The impact of migration on intergenerational solidarity types

ABSTRACT This paper aims to expand knowledge on the effects of an international migration on parent–adult child relationships. We develop a typology, include non-migrants in the country of origin for comparison, and consider transnational families. Analyses are based on the Turkish 2000 Families Study, using information of adult non-co-resident children about their relationships with their parents. The research questions are: Do intergenerational solidarity types in migrant families reflect the patterns prevalent in the origin context or migration-specific adjustments? Do solidarity types of migrants differ, depending on whether they are transnational, of first- or second-generation children? Are differences due to composition effects? Latent class analysis shows four solidarity types. Their prevalence differs remarkably across the migrant groups. The proportion of the full-solidarity type is larger and that of the autonomous type is smaller in the relationships of first- and second-generation children with their migrant parents than among stayer dyads in Turkey. In transnational relationships, there is less full solidarity, and autonomous relationships are more likely. All migrant groups display less advice-oriented and more material-oriented support relationships. These results indicate stronger intergenerational cohesion in non-transnational migrant families and few changes across migrant generations. The observed differences are not due to composition effects.


Introduction
In the study of demographic ageing, intergenerational solidarity has become a flourishing field of research. In family sociology, the intergenerational solidarity model (Roberts, Richards, and Bengtson 1991) has triggered copious research on parent-adult child relationships. This model was the first to systematically differentiate between six dimensions of solidarity, including aspects such as emotional closeness, agreement in values, geographic proximity, contact frequency and financial and instrumental support. Considering all or a selection of these dimensions, many studies found evidence for intense intergenerational solidarity in modern societies (Fokkema, ter Bekke, and Dykstra 2008;Nauck and Steinbach 2009). As well as the analysis of single dimensions, a number of studies applied clustering approaches which have stressed the intersection of the solidarity dimensions and the multifacetedness and complexity of parent-adult child relationships (Dykstra and Fokkema 2011;Silverstein and Bengtson 1997).
Migrant families are increasingly included in these research activities . Studies show overall high intergenerational solidarity among migrants in Europe (Baykara-Krumme 2008; Bordone and de Valk 2016). This outcome was supported by a recent study that considered migrant solidarity types beyond single dimensions of intergenerational solidarity (Rooyackers, de Valk, and Merz 2014). However, these studies are commonly restricted to migrants and non-migrants in European residence countries. Migrants' solidarity patterns are compared against the outcomes prevalent in the native population of the destination country. The residual effects are explained by migration-or origin-culture-related concepts, without actually attributing outcomes to one or the other set of mechanisms.
To better understand the adjustments in intergenerational solidarity following international migration, a different perspective is required, namely the consideration of nonmigrant compatriots from the same original regional-cultural contexts (Foner 1997;Nauck 1989;van Hook and Glick 2007). In this paper, we use data from the 2000 Families Study, which covers families from five regions in Turkey who either live in Western Europe today or never left their country (Guveli et al. 2016). We compare intergenerational solidarity patterns of stayers and migrants, focusing on the perspective of the adult child. In the process, we not only contribute to the understanding of family systems in non-Western transitional countries (Guo, Chi, and Silverstein 2012), but also offer a more adequate analysis of the impacts of migration.

The study of intergenerational solidarity types
The paradigm of intergenerational solidarity, also called the 'intergenerational solidarityconflict model', is the cornerstone of research on parent-adult child relationships (Roberts, Richards, and Bengtson 1991). Based on five dimensions of the original model (excluding agreement in values, i.e. normative solidarity), Silverstein and Bengtson (1997) were the first to develop a typology. The authors found five types of family relationships in the United States: tight-knit, sociable, intimate but distant, obligatory and detached. The tight-knit type (i.e. engagement on all indicators) was most common in mother-child relationships (about one-third of the sample), whereas the detached was most frequent (more than one-quarter) in father-child relationships, followed by the sociable type, lacking only support exchange, in both groups.
In subsequent years researchers followed this approach and studied solidarity types in different contexts, from the perspective of either the adult children or the parent and with varying dimensions (see overview in Guo, Chi, and Silverstein 2012). Findings show differentiations between families not only in terms of aspects like low and high support exchange, but also upward or downward direction of predominant support (Dykstra and Fokkema 2011). Predictors of solidarity types were identified, such as parents' and children's socio-demographic features (i.e. age, gender, marital status) and their needs and resources (education, health status, income) or life-course transitions (Schenk and Dykstra 2012). More recently, researchers broadened the perspective by studying solidarity types across national and cultural contexts. International research on single solidarity dimensions had already pointed to macro-level impacts based on national welfare regimes and regional family cultures (Albertini and Kohli 2013;Brandt and Szydlik 2008). Typological studies of intergenerational relationships have also found that solidarity types vary across countries, while at the same time heterogeneity within countries is large (Dykstra and Fokkema 2011;.

