Abstract
This study integrates theory and research on household fission (or partition) and migration to better understand living arrangements following marriage, especially in historically patrilocal and primarily agricultural settings. Using panel data from the Chitwan Valley Family Study to analyze the sequential decision-making process that influences men’s living arrangements subsequent to first marriage, we demonstrate the importance of distinguishing among extended family living, temporary migration, and the establishment of an independent household. We find that community economic characteristics, such as access to markets or employment, as well as household wealth affect the initial decision to leave the natal home. Household resources and use of farmland, along with the young men’s own education, media exposure, travel, and marital behavior, influence the decision to make the departure from the natal home permanent. Our findings explain why previous results regarding household fission and those focused on migration have provided such mixed results, and we establish a new framework for thinking about how families and individuals manage living situations.
Introduction
The life cycle perspective on family structure and processes formulated by Glick (1947, 1989) defines the first marriage of an adult child as a critical event in terms of the future composition and size of the familial household. Whether adult children live separately from their parents after marriage (or before) varies across societies, with the nearly immediate establishment of a separate household dating back centuries in Western European societies (Davis and Blake 1956; Hajnal 1982; Ruggles 2009; Smith 1993). More recently, in many developing countries—particularly in Asia—the proportion of nuclear families is growing as extended family households become less common (Jones 2012; Liu et al. 2003; Zhang 2002). As household fission increases, the population is spread across more households. Even controlling for overall population growth, increasing numbers of households mean a greater depletion of land and resources as well as a greater production of waste (Liu et al. 2005; Moran et al. 2005). Therefore, understanding factors behind the world population being increasingly spread across a greater number of smaller-sized households is important.
Household units are notoriously difficult to study because they are dynamic groups, alternating between growth and fission phases (Stevenson 1997). In spite of such flux, most of the prior research on patterns and changes in household structure is static in nature, focusing on the composition of households at a single point in time (Lavely and Ren 1992; Niranjan et al. 2005). Conclusions drawn from such research are problematic: the causal direction in relationships between household structure and factors such as land size and household wealth is difficult to disentangle.
However, a key mechanism for the growth in number of households in agricultural settings that have historically practiced extended family living is the increasing rate at which married sons establish an independent nuclear home with their wife and children. Therefore, an alternative and dynamic approach to explore household fission is to focus on one member of the household—a recently married male—and study the factors that contribute to whether and when he establishes his own household. Although a few studies have explored the factors that influence young adults in developing countries to leave their parental home, these studies have included single, recently married, and long-term married men and women (DeVos 1989; Johnson and DaVanzo 1998; Winters et al. 2009).
In addition, prior research on household structure and fission has implicitly defined the movement of members out of the household, measured at one particular point in time, as reflecting a permanent condition—for example, the establishment of a new household. However, a deeper understanding of household demography and living arrangements should take into account cyclical movement of adult members in and out of the household. The phenomena of individuals from developing countries temporarily migrating either within their home country or abroad to improve their family’s economic prospects by sending home remittances is widespread (Piotrowski 2008, 2013; VanWey 2003). Thus, individuals who cycle in and out of their household are still intertwined in the household economy. However, by living away from the household for substantial portions of time, they also transform relationship dynamics and the distribution of resources within the household.
Therefore, recently married males’ (and their wives’) decisions regarding living arrangements (in relation to his natal home) are more accurately understood as involving three general possibilities: (1) establishment of an independent household, (2) cycling in and out of the natal home, and (3) remaining continuously in the natal home. As early as 1976, Conklin raised the possibility that in South Asia, the necessity of searching for work beyond the natal household farm—and in some cases, beyond the local community—can result in either a temporary departure from the natal household or a split from the household. Conklin suggested that both socialization vis-à-vis the importance of the extended family as well as economic necessity influence the final outcome. However, to date, research has not simultaneously examined the factors leading to these three interlinked decisions about household living arrangements. Rather, research has generally examined these decisions as binary processes—(1) migrate versus remain home, or (2) establish an independent household versus remain in the natal household—combining those who cycle in and out either with those who remain in the natal home or those who establish a permanent independent household.
In this study, we examine the factors related to whether a married son establishes a separate household as opposed to remaining in or temporarily living away from his natal home. We focus on a sample of recently married men in the Chitwan Valley of Nepal, a predominantly patrilocal society where agriculture is the main form of livelihood and extended family living was the norm until very recently. We explore a comprehensive set of factors that may influence household living arrangements, including community and household socioeconomic characteristics as well as the married son’s own resources and experiences. The community- and individual-level factors we examine add considerably to the existing literature, which has focused mainly on household-level economic characteristics.
Background and Theory
Linking Household Formation and Migration
The new economics of migration theory posits that decisions to migrate are made not only by individuals seeking to maximize economic gain but also collectively by households with the intention of spreading financial risk by diversifying economic activities across family members (Chen et al. 2003; Massey et al. 1993; Sana and Massey 2005; Stark 1991). Decisions to leave temporarily or establish a separate household are thus clearly conceptually linked given that both occur within the larger context of the resources and needs of the household economy (Massey et al. 1993) and both options spread individuals across a greater number of households either temporarily or permanently.
