Contributions of family and neighbourhood factors to the mental health of migrant children in China: implications for policy and services

Abstract

Drawing upon a sample of 806 migrant children in Beijing, China, this study investigated how community social capital, family social support, family human capital and family financial capital influenced the mental health of Chinese migrant children. Using structural equation modelling with latent variables, the study results suggested that higher levels of family social support were associated with better mental health of migrant children. Family social support also mediated the effect of community social capital on children's mental health. Family human capital and family financial capital both contributed to children's mental health through their significant effects on community social capital. However, family financial capital was negatively associated with family social support, resulting in an unexpected, indirect negative effect on children's mental health. Implications of these findings for social policy and social services were discussed.

Keywords:

• community social capital
• family social support
• family human capital
• family financial capital
• mental health
• migrant children

Introduction

Children of migrant workers are a fast growing population in mainland China as a consequence of the huge population flow from rural to urban areas in the past three decades. The population growth has been especially drastic in recent years as the pattern of migration switches from personal migration to family resettlement, and from temporary inhabitation to permanent stay in the city (Li, 2006; Liang, 2007; Roberts, 2002). According to the 2000 census, there were 121,070,000 migrants in China (National Bureau of Statistics, 2000), 19.37% of whom were children under age 18. The number of migrants increased to 211 million by 2009, 20.8% of whom were children under age 14 (National Committee on Family Planning, 2010). In the city of Beijing alone, there were 477,000 migrant children by 2009, accounting for 23.83% of the city's entire child population (Duan & Yang, 2009).

However, under China's household registration system (hukou), which has been designed to control rural–urban mobility for economic and political purposes, migrant children are not granted equal access to a city's education, medical care and other social services, due to the lack of urban residence. Large numbers of migrant children are thus denied entry into public schools in the city and are forced to enrol in so-called ‘migrant children's schools’, which are of significantly lower quality. Recent policies have sought to address this problem and required public schools to accommodate migrant children. Nevertheless, the delay in policy implementation has perpetuated a harsh situation. A considerable proportion of migrant children still have to rely on the low quality private schools as their only choice for receiving education, given the institutional, geographical and financial difficulties of entering public schools (Xie, Zou, & Li, 2007).

The restrictions in the educational system as well as other structural barriers to easy adaptation to the new urban environment pose many threats to the healthy development of migrant children in the city, including their mental health. An increasing number of studies have found that migrant children did experience more mental health problems than their urban counterparts. For instance, Li, Zou, Wang, and Dou's (2008) study in Beijing found that the self-esteem of migrant children were significantly lower than that of their non-migrant counterparts, and levels of self-esteem of migrant children attending migrant children's schools were significantly lower than that of migrant children attending public schools. Hu, Fang, Lin, and Liu (2009) found the level of social anxiety and loneliness to be significantly higher among migrant children compared to native urban children in Beijing and rural children from five provinces where a majority of migrants in Beijing came from. Using SCL-90, Yang, Wang, Feng, and Tang (2007) suggested that migrant children suffered from obsessive-compulsive disorders, interpersonal sensitivity, anxiety and hostility. Migrant children were also found to exhibit more internalising behavioural problems (Li, Zou, Jin, & Ke, 2008) and more feelings of self-humiliation (Xu & Deng, 2010) than their urban counterparts. These are consistent with what documented in the Western literature that immigrant children are at higher risk for developing mental health problems either as a direct result of immigration stress or indirectly through parental distress and disrupted family context (Short & Johnston, 1997). They face a number of challenges that include adapting to a new school and social environment, segregation, prejudice, marginalisation and changes in family ecologies (Baptiste, 1993; Chan, Ip, & Yuen, 1996; Chiu, Feldman, & Rosenthal, 1992). In an overview of existing research on immigrant children, problems related to their mental health include lower self-esteem, more depressive symptoms and anxiety disorders, higher levels of stress, hostility and aggression, and deviant behaviours (Aronowitz, 1984; Bankston & Zhou, 2002; Baptiste, 1993; Jaycox et al., 2002; Padilla & Duran, 1995; Rumbaut, 1994).

