The roles of family resources and family structure in moving from the parental home and village among young Indonesians

This study assesses the effects of family resources and family structure among young Indonesians on the likelihood of leaving the parental home and village in general and for three types of reasons: work, education, and marriage. Using all five waves of the Indonesia Family Life Survey, we find that the effects of family resources and structure differ by reason for moving. For example, parental education is positively related to moving for education, but not to moving for work or marriage. We also find that being the oldest child is positively related to moving in general and belonging to an extended family is negatively related to moving for work. Our results suggest that moving from the parental village for work, education, and marriage are different processes. Furthermore, while some of the findings are in line with previous findings for leaving home in Western countries, other findings are typical of developing countries.


| INTRODUCTION
The migration of adolescents and young adults is distinct from the migration of other age groups because it occurs during a period when individuals face profound changes in their lives due to physical maturation and cognitive, social/emotional, and interpersonal changes.
Migration also adds another layer of complexity to the lives of young people, as it tends to coincide with life events that mark the transition to adulthood, such as transitions to work, marriage, and higher education (Juárez et al., 2013;Lloyd & Grant, 2005). Therefore, although migration may expand young people's opportunities, it can also lead to increased levels of vulnerability (Juárez et al., 2013).
Whether a young person who migrates is able to take advantage of these opportunities, while also dealing with new situations and potential adversity, very much depends on the individual's physical and emotional maturity, which is generally determined by his/her age at migration (Juárez et al., 2013). Indeed, studies have shown that the outcomes of migration depend upon the age at migration (McDonald, Utomo, Utomo, et al., 2013). Consequently, young people who are living in a new environment with less family and social support may be especially vulnerable (Juárez et al., 2013). This potential vulnerability associated with migration for young people motivates us to investigate the migration 1 of young people without their parents for the first time. Therefore, our focus in this study is on leaving the parental household while also leaving the parental village (or community within a city; from here "village").
In the literature on leaving the parental home in Western (developed) countries, it has been documented that the timing of leaving home is partly determined by family resources and family structure (De Jong Gierveld et al., 1991;Garasky, 2002;Mitchell et al., 1989;Mulder & Clark, 2000). The effects of family resources and family structure are likely to vary depending on the reason for leaving home, such as work, education, or marriage or household formation.
Whereas there is abundant evidence of an important role of family resources and family structure in leaving the parental home in Western countries, this is much less the case for developing countries. This is unfortunate, because the role of these factors is likely even more important in developing countries as many young people in developing countries live in poverty and vulnerability (Juárez et al., 2013). In the context of developing countries, the migration of young people is often the result of a collective family strategy. Families may send children away for a number of reasons, including to work in order to diversify the family's sources of income; to enable them to continue their schooling elsewhere due to a lack of educational facilities in close proximity; or even to marry as a strategy to secure their livelihood or to extend their economic resources (Juárez et al., 2013;Kaur, 2010;Stark & Levhari, 1991). In this sense, young people from more disadvantaged families who migrate may be particularly vulnerable. Family structure factors, such as birth order or whether a family is a nuclear or extended family (Bratti et al., 2016;Root & De Jong, 1991), may also play a greater role for migrants from developing countries than for their counterparts from Western countries.
The aim of this study is to assess the effects of family resources and family structure on the timing of moving out of the parental home and village in a developing country, using the case of Indonesia.
We contribute to two bodies of literature. First, we contribute to the literature on leaving the parental home 2 by investigating this topic for a developing country. Second, we contribute to the literature on youth migration in developing countries, by investigating the first moves of young people without either of their parents. For this purpose, we include indicators of family resources and family structure in our analysis that are typical of developing country contexts: involvement in a farm business, which entails high uncertainty of family income (Stark & Levhari, 1991), birth order, and whether a family is nuclear or extended.
By using the case of Indonesia, we follow up on a few previous studies of migration in Indonesia across much wider age ranges than ours (Muhidin, 2002;Wajdi et al., 2017) and on some studies that have focused specifically on young people (Malamassam, 2016;McDonald et al., 2013;Witoelar, 2008).