Building off the new economics of migration theory, we argue that the processes of cycling in and out of the natal household and establishing an independent household represent a series of sequential decisions. First, a recently married man and his wife and/or his natal family determine whether the existing economic resources of the natal household can support them or whether he needs to supplement those resources by involvement with employment, land, or businesses outside the natal household economy. This second option typically requires living away from the natal home for at least some period. We expect that this decision is based largely on economic needs and resources, although other factors may also play a role. Second, if he does decide to pursue resources outside the natal economy, the man and/or his natal family make a decision whether he will temporarily live and work away from the natal household or whether he and his wife will establish their own independent household. Although economic needs and resources may influence this second decision, we expect that noneconomic preferences, attitudes, and values will also play a significant role.
Although these two decisions are sequential conceptually, they likely occur very closely in a temporal sense. That is, it is not necessarily that a man temporarily migrates, and then he and his wife decide to establish an independent household. Rather, by the time he actually leaves the natal home, the decision about whether the move constitutes temporary migration or household fission has typically already been made. The sequential decision process is based on and incorporates the three types of household living arrangements: establishment of an independent household, cycling in and out of the natal home, and remaining continuously in the natal home. The first stage of the process is deciding between remaining continuously in the natal home versus either cycling in and out or establishing an independent household. The second stage of the decision distinguishes between the latter two arrangements.
A substantial literature has explored the relationship between household economic factors migration and household fission. Missing, however, is consideration of how community context and individual characteristics combine with household-level factors to shape the likelihood that a recently married adult son will temporarily migrate or establish an independent household. The extent to which marital processes work as intervening mechanisms in the relationships among household, community, and individual factors and living arrangements following marriage has been overlooked as well. This study addresses the limitations of prior research by investigating how community context, household factors, individual experiences, and marital processes collectively affect the decision to leave the natal household, temporarily migrate, or create a new household.
The Chitwan Valley
Our research is based in the Chitwan Valley, a primarily rural, agricultural setting in south central Nepal. Although the Chitwan Valley was largely isolated before 1970, expansion of Nepal’s networks of roads and highways during the late 1970s connected Narayanghat (Chitwan’s largest town) to Kathmandu and cities in eastern and western Nepal. Being centrally located, Narayanghat became a transportation hub for Nepal, sparking rapid development and change marked by the expansion of schools, health centers, and mass transportation. Comparing individuals born in the 1930s with those born in the 1970s, the percentage of the cohort who had attended school rose from 15 % to nearly 90 % (Brauner-Otto 2009). Although the economy is still predominantly agriculture-based, nonagricultural employment has expanded (Axinn and Yakibu 2001; Brauner-Otto 2009; Piotrowski 2013).
This setting is ideal for studying postmarital living arrangements for several reasons. Although patrilocal extended families have historically been the norm—wives move to live with husbands’ families following marriage—this norm is changing (Jones 2012). Using data from the Chitwan Valley Family Study (described shortly), we calculated that in 2008, 61 % of households in the Chitwan Valley were nuclear family households (a husband, wife, and their children). Still, elderly parents nearly always live with at least one of their married sons (most often the youngest), and property is typically equally divided among all sons (Bista 1991).
Historically in the Chitwan Valley, almost all marriages were arranged, with parents or other relatives selecting spouses for their children. However, recently, this pattern has changed: now the majority of marriages are a hybrid of arranged marriage and “love marriage” (Axinn et al. 2008; Ghimire et al. 2006), with young people and parents both having input. Also, a substantial portion of young men and women select their own spouse.
International out-migration from Nepal has also increased dramatically since the early 1960s and especially over the past 25 years. Even though the country opened up to the rest of the world in the 1950s (Thieme and Wyss 2005), it was not until the Nepalese government promulgated the Foreign Employment Act of 1985 that international labor migration to destinations other than India became viable (Shrestha 2008). This law licensed nongovernmental institutions to export Nepalese workers abroad and legitimized certain labor contracting organizations. This action ignited large streams of international migration to many countries besides India (Kollmair et al. 2006; Thieme and Wyss 2005). In 1961, the out-migration rate from the Chitwan district was less than 5 persons per 1,000; between 1985 and 1993, this rate oscillated between approximately 15 and 25 persons per 1,000. Over the past 20 years, the increase in out-migration has been more dramatic. Recent data for 2013–2014 from the Nepal Department of Foreign Employment shows that Nepalis migrated to 131 countries for work. The number of permits issued for international labor migration increased steadily and dramatically from 220,000 in 2008–2009 to 522,000 in 2013–2014, representing more than a 130 % increase within a six-year period (Government of Nepal 2014). More than 1,500 Nepalis move outside Nepal every day (Pattisson 2014).
Many of these migrants retain economic links to their households and eventually return to the Chitwan Valley (Piotrowski 2013). Based on the aforementioned trends regarding household structure and out-migration, both temporary migration and household fission appear to be on the rise. Understanding what is behind these trends is important for being able to predict future population dynamics and possible policy interventions, especially how households and communities shape living arrangement decisions through individual-level experiences and exposures.