Existing studies have also identified several protective factors in a child's immediate social contexts that contribute to the mental health of immigrant children, most notably in the family and neighbourhood contexts. At the neighbourhood level, social capital has been an increasingly popular concept in explaining the relationships between neighbourhood resources and youth development. Usually referred to as community social capital, social capital embedded in the neighbourhood refers to the social connectedness among resident adults and children which provides a base of potential resources that people could draw upon for their own benefits and for the good of the neighbourhood as a whole (Coleman, 1988; Putnam, 1993). It encompasses norms, trust, sense of belonging to the neighbourhood and civic engagement, all of which facilitate accumulation and transmission of resources as well as collective actions for public good (Putnam, 2000). Empirical studies suggest that community social capital establishes an extra-familiar social network that provides social control and monitoring functions for children and youths living in the neighbourhood (Brooks-Gunn, Duncan, Klebanov, & Sealand, 1993; Coleman, 1990; Sampson, 1997; Simons, Simons, Conger, & Brody, 2004) as well as increases the family's access to additional social resources through interactions at the community level (Furstenberg & Hughes, 1995; Teachman, Paasch, & Carver, 1996). For example, Stevenson (1998) found the level of depressive symptoms of low-income, inner-city African-American adolescents to be lower when they lived in neighbourhoods with higher social capital. Runyan and colleagues' (1998) study with a sample of high-risk preschool children suggested that community social capital had an impact on child well-being as early as the preschool years, especially for children from families with limited financial and educational resources. Drukker, Kaplan, Feron, and van Os's (2003) study in The Netherlands also demonstrated a significant link between community social capital and child mental health.

In the family sphere, social support from family members was consistently reported to exert positive effects on the mental health of adults and children, especially in families with immigrant backgrounds. Studies have suggested that family processes would have great impact on children growing up under economically or socially adverse conditions, possibly because supportive family relationships might be able to buffer the effects of environmental stressors on children in the family (Leidy, Guerro, & Toro, 2010; Masten et al., 1988). Numerous studies have indicated that parental support, or support from families generally, was consistently associated with healthier development of youth, such as better school adjustment (Dubow, Tisak, Causey, Hryshko, & Reid, 1991), higher self-esteem (Hoffman, Ushpiz, & Levy-Shiff, 1988) and lower depression (Colarossi & Eccles, 2003; Newman, Newman, Griffen, O'Connor, & Spas, 2007). For example, based on analysis of longitudinal data, Zimmerman, Ramirez-Valles, Zapert, and Maton's (2000) study on a sample of urban African-American adolescents suggested that parental support would help insulate these youths from anxiety and depression. Cheng's (1997) study on secondary school Chinese adolescents in Hong Kong found perceived social support from the family to be predictive of lower depression level, especially while under higher stress. Based on their longitudinal data collected in a large suburban middle school, Rueger, Malecki, and Demaray (2010) also reported that parental support from the family was a robust unique predictor of adjustment of both boys and girls in their early adolescence, including depressive symptoms, anxiety, self-esteem and academic performance. Some previous research on migrant children in China also indicated the protective role that perceived social support played in promoting the psychological well-being, fostering self-efficacy and reducing problem behaviours among migrant children (Liu, et al., 2010; Xie, Hou, & Xu, 2009; Zeng, 2011).

Two other family level variables, family human capital and family financial capital, were also documented to influence the developmental outcomes of children and youth through both direct and indirect pathways. According to Coleman (1988), family human capital refers to the cognitive skills of parents, usually measured as the parents' educational attainments, which are expected to influence the cognitive environment within a home. Children from families with higher human capital tended to achieve better in school, obtain successful employment and be less likely to present internalising and externalising behavioural problems (Coleman & Hoffer, 1987; Furstenberg & Hughes, 1995; Putnam, 2000; Runyan et al., 1998; Teachman et al., 1996, 1997). Family financial capital, reflecting the economic conditions and available financial resources of the household (Coleman, 1988), also demonstrated significant effects on various social and emotional outcomes of children and youth (Coleman & Hoffer, 1987; Furstenberg & Hughes, 1995; Putnam, 2000; Runyan et al., 1998; Sampson, Morenoff, & Earls, 1999; Teachman et al., 1996, 1997).