Because the effects of parental resources and family structure may differ depending on the reason for moving, we distinguish between different types of moves: that is, moves for work, education, or marriage. We use data from all five waves of the Indonesian Family Life Survey (IFLS). We perform discrete-time event history analysis to determine whether young adults aged 15-28 moved from their parental home and village and multinomial logistic regression of moving for specific reasons compared with not moving.
2 | THEORETICAL AND RESEARCH BACKGROUND 2.1 | Leaving home and leaving the village Villages in Indonesia represent the lowest-level administrative areas, which are called "desa" (rural or urban) or "kelurahan" (urban) (Kato, 1989; we use the term "village" to denote both desa and kelurahan). The topic of our analysis is leaving the parental home while also leaving the parental village (from here: "leaving the parental home and village" or simply "leaving the parental village"). As argued in the next section, we think leaving the parental home and village is a meaningful concept in the context of Indonesia.

| The role of the village in the lives of Indonesians
In the lives of Indonesians, the village traditionally has a role that is similar to that of a home. It is usually difficult to separate village or community ties from family ties (Hugo, 1981). The limits of an Indonesian village community are primarily determined by a sense of identity that is marked by identity symbols (e.g., a sacred building or a community house) and by the social activities centring around these symbols (Koentjaraningrat, 1967).
To a great extent, all individuals who live within the same neighbourhood in a village are referred to as kin (relatives) and often treated as such, without much differentiation being made between the roles of neighbours and kin (Jay, 1969). While modernisation has partly altered the social and cultural landscape of Indonesians in terms of the roles of village, community, kinship, and family, many dimensions of traditions in Indonesia tend to persist and affect the lives of the young (e.g., traditional marriage norms; see Buttenheim & Nobles, 2009). Young people in Indonesia still tend to develop a strong primary identity from place, kinship, and religion; this is particularly true for those living outside the cities (Nilan, 2008). Therefore, if young people leave the parental home without leaving the village, they are unlikely to experience the radical changes in terms of decreased parental support, disruptions of local ties, or challenges in adjusting to a new environment that young people who are moving longer distances might face (Leopold et al., 2012). We therefore think that, in the context of Indonesia, it makes sense conceptually to study leaving the parental home and village as an important event in young adulthood.
In some cases, the traditional boundaries of villages coincide with administrative units. But in other cases, the boundaries of villages were set by the governments of larger administrative units in order to standardise the local administrative structure. Larger villages are often the result of merging by the central government (Kato, 1989;Koentjaraningrat, 1967). Thus, today, the administrative unit of the village does not always represent the traditional sense of identity that it once did, especially in the cities and other urban areas. We therefore acknowledge that the geographical unit we use in our analyses is not ideal in all cases.

| Leaving the parental home and young adult migration: Developing versus Western countries
In Western cultures, leaving the parental home has long been considered as a marker of the transition to adulthood (Blaauboer & Mulder, 2010;De Jong Gierveld et al., 1991;Leopold et al., 2012). In developing countries, and particularly in Asia, leaving the parental home is not (yet) viewed as a common cultural marker of the transition to adulthood (Ting & Chiu, 2002). Rather, for young people in developing countries, leaving the parental home is often associated with marriage (Juárez & Gayet, 2014). In some contexts, young people continue to live with their parents after they marry or have a first child (De Vos, 1989;Witoelar, 2008). While living in the parental home after marriage was a common norm in Indonesia in the past (Buttenheim & Nobles, 2009), children are actually expected to establish their own household after marriage (Jay, 1969;Megawangi et al., 1995). Indeed, in the cultures of some developing countries, leaving the parental home before marriage seems to be unthinkable, except when it is driven by a need to migrate for work or to pursue higher education (Kõu et al., 2017). Furthermore, the marriage migration of young children can sometimes be used as a strategy to lessen the economic burden of the family (Kaur, 2010;Stark & Rosenzweig, 1991).
When a young adult in a Western country is ready to leave the parental home, the decision to move is often made not solely based on the wishes of the young adult, but on family or household factors (Blaauboer & Mulder, 2010). It has also been pointed out that in developing countries, particularly in Asia, young people's decisions to move are typically made in the context of a culture of familism (Ting & Chiu, 2002), which emphasises collective family welfare and kinship ties. Thus, a young person's decision to migrate may be heavily influenced by the family (Kõu et al., 2017, for India). However, children are not always the passive recipients of parental decisions.