Family-of-Origin Household Factors
Two commonly studied household-level factors related to household composition are wealth (which enhances the ability of the household to provide for household members) and the intensity of a household’s farming activity (which determines demand for household agriculture labor). Research in India, Bangladesh, China, Brazil, and Iran has indicated associations between household living arrangements and land ownership, and agricultural activities and wealth. Nuclear households are more likely when the household head is employed outside agriculture, while extended family household structure is more common for households with large landholdings. Ownership of land is also associated with a lower likelihood that men (married or not) will leave their natal home. Household wealth—as captured by a variety of measures, such as total value of household property, value of livestock, and ownership of consumer durables—is also associated with an extended family household structure, a larger household size, and a lower likelihood of household partition (Amin 1998; Conklin 1988; Foster 1993; Krishnaji 1995; Lavely and Ren 1992; Mir-Hosseini 1987; Niranjan et al. 2005; Sereny 2011; Sung 1971; VanWey and Cebulko 2007; Yadava 1966; Yi 1986). In sum, research suggests that owning land, agricultural activities, and household wealth promote extended family living arrangements and discourage household fission.
Findings from the migration literature are less conclusive. The probability of migration has been found to be highest for households with small landholdings (Bhandari 2004; Shrestha and Bhandari 2007); higher for households with middle-sized landholdings, relative to both large and small landholdings (VanWey 2003, 2005); and higher for households with large and small landholdings relative to middle-sized landholdings (Bilsborrow et al. 1987). Further, a household’s ownership of consumer durables is sometimes related to a decrease (VanWey 2003), sometimes related to an increase (Bhandari 2004; Massey et al. 2010a; Piotrowski 2013), and sometimes unrelated to the likelihood of migration (Piotrowski 2013; VanWey 2005).
Theoretical perspectives attempt to reconcile disparate findings regarding households in the migration literature. For instance, individuals from poor households with little land are believed to migrate to increase the wealth of their household, but households with more wealth and land are understood as viewing sending a household member to another community/country as an investment to produce more wealth. Results appear to vary across local and national contexts as well as depending on what other variables are included in models. One additional reason why the findings may vary across settings and models is that migration is rarely conceptualized or operationalized as having different contributing factors depending on whether that migration is temporary or permanent (as in the case of household fission). When the two types of migration are not distinguished but vary in relative frequency across settings, inconsistent estimates of common predictors could be produced as a result.
Based on the aforementioned evidence, we predict that owning farm land and consumer durables will lower the likelihood of leaving the natal home soon after marriage. We also predict that the more land is available, the more likely sons are to establish an independent household than to temporarily migrate for work.
Community Characteristics
The economic characteristics of the surrounding community are also likely to be important for choosing living arrangements. For instance, the presence of a local market enhances the ability of the household to engage in profitable small-scale commercial agriculture or off-farm employment. In addition, in economies that are transitioning from largely subsistence agriculture to incorporating manufacturing and service industries, households commonly use multiple economic livelihood strategies (Piotrowski 2008). For instance, a household may operate a small subsistence farm, while some members also work outside the household at wage labor jobs or running small businesses or shops. Men living in neighborhoods that lack economic resources, such as markets and/or nonfarm employment, will be more likely to leave the natal home (temporarily or permanently) in search of economic opportunities.
The notion of moving from a disadvantaged community to one with greater resources in order to improve household wealth is a key element of domestic and international migration theory (Bhandari 2004; Massey et al. 1993). Massey et al. (2010b) found that the presence of economic services—such as banks, buses, employment, and markets—in Nepalese communities lowered the probability of (temporary or permanent) migration out of the district. However, literature on household structure and formation has largely ignored the economic characteristics of the surrounding community. We predict that greater economic opportunity in one’s community, such as nonfarm employment and markets, will raise the likelihood that men remain in their natal home.
Individual Factors
Prior research has also focused on elements of human capital, such as educational attainment and employment, as factors shaping household composition over time. The more education a man achieves, the greater his employability for jobs other than farming. Therefore, increased education may result in his being more likely to move (temporarily or permanently) outside the agricultural setting to find a nonagricultural job in a city or another country. In addition to increasing human capital, schooling also serves as a vehicle for exposure to Western values, such as individualism (Thornton 2004). In fact, when analyzing the relationship between education and household living arrangements, it is difficult to separate the effect of human capital from the effect of exposure to Western culture and values per se (Axinn et al. 2008).
Several studies in India, China, and Malaysia have found an association between education (DaVanzo and Chan 1994; Niranjan et al. 2005; Wong 1975) or white collar occupations (Wong 1975) and nuclear or smaller families. Research on the living arrangements of young adults in Brazil, Malaysia, and Nicaragua has shown that educational attainment increases the likelihood of living outside their natal household (Johnson and DaVanzo 1998; VanWey and Cebulko 2007; Winters et al. 2009). These studies included both married and unmarried respondents, making no attempt to determine the permanence of the living arrangements. Findings from the migration literature have consistently indicated that higher levels of education are associated with a greater likelihood of both domestic and international migration (Bohra and Massey 2009; Piotrowski 2008; Piotrowski et al. 2013; Williams 2009; Winters et al. 2009).
Exposure to Western ideas and values in media such as newspapers, radio, and television (which tend to emphasize individual achievement and promote nuclear family households) or through travel to other countries (which provides exposure to alternate living situations) may influence household living arrangements (Axinn et al. 2008; Conklin 1976; DeVos 1989; Thornton et al. 1994). Although relationships between media exposure or travel and household structure or living arrangements have not been directly studied, research has examined the relationship between this exposure and other attitudes and behaviors, which are closely related to household living arrangements. Pienta et al. (2001) found that in rural Nepal, travel to Kathmandu was associated with less agreement with the attitude that a married son should care for his elderly parents. Also in rural Nepal, exposure to television, radio, and movies is associated with higher use of contraception, a preference for a smaller number of children, an increase in the belief that young adults should play a larger role in the selection of their spouse (Barber and Axinn 2004), and an increased likelihood of temporary and overall migration in Nepal (Piotrowski 2008). These behaviors and attitudes reflect the idea that young couples should have a greater say in the factors that influence their marital experience (including making conscious decisions about the number of children to have) and are likely to affect household living arrangements. Taken altogether, the evidence suggests that increased education, media exposure, or travel outside Nepal will result in men leaving the natal home and will probably also encourage the establishment of independent households.