In addition to the direct effects of these family and neighbourhood factors, previous studies have also implied potential indirect pathways among these contextual factors and youth development outcomes. As Coleman (1988) suggested, parents' educational and financial advantages cannot be transmitted to children automatically without enough parent–child interactions. That is, family social support, or parental support for children particularly, may serve as a mediator to realise the effects of family human and financial capital on child development. Moreover, research provided evidence that parents with more educational and financial resources were able to mobilise greater social capital in the neighbourhood, possibly because the family socio-economic status reflected by education and income would partly determine where the families live and how much they could invest in the place where they live, thus influencing the stock of social capital in the neighbourhood (Coleman & Hoffer, 1987; Furstenberg & Hughes, 1995; Putnam, 2000; Teachman et al., 1996, 1997; Wu et al., 2010). This suggested the possible mediator role of community social capital between family human and financial capital and the youth outcome. Importantly, neighbourhoods also provide a social context that outlined the resources of families and moulded the experiences of parents and children (Hughes, Furstenberg, & McDonald, 1998). Community social capital was thought to bolster the efforts of parents on becoming more attendant to their children and more caring for their adjustment (Dorsey & Forehand, 2003). It was documented that youth living in a poor neighbourhood with insufficient resources tended to receive less social support and supervision from parents than those who lived in safer and more resourceful neighbourhood (Klebanov, Brooks-Gunn, & Duncan, 1994). This indicated a potentially indirect pathway from community social capital to youth development through family social support. These hypothetical mediating effects of family and neighbourhood factors were all incorporated into the conceptual framework of the present study for a thorough examination.

Despite the burgeoning Chinese literature that documented the mental health difficulties of migrant children in the urban cities, as well as the well-established evidence in the Western literature that family and neighbourhood factors are influential for the mental health of immigrant children, to date it has rarely been examined how the family and neighbourhood factors contribute to the mental health of migrant children in the context of internal rural–urban migration in mainland China. The mechanism by which the family and neighbourhood factors work together on the child's mental health has yet to be revealed as well. Drawing upon a sample of 806 secondary school migrant students in Beijing, China, the present study was aimed to investigate how neighbourhood social capital, family social support, family human capital, and family financial capital work together to influence the mental health of migrant children in mainland China. Findings of this study will shed light on the mechanism through which familial and contextual factors work independently and in concert to exert synergistic effects on youth development. Most importantly, the study results will inform the development of social service programmes and social policy to better address the challenges in migrant children's adaptation to the urban environment and to provide adequate mental health services and interventions for this vulnerable population. The conceptual framework of this study is presented in Figure 1.

Figure 1 Hypothesised structural model of the effects of family and neighbourhood factors on child mental health.

Method

Participants and procedure

Participants of the study came from two privately run migrant children schools and one state-run public school in Beijing. Due to the sensitivity of the study issue and the consequent difficulty in gaining access to schools for research, selection of the three schools did not follow a random sampling procedure, but was more dependent on the connections of the investigators, which led to a relatively convenient sample for the study. However, the three schools are representative of the typical schools that the majority of migrant children go to in Beijing in terms of the quality of school, the composition of student body and the socio-economic status of students. The one public school was at a lower ranking amid the city's public education system and had over 30% migrant children in its student population (higher ranking schools usually do not admit migrant children). The two migrant children schools were privately run with much less resources and with teachers of limited qualifications. According to a recent report, the enrolment rate of migrant children in public schools was only 61% in Beijing (Xie et al., 2007), suggesting that a large number of migrant children still receive education in migrant children schools.

A total of 806 students at seventh to ninth grades from the three schools were recruited for the study and completed a paper-and-pencil survey in their classrooms during a single class period (45–50 minutes) under the guidance of trained research assistants. Prior to the administration of surveys, students were provided with consent forms so that they were fully informed that participation in the study was completely voluntary and they had the right to refuse participating and to withdraw from the study at any time. The surveys were identified only by a code number; student names or any other identifying information were not collected. Full descriptive statistics of the sample are presented in Table 1.

