Abstract
Social network data from class-based interviews of N = 257 10-year-old students in N = 4382 directed dyads show that the propensity for being friends is moderately affected by ethnic heterogeneity. However, the likelihood of acquaintanceship among children’s parents is strongly reduced in ethnically heterogeneous dyads. This has been referred to as ‘intergenerational openness’ of networks in the work of J.S. Coleman. It will be shown that intergenerational openness has a strong negative association with mutual invitations to birthday parties, which are assumed to play an important role in the process of cross-generational social integration. Accordingly, intergenerational openness indicates a ‘gap in social integration’ between generations and corresponds negatively with the social integration of immigrant children.
Introduction
Second generation immigrants grow up in a host country’s educational institutions and can establish social ties with natives right from the start. But results from social network analysis in the USA still reveal a strong ethnic segregation of friendship ties. Second generation Asian immigrants tend towards a pattern in line with segmented assimilation theory (Zhou, 1997): they form friendships with native whites, but have even closer ties with friends of their own ethnic group, indicating the importance of ethnic group solidarity (Quillian and Campbell, 2003). While ethnic segregation might still exist in networks of 10-year-old children, the question remains to what degree their parents’ social ties depend on ethnic origin. If there is an increase in assimilation over generations, the likelihood that children’s friendships are embedded in intergenerationally ‘open’ ties (Coleman, 1990) – meaning that children’s parents do not know or not even occasionally phone or meet up with each other – will be higher in ethnically heterogeneous dyads.
The current study investigates three dimensions of social assimilation among Turkish immigrants, who constitute the largest immigrant group in Germany. Using social network data of 10-year-old children in school classes in the cities of Bremen and Bremerhaven, social assimilation into network ties is analysed in both the children’s and the parents’ generation. At the children’s level, the study analyses determinants of friendship ties and birthday party invitations. At the parents’ level, it focuses on acquaintanceship ties. It examines whether intergenerational openness corresponds negatively with invitations to other children’s birthday parties. Birthday parties will be described as salient events in the lives of children which can, according to existing research, affect the process of social integration of children as well as that of the social integration of their parents.
The concepts of social assimilation and intergenerational closure are briefly introduced, and existing ethnographic research on intergenerational networks of visits at children’s birthday parties is discussed. The next section describes the collection of network data, instruments, measurements and statistical methods. For each dyad, that is for each pair of actors in the network, the data used in this study allow for the estimation of covariate effects on the presence of a link between two actors, e.g. if ego went to alter’s birthday, if there are network ties between parents and if children are friends. Empirical results of these estimations are then presented.
Friendship Networks of Children and ‘Intergenerational Openness’
The concept of social assimilation simply means that the ethnic composition of network ties does not differ substantially between immigrants and natives (Esser, 2004; Gordon, 1964). Friendship networks consisting of both immigrants and natives might encourage immigrant children to improve their language skills and can offer productive learning environments for adopting a host country’s behavioral routines and daily practices (Martinovic et al., 2008). But not all friendships provide the same level of social capital. Coleman’s (1990) analysis of intergenerational closure of networks, which refers to social integration in general and not just to the social assimilation of immigrants, has revealed that networks provide less social capital if they are intergenerationally open. Intergenerational closure, in contrast, means that the acquaintanceship or friendship of children’s parents facilitates parents’ mutual agreement on social norms and social control over children. In the intergenerationally closed network {A;a;B;b} in Figure 1, both parents {A;B} and children {a;b} are connected to each other. Parents can exchange information about their children’s actions and consult with each other on the validity of norms. In contrast, the network {C;c;D;d} indicates intergenerational openness since vertices (or ‘nodes’) in the children’s generation are connected, whereas vertices in the parental generation are not. In intergenerationally open networks parents cannot coordinate the enforcement of social norms.
James Coleman’s concept of intergenerationally open and closed social networks: (1990: 319)
The concept of open and closed networks is based on Coleman’s (1990: 57) idea of rational actors holding the rights to carry out certain actions in a system of exchange of mutual obligations, and also transferring authority over their actions to significant others. Since many actions impose externalities on others, one benefits from refraining from such actions if others do so as well. As Coleman argues, it is only when parents are linked to each other that they can mutually exchange control over the right to refrain from imposing sanctions on their children. However, assimilation theory suggests that second generation immigrant children are better assimilated with regard to education and social networks than their parents. The advantage of younger generations results from children’s early socialization in the host country. In the parental generation, incomplete acculturation often accompanies low language proficiency and distance in habitus and culture. Additionally, since many immigration processes are channeled by chain migration, ethnic capital is more important in the first generation, and this usually corresponds with close intra-ethnic ties. Due to this ‘intergenerational gap’ in social assimilation, intergenerational openness will be more prevalent in ethnically heterogeneous dyads, that is, in dyads composed of immigrant families on the one hand and native families on the other. This holds all the more true if some immigrant groups retain strong social bonds to their co-ethnics, as highlighted by segmented assimilation theory (Zhou, 1997). The intergenerational closure or openness of children’s friendship dyads is thus expected to depend on ego’s and alter’s ethnic backgrounds.