Indonesians, for example, encourage children to be independent and self-reliant, albeit not at the expense of social bonding (Megawangi et al., 1995). Furthermore, when children have independence, and particularly financial independence, they are better able to fulfil their family obligations and contribute to the goals of the family, such as providing for their parents' old age-security and for family continuity (Albert, Trommsdorff, Mayer, et al., 2005;Jay, 1969;Megawangi et al., 1995;Naafs, 2013).

| Family resources and moving
Following the literature on leaving the parental home in Western countries, we regard family resources as important determinants of leaving home (Avery et al., 1992;Blaauboer & Mulder, 2010;Whittington & Peters, 1996). The family and the young adult can use such resources in two ways: to facilitate the young adult's departure from home (if they view it as desirable) or, instead, to prevent it (if they view it as undesirable). The main types of family resources are material resources, such as those derived from income, employment status, 3 and assets (including homeownership); and nonmaterial resources, such as the parents' education and health. According to De Jong Gierveld et al. (1991), family resources affect leaving home in different ways, depending on the types of resources involved and the reason for leaving. They hypothesised that several material and nonmaterial resources, such as the parent's income, economic status, and education, are positively associated with leaving the parental home in general and with leaving the parental home for work and education, whereas the opposite is the case for leaving the parental home to live with a partner.
Parental income is an important material resource enabling the young adult to bear the costs of moving and of setting up a new household. At the same time, it helps parents prevent the child from leaving home too early, and, from the child's perspective, it could cause reluctance to leave a so-called feathered nest-a comfortable home situation (Avery et al., 1992). In line with these different potential effects of parental income-facilitation or prevention-the effect of parental income on the likelihood of leaving the parental home has been argued to change from negative to positive with the increasing age of the child (Blaauboer & Mulder, 2010;Whittington & Peters, 1996). While the child's own income has been described as a resource that enables him/her to leave the parental home (Mulder & Clark, 2000), others have argued it is endogenous to the decision to leave and should not be taken into account (Whittington & Peters, 1996). In the context of Indonesia, we take the view that family income is a pooled resource. We therefore expect to find that family income is negatively related to leaving the parental village at younger ages, but is positively related to leaving the village at later ages.
Next to income, a better employment status of the parents can also be a source of material resources (De Jong Gierveld et al., 1991), particularly in relation to the economic dimension of a job (De Graaf & Kalmijn, 2001). A better employment status of the parents is therefore expected to affect leaving the parental village positively.
Related to the effect of family income, we include in our analysis the involvement of the family in agricultural work. The high level of risk and the uncertainty of income associated with agriculture tend to drive the migration of the members of farming families in order to smooth consumption (Stark & Levhari, 1991). Therefore, we expect to observe that the involvement of family members in farm businesses increases the likelihood of moving in general and of moving for work, but is negatively associated with moving for education. As marriage could be more prevalent among families involved in agriculture (Dixon, 1971, for the case of Eastern Europe, Middle East, and Asia), we expect to find that participation in agricultural activities has a positive effect on moving for marriage.
The effect of homeownership may differ somewhat from the effects of other material family resources. On the one hand, homeownership is usually the result of having a good financial situation (Blaauboer & Mulder, 2010). On the other hand, homeownership may delay the timing at which the children leave home because it provides them with stability, including the opportunity to inherit the house (Ting & Chiu, 2002). Thus, homeownership might be positively or negatively related to leaving the parental village.
The parents' education is another important nonmaterial resource. Particularly for developed countries, it has been put forward as an important determinant of whether young people adhere to the values that promote leaving home, such as autonomy and openness to innovation (De Jong Gierveld et al., 1991). Moreover, young adults whose parents are highly educated will have a greater tendency to pursue higher education or job opportunities and be more willing to move to do so. In line with these ideas, parental education is expected to be positively related to moving for work or education. Conversely, moving for marriage can be viewed as related to having a traditional upbringing that determines what behaviour is socially acceptable upon reaching adulthood, and it would therefore be expected to be negatively related to parental education (De Jong Gierveld et al., 1991).