Marital Processes
In addition to the aforementioned household-, community-, and individual-level factors, the characteristics of the marital process itself may influence postmarital living arrangements. A man’s level of influence (compared with his parents’) in selecting his own spouse may reflect the degree of internalization of Western cultural ideas by a man and his parents. Thus, if a man chooses a spouse on his own, he may be more likely to desire independence from his natal home through either temporary migration or establishing a new household. In addition, if a man chooses his spouse without parental involvement, and the parents are unhappy, the couple sometimes elopes, which raises household acrimony (Allendorf 2013). Such a scenario would certainly increase the likelihood that a couple would form an independent household. Further, the level of participation in spousal selection may influence and reflect the initial closeness of the emotional bond between a man and his wife, which may affect decisions to establish a separate household (Macht 2008). For all these reasons, we predict that having a greater say in the spouse-selection process will make a man more likely to establish an independent household with his wife.
Another important factor in the marital process is age at marriage. For instance, in Nepal (as elsewhere), individuals with higher education tend to marry later (Yabiku 2005). Individuals with greater media exposure or travel experience are more likely to delay marriage (Axinn et al. 2008). Respondents who marry later may have gained emotional maturity and increased independence, which may translate into greater likelihood of establishing a household independent of one’s natal home. However, older respondents may also have greater desire for stability in living arrangements and may thus be less likely to leave the natal home.
To summarize, our theoretical framework suggests that community context (market access and employment availability), household context (land holdings and household wealth), individual experiences (travel, media exposure, and education), and marital processes (spouse selection process and age at marriage) combine to influence the choice between three postmarital living arrangements: staying in the natal home, temporarily migrating, or establishing an independent household.
Data and Methods
The data we use to test our predictions regarding postmarital living arrangements are from the Chitwan Valley Family Study, a longitudinal, multilevel mixed methods study of 151 neighborhoods (naturally occurring clusters of 5–15 households throughout the valley) and all the residents therein from 1996 to 2008. We draw from an individual-level survey of 5,271 respondents in 1996, which had a 97 % response rate and included a life history calendar collecting information on prior travel and employment experiences. We also draw from household surveys conducted in 1995, 2001, and 2006 (response rates from 98 % to 100 %). Finally, we use data from a neighborhood history calendar collected in 1996, which assessed the distance from each neighborhood to institutions, such as schools, hospitals, marketplaces, and places of employment. To assess temporary migration and household fission, we use monthly demographic household registry data collected from 1997 through 2008. We focus on the 344 men who were living in their natal homes in 1996 and married between 1996 and 2006. The ages of the men at the time of marriage ranged from 17 to 36 years.
Outcome Measure: Postmarital Living Arrangements
We draw from a conceptual framework of the life course that considers marriage an event that is integrally and temporally linked to decisions regarding household formation (Glick 1947, 1989). Therefore, the years immediately following marriage are particularly appropriate for examining the factors related to household living arrangements. We thus use data from the first 24 months of marriage to measure our outcome variable, sons’ postmarital living arrangements. Our dependent variable has three categories. Men who were living in their natal household for each of the 24 months are defined as having remained in the natal home. Men who live for at least six continuous months away from the natal home, and whose wives live with them for at least three of those months, are considered to have established an independent household. Men who live away from the natal home for less than six months, or migrate for longer periods, but without their wife, are considered to have temporarily migrated away from the natal home.1
Our definition of temporary migration is based on theories of household formation and departs somewhat from standard definitions found in the literature on internal and international migration. If a man migrates from his natal home, even for a lengthy period, but his wife and children continue to live in that home, he will remain an integral part of the household economy of his natal home through remittances and other contributions (Piotrowski 2008). Therefore, we do not consider this the establishment of an independent residence. However, because our focus is on household formation rather than migration per se, we do not factor in distance from the natal home into our definition of temporary migration, which may include temporarily living in a community not far from the natal home. Although our definition of migration is based on movement away from the household, all but one of the men in our sample who temporarily migrated did so to locations outside of their local neighborhood. Men who established an independent household did so outside of their natal neighborhood in 88 % of cases.
Furthermore, because our interest is in household formation—and therefore, men’s relation to their natal home—we do not separately consider subsequent migration from a new independent household that the man has established with his wife. Whether the man currently lives in this new household or may be temporarily living away, he is considered to have established an independent household. Table 1 illustrates both the full set of possible living arrangements as well as the three living arrangements used in our analysis.
Predictors
Community-Level Measures
For community-level economic factors, we use measures of whether in 1996 there was (1) a workplace within a 10-minute walk and (2) a marketplace within a 5-minute walk of the neighborhood in which the respondent’s natal home is located. Workplace includes any organization or business, such as a school, hospital, store, or factory, employing at least 10 persons. We chose to measure distance in terms of travel time by foot because that is the primary mode of travel in the Chitwan Valley. We use a shorter distance to market than workplace because local markets are more geographically dispersed relative to workplaces and because travel to a market tends to be a more strenuous activity given that goods are usually also transported.