Table 1 Descriptive statistics of sample characteristics (N = 806).

Characteristic	Frequency	%
Gender
Male	450	56.4
Female	348	43.6
Education (father)
Didn't receive formal education	8	1.0
Elementary school	150	19.3
Secondary school	384	49.3
High school	167	21.4
Professional high school	32	4.1
Diploma or higher	38	4.9
Education (mother)
Didn't receive formal education	50	6.3
Elementary school	274	34.6
Secondary school	320	40.4
High school	104	13.1
Professional high school	22	2.8
Diploma or higher	22	2.8
Employment status (father)
Full-time	622	82.5
Part-time	74	9.8
Unemployed/laid off	41	5.4
Retired	17	2.3
Employment status (mother)
Full-time	467	60.1
Part-time	72	9.3
Unemployed/laid off	45	5.8
Retired	10	1.3
Housewife	183	23.6
Monthly household income
0–500	18	3.2
501–1000	111	19.7
1001–1500	114	20.2
1501–2000	98	17.4
2001–3000	92	16.3
3001–4000	69	12.2
4001–8000	38	6.7
8000 above	24	4.3
	Mean	SD
Age (years)	14.46	1.22
Length of residence in the city (years)	6.30	4.21

Measures

Mental health was measured by the 12-item General Health Questionnaire (GHQ-12), a widely used instrument that assessed both the positive and negative aspects of mental health and had been applied in many countries (Goldberg, 1978; Jackson, 2007). It consists of 12 items, each assessing the severity of a mental health problem and responded on a four-point scale ranging from 0 to 3. The total composite score of GHQ ranges from 0 to 36, with higher scores indicating better mental health (Goldberg & Williams, 1988). The Chinese version of GHQ-12 has also been validated in previous research (Chan & Chan, 1983; Pan & Goldberg, 1990).

Community social capital was assessed by four indicators that captured the degree of cohesion and caring for children and youth in the neighbourhood, which implied the social resources embedded in the neighbourhood for their development. The four indicators included: (1) the number of parents' friends in the neighbourhood; (2) frequency of visiting friends' home or hosting friends at home, which was responded to on a five-point Likert scale from 1 = ‘never’ to 5 = ‘very often’; (3) how many activities are there in the neighbourhood for children and youths, which was responded to on a four-point scale from 1 = ‘never’ to 4 = ‘often’; and (4) number of friends of children themselves living in the neighbourhood, which ranged from 1 = ‘none’ to ‘4 = ‘many’. The latent construct of community social capital was thus formed by these four observed indicators reflecting resources in the neighbourhood.

Family social support was assessed by the family support subscale of the Multidimensional Scale of Perceived Social Support (MSPSS; Zimet, Dahlem, Zimet, & Farley, 1988), which has been used widely across cultures and a variety of populations, including Chinese adolescents (Chou, 2000). The subscale included four items asking the respondents whether their family members are willing to help them, can provide them with emotional support, can be talked with about their problems and are willing to help make decisions together with them. Each item was responded on a seven-point scale, ranging from 1 = ‘very strongly disagree’ to 7 = ‘very strongly agree’. These four observed indicators thus formed the latent construct of family social support.

Family human capital was assessed by parents’ educational attainments, ranging from 1 = ‘never received formal education’ to 6 = ‘tertiary education or above’; the two indicators of father and mother's educational attainments formed the latent construct of family human capital. Family financial capital was assessed by four indicators: monthly household income, number of rooms per person in the household, monthly rent of the apartment and type of floor of the apartment. Given the possible inaccuracy of children's reports on family income and rent, the number of rooms per capita and the type of floor (ranging from 1 = ‘soil’ to 6 = ‘wood’) were used as supplementary, and probably more accurate, indicators that reflected the economic condition of the family. The four observed indicators formed the latent construct of family financial capital in analysis.

Other socio-demographic variables included in the test of the hypothesised model included gender (1 = female, 0 = male), age (in years) and length of residence in the city (number of years that the child have lived in the city).

