Chinese parent-child relationships in later life in the context of social inequalities

This paper examines how parent-child relationships vary against the backdrop of socio-economic inequalities evident in China. China is both an increasingly unequal and rapidly ageing country. Understanding how the relationships that older Chinese have with their children are associated with social inequalities is therefore of paramount importance. We do this by examining the effect of socio-economic indicators of the parent and child on their relationship in a multilevel, multinomial logit model of parentchild dyads using data from the Chinese Family Panel Study. First, the relationships we observe are not unidimensional and display complex patterns which deviate heavily from a ‘strong versus weak’ description of family ties. The results do not support a family displacement perspective of parent-child relationships but instead suggest that educational and financial resources facilitate support that is associated with greater emotional closeness and negates the need for support which places an emotional strain on the parent-child relationship.


Introduction
China presents a fascinating opportunity to examine parent-child relationships given rapid economic development, increasing inequalities, and a unique and highly specific social and cultural context with regards to family ties. For example, incomes more than doubled in the last ten years (IMF 2016), leading to considerable inequalities within society but also within families (W. Cheng/Wu 2015;Kanbur/Zhang 2009). Educational opportunities such as access to higher education have also expanded rapidly over the past 20 years, meaning that educational inequalities within and between generations is an ingrained feature of modern Chinese families (Emran/Sun 2011). In addition, the mass migration of individuals from rural to urban areas, particularly amongst younger generations, has led to increasing proportions of aging parents who are separated from their adult children (Cong/Silverstein 2011;Connelly/Maurer-Fazio 2015;Guo/Chi/Silverstein 2012). This trend of modernization has transformed Chinese society and, given the relative shift in resources between generations, the impact on Chinese families is not likely to be immune to this change. This paper seeks to examine how Chinese parent-child relations in later life are situated in this unique socioeconomic context through the creation of a relationship typology and scrutiny of the typology's association with a variety of socio-economic factors.
The multiple dimensions of intergenerational ties are best understood as a complex multi-faceted set of arrangements within the context of a longstanding and often at times mixed affectual and emotional relationship (Dykstra/Fokkema 2011;Ferring/Michels/ Boll/Filipp 2009;Silverstein/Bengtson, 1997;Silverstein/Gans/Lowenstein/Giarrusso/ Bengtson 2010;Steinbach 2008;Van Gaalen/Dykstra 2006). Empirical studies of parentchild relations have revealed complex, multi-dimensional relationships described by the intergenerational solidarity model, which identifies several underlying dimensions including, but not limited to, affectual and functional solidarity (Bengtson/Roberts 1991). To consider the different dimensions of parent-child relationships independently, focusing only on separate elements of the solidarity model (for example, associational or the structural part of the relation), neglects many of the nuances that exist within parent-child relationships that cannot be captured by the idea of strong versus weak family ties (Reher, 1998). Research on modernity and family relations in later life is relatively less common but there is strong empirical evidence in support of the intergenerational solidarity model, including from China (Ikels 2006;Lei 2013;Ruggles 2007).
Modernization theory identifies greater economic resources as the most important reason for a variety of social outcomes and sees higher incomes and independence as juxtaposed to complex family arrangements and interdependence in traditional societies (Parsons 1960;Slater/Goode 1964). With regards to intergenerational relations, it can be argued that they are "situationally dependent and shaped by local circumstances of history, economics, social organization, and demography and by personal circumstances of wealth, gender, and family configuration" (Ikels 2004: 2). Previous empirical research on modernization theory and the family has focused on its impact on family formation, the distribution of household work and the first and second demographic transitions, whereas research on later life too frequently uses filial piety and cultural norms to explain findings that vary in the separate dimensions of the solidarity model.
In this paper, we seek to examine whether higher incomes, education and geographical mobility that result from rapid and extensive modernization are indeed associated with 'modern family behaviours' identifiable by withering parent-child relations or whether these relationships are still prescient. How do Chinese who have benefitted from sustained and rapid economic development differ in their family relationships from those who are less fortunate? What types of parent-child relationships can we expect to find when a child of a rural labourer has a university degree? How do the parent-child relationships of the 245 million urban migrants differ from those who stayed behind? Do the social inequalities arising out of modernization result in inequalities in parent-child relations? In our analysis, we use representative data of the Chinese population over 60 from the Chinese Family Panel Study (CFPS). Educational mobility, economic prosperity and migration patterns are not evenly distributed across the population of China and there are large differences between urban and rural areas as well as across provinces. The aim of this paper is to give a comprehensive overview of how Chinese parent-child relationships in later life compare for different groups within China and therefore capture the true depth and breadth of social transformation within China. The representative sample of the Chinese population allows for a clear identification of how groups with diverse socioeconomic circumstances maintain family ties, and to our knowledge is the first study to encompass both rural and urban Chinese populations.