Household-Level Measures
For household-level economic characteristics, we use the total land area farmed by the household. In the Chitwan Valley, land area is measured in kattha, with 30 kattha equal to one hectare of land. The land area farmed taps into the amount of person-power needed to conduct household tasks as well as the resources available to support members of the household. Because the effect of land size may not be linear, we separate land size in kattha into the following categories: 20 kattha or fewer, 21–30 kattha, and more than 30 kattha. The middle category of 21–30 kattha is considered sufficient to support a household of four to six persons in the Chitwan Valley; approximately 20 % of the sample falls into this category.2 In the Chitwan Valley study area, nearly all households own some land. More than 96 % of our respondents come from natal landowning households. Therefore, we do not include a variable for land ownership.
We also include a more general measure of household resources/household wealth: specifically, the number of consumer durables owned by the household, including items such as radios, televisions, and bicycles, as well as agricultural items such as tractors, carts, and farm tools.
In addition, we include a measure of the number of adults aged 15–59 in the household in the month the respondent was married. This variable serves to put other household-level variables on an approximate scale, representing the amount of labor available to the household.3 We also include a measure of the number of dependents in the household: the total number of individuals under age 15 or over age 59. These two measures together capture the demand for available household resources.
Individual-Level Measures
For measures of exposure to Western ideas and values, we first include a media exposure index constructed as follows. In the 1996 survey, respondents were asked how often they read newspapers, listen to radio, and watch television. For each form of media, responses are coded as 0 if the individual had never been exposed. For newspaper and radio, responses were coded as 1 if the individual used the respective form of media less than once per week; a value of 1 indicates viewing television less than once per month. Responses were coded as 2 for individuals who watched television at least once per month and read the newspaper or listened to the radio at least once per week. The media exposure index is the average of the individual scores for the three forms of media; the range of the index is thus 0 to 2. We also include a measure of whether a married son had ever traveled and/or lived outside Nepal prior to the end of 1996. Responses were coded 1 if the individual ever traveled or lived outside Nepal and coded 0 if not.
We use the respondent’s educational attainment to measure the respondent’s human capital and as a third measure of exposure to Western attitudes and values. It is reasonable to expect that the effect of educational attainment on household living arrangements may not be linear. For instance, attitude changes that are frequently associated with increasing education may not transform unless several years of schooling are completed and may eventually taper off at higher levels of educational attainment. In addition, employment opportunities are often available for only those with a significant amount of schooling. Therefore, it is reasonable to expect that the effect of education on household living arrangements operates as a threshold effect. Because the highest grade in Nepali high schools is grade 10, we use a measure of education indicating whether the respondent had completed at least 10 grades of formal schooling by 1996.
To consider the role of the marriage characteristics, we include a five-category variable for the relative degree of participation of the respondent and his parents in choosing his spouse: (1) the parents had sole authority in choosing the spouse; (2) the parents had primary authority; (3) the respondent and his parents had equal say; (4) the respondent had primary say; and (5) the respondent chose his spouse entirely on his own. We also use a measure of the age at which the respondent married.
Finally, we include several control variables that influence a wide variety of household and individual behaviors in Chitwan Valley (Axinn et al. 2008). We include variables for the primary caste/ethnic groups present in the Chitwan Valley—Brahmin/Chhetri (upper caste Hindu), Dalit (lower caste Hindu), Newar, Hill Janajati (Hill indigenous), and Terai Janajati (Terai indigenous)—as well as the education level (in years) of the respondent’s father.
Analytic Strategy
Our analytic strategy has two stages. In the first stage, we consider migration and household fission separately, similar to how these concepts are typically measured in the literature. Our outcomes in this stage are two binary variables. The first outcome, migration, occurs when the man either temporarily migrates from his natal home or establishes an independent household with his wife. The second outcome measures household fission: whether or not the man and his wife have established an independent household; temporary migration and remaining in the natal home are collapsed into a single category. We use separate probit models to estimate the likelihood of (1) migration and (2) household fission.
In the second stage of our analytic strategy, we jointly model the full set of possible household living arrangements as a sequential series of two decisions. We use a censored (two-step) probit model, which is an extension of the Heckman selection model to binary outcomes (Dubin and Rivers 1989). The outcome for the first equation is the same as for the binary model of migration described in the previous paragraph: whether or not the man leaves his natal home. The first equation is thus conceptually equivalent to the model of migration in the first stage of our analysis, although the estimation processes are not mathematically identical.4 The outcome for the second equation is whether, after the decision to leave the natal home has been made, the man also decides to establish an independent household. In other words, the possible outcomes for the second equation are establishing an independent household and temporarily migrating; individuals who remain in the natal home do not have a value for this outcome and are not included in this equation.
The censored probit model is preferable to simply modeling the likelihood of establishing an independent household versus temporary migration for the subsample of those who have made the decision to leave the natal home, because the censored probit model accounts for the correlation of unobserved characteristics that may influence both the initial decision to leave the natal home and the second decision to establish an independent household. To accurately estimate a censored probit model, it is necessary to include at least one predictor that influences the first equation but not the second equation. Based on the results from the binary models, we select the community economic characteristic variables (presence of a market and presence of a workplace) as the predictors that influence the decision to leave the natal home but not the decision to establish an independent household. For all our models, we use adjusted standard errors to correct for correlated errors due to the clustering of men within natal households and neighborhoods (Hsiao 2003).