Data analysis

Structural equation modelling (SEM) was conducted using MPlus 5.0 (Muthén & Muthén, 2007) to test the hypothesised model. SEM permits the use of latent constructs and allows for estimating the relationships among latent constructs while providing explicit estimates of measurement errors (Byrne, 2001). Moreover, it permits the simultaneous estimation of direct and indirect paths, estimating each path after accounting for the effects of all other paths. These features render SEM an especially well-suited technique for this study, given the hypothesised direct and mediating effects of latent variables on the educational outcomes of migrant children. The study followed the two-step approach to SEM (Kline, 2005): testing the measurement model at the first step to establish a statistically reliable measure for each latent construct, and then testing the structural model to examine the multivariate relationships among latent constructs.

Multiple indices were used to assess the model fit while testing both the measurement and structural models, including: (1) the likelihood ratio test statistic (χ²), where an associated probability value showing non-significant χ² represents a closer fit of the hypothesised model to the perfect fit; (2) the Comparative Fit Index (CFI), where values above 0.90 denote a good model fit; and (3) the Root Mean Square Error of Approximation (RMSEA), where values less than .05 indicate a ‘close fit’ (Kline, 2005).

Results

Test of measurement model

The measurement model of the four latent constructs (i.e., community social capital, family social support, family human capital and family financial capital) was evaluated first via confirmatory factor analysis before testing the hypothesised structural model. The model provided a good fit to the data (χ² = 85.433, df = 63, p < .05, CFI = 0.984, RMSEA = .023) with CFI greater than 0.90 and RMSEA smaller than .05. All observed variables were significantly loaded on the corresponding latent constructs in the expected directions, suggesting that the selected indicators reasonably represented the underlying constructs in a statistically reliable manner. The standardised factor loadings of all indicators on each latent construct are presented in Table 2.

Table 2 Standardised factor loadings of observed indicators on the latent constructs.

Construct	Indicator	Factor loading
Community social capital	Number of friends of parents in the neighbourhood	.301
	Frequency of visiting friends' home or hosting friends at home	.390
	Activities for children and youths in the neighbourhood	.470
	Number of friends of children themselves in the neighbourhood	.323
Family social support	Family members willing to help	.590
	Family members can provide emotional support	.638
	Talking with family members about your problems	.349
	Family members willing to make decisions together with you	.681
Family human capital	Father education	.703
	Mother education	.743
Family financial capital	Monthly household income	.730
	Number of rooms per person in the household	.426
	Monthly rent	.560
	Type of floor	.467

Test of structural model

Test of the hypothesised structural model yielded a good fit to the data. Although the chi square value is large and significant (χ² = 135.202, df = 104, p < .05), due to its sensitivity to large sample size, other goodness-of-fit indices demonstrated satisfying results, with CFI (0.978) greater than 0.90 and RMSEA (.021) lower than .05. The standardised solution for the test of the structural model is presented in Figure 2.

Figure 2 Standardised solutions for the effects of family and neighbourhood factors on child mental health. *p < .05, ***p < .001.

As hypothesised, the protective factors in the family and community showed significant positive effects on the mental health of migrant children. Mental health appeared to be better when there were higher levels of family social support (β = 0.200, p < .05). Community social capital did not present a significant direct effect on children's mental health (β = − 0.004, p>.05), but higher levels of community social capital was associated with stronger family social support (β = 0.777, p < .001), which, in turn, predicted better mental health. That is, community social capital promoted the mental health of migrant children through significantly boosting the level of family social support.

The two other family level factors, family human capital and family financial capital, showed somewhat complex effects on the mental health of migrant children. Both family human capital (β = 0.016, p>.05) and family financial capital (β = − 0.049, p>.05) appeared not to be related to children's mental health in a direct way. However, higher levels of family human capital was associated with higher levels of community social capital (β = 0.231, p < .05), which fostered stronger family social support as reported earlier, thus leading to better mental health of children indirectly. Similarly, family financial capital was strongly associated with community social capital (β = 0.342, p < .001), and therefore predicted children's mental health through the same indirect path. What worth special attention here was that, in addition to its positive association with community social capital, family financial capital was also significantly associated with family social support but in a negative direction (β = − 0.243, p < .05). This result suggested that family financial capital might have a potential negative effect on the mental health of migrant children through decreasing the level of family social support.