Parent-child relationships in China
Research on later life parent-child relationships in China has been developing rapidly in the past few years. For example, Guo et al. (2012) analysed the parent-child relations of the over 60's in rural parts of Anhui province, situated in Eastern Central China on the Yangtze River. Their findings suggest that the considerable number of migrant children providing remittances to their parents reflect the strong filial obligations that Chinese adult children have toward their parents and that many migrant children engage in complex exchanges of support which reflect collaborative and mutually beneficial parentchild relations in the context of massive rural-to-urban migration. These findings emphasise that, in contrast to the European and North American context, Chinese parent-child relations are greatly shaped by migration. This is exacerbated by the Hukou household registration system (T. Cheng/Selden 1994), adding dynamics to Chinese families that are not found elsewhere. Each individual is registered with either an urban or rural hukou. If you have a rural hukou, you can live in an urban area but you do not have the same status as someone with an urban hukou. The strict hukou system denies migrants access to many of the social services in urban areas such as schools, welfare systems and certain forms of employment (Cao/Liu 2015;Y. Chen/Feng 2013;Han et al. 2014;Li et al. 2014;Y. Wen/Hanley 2015) and therefore Chinese who migrate to urban areas maintain complex family ties with those they left behind. When their own children migrate, old parents frequently continue to live in a rural area, possibly by themselves or taking care of grandchildren (Biao 2007;M. Wen/Lin 2012). Thus, intergenerational support is related to migration also in the form of childcare for children who are left behind with grandparents in rural areas. Urban migrants therefore maintain close relations, often sending financial support back to their parents (Cai 2003;Secondi 1997). Regarding gender, women represent approximately half of all migrant workers (National Bureau of Statistics of China 2016). The large-scale migration of women into urban areas where they have limited access to childcare has therefore led to complex intergenerational arrangements. Older Chinese care for their grandchildren in what appears to be exchange for both long and short term financial support (Cong/Silverstein 2011). An open question remains, however, as to how this exchange fits within the wider parent-child relationship and particularly its association with emotional aspects (Silverstein/Bengtson 1997;Tu, 2016).
Despite the attention that exchange-based relationships with migrant children have received, they are less prevalent than relationships where children provided financial support to family left behind in rural areas without any identified reciprocation (Cong/Silverstein 2011). This particular type of relationship is possibly attributable to two factors: the absence of social support mechanisms for the elderly which means that older persons are heavily reliant on family networks for subsistence (Connelly/Maurer-Fazio 2015), and the high degree of filial piety within Chinese society, traditionally associated with Confucian patrilineal, family values (Hamilton 1990;Zhan/Montgomery 2003).
Parent-child relations are also influenced by the changing policy context for Chinese families. China is currently undergoing the implementation of nationwide pension reforms (OECD 2015). These include the introduction of the New Rural State Pension as well as expansion of individual accounts alongside the urban pension system. Yet these pension reforms vary in the implementation across provinces and between rural and urban areas creating further inequalities in the incomes of Chinese older persons (Wu 2013). This could potentially impact upon parent-child dynamics with a reduced reliance on the support provided by children (Kohli 1999), and the reduced dependence on the financial contributions of children could have various implications. First, it could reduce the levels of remittances by children whose parents are now more financially independent. Second, rural children may be less inclined to migrate in the first instance, given that there is potentially a reduced need for remittances. Finally, it could be largely neutral in its effect on parent-child relations, indicating that parent-child relations are not a reflection of parental dependence but instead rooted in short or long-term exchange mechanisms or traditions of filial piety. These questions are complex and go beyond the scope of this paper, but they highlight the need for analysis that is nationally representative and accounts for the diverse contexts within China with regards to family relations.