Family relationships in a migration context
Most studies have included only the non-migrant majority population. At least one recent study, however, explicitly studied solidarity types among migrants. Rooyackers, de Valk, and Merz (2014) considered mother-child relationships in the Netherlands, comparing Dutch to Mediterranean and Caribbean migrant families. Accounting for eight solidarity dimensions, they identified five types, ranging from the full-interdependent to the independent type with three intermediate types, reflecting different kinds and directions of support (upward-interdependence, downward-interdependence and emotional-interdependence). As expected, the prevalence of the full-interdependent type was higher among migrant families, whereas the intermediate types were more common among the Dutch. The independent type was rare and, against expectations, equally common among the Dutch and the migrants. Socio-demographic compositional differences were not accounted for.
A number of studies address family solidarity among migrants in Europe in more detail, but focus on single dimensions and often single migrant groups only. Findings indicate strong family solidarity norms among migrants compared to non-migrant Northwestern Europeans even when demographic and structural differences are taken into consideration (Arends-Tóth and van de Vijver 2008; Carnein and Baykara-Krumme 2013). Behavioural differences between migrants and non-migrants are smaller and less clear. While among migrants co-residence, contact with and financial support to parents seem to be more common (Baykara-Krumme 2008; Bordone and de Valk 2016;Steinbach 2013) and financial support to children less frequent, findings on non-material kinds of support are inconclusive (Baykara-Krumme 2008; Bordone and de Valk 2016).
In all these studies, migrants were compared to non-migrants in the destination context. By highlighting differences, the findings have drawn a specific picture of families in contexts of adaptation and acculturation. This includes Turkish families, who are present in most studies as they constitute a large migrant group in Northwestern Europe and display strong family cohesion. Existing research, however, cannot resolve whether evident differences between migrants and non-migrants can be attributed to culture-of-origin or to migration-related influences. An international migration is a 'natural experiment' in which families change their socio-ecological context in a profound way (Nauck 1989). The new context provides new opportunities and constraints as well as new sets of values, beliefs and standards; the families may change accordingly, while pre-migration cultural frameworks are likely to remain influential to a certain extent (Foner 1997).

Theoretical framework and hypotheses
Individual resources and needs are major predictors of intergenerational solidarity, e.g. financial, health and social resources (Albertini and Kohli 2013;Szydlik 2008). For instance, research shows that children are more likely to support their parents when the need arises due to illness, and parents have more contact with non-co-resident children who are not yet married. Important broader contextual factors are the dominant cultural context and welfare regimes which shape prevalent norms and provide varying sources of support outside the family (Albertini and Kohli 2013;Brandt and Szydlik 2008;Dykstra and Fokkema 2011).
The main focus of this paper lies in the specific features concerning international migration. We pay less attention to varying socio-demographic and socio-economic composition among the migrant and stayer groups, which may account for observed differences in intergenerational relationships (we will control for these factors in the analyses; see below). Rather, given the sparse knowledge about stayer comparisons and the explorative approach implied in the analyses of solidarity types, we aim to generate a framework that outlines migration-specific conditions and two related sets of hypotheses. Broadly, two divergent outcomes have been suggested for the effect of international migration on family solidarity, which derive from different underlying mechanisms (Baykara-Krumme 2008;Nauck 1997Nauck , 2007. The first outcome implies lower family cohesion among migrants compared to their stayer peers, and this is attributable to changes of the cultural and institutional context in the course of migration. The second outcome implies higher family cohesion due to the migrants' specific experiences as ethnic minority members.

Cultural and institutional context
With respect to cultural context, theories of 'cultural family systems' refer to historically different family and kinship structures (Reher 1998). Our focus country, Turkey, is characterised by a patrilineal regime of descent, which is widespread among the rural population and among the less educated in the larger cities (Nauck and Klaus 2008). This becomes empirically evident in the high expectations the elderly have of intergenerational support from their descendants. By contrast, in the affinal kinship system of Western European countries prime solidarity lies between the spouses, and intergenerational expectations are less prevalent (Nauck and Suckow 2006). In a similar vein, in their concept of family change, Kagitçibasi, Ataca, and Diri (2010) describe Turkey as an example of the 'family model of (emotional) interdependence', which differs from Western societies and their proposed 'family model of independence'. Accordingly, Turkish society is characterised by a culture of relatedness and a higher prevalence of collective rather than individual norms, with specifically strong emotional ties yet decreasing material dependencies (Kagitçibasi and Ataca 2005).