Of the socio-demographic variables, gender had a significant direct effect on children's mental health (β = − 0.079, p < .05); that is, girls were more likely than boys to present poorer mental health. Age was not related to children's mental health (β = − 0.072, p>.05). With regard to the effect of length of residence, it did not affect children's mental health directly (β = 0.014, p>.05). However, longer residence in the city was associated with higher community social capital (β = 0.215, p < .01), but weaker family social support (β = − 0.155, p < .05), which exhibited contrasting indirect effects on the mental health of migrant children. The standardised direct, indirect and total effects of major predictor variables on children's mental health are presented in Table 3.

Table 3 Standardised direct, indirect and total effects of major predictor variables on children's mental health.

	Mental health
Major predictor variables	Direct	Indirect	Total
Community social capital	− .004	.155*	.151*
Family social support	.200*	—	.200*
Family human capital	.016	− .018	− .002
Family financial capital	− .049	− .050*	− .099*
Gender	− .079*	.011	− .068*
Age	− .072	.002	− .070
Length of residence	.014	− .032	− .018
Note: *p < .05.

Discussion

This study draws upon a sample of migrant children in Beijing to investigate how family and neighbourhood factors contribute to the mental health of this vulnerable child population in the context of internal migration in mainland China. It is in line with the paradigm shift in youth studies which switches from a deficit-based perspective to a strength-based perspective, focusing more on the internal and external assets of children and youth that may lead to desirable developmental outcomes (Scales, Benson, Leffert, & Blyth, 2000). Results of the study revealed the mechanism by which community social capital, family social support, family human capital and family financial capital work together to influence children's mental health. Major findings of this research are summarised later, followed by further elaboration and discussion of their implications for practice and policy.

First, it was not surprising to find that stronger family social support is significantly associated with better mental health of migrant children. This echoes what has been consistently documented in the literature that perceived social support from family members can serve a protective function effectively that buffers the negative effect of adverse economic and social conditions (Leidy et al., 2010; Rueger et al., 2010; Zimmerman et al., 2000). According to Bronfenbrenner's (1986) ecological theory, family is viewed as a critical context for youth development. Positive family relationships, together with positive communication and family standards, would create a favourable environment for successful youth adjustment (Steinberg, 2001). Identification of the positive effect of family social support on children's mental health in the present study is consistent with these commonly accepted contentions. Particularly for migrant children in China, given their marginal status in the urban community, family is likely to be the most important, if not the only, source of social support when they have any instrumental or emotional need. Therefore, among all the family and neighbourhood factors tested in this model, family social support is the only one that significantly contributes to children's mental health in a direct manner. The importance of maintaining adequate level of social support from the family is also emphasised in previous research on Chinese migrant children (Yang, Li, & Peng, 2009).

In addition to its significant direct effect, family social support also mediates the effect of community social capital on the mental health of migrant children. Although there is no direct association between community social capital and child mental health, higher levels of community social capital are associated with stronger family social support, which, in turn, predicts better mental health status of migrant children. As suggested by previous research, family processes and management are contingent on the larger neighbourhood context (Beyers, Bates, Pettit, & Dodge, 2003; Wu, Palinkas, & He, 2011). Family characteristics may be the intermediate that link neighbourhood context to child internalising and externalising behaviours (Roosa et al., 2005; Wu et al., 2010). It has been found in previous research that youth living in poor neighbourhoods with a higher crime rate tend to receive less social support and supervision from parents than those who live in safer and more resourceful neighbourhoods (Klebanov et al., 1994). Findings of the present research support these contentions by showing that, when there are stronger social bonds, clearer social norms and higher levels of trust among people living in the neighbourhood, each member of the community also tends to provide more social support to their own family members in order to be congruent with the community atmosphere. Consequently, the increased level of social support will benefit the development of children, especially those who have access to limited resources in other social contexts such as children from migrant families. In the urban Chinese communities, migrant children may not have sufficient direct interactions with neighbourhood adults and children due to their peripheral status, but migrant families are inevitably influenced by the resources and overall atmosphere of the neighbourhood. Therefore, the community level resources would go through family processes to shape children's mental health, rather than functioning on children directly.