A drawback of existing research on intergenerational ties in China is the limited frame of reference which has been used, focusing predominantly on a select population in isolation such as rural residents in a specific province or urban migrants within specific cities. As stated at the outset, considerable socioeconomic differences exist within China and by investigating the population of older Chinese persons together, it is possible to examine how parent-child relations differ between urban and rural, rich and poor, highly educated and low educated and those with access to social services and those with none. For example, per capita disposable incomes of urban residents are three times that of rural residents (China Statistical Yearbook 2014). This is particularly important considering the significant social inequalities evident across China. In addition, the limited frame of reference within existing research is not only empirical. The analytical frameworks used in the analysis of Chinese family ties have tended to do so in a comparative approach, focusing on what separates Chinese families from North American or European families. In this paper, we instead aim to understand how families with diverse social circumstances, differ in their parent-child relationships.

Family relations and social inequalities
The process of modernization in China has been rapid, extensive and reached into all areas of Chinese society resulting in significant inequalities in resources. Following theories of modernization (Offer 1998;Slater/Goode 1964), it is tempting to expect that family relations will be more intense when there are strong resource constraints necessitating mutual loyalty and support. These are commonly referred to as traditional family relations. Conversely, economic growth and social upheaval are thought to lead to weakening of family interdependence in what are commonly referred to as 'modern family relations' (Yan 2010;Zore-Maver 2002). It might be expected that modernity within society (i.e. higher incomes, higher education and higher urbanicity) will lead to more 'modern' family relations. Yet this is an assertion that has been challenged by scholars who argue that a more complex interplay between social context and parent-child context exists (Szydlik 2008).
Among the old, the impact of modernization is evident in higher income levels. These higher incomes amongst older persons in China have been through two processes. (1) The expansion of state capacity has included the development of a range of first and second pillar pension arrangements. (2) The general increase in wages and prosperity over the past 30 years has raised living standards across China and enabled higher savings levels and development of third pillar pensions. The number of urban residents in receipt of a basic pension doubled between 2000-2010 and the number of rural residents in receipt of a pension doubled between 2009-2012 alone (China Statistical Yearbook 2013). This represents a large shift in basic income provision amongst China's older population but these developments have been uneven. Following a modernization perspective, we hypothesize that parents with higher income levels are less likely to have traditional relationships in which they are the recipient of substantial transfers and support (H1).
Among the young, modernization is evident in inequalities in levels of education. The percentage of school leavers going on to university rose from just 3% in 1993 to 27% in 2013 (China Statistical Yearbook 2014). There are several consequences for intergenerational relations of this: (1) increases in education lead to a shift in the balance of resources between generations over time with younger generations outperforming their parents on the labor market, sometimes by significant amounts (Emran/Sun 2011); (2) tertiary education often requires the child to move away from their parents at least for the duration of their studies and greatly increases the chances that they will stay away once their studies end (geographical mobility) (Kalmijn 2006), and (3) tertiary education itself is potentially associated with extensive value differences across generations, potentially leading to normative discordance including familial values and potentially notions of filial piety (Inglehart/Baker 2000). We argue that children with higher levels of education will be more likely to have modern parent-child relationships with little exchange up and down generational lines (H2).
In addition to a reduction in exchange between parent and children, it is anticipated that modernization could lead to a reduction in exchanges that qualify as emotionally strained. These strained parental-child relationships are marked by intergenerational support alongside a weaker emotional relationship and reflects existing findings which suggest that intergenerational support does not always exist alongside close emotional ties (van Gaalen/ Dykstra/Komter, 2010). The lower levels of dependence for higher income parents and highly educated children should allow for relationships with less resource constraints and thus potentially less emotional conflict or tension. This leads to the hypothesis that parents with higher incomes and children with higher education levels are less likely to give and receive intergenerational support alongside emotional distance (H3).
In addition to inequalities in income and access to higher levels of education, socioeconomic development in China has resulted in approximately 245 million individuals moving to urban areas but remaining formally registered in rural areas under the hukou system (China Statistical Yearbook 2014). Around 55% of older adults (ibid.) live in urban areas, yet their intergenerational ties are largely absent from the literature which focuses mainly on rural residents.