Why should intergenerational closure be important for the integration of immigrant children? As Wilson (1987: 56) argued, adult role models in the neighborhood provide a basis for collective socialization, which is particularly important for children who grow up in poor female-headed families (Jencks and Mayer, 1990: 114f; Pong and Hao, 2007). Representing daily work and career ambitions, these ‘external’ role models can partially compensate for the absence of role models in disrupted families. ‘External’ role models can represent the habitus of highly qualified service sector occupations where parents of immigrant children are under-represented (Seibert and Solga, 2005). During visits to their native friends’ households, immigrant children often get exposed to host country habitus forms, including correctly spoken native language and upper and middle class behavioral standards. In the left column of Figure 1, the dashed line represents the effect of adult role models on acculturation if A is a native parent and b is an immigrant child. Following Jencks and Mayer (1990) and Pong and Hao (2007), it is essential whether immigrant children have contact with their native friends’ parents or not. Although inter-ethnic adult role models cannot be measured directly in the following analysis, visiting native friends’ households to join their birthday parties is an approximate indicator since children’s birthday parties are, in most cases, organized by the hosts’ parents, who are usually present during the event. But while the expected utility of ties to natives seems to be an incentive for establishing them, it is known at the same time that ‘Homophily in race and ethnicity creates the strongest divides in our personal environments’ (McPherson et al., 2001: 415; see also Martinovic et al., 2008). Regarding children’s parents who do not regularly meet others at focal points, such as their children’s classrooms, it is thus an empirical question how strong the ethnic segregation of ties in networks is. But why are birthday parties important for the social integration of both children and their parents?
Birthday Parties as a ‘Potlatch’ and Their Impact on the Integration of Children and Parents
As shown by Clarke’s analysis of children’s birthday parties from an ethnographic and feminist perspective, the exchange of gifts ‘. . . is a highly symbolic event of conspicuous consumption, in which those that have received goods and gifts strive to give more in order to preserve and increase their social power and standing’ (2007: 270). Birthday parties comprise practices of social exchange akin to the famous ‘potlatch’ in classical anthropology (Mauss, 1990). The central activity in the potlatch ritual was the giving of luxurious presents, not primarily to satisfy material needs, but rather to maintain the community’s prestige (1990: 90). In his introduction to ‘The Gift’, Mauss (1990) points to modern societies where Christmas and birthday presents, weddings and any kind of invitation imply obligations, and the gift’s recipient feels strongly compelled to reciprocate. There is no rational explanation for the value of gifts and the luxuriousness of children’s birthday parties. Many mothers complain about the time and efforts involved in arranging birthday parties and arranging presents, but join these gift-economic circles nevertheless. Clarke explains this by their intention to ‘. . . ‘fit-in’ to the local modes of mothering’ (2007: 272). If there is actually a difference in culture and behavioral standards between natives and immigrants – as assumed by assimilation theory (Alba, 2008; Esser, 2004) – being integrated in these exchange communities will have an enormous impact on the acculturation of immigrant children and of their parents.
As argued above, the second generation should be better socially assimilated than the first (Alba, 2008). If that were so, then school children would pay rather little attention to ethnic origin as a criterion for friendship choice, while parents of immigrants would remain more segregated and maintain stronger ties to co-ethnic immigrants. Results for Germany indicate that immigrants from Turkey, who constitute the country’s largest immigrant group, have comparatively high rates of intra-ethnic friendship ties (Haug, 2003: 726). Since social cohesion of immigrants is comparatively high in the Turkish group, intergenerational closure might be even more prevalent in ties of Turkish immigrants than for native Germans.