The two other family factors, family human capital and family financial capital, both influence children's mental health through their significant effects on community social capital. Higher levels of family human capital and financial capital are associated with higher levels of community social capital, which potentially predict the mental health of migrant children thorough the indirect pathway elaborated earlier. This is conceivable, because migrant parents with higher educational attainment and better financial situation may have more consciousness and flexibility in choosing the neighbourhood where they want to reside. Their decision on the place of residence actually determines how much social capital they and their children could have access to in the community. It is through this mechanism that the mental health of migrant children is influenced by the educational and financial resources of their families. The significant effects of family human and financial capital mediated by community and family factors have also been documented in earlier research on youth academic achievement, successful employment and deviant behaviours (Davis-Kean, 2005; Furstenberg & Hughes, 1995; Runyan et al., 1998; Teachman et al., 1996, 1997).

Last, a most unexpected finding of this research that is worth interpretation is the negative effect of family financial capital on the level of family social support. Despite its positive effect on community social capital, which potentially leads to children's better mental health, family financial capital is also associated with lower levels of family social support, resulting in a negative effect on mental health. In other words, migrant children from wealthier families tend to perceive less social support from parents and other family members, and the decreased family social support would degrade their mental health status. This is contrary to many research findings in the literature that typically delineate family wealth as a protective factor for child development (Brooks-Gunn & Duncan, 1997; Davis-Kean, 2005; McLeod & Shanahan, 1993; McLoyd, 1998). However, this surprising result is consistent with some earlier research on migrant children in China, which also demonstrates a negative effect of family financial capital on children's psychosocial adjustment through its association with poorer parent–child interactions (Wu et al., 2011). Therefore, the undesirable effect of family financial capital might portraits a dilemma that is common among migrant families in China: Families with higher financial capital are usually those with both parents working outside and working for longer hours; as a result, the time they could spend with children and provide immediate support to them become very limited. This explains why family financial capital is negatively associated with the level of family social support. What this dilemma leads to is the ironic circumstance that adults in the family make great endeavours for a better financial situation in order to facilitate the healthy development of children, but at the cost of spending quality time with and providing needed support to children, which unfortunately results in children's development problems.

In summary, the present study identified the significant positive effect of family social support and the indirect effect of community social capital on the mental health of migrant children in Beijing, China. It also reveals the mechanism through which family human capital and financial capital influence children's mental health via various complex pathways. However, findings of this research should be interpreted with caution because of several limitations of the research design. First, due to the sensitivity of the study issue, participants of the study were not selected through a random sampling procedure. Migrant children in the study sample came from three schools that are willing to participate in the research and allow access of researchers. Therefore the generalisability of the research findings is constrained by the non-probability sampling strategy. Second, with a cross-sectional design, it is hard for the present study to ensure the direction of causal relationships among the major variables tested in the model. For example, there exists the possibility that children with better mental health status tend to perceive more positively of social support from family members and be more active mobilising resources in their community. Therefore, longitudinal data are needed in future research to further examine whether children's mental health is a function of family and community factors, or vice versa. Third, this study only examines the family and neighbourhood factors that influence children's mental health. But school constitutes another important social context where children and youth spend a great deal of their time in the day, thus bearing strong potential to have an impact on their mental health as well. It will help reach a more comprehensive understanding of the social determinants of children's mental health in future research to incorporate school factors into the investigation.