Urban Chinese maintain relations with their family but potentially in forms that differ markedly from the rural population, particularly given the hukou system. Amongst these urban Chinese only 56% (CFPS 2010) have an urban hukou and it should be expected that their circumstances are very different, given that those without an urban hukou lack access to public and private services. Hence, in line with theories of modernization we expect that children residing in an urban area that also have an urban hukou will be more likely to have modern parent-child relationships with more limited exchange than both rural residents and rural-urban migrants (H4). This assumption is based on the vast differences in social services found between urban and rural areas of China. With poorer social services available to urban migrants and those family left behind, it is anticipated that these groups will continue to rely on family support mechanisms. Poor social services in rural areas will require migrant children to send remittances home to their parents, and poor access to educational and childcare services will lead to greater support provided by the family left behind. This view of modernisation is closely aligned with crowding-out theories of family relations which see public services as a substitute to family provided care. However, we caution against a strict interpretation of crowding-out theory and instead anticipate that family relations will allow family members to specialize in the particular forms of care and support that they are best placed to provide (Igel/Brandt/Haberkern/Szydlik, 2009). This ability to specialize in the types of provision offered could enable support exchange which places less of an emotional strain on the relationship. Therefore, we expect that the prevalence of intergenerational support alongside emotionally distant relationships to be lower amongst urban residents than migrants or rural residents (H5).

Data and methods
We use data on parent-child dyads from the first wave of the Chinese Family Panel Study collected in 2010. The Chinese Family Panel Study is a household panel study conducted in 33 Chinese provinces (Xie/Hu 2014). The response rate in 2010 was 81.3 percent at the household level and 84.1 percent at the individual level. We restrict our sample of parents (the anchor) to the respondents over 60, given that questions about intergenerational support were specifically asked to this subpopulation. This left a sample of 5,412 individuals. Once individuals with missing values on variables of interest were excluded, 4,673 individuals were left in the analytical sample. The impact of this is presented in table 2. The average number of children for respondents is 2.7, and there are a total of 12,958 parentchild dyads. The sample consists of 4,568 parents and each parent is asked specifically about their relations with each child in turn. 30 respondents refused to provide an answer regarding at least one of the indicators for at least one of their children and were thus excluded from the analysis. This provides us with a multi-level structure for our analysis such that parent-child dyads are nested within parents, therefore expanding on previous research that has focused on individual parent-child dyadic relations (Van Gaalen /Dykstra 2006), or that has aggregated sibling information to form a general family pattern (Dykstra/Fokkema 2011). The high number of children per respondent is due to the fact that those over 60 were unaffected by China's one-child policy.
In order to measure the various dimensions of the parent-child relationship, a total of 7 survey instruments were used (please refer to Table 1). The first instrument indicates emotional closeness where 1 = Not Close; 2 = Close; 3 = Very Close (Question F1) 1 . The remaining indicators are dichotomous indicators of whether support was exchanged with the specific child and include (1) giving grandparental childcare, (2) receiving care and support in the household; (3) giving financial help; (4) receiving financial help; (5) helping the child with housework; (6) receiving help from the child with housework. Respondents were asked to indicate whether they had engaged in any of these activities over the past 6 months with their children and then asked to stipulate which children (Question F2) 2 . Parental income was measured as the log of the equivalized household income in the last 12 months. The education level of the parent was coded into three levels (low, medium and high), reflecting those who have not graduated from high school, those who have only graduated from high school and those who have had education beyond high school respectively. The same coding was used to establish the child's education. However, in the analysis we use different reference categories given the differing distributions between parents and children. Amongst the parents, 55% didn't graduate from high school. Amongst the children, 73% graduated from high school, but went no further. Given this, for the parent's we use 'low education' as our reference and amongst the children we use 'medium education' as the reference category. Residence and migration is captured through an indicator of whether an individual is a rural resident, urban resident or urban migrant (living in a city without an urban hukou which is a combination of variable Urban and qa2). Parents did not specifically report whether children lived in an urban area. We assumed that if a child is living in a county different from the parental home and their parents do not have an urban hukou, then they are living in an urban area. We checked this using the general sample from the CFPS, where 94% of those individuals not living within 50 km of their county of birth and whose parents did not have an urban hukou, lived in urban areas. We also assumed that a child was living in an urban area if they were living in the same county as their parents and their parents were themselves urban migrants. Weighting of the data was applied throughout the analysis using 'rswt_nat' to provide a representative sample of parents over the age of 60 in the 33 provinces for which the CFPS provides data. The first stage of the analysis consists of a Latent Class