With respect to the related institutional context, the family remains an essential source of security against the life time risks of individuals in Turkey's social insurance system (Grütjen 2008). Although it is not included in research on European welfare regimes which differentiates between Scandinavian, Continental and Mediterranean welfare regimes (e.g. Albertini and Kohli 2013), Turkey has been assigned to the Mediterranean regime, characterised by low public support in terms of transfers and services and a corresponding high welfare burden on families (Gal 2010). In the other regimes, welfare state provisions are institutionalised differently and more generously, making it less necessary for family members to rely on each other and allowing for specific patterns of public-private task complementation, substitution and specification (Brandt and Szydlik 2008). Within Europe, resulting differences in intergenerational solidarity patterns have been described in terms of a 'north-south gradient'. Family obligation norms, co-residence and contact rates are lower in the more generous welfare regimes of the Northern countries; transfers of time and money from parents to children occur more frequently but less intensely than in Southern Europe (Albertini and Kohli 2013;Brandt and Szydlik 2008). In Western and Northern European countries the ascending type with prime support from children to parents is less prevalent, and Southern European countries display a lower proportion of the autonomous type (Dykstra and Fokkema 2011).
Consequently, Turkish migrant children and parents are assumed to adjust their relationships to the cultural and institutional patterns of the destination regions of Northern and Western Europe. The emerging solidarity outcomes may be a situational response to better welfare systems that presuppose less reliance on family support. It may also be due to cultural learning and adoption of less family-oriented value patterns in the course of intercultural contact (Berry 1997;Nauck 1989). More specifically, acculturation and assimilation theories suggest differences between first-generation migrants and their European-born offspring. Given the different cultural socialisation context and the better opportunities for cultural and structural integration in subsequent generations (Portes and Rumbaut 2001;Thomson and Crul 2007), second-generation members are assumed to adapt faster and diverge further from stayers in Turkey than first-generation migrants. Existing findings on differences in family solidarity of firstand second-generation members remain inconclusive. Some suggest variations in family closeness and support for family norms between immigrant generations (Arends-Tóth and van de Vijver 2008; Merz et al. 2009), others stress the high normative intergenerational similarities (Carnein and Baykara-Krumme 2013).
Transnational families with parents in the home country are a specific case. Crossborder ties have still scarcely been addressed in research on intergenerational relationships (Karpinska and Dykstra 2019). Living transnationally can be seen as a threat to family solidarity due to geographical distance (King et al. 2014), yet innovative technologies facilitate new forms of contact and support exchange (Baldassar et al. 2016), and detached relationships are quite rare (Rooyackers, de Valk, and Merz 2016). Applying structural and cultural explanations, transnational families are likely to display different patterns from stayers. As numerous studies show, the migrant child may adapt to the better welfare regimes and to the more individualistic culture of Northern and Western Europe with lower reliance on the family, but at the same time engage with the specific needs and expectations of parents in the home country, most likely through financial support (Silverstein and Attias-Donfut 2010; Wolff 2019).
Based on these considerations of different levels of cohesion and autonomy in intergenerational relationships, we formulate a first set of hypotheses: H1a. Given the different cultural and institutional settings and related adaptation and acculturation processes, relationships of children of the first and the second generation and their migrant parents are less cohesive and more autonomous than child-parent relationships among stayers. Due to more advanced adjustment, these differences become more pronounced in the second generation.
H1b. Transnational relationships between children in Europe and their parents in the home country are less cohesive and more autonomous than relationships of stayers, but involve more upward material support.

Ethnic minority status
The second line of reasoning considers mechanisms related to the challenges that migrants face in a minority context. This argument has been traced back to observations in family sociology (Nauck 1997). The refugee family after World War II was described as a specific family type, characterised by high stability, strong social isolation and retreat from the outer world (Schelsky 1950). Other exceptional historical circumstances that strengthened family ties were the collapse of economies and political systems (Szydlik 2008) or international migrations. The focus now is on how families develop strategies for survival and assist their members in the process of adjustment and advancement in a foreign setting (Foner 1997; van Hook and Glick 2007). As discussed above, studies show that interdependent family relationships still play an important role among migrants; the gap in the research concerns whether this pattern is migration/minority-specific or a reflection of patterns in the original context. At least three minority-specific mechanisms can be found in the literature.
First, experiences of discrimination and social exclusion in the majority society may emphasise the salience of the family as a refuge in times of trouble. Such feelings of insecurity may last well into later life (Cela and Fokkema 2017). Kin support can serve as a buffer to stress (Bengtson and Martin 2001), and family ties may help cope with uncertainty and economic scarcity. While the family is certainly not always a 'safe haven' but a source of conflict too, its significance may be reinforced in the minority situation, as migrants draw on pre-migratory cultural frameworks of strong family orientation (Foner 1997). Studies show, for instance, that discrimination experiences often go hand in hand with language retention in families (Nauck 2007) and traditional family values (Arends-Tóth and van de Vijver 2008).