To check the robustness of our measures of establishing an independent household, we ran two sets of sensitivity tests. First, we estimated the aforementioned models, using an outcome measure that defined establishment of an independent household as occurring if the man lived away from his natal home continuously (with his wife) for at least three months. These results were very similar to the results using the six-month criteria and are available upon request. In fact, the number of men establishing an independent household increases only by 6 (from 85 to 91) when the three-month instead of the six-month criteria are used. This finding suggests that the measure of establishing an independent household itself is stable and robust to the exact number of months living continuously away from the natal home that are used as criteria.
As a second sensitivity test, we also estimated the models, using only the first 18 months after marriage. Again the results are very similar to the results using the 24 months and are available upon request. Also, the changes in the percentages of respondents in each of the three outcome categories are not large. When the 18-month criteria are used, 22 % of respondents establish an independent household; 43 % temporarily migrate, and 35 % remain continuously in their natal home. When the 24-month criteria are used, the figures are 25 % of respondents establish an independent household; 45 % temporarily migrate, and 31 % remain continuously in their natal home.
To more clearly visualize and interpret the magnitude of effects of the statistically significant coefficients from the censored probit models, we calculate a series of simulated predicted probabilities of migration and/or establishing an independent household for each predictor that showed a statistically significant effect in the equation modeling that decision process (leaving vs. remaining in the natal home; establishing an independent household vs. temporary migration).5 We use the multivariate delta method to calculate variances and covariances for these predicted probabilities, allowing us to test whether changes in each predictor have a significant effect on the value of the probability of migrating away from the natal home and/or establishing an independent household. In other words, the results we present are the independent relationships that a set of community-, household-, individual-level characteristics and marital processes have with probabilities of leaving one’s natal home and establishing an independent household in the two years following marriage. The values of the predicted probabilities of establishing an independent household are absolute rather than conditional on having left the natal home.
Results
Sample descriptive statistics are presented in Table 2. Nearly one-half of the men come from households in a neighborhood within a 5-minute walk of a marketplace and a 10-minute walk of a workplace in 1996. Approximately 20 % of households fall into each category of land holdings. The average number of consumer durables owned by the household is slightly more than two; on average, slightly less than four adults live in the natal home. In addition, 37 % of the men completed at least 10 years of formal education, and 30 % had traveled outside of Nepal. The average media exposure is 1.57 on a scale of 0 to 2. The largest proportion of men selected their spouse on their own (44 %), and the average age at marriage was 24 years old. Finally, the largest ethnic group represented is Brahmin/Chhetri, making up 54 % of the sample.
Probit Models
The probit models of migration and household fission, (displayed in Table 3) indicate that different factors are important for the two decisions. From the probit model of migration, where temporary migration and establishing an independent household are collapsed into one category, we see that the community economic characteristics, number of consumer durables, and age at marriage influence the likelihood of migrating versus remaining in the natal home. From the model of household fission, where temporary migration and remaining continuously in the natal home are collapsed into one category, we see that number of consumer durables, size of landholdings, media exposure, and the spousal-selection process influence the likelihood of household fission. Thus, estimating either of these models in isolation does not provide the complete picture of the relationships among individual, household and community characteristics, and household living arrangements.
Censored Probit Model
The censored probit model incorporates all three types of household living arrangements into a sequence of two decisions, each modeled by a separate equation: (1) leave versus remain in the natal home, and (2) establish an independent household versus cycle in and out of the natal home (temporary migration). Not surprisingly, the results from the first equation are very similar to the results from the probit model of out-migration given that the estimation strategy for the probit model of migration and the censored probit leave the natal home equation are conceptually equivalent, although not mathematically identical (see Analytic Strategy section for details).6 Results from the censored probit model, presented in Table 4, suggest that different factors influence the initial decision about whether to leave the natal home and the subsequent decision whether to establish a permanent independent household. The initial decision to migrate is driven by community economic characteristics, household wealth (although not land holdings), and the age of the man at marriage. The subsequent decision to establish a permanent independent household is also influenced by household wealth in addition to individual characteristics, including education level, media exposure, and the spousal selection process.
Community Economic Context
The presence of a market and workplace nearby have similar impacts on the probability of leaving the natal home. As shown in Fig. 1, the discernible role of a market within a 5-minute walk is that the predicted proportion of men who leave the natal household decreases (from 77 % to 62 %; p < .05). The availability of a workplace within a 10-minute walk is also associated with an decrease in the predicted proportion of men who leave the natal household (from 77 % to 61 %; p < .05).
Household Economic Context
Figure 2 shows that as the number of consumer durables increase, both the probability of leaving the natal home and the likelihood of establishing an independent household steadily decline. As the number of consumer durables increases from zero to six, the predicted probability of leaving the natal home decreases from 81 % to 48 % (p < .05). The predicted probability of establishing an independent household also declines from 40 % to 9 % (p < .05). Clearly, an increase in household resources (reflected here by owning consumer durables) helps keep married sons living in their natal homes, while the lack thereof makes it more likely that the man and his wife will establish their own household.