Analysis, a statistical approach used to find groups or subtypes of cases in multivariate categorical data. A latent class approach was used in order to examine the multidimensional nature of parent-child rela-tions. All parent-child dyads are analysed and assigned to a class, and the Bayesian Information Criterion is used to assess best model fit, with lower BIC indicating a better fit. The best fitting model was used to determine the number of classes and the analysis was conducted with the R package poLCA. Once parent-child dyads are assigned a class, a multilevel, multinomial random intercept logit model is fitted using MLwiN to investigate the associations between the class membership of each parent-child dyad with sociodemographic indicators of parent and child used as controls: gender, age, marital status, income, residence status, hukou status, education level and the child's birth order. These were selected based on the existing research on intergenerational relations (Bengtson/Roberts 1991;Cong/Silverstein 2011;Dykstra/Fokkema 2011) and the hypotheses laid out in the previous section. A multilevel logit was used given that for each parent, there were several parent-child dyads representing each of their children and they were therefore nested within the parental level. Please refer to Table 2 for the descriptive statistics of all used indicators. The Latent Class Analysis (LCA) identified six classes of parent-child relationships. The LCA propensities are presented in table 3. The results show a large dominant group of parent-child dyads through which little exchange is identified, labeled as "autonomous". Previous research, based on rural parents only, had suggested that this constituted around 30% of parent-child dyads (ibid). Our analysis, inclusive of urban residents and urban migrant parents, suggests that this is far higher amongst the wider Chinese population with 50% of parent-child relationships, exhibiting little to no exchange. We would refrain, however, from presenting these relationships in a negative light and have thus opted for the label 'autonomous' to reflect the lack of exchange without attaching negative connotations. These relationships may represent a 'dormant' relationship rather than one in which the parent and child are averse to exchange. Nevertheless, the relationships also exhibit the lowest propensity for emotional closeness, suggesting that a lack of exchange is associated with emotional distance. Standing in direct contrast to this class is the 'intensive exchange' class (2%). These are parent-child relations in which there is a high propensity for exchange across all the indicators included. This suggests a very complex and intricate pattern of interdependence in which parents and children are simultaneously providing help to each other, often in the same form (e.g. upward and downward financial support or upward and downward care giving), suggesting a considerable degree of resource pooling and close cooperation. These relationships are marked by a high propensity for emotional closeness, suggesting that the functional exchange operates in a cooperative and emotionally positive relationship.

Latent Class Analysis
The third and fourth classes represent uni-directional support relationships and are differentiated by whether the support is provided by the child to the parent (upward support, 12%) or whether the support is provided by the parent to the child (downward support, 10%). The propensity for each activity is more balanced than in the second class, suggesting a greater diversity in arrangements of support. For example, the propensity to provide upward care in class 3 is 68%. This is high, suggesting that this is a common, but not so high as a defining factor of such relationships. Instead, these arrangements indicate general support that comes as a combination of some or all of financial help, housework help or care. The propensity toward emotional closeness in these two classes which is similar to that in the second class suggests very little emotional tension. These types of relationships are far more common that intensive exchange based relationships and together represent 22% of all the parent-child relationships in the analysis.
These classes stand in contrast to the final two classes which share two specific features. Firstly, they are focused on two specific forms of support. Class 5 (upward financial support, 16%) is dominated by an absolute propensity to provide upward financial support and a low propensity to provide or receive anything else. Class 6 (care dependent, 10%) is a group dominated by an absolute propensity for the provision of care by the child to the parent. The second commonality is the very low propensity for emotional closeness in these classes. The propensity is not quite as low as that in the 'autonomous' class but it is far below the levels observed for "intensive exchange", "upward support" and "downward support" classes. Taken together, this suggests that the relationship is focused around a very narrow functional dependence and that this is associated with a lack of emotional closeness. We argue that these two classes therefore represent relationships where functional support is observed alongside a lack of emotional closeness (Luescher/Pillemer 1998).
Latent Class Analysis provides a way in which to simplify, cluster and organize parent-child relations. The analysis provides a clear typology of six classes which can help more effectively assess the degree to which parent-child relations vary within China. These six classes fit within three broad patterns: (a) extensive exchange and support alongside an emotionally close relationship [Classes 2, 3 and 4 = 24% of dyads], (b) autonomous and largely independent parent and child [Class 1 =50% of dyads] and (c) relationships distinguished by exchange upon a singular vector and relatively low propensity for emotional closeness [Classes 5 and 6 = 26% of dyads]. These three patterns help elaborate on the hypotheses in that we would expect indicators of modernization to be associated with a greater number of autonomous relationships (H1, H2, H4) and a lower number of relationships with support but lacking emotional closeness (H3, H5).