A second mechanism refers to the (non-)availability of ties outside the family. Migrants turn to their family to compensate for the loss of wider social networks that were available back in the country of origin (Foner 1997). Fewer non-family social ties in the country of destination result in a higher reliance on the family for meeting emotional and practical needs, and this holds true not only for recent migrants but also for migrants who age in place (Cela and Fokkema 2017). Among the few studies that have explicitly examined the composition of Turkish migrants' network ties, Nauck and Kohlmann (1998) observed the strong role of kinship in migrants' social networks.
The third mechanism entails the intergenerational transmission of family solidarity. Cultural transmission is the foundation for family solidarity and intergenerational continuity specifically in a minority situation (Phalet and Güngör 2009). If other culture-transmitting institutions (such as schools or religious bodies) are missing, the importance of intergenerational transmission in families increases. The fact that this is more difficult and yet more necessary after migration has been called the 'paradox of the migration situation' (Nauck 2007, 48). Parents may have a strong intention to transmit traditional values, in order, for instance, to safeguard their return projects or ensure filial support in old age; early findings show that value transmission was somewhat stronger, i.e. attitudes were more similar, between children and parents in Turkish migrant than in stayer families (Nauck 1997). Additional studies show high similarities in the preferences and behaviours between migrant children and their parents (Phalet and Güngör 2009). This reflects a continued orientation towards the heritage culture (or parts of it), which tends to be stronger than in the home country where families have experienced intense social change (Foner 1997;Guveli et al. 2016).
These considerations imply higher family cohesion among Turkish migrants than stayers as a situational response to their minority situation. Stayers are able to rely on wider social networks and non-family institutions for cultural transmission, rendering family ties less important. The higher family cohesion is likely to apply to both migrant generations alike, because experiences of exclusion, fewer non-family ties and parental interest in strong transmission often persist into the second generation (Portes and Rumbaut 2001;Thomson and Crul 2007). The parents of children born in the host country may be specifically interested in cultural transmission. The children themselves may havehigh expectations of inclusion and equal participation in the residence country, but turn back disappointed to their families when their aspirations are not met. The experience of discrimination and the lack of social ties may also lead migrants to take refuge in intense mutual exchange creatingtransnational social space (Mazzucato 2011). The family members back home may be specifically interested in transmitting family norms to maintain cross-border emotional and supportive bonds.
Accordingly, a second opposing set of hypotheses can be formulated with regard to divergent patterns of cohesion and autonomy in intergenerational relationships: H2a. Given the specific minority situation following migration, relationships of children of the first and the second generations and their migrant parents are more cohesive and less autonomous than child-parent relationships among stayers.
H2b. Transnational relationships between children in Europe and their parents in the home country are more cohesive and less autonomous than relationships of stayers.

Data and method
Sample Analyses are based on the 2000 Families Study, collected between 2010 and 2012 (Guveli et al. 2016). This contains 1992 multi-generation genealogies of male labour migrants and their non-migrant peers from five regions in Turkey: Acıpayam (Denizli), Akçaabat (Trabzon), Emirdağ (Afyon), Kulu (Konya) and Şarkışla (Sivas). In each region, a clustered probability sample was selected using address registers from the Turkish Statistical Institute to identify primary sampling units. The randomisation included a random walk through the neighbourhoods to identify migrant and non-migrant men (the 'anchor' persons). Interviewers had to identify four migrants before they could ask about a nonmigrant, resulting in a quota sample of 80% migrants and 20% non-migrants. The anchor was defined as a male relative of the household member, who was born between 1921 and 1946, originated from the region, was a labour migrant to Northwestern Europe between 1961 and 1974 and remained there for a minimum of five years ('migrant'), or who might have migrated, but did not ('stayer'). For each eligible anchor, basic data of all his male and female descendants were recorded at the doorstep.
For the personal interviews, carried out face-to-face or by phone, the anchor was selected. Subsequently two of his adult children-each with two of their adult children (the grandchildren)and, in very few cases, adult great-grandchildren were selected randomly and also interviewed (N = 5980).
In this analysis, we draw on personal interviews with the adult descendants of the anchors, who answered questions about their relationships with their non-co-resident parents. Based on information of the adult focus child and his/her parents, we differentiate the following dyads: 'stayers', i.e. adult children with parents who live in Turkey and never left their country before the survey; 'transnationals', i.e. adult migrant children who live in Europe but whose parents live in Turkey (the parents may have previously lived in Europe temporarily); and 'migrants', i.e. adult children with parents who live in the same European country (we excluded 40 cases in which the child lived in a different European country from their parents). Based on the child's country of birth, the latter group was further divided into dyads with 'first-generation migrants' (the focus child was born in Turkey, and mostly belongs to the so-called 1.5-generation, members of which migrated with their parents during their childhood or youth) and dyads with European-born second-generation children whose parents were first-generation migrants . Excluding those cases which had missing values on the solidarity variables of interest (e.g. due to parents' death), lived co-residentially or had incomplete migration-related information, we achieved a final sample of N = 2282.