Implications for policy and practice

Despite the previously discussed limitations, findings of this research have important implications for policy and social services targeting on migrant children in China. With the continuous rapid growth of its population size, migrant children have gradually become a major component of the child population in urban China. This being considered, the growth and development of migrant children would in a large part determine the quality and competitiveness of the nation's future generation. Maintaining good mental health of migrant children and promoting their well-being are by no means ignorable by policymakers and should be one focus of social policy and social services, if they are to contribute to the well-being of the entire society. Given the research finding that community social capital plays a significant role in predicting the mental health of migrant children, policymakers may consider taking proactive measures at the structural level to facilitate the accumulation and mobilisation of social capital in the neighbourhood of migrant families. Feasible strategies include building more public facilities and providing safe public spaces in the migrant communities, which create more opportunities for people living in the neighbourhood to interact with each other in natural means. Increasing meaningful interactions among neighbourhood adults and children will not only strengthen their informal social networks and the level of trust among one another, but also fortify their sense of belonging to the neighbourhood, both facilitating the construction and accumulation of social capital in the community. Government could also allocate more resources or encourage NGOs to establish community centres in the migrant neighbourhood, which will serve as a platform for neighbourhood adults and children to engage in collective activities and discuss issues about the community together. Participation in social activities organised by the community centres will naturally help forge stronger ties among neighbours and give each member a role to play in the community, thus stimulating their will and effort to contribute to the community. This would constitute a solid base for the growth of community social capital. As Putnam (2000) suggests, neighbourhood organisations with recreational facilities and other programmes may serve both a bonding function that brings together individuals with common backgrounds or interests and a bridging function that ensures the sharing of limited resources and information. Particularly in the urban Chinese neighbourhood, community centres may also facilitate the integration of migrant families into the host community through connecting them to non-migrant neighbours, which generally creates a more tolerant and friendly atmosphere for migrants in the neighbourhood that is desirable for the emotional and social adjustment of migrant children as well.

With regard to social services, given the significant finding that stronger family social support is associated with better mental health of children, services provided to families are expected to generate desirable outcomes for migrant children. Family education workshops, especially through the schools, may be offered to inform participants about the importance of social support among family members and the strategies to strengthen social support in the family system. For families in which adults must work for longer hours in order to maintain the financial stability of the household, these education workshops would focus on teaching parents how to enhance the quality of the limited time they could spend with children and other family members, thus strengthening the perception of social support among themselves, even though no tangible support is provided. Importantly, given the unexpected finding that family financial capital exhibits a negative effect on family social support, educational programmes must be designed for parents to realise the importance of balancing their efforts on making a living and their support to children and other family members. Taking into account the difficult situation of migrant families, feasible strategies should be suggested on how to achieve that balance when they have to manage various challenges in their economic hardship, family relationships and parenting practices. For those parents who cannot attend such workshops, schools could consider some outreach programmes in collaboration with neighbourhood community centres to reach these families and to provide as much as possible assistance for them to be able to exercise their supportive roles for the children.

In addition to educational programmes focusing on the importance of family, given the indirect effect of community on youth mental health outcomes, another direction of social services could be organising structured youth programmes through the community centres. For example, a community centre could consider hosting adult volunteers to develop tutoring or mentoring programmes. These kinds of programmes have been demonstrated to be effective in promoting youth competencies, especially among low-income minority youth living in disadvantaged neighbourhood (Benson, 2002; Quane & Rankin, 2006). Positive relationships with staff and peers in the community youth programmes will not only foster positive developmental outcomes of youths, such as increased self-esteem, decreased problem behaviours and better psychosocial functioning (Hirsch, Roffman, Deutsch, Flynn, & Pagano, 2000; Loder & Hirsch, 2003; Pierce & Shields, 1998), but also allow youths to bring back home the culture of support and to be more active in seeking and providing support in their family spheres, thus generally creating a supportive family environment that is beneficial for their own health and well-being.

To conclude, children from migrant families constitute a vulnerable group of young people in the context of rural–urban migration in mainland China. Factors embedded in their immediate social contexts, such as in the family and the neighbourhood spheres, are found by the present study to be significantly predictive of their mental health status, an important indicator of their overall development. In an effort to promote the mental health of migrant youths, policymakers and social service professionals should take active measures at the community, family and individual levels to foster supportive relationships in the family environment and to facilitate the accumulation and mobilisation of social capital in the neighbourhood context.