Multilevel modelling
The multilevel, multinomial, random intercept logit model shows the odds of class membership relative to the class 'Autonomous'. 'Autonomous' was chosen as the baseline category because it is the most numerous and also represents a situation in which there is an 'absence' of a relationship which makes the interpretation of the odds more meaningful. From table 4, we observe that parents with higher income have markedly different parent-child relationships than those with lower income, but not in the direction hypothesized (H1). A parent in the highest decile of income is 54.5% more likely to have a relationship of intensive exchange when compared with someone from the lowest decile 3 . This is a considerable effect size and suggests that, whilst such relationships are rare, they are rarest amongst parents with low income. Similarly, high income parents are 19.3% more likely to have 'Downward Support' relationships relative to 'Autonomous' relationships than low income parents. By contrast, high income parents are 23.9% and 14.7% less likely to have a relationship exhibiting 'Upward Financial Support' or 'Care Dependent' respectively relative to an 'Autonomous' relationship than low income parents, supporting H3. With regards to a child's educational attainment, those with children with higher education are 28% more likely to be in an 'Upward Financial Support' relationship relative to an autonomous relationship than someone who has completed high school (medium education). This does not necessarily mean that they are more likely to provide financial support. Upward financial support can also be observed in other classes such as classes 2 and 3, where it represents a wider set of support giving. The observation is therefore that higher educated children are more likely to have a relationship that is marked by the provision of financial help to parents alongside a lower propensity for emotional closeness and lower propensity for other forms of exchange. In short, they provide money but little else and lends some support for H2. This is the only statistically significant difference between those with higher education and those who have completed high school. When comparing those who did not complete high school with those who did, we can see that they are less likely to have 'Upward Support', 'Downward Support' or 'Upward Financial Support' relationships relative to an 'Autonomous' relationship. This runs counter to the direction hypothesized (H2).
Migration status is the third indicator identified as potentially shaping parent-child relations. First, we compare urban migrants with rural residents. Unsurprisingly we find that urban migrants are far less likely to have relationships that are marked by functional exchange (apart from 'Upward Financial Support') relative to an 'Autonomous' relationship than rural resident children, supporting H4. Somewhat surprisingly, we find that there appears to be no statistical difference in the likelihood of an urban migrant child and rural resident child having an 'Upward Financial Support' relationship despite strong evidence in the literature that remittances are a primary part of migrant intergenerational relationships. To examine this further, the model was run without child's educational status and the Urban Migrant coefficient then became positive, large and highly significant poten-tially indicating that the relationship between migration and remittances is heavily mediated by the child's educational status and subsequent labor market earnings. When we compare urban resident children (with urban hukou) with rural resident children, it is shown that they are respectively 23.4% and 30.4% less likely to have 'Upward Financial Support' or 'Care Dependent' relationships relative to an 'Autonomous' relationship. This indicates that relationships with support but lacking emotional closeness are far less common for children with an urban hukou. Nevertheless, children with an urban hukou are not less likely to have 'Upward Support', 'Downward Support' or 'Intensive Exchange' based relationships relative to an 'Autonomous' relationship than a rural resident child, thus supporting H5. With regards to other predictors of parent-child relationships, for older parents the parent-child relationship is one of receiving greater support, particularly with regards to classes 'Upward Support' 'Care Dependent' and 'Upward Financial Support'. In line with general findings on intergenerational relations, fathers appear to have less support relationships than mothers and are more likely to have 'Autonomous' relationships (Luo/ Waite 2014).
With regards to the characteristics of the child presented in table 4, sons are far more likely to be in relationships shaped by the receipt of support, and daughters are 70.4% less likely to be in a 'Downward Support' relationship relative to an 'Autonomous' relationship. This is in line with existing literature indicating the preferential support offered to sons in China (Cong/Silverstein, 2011). It would not, however, appear that daughters are more likely to be in upward support relationships such as 'Care Dependent' or 'Upward Support' as is generally the case in Europe and North America (Dykstra/Fokkema 2011). This is potentially because women provide support to their in-laws in China and this is not covered by the dyadic parent-child approach used here (F. Chen 2004). With regards to birth order of the child, we see a relatively consistent effect that children lower down the birth order are more likely to be 'Autonomous' than any other parent-child relationship type. This supports the existing literature which suggests that support relationships both upward and downward in China are concentrated upon the eldest child (Das Gupta et al. 2003).