Method
The empirical analysis was carried out in three stages. First, we developed a typology of intergenerational relationships, based on three solidarity dimensions (see below). To identify the number and nature of relatively homogeneous groups among the dyads, we conducted latent class analysis (LCA) using Latent Gold 4.0 (Vermunt and Magidson 2005). Selection of the optimal number of classes was based on several statistical fit indices, including the likelihood ratio chi-square (L 2 ), Bayesian information criterion (BIC) and Akaike information criterion (AIC). L 2 measures how well a latent class model fits the observed data; a non-significant L 2 indicates a good fit. BIC and AIC are then used to compare competing models; lower observed values indicate a better fit. Of these fit statistics, BIC has been identified as the most reliable (Williams and Kibowski 2016). Preference is given to the best-fitting model with the fewest classes. We also estimated separate latent class models for the different groups (stayer, firstand second-generation migrant, and transnational) to determine the robustness of the selected latent class model. Furthermore, taking into account the nested structure of the data, we estimated latent class models with one randomly selected child-parent dyad per lineage. The results were robust, yielding the same class outcomes (results available upon request). Second, the hypotheses described above were tested by examining the difference in distribution of the derived types between adult children and their parents across the four interest groups. Finally, we considered the associations between the types and socio-demographic characteristics of children and parents to test whether differences in the distribution of the in-between types were mediated or suppressed by differential composition. In these analyses, we accounted for the nested data structure by using the cluster command in Stata. We performed multinomial logit regression analysis. For easier interpretation, we estimated and present marginal effects, reporting the change in probability by one unit change in an explanatory variable when all other variables are held constant at sample mean values. For each variable, the marginal effects across the models add up to zero. We dealt with missing values (which ranged between 1% and 9% for single variables but would have reduced the overall data with listwise deletion by 16%) by estimating our models on multiple imputed data. The results obtained via listwise deletion and with imputed data were largely similar (results available upon request). Below we present the outcomes based on data which we imputed with chained equations using Stata 14.1's mi impute suite, and including all variables (White, Royston, and Wood 2011). We imputed complete sets of responses for 20 imputed data sets and calculated our models in Stata with the mi estimate and mimrgns commands.

Measurements
The survey included seven items on relationships of children with their parents, corresponding to three dimensions of intergenerational solidarity: contact frequency, strength of solidarity obligations and support exchange. All refer to forms of intergenerational solidarity that do not require geographical proximity. The children were asked to supply answers for both parents if both were alive and living together, for the father or mother if only one parent was alive, or for the mother if both parents were alive but separated. Table 1 presents the indicators used in the analyses. The socio-demographic factors included the age (18-25, 26-35, 36-45 and 46 or older) and gender of the child. Children's resources were expressed by educational level (primary, secondary or higher; values 1-3) and activity status (five categories: employed, unemployed, housework, student and other). Children's needs and their support potentials were gauged by health problems, indicated by the onset of an illness/disability and whether this limited their daily life. Family structure signals availability of alternative relationships, support potential and available time, and was measured by partner status (three categories: married/cohabiting; single, never married; single, separated/divorced/ widowed) and having children (1) or not (0). Non-family friends were included as another indicator of the availability of alternative relationships. Respondents were asked to give the number of people who are important to them and whom they feel close to, excluding parents, partner and children. Four categories were distinguished: no friends (0), up to three friends (1), 4-10 friends (2) and more than 10 friends (3). The cultural background of the child was identified with information on the importance of religion, based on the question 'How important is religion to the way you live your life?', with response options 'totally unimportant' (1) to 'very important' (5). Extent of agreement with egalitarian gender values was based on two items: 'A university education is more important for a boy than for a girl' and 'On the whole, men make better business executives than women', with response categories ranging from 'strongly agree' (1) to 'strongly disagree' (5): scores were summed with a higher score indicating more egalitarian views. Agreement with values supporting independence from parents in life-course decisions was based on two items: 'Parents should have the final say in their children's choice of a partner' and 'Whatever their parents think, children should be able to choose their own educational path', with response options 'strongly agree' (1) to 'strongly disagree' (5). Scores for the second item were reversed and then both were summed, with a higher score indicating support for independence from parents. For the parents, we included information on who was still alive (mother only, father only or both) and number of children (number of siblings of the child) as a continuous variable. 1 Table 2 presents the descriptive statistics for the study sample. Note the higher proportion of women among stayers, and the younger age structure of European-born second-generation children. Educational level also varies significantly across the groups, with a higher level of education in the second generation and the highest proportion of lower-educated among the stayers. Fewer stayers than migrants were employed, with almost half of the stayers doing housework. Health problems were least frequent among the comparatively young group of second-generation children, the group with the highest proportion of childlessness. Migrants in transnational dyads were most frequently married. Migrants had fewer friends than stayers and the proportion of persons without any friends was lowest among stayers. In cultural terms, religiosity was highest and support for egalitarian gender values was weakest among stayers. No difference exists for independence values. Few migrants of the second generation had already lost one parent. The average number of children (i.e. siblings) was significantly lower for the European-born.