However, De Jong Gierveld et al. did not find support for this expectation, and one might also see parents' education as a resource facilitating marriage. Findings for the United States and the Netherlands have indeed shown that parental education is positively related to leaving the parental home not only to live alone, but to live with a partner (Blaauboer & Mulder, 2010;Mulder & Clark, 2000). Therefore, an alternative expectation is that the effect of the parents' education on moving in general, and on all types of moving, will be positive.
Finally, the parents' health can be seen as a nonmaterial resource.
In the Indonesian context, children are usually expected to take care of their parents. We therefore expect to find that having unhealthy parents decreases the likelihood of moving from the parental village, regardless of the reason for the move. Indeed, having an older parent in close proximity who was in poor health has been found to negatively affect the migration of working-age adults in Indonesia (Rammohan & Magnani, 2012).

| Family structure and moving
Opportunities to migrate may not be equal across the children within a family. In the context of developing countries, older, particularly adult children, are beneficial for the family in terms of supply of labour and providing assistance with personal care for parents or young siblings (Johnson & DaVanzo, 1998). In Indonesia, older siblings are normally expected to take care of their younger siblings when they are small or to support them financially when necessary (Jay, 1969;Megawangi et al., 1995). Thus, children of higher birth order may stay at home to provide assistance, or even sacrifice their opportunities to pursue higher education for the sake of the education of their younger siblings. They may also move for work to support the family. We therefore expect to find that older siblings are less likely to leave the parental village and less likely to move for education than younger siblings, but more likely to move for work. Some evidence indeed shows that being the first-born has negative effects on nest-leaving in Malaysia (Johnson & DaVanzo, 1998), but there is also evidence that older siblings are more likely to migrate than younger siblings (Bratti et al., 2016, for Mexico).
The number of siblings may also affect leaving the parental home. A larger number of children may indicate that a family has greater economic needs and a greater opportunity to send a child elsewhere to pursue economic gains (Lauby & Stark, 1988), whereas the other siblings can assume the family and household responsibilities if the older child has to move out (Johnson & DaVanzo, 1998).
Furthermore, the larger the number of siblings in a family, the less parental support per child is available because the parents' resources have to be shared with other siblings (Leopold et al., 2012;Sandefur et al., 2006).
In developed countries, a greater number of siblings in a family implies there is less privacy and space in the parental home, and thus, a greater incentive to leave. Studies for developed countries have indeed found that having more siblings is related to a higher likelihood of leaving home (Blaauboer & Mulder, 2010;De Jong Gierveld et al., 1991;Mitchell et al., 1989). In the context of developing countries, privacy issues may be less important than economic issues in driving leaving the parental home. Overall, we expect that the likelihood of leaving the village increases with the number of siblings.
Following a similar line of argument as for siblings, belonging to an extended family could be a source of labour supply and intrafamilial care that makes young people reluctant to leave the village.
However, it could also mean that there are fewer resources available for each of the members, particularly if the extended family arrangement is the result of economic difficulties (Aquilino, 1991). Thus, depending on what counts more-labour supply and familial care, or resources-the effect of having an extended family on leaving the parental village could be positive or negative. For the case of the Philippines, Root and De Jong (1991) found that in extended families, the probabilities that some of their members would migrate were higher than in nuclear families.
Good intra-familial relations may prevent children from leaving the parental home early (De Jong Gierveld et al., 1991). Children from single-parent families have likely grown up in a less stable family situation. Presumably, because of this lack of stability, they have a higher likelihood of leaving home (Blaauboer & Mulder, 2010;Mitchell et al., 1989). We therefore expect to find that living with one parent is positively associated with leaving the parental village. Some studies have shown that in the context of developing countries, having divorced or separated parents indeed increases the likelihood of nest-leaving (Johnson & DaVanzo, 1998, for Malaysia).

| Other factors to account for: Individual and regional characteristics
In addition to family resources and family structure, it is important to account for individual and regional characteristics that are likely to affect the likelihood of moving from the parental village. First of all, we account for age differentiation in the likelihood of moving (Chae et al., 2016;Malamassam, 2016;Wajdi et al., 2017). We also expect to see gender differences in the likelihood of moving (Malamassam, 2016). For example, females are more likely than males to move for marriage, whereas males are more likely than females to move for work (Beegle & Poulin, 2013, for Malawi). Level of education is known to be positively related to leaving the parental home (Johnson & DaVanzo, 1998;Mitchell et al., 1989;Mulder & Clark, 2000). We also account for marital status, as being married was found to positively affect leaving home (e.g., Garasky, 2002). Because one of the reasons for moving is for marriage, we do not assess the effect of marital status on different types of moves.