Discussion and conclusion
The aim of this paper was to establish how the diverse and dynamic Chinese social context impacts parent-child relationships and investigate a representative sample of the Chinese population that includes both urban and rural parent-child dyads. We paid special attention to income, education and migration status as key characteristics that shape intergenerational relations in contemporary China. Contrary to (H1), parental income was positively associated with membership of two classes relative to autonomous relationships ('Downward Support' and 'Intense Exchange'). Both of these groups appear to describe family relationships that have strong provision of help and support including financial help, childcare and housework. The analysis suggests that it is parents from higher income groups that are more likely to be engaged in such patterns of exchange. This is in contrast to traditional theories of modernization where economic development and the increase in incomes at older ages provided for by pension systems and greater income over the life course are generally considered to reduce intergenerational interdependence (Kohli 1999). The notion that a higher income means that there is less exchange with children at older ages, is not supported in our findings. And yet, higher income parents are less likely to have the two forms of relationship identified as being marked by intergenerational support alongside a lack of emotional closeness: 'Care Dependent' and 'Upward Financial Support'. This suggests that for the most intense forms of dependency, parents are able to use their resources to maintain some degree of autonomy, particularly from forms of dependence that are associated with lower emotional quality (Albertini/Kohli, 2013). This supports (H3) in suggesting that the reduction in dependency brought about by higher incomes is associated with a move away from support accompanied by emotional distance and interdependency.
At the outset of the paper, we suggested that the rapid expansion in education for younger generations and the subsequent social inequalities would likely be associated with dramatic differences in intergenerational relations. Modernization theory suggests that educational mobility lessens the degree of intergenerational exchange and support. When we compare those with higher education to those who graduated only from high school, however, we see very few differences and little evidence in support of this hypothesis. We do, however, see a greater prevalence of 'Upward Financial Support' relationships in this group reflecting their ability to provide remittances. This in itself may suggest that remittances should be viewed, as has been seen in findings from Anhui, as a return on the investment in the child and the closeness of the relationship (Ji/ Guo/ Feldman, 2015). Our typology offers a further insight, however, as it should be noted that this relationship is one that we identify as marked by greater emotional distance. This could reflect a few factors such as a discordance in normative values that is brought about by higher educational attainment, greater physical distance associated with higher educational attainment or even a resentment of the implicit 'pay back' involved in rapid social mobility and familial remittances. What the results appear to suggest, however, is that it should not be assumed that remittances are a sign that a parent-child relationship is one based on emotional closeness and that a multidimensional approach to familial relations is necessary.
When we look at the difference between those that have not completed high school and those who have, the results do not necessarily support the simple narrative offered by theories of modernization and their impact upon intergenerational relations. Contrary to (H2), the low educated were far less likely to be involved in exchanges of support with their parents. Not only that, those who had graduated high school were far more likely to be in relationships typified by a high propensity toward emotional closeness (Class 2, 3 and 4). It is unclear as to precisely why this might be and it lies beyond the power of this analysis to determine this. However, there are several theories that offer potential explanations. First, it could be that investment made in the child through better education leads to a more supportive and closer relationship between parent and child. Second, it could be that the resources afforded to the child by their education enable them to participate in and benefit from support and exchange with their parents. Third, it could be unobserved effects such as serious exogenous shocks (i.e. crop failure, death in the family, ill health) that disrupt both the child's educational development and the development of their relationship with their parent. The evidence is, however, not in support of the hypotheses that the higher educated would be less engaged in support (H2).
One further reason for this may be that the pace and scope of educational expansion in China has defused the effect of social mobility. Previous research on social mobility's effect on family ties has indicated that intergenerational support is only reduced when intergenerational mobility is the result of individual achievement and not the product of structural change such as an upward shift in the occupational distribution or the expansion of higher education (Kalmijn, 2006). The negative impact of increases in income, higher educational attainment and geographical mobility would only occur when they are brought about by individual differences in achievement or through choice. The inequalities in China we observe are structural and it may therefore be that children are increasingly distant from their parents in terms of income, education and geography not through choice and effort but by forces that are out of their control. Family ties could therefore be as relevant for those affected by these structural factors as for those who are not. These arguments suggest that parental income and offspring educational attainment would not differentiate parent-child relations.