Figure 3 illustrates a U-shaped relationship between landholdings and the probability of establishing an independent household. The highest predicted probability of establishing an independent household (29 %) is for households with the largest landholdings (>30 kattha). The second highest predicted probability (26 %) is for households with the smallest landholdings (≤20 kattha), and the lowest predicted probability of establishing an independent household (15 %) is for households with medium-sized landholdings (21–30 kattha). Predicted probabilities of establishing an independent household are significantly different (p < .05) between households with >30 or ≤20 kattha and households with 21–30 kattha. It seems that a household’s land plays a key role in married son’s living arrangements with low or high amounts of land pushing sons and their wives to establish an independent household.
Individual Characteristics
As illustrated in Fig. 4, having at least 10 grades of formal education increases the probability that a man and his wife will establish an independent household. For those who have at least 10 grades of education, the predicted probability of establishing an independent household is 32 % compared with 21 % for those who have fewer than 10 grades of formal education.
Media exposure also increases the likelihood that a man and his wife will establish an independent household. As Fig. 5 illustrates,7 as media exposure increases from 0.67 to 2, the predicted probability of establishing an independent household increases from 14 % to 31 %; each change of 1.0 units on the media exposure index results in a significant (p < .05) increase in the predicted probability of establishing an independent household. Taken together with results for education, it appears that exposure to Western values, whether through media or education, increases the likelihood of a man and his wife establishing an independent household. However, the effect of international travel is insignificant.
Regarding the spousal selection process and living arrangements after marriage, Fig. 6 reveals a U-shaped relationship. Men who shared equal decision-making power with their parents are much less likely to establish a separate independent household (7 %) than either men who chose the spouse on their own (33 %) or men whose parents had sole decision-making authority (26 %); both differences are significant at p < .05.
Finally, marrying later is associated with a decreased likelihood of leaving the natal home. In fact, as the age of marriage increases from 20 to 30, the predicted probability of leaving the natal home decreases from 77 % to 55 % (p < .05).
Conclusions
The results generally confirm that community economic characteristics influence household living arrangements, as predicted. In particular, the presence of a local market and/or workplace is associated with an increase in the likelihood of remaining in the natal home. This finding is not surprising given the findings from the migration literature that a common reason for out-migration is the search for better economic and employment opportunities (Massey et al. 1993). The findings regarding community economic resources highlight the importance of examining all three of the types of household living arrangements through the sequential decision-making process discussed in this article: community economic characteristics do not appear to play a role in the decision to establish an independent household per se.
Household economic characteristics also influence whether a man remains in his natal home. However, household land holdings primarily operate through influencing whether a man decides (with his wife) to establish an independent household or cycle in and out of his natal home after the decision has been made for him to leave his natal home. We did not find support for our prediction that landholdings would influence the initial decision to move out of or remain in the natal home. Therefore, we see an interesting pattern of community economic resources (or the lack thereof) driving the initial decision to leave the natal home, whereas household economic resources, and the associated labor demand, influence whether a man and his wife establish a households independent from his natal home.
We theorized that because of increased resources and higher labor demand, the greater the landholdings of the natal home, the less likely the man and his wife would be to establish a separate household. However, the actual findings suggest a more complex relationship between landholdings and household living arrangements. Men whose natal households have large or small landholdings are more likely to establish an independent household than men from households with medium-sized landholdings. One explanation for the somewhat surprising finding that large natal household land holdings lead to the formation of independent households is the possibility that parents with particularly large land holdings are choosing to divide their land among their sons.8 This results in household fission, which nevertheless uses the resources of and concurrently fulfills the labor demand of the natal household. Households with medium-sized land holdings do not have the capability to divide the land among their sons and still provide each son with sufficient land on which to support his family. In the case of households with minimal to no land holdings, married sons do not have much to either help with or inherit. In fact, they may also feel like a burden to a low-resource household; therefore, when they move out of their natal home, they are more likely to establish a permanent independent household with their wife rather than eventually cycle back into the natal home.
We also find that household wealth in the form of consumer durables is associated with a decrease in the likelihood of both deciding to leave the natal home and subsequently deciding to establish an independent household. We can understand these patterns as indicating that household wealth serves as a strong link between a man and his natal home. However, we do not find evidence that a lack of household wealth drives men to cycle in and out of their natal home in search of opportunities to increase the wealth of this household.
We find that greater media exposure and education are both associated with a higher probability of establishing an independent household. The highly Western-centered or upper-class–focused media as well as the educational system likely portray the individualistic idea of a nuclear household as an option. However, we did not find that media exposure or education drive the initial decision to leave the natal home. Here again, we see the importance of exploring forms of household living arrangements as a series of two distinct decisions: the set of experiences that have a strong impact on the initial decision may have a negligible impact on the subsequent decision, and vice versa.
We hypothesized that greater involvement of men in the selection of their spouse would be associated with a greater likelihood of establishing an independent household. The actual results were more complex. We found that men who shared equal decision-making authority with their parents in the selection of their spouse were the least likely to establish a separate independent household relative to both men who chose their spouse on their own and men whose parents chose their spouse for them. It is possible that the key factor here is neither tradition nor emotional nucleation, but rather the greater harmony that may exist when the son and his parents contribute equally to the spouse-selection process (before and/or after that process). Parents may be more likely to get along well with a spouse whom they helped select. It is also possible that relative to situations in which the parents chose his spouse for him, the son may feel more respected by his parents and appreciative of the role that he was allowed in the spousal selection process, which may in turn lead to a greater desire to remain connected to his natal household. If this is the case, it should be understood in the context of changing family structure and family dynamics in a historically patrilocal setting. Because nuclear families are becoming more common, the primary driving factor related to decisions regarding household living arrangements may not be the level of traditionalism of the family but rather the extent to which changing needs and expectations on the part of youth can be incorporated into long-standing family practices.