We also account for ethnicity and religion. For example, the Minangkabau and the Batak had a higher number of migrations over the lifetime than the Javanese, whereas the Sundanese, the Betawi, and the Balinese had a lower number of migrations (Pardede et al., 2019).
Because norms regarding the age at marriage vary between ethnic groups (Buttenheim & Nobles, 2009), leaving the parental village for marriage may differ by ethnicity. In a similar vein, there may also be differences between religions in norms and behaviour with respect to leaving the parental village.
More developed regions and more urbanised areas usually offer better job opportunities and better educational facilities (Leopold et al., 2012), and are therefore likely to attract migration. Given the regional differences in levels of development in Indonesia, which is the highest in Java, followed by Sumatra, and then the rest of Indonesia (Firman, 1997), we expect to find that the likelihood of moving from the parental village is highest for young people living outside of Java and Sumatra, followed by young people living in Sumatra, and then in Java. We also expect to find that the likelihood of moving from the parental village is higher for young people living in rural areas than for young people living in urban areas, especially if they are moving for work or education. household members. Of the 33,081 household members, 22,019 were randomly selected for a detailed interview as IFLS main respondents (Frankenberg & Karoly, 1995). Reinterview rates were high: 82% of the main respondents in 1993 were interviewed in 2014 (Strauss et al., 2016).
The respondents who had left the original 1993 IFLS households were tracked, primarily to find these main respondents. In 1997In , 2000In , 2007In , and 2014, all household members aged 15 or older were interviewed. Due to new entrants and children born into the original IFLS households, and after some attrition, deaths, and the re-joining of household members, the number of household members had grown to around 50,000 at the latest survey (Strauss et al., 2016).
From the main respondents, detailed individual information was obtained on characteristics such as education, employment, and migration history (from age 12). The migration history module included questions about the time of the move, the reason for the move, and with whom the respondent moved.
Among these main respondents, we selected all of the children of the head or the spouse, including the adopted children, who were living in the household and were aged 14 or 15 at the time of one of the surveys conducted in 1993, 1997, 2000, or 2007. We followed these respondents from age 15 until their first migration from the parental village or censoring. We censored the observations at age 28 because there were few cases left for the analysis at that age; or if the young person died. We also censored the cases of young people who did not leave their village but were no longer living with either of their parents. In these cases, it cannot be determined whether the parents were still living in the same village. These respondents may have left home but without leaving the village, or their parents have died or moved out at the time of a particular survey. Our final sample consisted of 3898 individuals in the format of 29,180 person-years. The reason for moving was taken from the respondents' answers to the question about the main reason for migrating (only one possible answer per move). We grouped the reasons into four categories:

| Dependent variable
(1) moving for work, (2) moving for education (including training, as it was not separated in the questionnaire), (3) migration for marriage, and (4) other reasons, including migration for family (to be closer to family, to move with family, death, divorce, sickness), and other types of moves (natural disaster, transmigration [a government programme to move people mostly from Java to the other islands], housing, independence). Not moving was the reference category. If a respondent indicated that s/he had moved for work, s/he was asked for whose work the move was made. If the move was made for the work of his/her spouse or of another household member, the move was categorised as a move for other reasons instead of for work.

| Independent variables
All of the independent variables-that is, family resources, family structure, individual characteristics, and regional characteristics-were measured at the time of survey t, updated with each survey, and assumed to remain constant until the next survey. In cases in which information was missing because a respondent did not participate in a certain wave, we imputed the values of the variables at time t with their values at time t À 1.
To measure family resources, we used indicators of household income, parents' employment status, whether any household member was working in a farm business, homeownership, parents' education, and parents' health. To measure household income, we used the standardised household per capita expenditure by dividing it by the urban or rural national poverty lines (we used the term "real expenditure" to refer to this indicator). We used national poverty lines because provincial poverty lines were not available for some provinces in 2000. As no poverty lines were available for 1997, we used the 1998 national poverty lines.