In support of (H4), we saw a dramatic impact of a child's migration on parent-child relations. This is unsurprising given that many of the indicators of support require a degree of physical proximity. Compared to both urban and rural residents, urban migrants have less exchange and support and this is reflected in relationships with less emotional closeness. From descriptive statistics alone it is clear that urban migrants are less likely to be 'very close' to their parents emotionally than non-migrants (24% v 33%). There is a degree to which this is a selection effect, with children migrating due to a lack of emotional closeness and so interpretations must treat causal mechanisms with caution.
Surprisingly, we find little support for modernization theory's suggestion that urban residents give and receive less support as there is no difference in the likelihood of having supportive relationships between urban and rural residents. Where we do see a difference, however, is in the prevalence of supportive relationships with a lack of emotional closeness such as 'Upward Financial Support' and 'Care Dependent' which are far less common amongst urban children. This supports the hypothesis that those in urban areas are less likely to have parent-child relations marked by functional support alongside emotional distance (H5) in suggesting that those in urban areas with access to support services are able to avoid intergenerational relations that make the parent strongly dependent on the child and potentially erode the emotional closeness within the relationship.
Overall, the picture of parent-child relationships in China shares two distinct features that are also found in European and North American research. First, the relationships are not unidimensional and display complex patterns which deviate heavily from a 'strong versus weak' description of family ties. This finding supports the use of a latent class approach above using single individual behaviors as indicators of parent-child relationships. Indeed, we found 26% of the dyads to exhibit characteristics that are associated with relationships in which there is functional support but emotional closeness is below average. Second, the association between the type of parent-child relationship and key socio-economic variables is not one that reflects standard modernization hypotheses. Instead, it would appear that relationships are adaptive to context and parent-child relationships are structured in order to navigate a complex and dynamic social context. This is in line with findings from Europe where parentchild relationships adapt and reflect shifts in resources, rather than merely retreating when other forms of support are not necessary or possible (Igel et al. 2009;Kalmijn 2006).
Higher parental income, high school graduation for the child and residence in an urban area were all hypothesized to reduce interdependence and therefore reduce both the giving and receiving of intergenerational support. However, in both regards we observed more support based relationships characterized by greater emotional closeness in these groups of dyads. We also saw that high income parents, highly educated children and children with urban residency were less likely to have relationships that are associated with support alongside a lack of emotional closeness. We conclude that modernization theory only partially allows those elements of a relationship that strain emotional ties to be circumvented and greater independence to be established and yet also provides resources and capacities that facilitate support. Furthermore, the support that is facilitated is closely associated with greater emotional closeness. Research regarding modernization and intergenerational relations would therefore be well-minded to incorporate a multidimensional view of intergenerational relations within their theoretical frameworks and analytical approach.

Future research
The analysis offered indicates that there is a complex association between intergenerational relations and modernization processes, with several findings pointing towards potentially interesting and fruitful avenues of research, especially given the rapid nature of developments in China's socio-economic context and social policy arrangements. Such future analysis of parent-child relationships will be better supported in the future by the continually improving data landscape within China. Future waves of the CFPS and surveys such as CHARLS (China Health and Retirement Longitudinal Study) will offer researchers opportunities to utilize more complex modelling techniques which can better isolate the causal processes at play and provide a more detailed description of developments. Some of the questions raised by this analysis such as the observation of an apparent educational 'pay back' role and the positive association between increased economic opportunities and greater intergenerational support giving and receiving are puzzling in the context of existing research on parent-child relationships and deserve further attention. This is enabled by rapidly expanding data collection in China but is also made possible by a shared conceptual and theoretical framework with North American and European research. Comparative research which incorporates European, North American and Chinese contexts should therefore be high on the research agenda of those examining family relations in later life as such comparisons offer the possibility of extending and refining our understanding of family dynamics in later life. We conclude by asserting that a narrow conceptualization of intergenerational support that identifies only functional support behaviors, presents an incomplete theoretical and analytical view of intergenerational relations in the context of modernization. The multidimensional approach to intergenerational relations adopted here is in line with many recent findings in the sociological literature which emphasize the complex interplay between intergenerational relations and their social context (Brandt/Deindl 2013;Emery 2016;Mudrazija 2016). This is essential if both social researchers and policy makers are to meet with the challenges that are presented by social inequalities currently observed within China.