Results
The analyses yielded an optimal number of four child-parent relationships (see Table 3). 2 The four types are not spread evenly across the sample. Most common is Type 1, which we called 'full solidarity' (see Table 4). More than a third of all child-parent relationships (35%) belong to this solidarity type of high cohesion. The coefficients in the columns of Table 4 represent the conditional probability that a relationship is characterised by the indicator of solidarity (given that it belongs to that type). For Type 1 there was an 86% probability of more than weekly contact and almost equally high probabilities on all other indicators of solidarity. Whereas this solidarity type reflects high cohesion on all dimensions considered, the two intermediate Types 2 and 3 indicate only partially cohesive patterns. Adult children and their parents in Type 2 show a high probability of frequent contact. Moreover, probabilities of strong family obligation norms and advice support exchange are high too, yet members of this type are significantly less likely to exchange material support. Hence we assigned the label 'advice-oriented' to Type 2, comprising 16% of the cases. Adult children and their parents in Type 3, containing 23% of the sample, feature a high probability of material support in both directions with a lower probability of frequent contact and strong norms and specifically low probabilities of advice exchange. We assigned the label 'material-oriented' to Type 3. Type 4 stood out from the others by having low probabilities on all indicators of solidarity, so we labelled it 'autonomous'; 26% of the adult children and their parents belong to this solidarity type of low cohesion. Table 5 presents the distribution of the four solidarity types by migrant status. The proportion of the full-solidarity type (Type 1) is most prevalent among first-generation migrants' and European-born second-generation children's dyads, and less frequent in  transnational dyads when compared to stayers. On the other side of the spectrum, the autonomous type (Type 4) is far more common among transnational child-parent relationships and much rarer among firstand second-generation children and their migrant parents than among stayers. The advice-oriented type (Type 2), characterised by high probabilities of advice support in both directions along with high probabilities of frequent contact and strong family obligation norms but low probabilities of material support exchange, is most widespread among stayer dyads. The material-oriented type (Type 3) that displayed high probabilities of both upward and downward financial support is most common in child-parent relationships of second-generation children and in transnational dyads. Overall, these findings give support to the hypotheses H1b and H2a, but not to H1a and H2b. We expected transnational child-parent relationships to be characterised by lower family cohesion due to geographical distance, yet more upward material support due to the role of remittances in transnational settings, and indeed do find a higher prevalence of the autonomous and material-oriented types in those dyads than among stayers (H1b). In line with the hypothesis H2a, formulated with reference to minority-context approaches, family cohesion is higher in a migration situation and few differences exist between the migrant generations. The results of the multinomial logit analysis are presented in Table 6. As regards the main effects of the control variables, children's sex and age made little difference to the distribution of solidarity types, with one exception: daughters were more likely to belong to the advice-oriented type than sons, and all older children, but specifically those aged 36-45, were more likely to have an autonomous type of relationship with their parents than the youngest age group. Children with a secondary education were more likely to have the most cohesive relationships (full solidarity) and particularly less likely to belong to the material-oriented type than the lower-educated, while there were no differences in the likelihood of being part of a specific type between lower-and higher-educated children. Being a student decreased a child's likelihood of belonging to either the full-solidarity or autonomous type, but increased the likelihood of having an advice-oriented type of relationship with her or his parents. Having an illness or disability that limits daily life and being single after a relationship also increased the likelihood of having an advice-oriented type of relationship with one's parents. Having friends was only partly associated with relationship types. Compared to children with no friends, those with four or more friends were more likely to belong to the advice-oriented type, and less to belong to the autonomous type. Cultural-religious patterns made a difference, as higher religiosity was positively associated with the full-solidarity type and negatively with the material-oriented type. With more egalitarian gender attitudes, the likelihood of belonging to the material-oriented type was greater but the likelihood of having an autonomous type of relationship with parents was less. Adult children who strongly agreed with values supporting independence from parents in life-course decisions had a lower likelihood of belonging to the material-oriented type. Finally, compared with children both of whose parents were alive, those without a mother were more likely to be part of the autonomous type and less likely to be part of the full-solidarity type. For the family type, the number of children of the parents (i.e. the number of siblings) made no difference.