For the parents' employment status, we used the father's employment as the priority because the father was usually the main breadwinner. The highest level of employment status was formal employment, followed by informal employment, and then not employed. If no information was available about the parents' levels of education or employment, we used the information of the head of the household or of his/her spouse. We used the mother's years of schooling as the priority for the parents' education if the information on both parents was available.
Parents' health status was measured using the self-reported scale from one (healthy) to four (unhealthy). Parents who reported being somewhat unhealthy (3) and unhealthy (4) were grouped as "unhealthy." Family structure was measured using the birth order of the respondent (oldest, middle, youngest or only child) and the number of siblings, only taking into account the children living in the household; whether the family was nuclear or extended, and whether the respondent was living with both parents or with one parent. If at the time of a certain wave the parents of the young adult were no longer the head of the household or the spouse, we used the previous birth order and number of siblings. For other household variables, the information about the household was used, regardless of whether the parents of the respondents were the head of the household or the spouse.
Age was grouped into three groups for Model 1 (15-17, 18-24, and 25-28) and into two groups for Model 2 (15-17 and 18-28). The youngest age is 15 because, in Indonesia, this is the usual age at which young people complete junior high school (9 years of education) and the minimum age for admission to employment (Suryahadi et al., 2005). Age 18-24 is the age group of the late adolescent phase (Lloyd & Grant, 2005) To assess the effect of location, we divided Indonesia into three regions based on the level of development (Firman, 1997): (1)

| Analytical strategy
We used two models for discrete-time event history analysis (see Yamaguchi, 1991): (1) a logit model to assess the effect of family resources and family structure on moving from the parental village; and (2) a multinomial logit model to assess the effect of family resources and family structure on moving from the parental village for work, education, marriage, or other reasons, compared with not moving. Both models were estimated with standard errors corrected for the clustering of young people who belonged to the same household at the beginning of observation (2884 households), using the Huber and White "sandwich" or robust estimator of variance (StataCorp, 2019). As we regard "other reasons" as a residual category, we do not discuss the results for it. To detect whether the effects of parental resources on migration varied by age, we included interaction terms for age group with real expenditure. The descriptive results in Figure 1 show that among moves for work, education, and marriage, moves for work were the most common. The percentage of young males moving for work was almost twice that of young females, whereas the percentage of young males moving for marriage was around half of that of young females. With regard to age group, those aged 18-24 had the highest rate of leaving the parental village, followed by the oldest age group, and then by the youngest age group, 15-17 (Table 1). The leading reason for migration among the 15-17 age group was education, whereas the leading reason for moving among the 18-28 age group was work ( Figure 1).  , 1993, , 1997, , 2000, , 2007, , 2014 The main effect of real expenditure on leaving the parental village was negative but only statistically significant at the p < 0.10 level,

| The effects of family resources
whereas the effects of this variable were more negative (b = À0.443) on moving for work but not significant on moving for education or for marriage (rather than not moving). Given that we also estimated an interaction term, the negative main effect of real expenditure on moving for work held for ages 15-17. For ages 18-28, the negative coefficient was almost offset by a positive interaction term (b = 0.398). The negative effect of income on moving for work for the young ages was in line with the expectations we derived from the Western literature in particular (Blaauboer & Mulder, 2010;Whittington & Peters, 1996).
Only a few of the effects of other family resources on the likelihood of leaving the parental village were statistically significant. We found weak evidence (p < 0.10) of a negative effect of the parents' employment in the formal sector on moving for marriage and a positive effect of parents not working on moving for work. Neither of these results confirmed our expectation of positive effects of better employment status on leaving the parental village. At the same time, the effect of parents not working on moving for education was negative and highly significant, which confirmed our expectation.
The involvement of any of the household members in a farm business was not significantly related to leaving the parental village, whereas it was positively related to moving for work or for education (at the 0.10 level). However, it was not found to significantly affect the likelihood of moving for marriage. This positive effect on moving for work confirmed our expectation that families who relied on agricultural activities would be more likely to send their children to work elsewhere than families who were not involved in agricultural activities. Our finding that engagement in farm work was positively associated with moving for education-which was the opposite of what we expected-might be related to the fact that farms tend to be located farther away from higher level educational facilities. No substantial effects were found of whether the home was self-owned or rented on the likelihood of leaving the parental village or of moving for a specific reason.
