance is greater understanding of the gender division of labor that leads family obligation attitudes to affect migration differently for women and men. Also of importance is more insight into how and why family obligation attitudes affect the choice of migration between domestic and international migration. With respect to Nepal, further inquiry into the relationship between caste/ethnicity and endorsement of particular attitudes and other cognitive preferences and beliefs is especially needed. In addition, we acknowledge that this quantitative study necessitated a binary conceptualization of gender as the CVFS only recorded two possible options: male and female. Future research on how gender attitudes matter for migration and other behaviors would benefit from looking beyond the gender binary to understand how gender is understood and negotiated across a broader spectrum within family life in Nepal, especially given that the country houses many indigenous populations that recognize a third gender community. Lastly, it would be useful to investigate these issues in other settings around the world.

Biography

Christina Hughes is a doctoral candidate at the Department of Sociology at the University of Washington. Her research broadly focuses global-transnational processes of race, class, and gender formation, particularly as they are shaped by social policy. Her dissertation explores how people working in the illicit marijuana industry navigate the legalization process currently unfolding in the state of California.

Prem Bhandari is a Social Demographer working at the Population Studies Center, University of Michigan. His research focuses on the socioeconomic and cultural influences on human fertility, migration and remittances, and population health in Nepal. Other areas include rural social change, population and environment relationships, and social research methods.

Linda Young De-Marco is a Lead Social Science Research Area Specialist at the University of Michigan’s Institute for Social Research. She collaborates with researchers in the Family and Demography Program to study people’s values and beliefs about development and the ways in which these beliefs and values influence subsequent behaviors. Young-DeMarco has co-authored articles and book chapters addressing this topic using data collected from a wide variety of locations around the world including Africa, Argentina, China, Eastern Europe, Nepal, the Middle East, and the United States.

Jeffrey Swindle is a doctoral candidate in the Department of Sociology and a Population Studies Center trainee at the University of Michigan. He studies global social change, focusing on the spread and influence of cultural messages about societal development and the good life. His dissertation applies these theoretical interests to the empirical case of interpersonal violence in Malawi.

Arland Thornton is Professor of Sociology and Research Professor at the Institute for Social Research at the University of Michigan. His research focuses on trends, causes, and consequences of marriage, cohabitation, divorce, fertility, gender roles, and intergenerational roles. Current work explores the ways in which values, beliefs, and people have been and are being distributed around the world. He is the author of Reading History Sideways: The Fallacy and Enduring Impact of the Developmental Paradigm on Family Life (The University of Chicago Press, 2005).

Nathalie Williams is Associate Professor at the Jackson School of International Studies and the Department of Sociology at the University of Washington. She is also Research Affiliate at the Center for Studies in Demography and Ecology. Williams’ research focuses on migration during armed conflict, natural disasters, climate change, and during periods of rapid social change. She also studies other behavioral responses to macro-level change, including marriage, childbearing, fear of violence, and mental health disorders. In addition, Williams has worked to develop innovative data collection, measurement, and analysis methods for research on migration and armed conflict, topics that are notoriously difficult to analyze with quantitative methods.