To summarize, the results indicate that community-level characteristics, exposure to Western values, and the marital experience—all of which have been neglected in the literature on household structure and fission—are important factors contributing to the living arrangements of recently married men in historically patrilocal settings. The results also indicate that the relationships between the spousal selection process and the size of landholdings of the natal household are more complex than is often assumed.
Finally, the results indicate the importance of analyzing decisions related to postmarital living arrangements in a framework that simultaneously considers the three outcomes of remaining in the natal household, temporary migration, or establishing an independent household through a series of decisions: (1) whether to leave the natal home, and (2) whether to do so temporarily or permanently. Nearly all the community, household, and economic characteristics influenced only one of the two decisions, showing limited or no effect on the other decision. Therefore, research that explores only two of the options for household living arrangements—lumping together temporary migration and remaining in the natal home (common in the literature on household structure), or collapsing household fission and temporary migration (common in the migration literature)—will not fully capture the complex set of processes that lead to decisions regarding household living arrangements.
Acknowledgments
We thank two anonymous Demography reviewers for helpful comments and suggestions. This research was supported by generous grants from the National Science Foundation, Partnerships in International Research and Education (OISE 0729709), and the National Institute of Child Health and Human Development (P2C HD050924 and T32 HD007168).
Notes
Our choice of the six-month criteria is as follows. Household membership in the Chitwan Valley is constantly in flux. Many men and their wives may spend a month or two away from their natal home with a relative or for temporary work. Even after establishing their own home, they may return to the natal home to help with the harvest, for example. Thus, the challenge is to determine a threshold of a continuous number of months, which most likely indicates that the new household is permanent but allows for the fact that there will be time spent away from the new household after it is established (e.g., the number of months cannot be too high). Based on residence and migration patterns in the Chitwan Valley, we determined that the minimum threshold of continuous months needs to be at least three because numerous couples spend one to two months temporarily in another household before returning home. Comparing the outcome measure using between three and six months, we found a high degree of stability given that only six respondents who spent at least three continual months away from the natal home did not spend at least six months away from the natal home. This suggests that the six-month measure is likely an accurate measure of the establishment of permanent independent households. We used the six-month criteria because it is slightly more rigorous: it excludes those individuals who returned to the natal home after spending three to five months away. We also ran sensitivity tests, estimating our models using outcomes based on the three-month criteria. The results are very similar to the models using outcomes based on the six-month criteria and are available upon request.
Immediately following marriage, two broad factors are of critical importance in influencing whether the man remains in his natal home, cycles in and out of that home, or established an independent household. First, can the natal household cope with the resources required to support an additional household member, on top of the wedding costs the household has already incurred? Second, what is the quality of the interpersonal relationships between the newlywed wife and the other members of the natal household; for example, are these characterized by harmony, acrimony, and so on? We chose to use the 24 months following marriage as our window during which to measure men’s household living arrangements because this is a reasonable amount of time for the man, his wife, and the members of his natal household to answer these questions and to determine whether it is most appropriate for him and his wife to remain in the natal home, cycle in and out, or establish an independent household. We also ran sensitivity tests, estimating our models using only the first 18 months after marriage. The results are very similar to results using the full 24 months and are available upon request.
We had originally explored using a five-category measure: 10 kattha or fewer, 11–20 kattha, 21–30 kattha, 31–50 kattha, and more than 50 kattha. These five categories approximate the quintiles of the distribution of land size in kattha. However, preliminary analysis suggested that the effect sizes for the smallest two categories were very similar and that the effect size for the largest two categories were very similar. Because our sample size is not large, we thus combined the categories to create more robust estimates less likely to be influenced by noise in the data and outlying cases.
More than 80 % of the population of our study area are farmers, and farming requires heavy physical work. Therefore, most people in our sample do not actively participate in farming or other economic activities after age 60, although they may help with childcare and other household chores. The harsh living conditions and heavy physical work that individuals perform during the early years of their life contributes to an accelerated aging process relative to industrialized nations.
The equations are as follows (y1* and y2* are latent variables; y1 and y2 are observed variables): y1* = x1β1 + ε1; y1 = 1 if y1* > 0; y1 = 0 otherwise. y2* = x2β2 + ε1; y2 = 1 if y2* > 0; y2 = 0 otherwise. The likelihood of the censored probit is [Φ(x2β2, x1β1, ρ)]y1y2 [Φ(x1β1) – Φ(x2β2, x1β1, ρ)]y1(1–y2)[1 – Φ(x1β1)](1–y1). ρ is the correlation between ε1 and ε2.
Probabilities are calculated for each respondent, who retains his or her own value for each of the other covariates. The simulated probabilities of each living arrangement are then averaged across respondents.
The model of household fission approximates fairly well the results from the second censored probit equation, although the significant association between education and establishing an independent household is not visible in the probit model. However, the probit model of household fission and the second equation in the censored probit model are not equivalent.
The lowest value of media exposure in Fig. 5 is 0.67 because only two respondents have media exposures of less than 0.67.
Roberts (1997) found this practice of households splitting landholdings among their sons in rural China, particularly among households where one or more sons were also engaged in employment outside household-based subsistence agriculture.