The most important finding is that the observed differences in the distribution of the four solidarity types by migrant status was not altered when the socio-demographic characteristics were accounted for. Compared with stayers, first-generation migrant and European-born second-generation children were more likely to display the most cohesive type of relationship with their migrant parents and were less likely to be part of the autonomous type. At the same time, the material-oriented type occurred more often and the advice-oriented type less often among the migrants. Migrant children in transnational relationships had a greater likelihood of being part of the autonomous solidarity type. However, no significant difference remained between transnational and stayers' relationships with regard to the full solidarity type. The former had a higher likelihood of displaying the material-oriented type than their stayer compatriots.

Conclusion and discussion
This research used the 2000 Families Study to broaden knowledge on intergenerational relationships among Turkish migrants. It is the first study to consider multiple domains of intergenerational solidarity with reference to parent-child relationships in the country of origin. In this way, we aimed to contribute to a better understanding of the impact of origin-culture versus migration on intergenerational solidarity. We also included members of transnational families to account for the specific contexts in which many migrants manage their family affairs (Karpinska and Dykstra 2019).
Our analyses of the associations between the solidarity dimensions of contact, norms, advice and material support exchange revealed four types of solidarity, namely the 'fullsolidarity' and 'autonomous' types and the two intermediate support types, 'adviceoriented' and 'material-oriented'. The full-solidarity and the autonomous types correspond with the tight-knit and the detached types in the typology of Silverstein and Bengtson (1997) and the full-interdependence and independence types found by Rooyackers, de Valk, and Merz (2014). Here adult non-co-resident children and their parents engage in either all or none of the contact and support dimensions in question, while both types are characterised by strong normative support to family obligations. Although these two types are common in solidarity typologies (Guo, Chi, and Silverstein 2012), not all analyses actually identify the full-solidarity type (Dykstra and Fokkema 2011). By identifying this solidarity type, our data underline the important role of the family previously shown in research on Turkish migrants. The full-solidarity type indeed is the most prevalent, comprising more than a third of the sample. That said, the finding that more than a quarter of all intergenerational relationships belong to the autonomous type provides a differentiated picture of the high heterogeneity of this ethnic group, which is often overlooked.
Typologies tend to vary for the intermediate types (Guo, Chi, and Silverstein 2012). While others differentiated between the directions of support (upward versus downward: e.g. Dykstra and Fokkema 2011;Rooyackers, de Valk, and Merz 2014), in our case the kind of support (advice versus material) was more decisive. Our findings suggest an exchange pattern of 'immediate reciprocity'; support is exchanged between children and parents in both directions. This pattern of mutuality contradicts the traditional bottom-up direction of support to the (needy) elderly implied in the patrilineal regime of descent which is expressed in parents' utilitarian expectations of their children (Nauck and Klaus 2008). This outcome may be a sign of (family-related) ongoing changes (Kagitçibasi and Ataca 2005;Kagitçibasi, Ataca, and Diri 2010) which highlight (new) patterns of parental financial support to their children .
Our main focus, however, was on the differences between the intergenerational relationships of stayers and migrants. We developed two opposing sets of hypotheses on the effects of context change (H1) and minority experiences following migration (H2). We found that migrant children (regardless of whether they were born in Turkey or in Europe) with their migrant parents are more likely to belong to the full-solidarity type and less likely to belong to the autonomous type than stayers in Turkey, even after controlling for compositional differences. The material-oriented support type is also more common among the migrant group, whereas the advice-oriented type is more likely among non-migrants. These findings contradict common assumptions of more intense acculturation and situational adaptation in the second generation (H1a), but they support evidence for persistently strong family solidarity (H2a) and thereby confirm previous results (Carnein and Baykara-Krumme 2013) and challenge other findings (Arends-Tóth and van de Vijver 2008).
Overall, the significant differences we find in the comparison with non-migrants in the country of origin suggest that, for the explanation of the specific child-parent relationships among migrant families in Europe, migration-specific rather than origin-culture-specific mechanisms are at work. Plain references to 'traditional (Turkish) family culture' to explain remaining differences between migrants and natives in the destination context are misleading (Foner 1997;van Hook and Glick 2007). What we observe in migrant families in Europe seems to be a specific pattern of adjustment in the migration and minority context, not in the sense of acculturation to the Western (welfare) context of weakening intergenerational solidarity but in terms of an intensification and strengthening of strong intergenerational ties.