Parents' education was found to affect the likelihood of leaving the parental village and of leaving the parental village for education.
This was in line with our expectations. Finally, there was no evidence to support rejecting the null hypothesis that the effect of parents' education on moving for work or for marriage was different from zero.
Compared with the young people whose parents were healthy, the young people whose parents were unhealthy had a lower likelihood of moving for work rather than of not moving. Although the effect was only statistically significant at the 0.10 level and was only found for work-related moves, this result confirms our hypothesis (see also Rammohan & Magnani, 2012).

| The effects of family structure
Contrary to our expectation, those who were the oldest child in the household were more likely to leave the parental village than the youngest children. On the other hand, being the oldest child did not significantly affect leaving the parental village for education. Instead, those who were the middle child were less likely to leave the parental village for education than the youngest children (p < 0.10). These results confirmed our expectation that being an older child was negatively related to moving for education, but not necessary for the oldest child. These findings could indicate that the responsibilities of oldest children tend to be expressed by moving away rather than providing in-home care and that families tend to prioritise sending the youngest or the oldest child away for education. No support was found for our expectation of positive effects of the number of siblings on leaving the parental village.
Living in an extended family did not significantly affect the likelihood of moving from the parental village, whereas it was negatively related to moving for work. These results are more in line with our hypothesis derived from the idea that an extended family could represent a source of supply of labour or intrafamilial care than our alternative hypothesis derived from the idea that an extended family would offer fewer resources. The results for the variable of living with one versus with both parents were not significant for moving in general.
As the number of young adults living in a one-parent family was too F I G U R E 1 Distribution of reasons for migration by gender and age group. Source: Author's calculation, 1993, 1997, 2000, 2007, 2014IFLS (observations = 1617 T A B L E 3 Logistic regression estimates of moving from the parental village (Model 1) and multinomial logistic regression results of moving for a particular reason vs. not moving (Model 2), standard errors in parentheses The results also show evidence of a gendered moving pattern. For example, young females were less likely to move for work than young males, but were more likely to move for marriage.
Young people's level of education was positively related to moving, moving for work, and moving for education, but education of primary school level was negatively related to moving for marriage  , 1993, , 1997, , 2000, , 2007, , 2014 (p < 0.10). Young people who were married were less likely to move from the parental village than those who were not married, which was contrary to our expectations and to findings on patterns of leaving the parental home in the Western world (Garasky, 2002). An explanation for this finding could be that it is still normatively acceptable for young couples to reside in the parental home after marriage in Indonesia (Buttenheim & Nobles, 2009).
The Sundanese, Bantenese, Betawi, and Cirebonese were less likely to leave the parental village than the Javanese and Madurese.
The Batak, Nias, and Minangkabau were more likely to move for education and less likely to move for marriage compared with the Javanese and Madurese. Those who belonged to Christianity were more likely to leave the parental village than those who belonged to Islam.
Those who belonged to Christianity or "Other" religions were also more likely to move for work than those who belonged to Islam.
Young adults living outside Java were more likely to move than those living in Java. Those living outside Java and Sumatra were more likely to move for education than those living in Java, likely because of a lack of educational facilities in these areas, but this finding was only significant at the 0.10 level. Interestingly, we found that the young adults who were living in Sumatra or the rest of the provinces were more likely to move for marriage than those who were living in Java. It is possible that this finding is related to the lower population density in these areas compared with Java, which could mean that the young adults in these areas needed to go outside of their parental village to find a spouse. As we expected, living in a rural area was positively related to moving and to moving for work and for education, but no relation was found for moving for marriage. in 1998, and the onset of fiscal decentralisation and regional autonomy in 1999 (Firman, 2004).

| CONCLUSION AND DISCUSSION
In this paper, we assessed to what extent family resources and family structure affected the likelihood of leaving the parental home and village in Indonesia; in general, and for three reasons: work, education, and marriage. Whereas leaving home has been investigated in Western countries numerous times, our study is among the first to address this topic in the context of Indonesia. The novelty of our findings is indeed connected with this different context of a developing country.