Although our findings support the argument of increased family cohesion in migration, we were not able to adequately test the proposed underlying migration-specific mechanisms by reference to experiences of exclusion, lack of alternative social ties and intergenerational transmission of family solidarity. Regarding the second mechanism, we were able to consider an indicator measuring non-family ties which showed some effects, but did not substantially mediate the impact of migrant status. We cannot, however, rule out that the earlier lack of alternative ties caused children and parents to become close and supportive of one another. Unravelling the underlying causes for the observed patterns seems a challenging task for further research, and more than the three mentioned explanations may be relevant. For instance, one major unsolved factor in migration research, specifically in studies like ours which compare migrants and stayers, is selectivity. We cannot rule out that migrants may be selected on the basis of family-related behaviours or other unobserved variables, i.e. the observed divergences may have prevailed before the migration of the parents instead of evolving in the course of migration. A further explanation considers change in Turkey. Our findings that the autonomous and the advice-oriented solidarity types are common among stayers seem to relate well to research in Turkey, according to which social change promotes more autonomous relationships and less material support exchange, but strong emotional bonds with related advice exchange (Kagitçibasi and Ataca 2005;Nauck and Klaus 2008). Migrants, by contrast, may use 'traditional' accounts of strong family solidarity in the minority context and develop 'new' patterns of non-co-resident family reliance (Portes and Rumbaut 2001). This may involve the specific adaptation to European welfare regimes. In contrast to our proposition H1a, the higher prevalence of the full-solidarity and material-oriented types among migrant families may result from a crowding-in process in the context of better public welfare provisions in Europe (Albertini and Kohli 2013;Brandt and Szydlik 2008). Such provisions may enable migrants to actually practice the kinds of solidarity that they adhere to normatively or due to specific needs in the family, thereby complementing existing and stimulating new forms of family solidarity. Finally, as we focus on non-co-resident child-parent relationships, we have taken no account of the cohabiting dyads (de Valk and Bordone 2019). Possible different patterns of cohabiting and leaving the parental home (including moving in with in-laws) between Turks in Turkey and Turkish migrants in Europe may account for the observed differences.
Transnational relationships display patterns divergent from those among stayers. In line with our expectations (H1b), the autonomous type and the material-oriented type are more common in transnational contexts. In terms of family cohesion, transnational dyads are in a disadvantaged position. Our findings thus provide an alternative to the often rather optimistic pictures that other, specifically qualitative studies draw of transnational families (Baldassar and Merla 2014). Given that none of the considered domains of solidarity requires geographical proximity, the picture would be even less rosy if hands-on-care was included. The analyses suggest that it is not the migration context as such but the specific conditions, notably the larger geographical distance, which are influential. Again, we must keep in mind selectivity effects: migration of children may be a result rather than a cause of distanced relationships. Moreover, even if advice support could be exchanged at a distance, children might refrain from it in order to avoid parental worries (Senyürekli and Detzner 2008). In any case, any firm conclusions on how 'harmful' transnational relationships are would be premature, as others (e.g. siblings in Turkey or friends and relatives in Europe compensating for the parental role) may fill the gap.
The limitations of this study include the cross-sectional nature of the data. The snapshot precludes us from drawing any conclusions about the stability of the differences in family solidarity types over the life course. The solidarity types remain robust after controlling for many needs and resources of the children, but we cannot rule out composition effects due to diverging parental characteristics (Dykstra and Fokkema 2011;Szydlik 2008). Needs, resources and attitudes of parents were hardly acknowledged in the models due to lack of data. The lack of information about the parents is an important drawback, as we know that (elderly) migrant Turks belong to the most disadvantaged groups in terms of socio-economic resources and healthat least when compared to natives in the host country (Fokkema and Naderi 2013; Liversage and Jakobsen 2016). We will also need more information to adequately test the impact of migration and minority-related mechanisms in order to better explain the observed patterns. This might involve an extension to other migrant groups and origin and destination contexts. It would certainly help further theorisation of the impact of migration on family solidarity, which seems one of the most urgent tasks in this field of research. Notes 1. We also considered parental health and activity status based on interviews with parents which were available for about 75% of all dyads. However, as only one parent was interviewed (the parent belonging to the lineage), this information is far from optimal. Moreover, these factors did not show any significant direct effects and hardly mediated the impact of the migrant status.
2. Moving from a one-to a three-class model, the L 2 decreased but was still significant; in the four-class model the L 2 corresponded with a borderline significant p-value of .05. Although a five-class model showed a p-value far above .05 (hence the L 2 and degrees of freedom were closer in value) and a lower AIC value than the four-class model, the latter had the lowest BIC value. Moreover, the results of the separate subgroup analyses revealed the superiority of the four-class model with the same general typology (results not shown; available upon request). Based on the L 2 , BIC and AIC statistics, this model was the best-fitting for all subgroups except first-generation migrants, for which selection of a three-latent class model is optimal.