We found evidence that some of the family resources and some of the indicators of family structure affected the likelihood of leaving the parental village, but not always in the expected ways. We also found that family resources and family structure affected leaving the parental village for the three different reasons in different ways, which supports the notion that leaving the parental village for different reasons involves different processes.
Our findings resonate with De Jong Gierveld et al.'s (1991) argument-and findings-for the Western world that material family resources affect leaving the parental home in a different way than nonmaterial family resources. These influences would also differ between different motivations for leaving home (De Jong Gierveld et al., 1991) and between younger ages-at which parents tend to prefer to keep the child at home-and older ages-at which parents tend to facilitate leaving home (Avery et al., 1992). As an indicator of material resources, family income negatively affected leaving the parental village, and for the younger ages, it also negatively affected leaving for work. Having nonworking parents negatively affected moving for education. However, the negative effect on moving for marriage of parents working in the formal sector compared with working in the informal sector may indicate that working in the formal sector may also reflect nonmaterial resources: it could play a role similar to parental education (which was expected to affect moving for marriage negatively). Those working in the formal sector tend to have a higher level of education than those working in the informal sector. The distinction between the formal and the informal sector has not been addressed in the literature on leaving home in Western countries.
Parental education-indicating nonmaterial resources-positively affected leaving the parental village in general and moving for education.
A finding regarding resources that is typical of developing countries was the positive effect of involvement in a farm business on moving for work. Different from our expectation, this variable also affected moving for education positively. Possibly, the notion that the movement of the members of families working in agriculture tends to be driven by the desire to smoothen family income (Stark & Levhari, 1991) needs to be extended to include the idea of moving for social mobility as well. Families involved in agriculture work may aspire to send their children to pursue higher education to increase their opportunity to get higher level occupations. However, it may also just be that such families tend to live further away from schools offering higher education.
Our findings also underlined the role of family structure in leaving the parental village. In line with the idea that birth order is important for leaving home in developing countries (Bratti et al., 2016;Johnson & DaVanzo, 1998), we found an effect of being the oldest in the household on leaving the parental village. However, in contrast with our hypothesis, this effect was positive. We derived our hypothesis from the idea that oldest children would tend to remain in the parental home to look after siblings, but speculated that the oldest children's familial responsibility could also lead to providing for siblings from elsewhere. We also found that living in an extended family was negatively related to moving for work. This could be because the extended-family members supported the family and thus lowered the chance of the children to move for work. In studies on leaving home in the Western countries, birth order and living in an extended family have been largely ignored. The roles of these factors may be worth investigating in these countries, for example, among immigrant families coming from developing countries. Such an investigation could also be relevant for the contexts of Southern European countries, as the role of the family on leaving the parental home is more decisive in these countries (Mandic, 2008 interesting line of enquiry to assess whether people who moved for work at such a young age turn out to be economically worse-off than those who moved for education or who stayed with their parents for a longer time. Likewise, it would be interesting to investigate the effects of leaving the parental village on the well-being of these children and how these effects differ by, for example, age and birth order. Several limitations of our data are worth mentioning. First, our data pertain to the 13 IFLS provinces in the western and middle parts of Indonesia that have higher levels of (economic) development than the rest. We could have obtained somewhat different results if we had included young people living in the middle and eastern parts and in the more remote, rural areas of Indonesia. Second, it would be useful to have data on leaving the parental home rather than on leaving the parental village. As "modernisation" continues in Indonesia, the role of the village as a community and as a source of social support could also decline. If such a shift occurs, leaving the parental village would not accurately reflect the chances that young people had become detached from parental support when they moved.
Lastly, we have shown a study on leaving the parental village in Indonesia where modernisation coexists with adherence to traditional values (Nilan, 2008). In this sense, we contribute to the youth mobility studies in Asian and developing countries (e.g., Ting & Chiu, 2002).
Our study is also useful for the debate on whether the transition to adulthood in the developing countries will converge to what has occurred in with the developed Western world, within which leaving the parental home is a part, studies regarding leaving home need to pay attention to the different process that may occur for those who diverge from the general patterns.