In summary, these theoretical considerations lead to the following line of argument: while networks of children are only moderately ethnically segregated, segregation between Turkish immigrants and native Germans should be far more pronounced in the parental generation. Segregation at the parental level, in turn, corresponds with reduced chances of attending birthday parties because sending children to these parties usually requires parental involvement and at least some degree of trust among the families – which, in turn, depends on social ties between them. While there is a rather negligible degree of social segregation of friendships, intergenerational openness accompanies a reduced quality of friendships. This holds true even for the subset of dyads in which children themselves report a friendship tie. However, a strict causal statement that parental contact increases the likelihood of attending birthday parties would neglect the reversed relationship – that regular parental contact can be an additional outcome of children’s birthday parties – which might actually be true in some cases. This suggests that there is a negative association between openness and birthday parties, rather than a clear causal impact.
If immigrant children have only limited opportunities to meet their native friends’ parents, there will be minimal learning from adult role models. Although the socializing effect of adult role models cannot be tested directly in this article, it can be taken as given according to previous research (Jencks and Mayer, 1990; Pong and Hao, 2007; Wilson, 1987). Thus, the empirical analysis focuses on the following hypotheses:
H1: Friendship choice of primary school children depends only moderately on ethnic homo- or heterogeneity.
H2: Network ties between migrants and natives have a higher likelihood of being intergenerationally open, especially if immigrants are of Turkish origin.
H3: The prevalence of intergenerational closure is even higher in dyads of two Turkish students than in dyads of two German ones.
H4: Intergenerational openness, in turn, corresponds with a lower likelihood of mutual visits to birthday parties because, in most cases, these visits require parental involvement and trust.
Even if two children consider each other as friends, the intergenerational openness of their network might coincide with a reduced quality of their friendship. This should be indicated by a negative effect of intergenerational openness on invitations to birthday parties, which reduces the opportunity for social learning from native role models – although causality is surely not as unambiguous as the following regression model suggests.
H5: Within friendship dyads, too, intergenerational openness corresponds with a lower likelihood of mutual visits to birthday parties.
Since native Germans and immigrants differ with regard to their socio-economic backgrounds, these hypotheses will be tested by controlling for indicators of social homophily, including spatial proximity, similarity in parents’ cultural capital (number of books at home), and the living situation at home (child has own room at home).
Data, Methods and Measurements
Social network analyses have only recently begun to be used in immigrant integration research (Martinovic et al., 2008; Mouw and Entwisle, 2006; Quillian and Campbell, 2003; Vermeij et al., 2009). The present study is based on a school survey conducted in the federal city state of Bremen, Germany, in the spring of 2008. The data included information on 263 elementary school students in 15 school classes. On average, 3.5 of 21 students were missing, either because of unattainable parental permission, illness, or other reasons. Due to item non-response, only network data on 257 students could be used in the empirical analysis, resulting in 4382 directed dyads as units of analysis. The questionnaire focused on family situation, migration background, leisure time activities, school attachment and achievement.
A visible number was placed on each student’s seat in the classroom. By entering their own numbers and the numbers of their fellow classmates into the questionnaire, students could report all their network links for each of the 11 network-related items. Among other things, these dimensions covered whom they considered as friends, whom they had as guests at their birthday party, whose parties they were invited to, and whose parents knew each other so well that they sometimes met or spoke on the phone. Considering visits to birthday parties as an objective event, the reliability of the network generator could be assessed by comparing ego’s information on who visited him at his birthday party with alter’s information on whose birthday he or she attended. This procedure yields a Cohen’s Kappa of 0.735 and an accordance rate of 92.3 percent, which indicates a good inter-rater reliability.
Recently, the use of cognitive maps was suggested as a method to avoid incomplete networks. This approach ‘. . . relies on a subset of actors to provide information about subgroups in the network system’ (Neal, 2008: 143). However, dimensions like interactions between children’s parents are invisible for many other students and can hardly be measured by cognitive maps. We thus opted for the traditional way of nominating actors who are present during the interview. Previous research has shown that most network measures are robust under small amounts of error; that is, 10 percent or less (Borgatti et al., 2006). Even though the average rate of absence in our study is 16.7 percent, the analysis can be justified by the specific approach of estimating odds of friendships from dyadic data conditional on covariates.
Subsequently, the effects on three dependent variables are estimated. First, we asked the children to report the number of classmates they consider as friends, without restricting the total amount of nominations. Second, children reported the number of classmates whose parents are acquainted with their own parents, so that they at least occasionally phone or meet with each other, which is the indicator of intergenerationally open or closed networks. Third, children reported the number of classmates who had invited them to a birthday party.
Mouw and Entwisle (2006: 397) argued that children living close to each other meet at focal points in their neighborhood, like bus stops, and are thus more likely to be friends. In our data, the spatial proximity of places of residence was captured by the variable alter lives close to ego, consisting of one item in the network generator: ‘Who lives so close to your home so that you can walk to him or her within a few minutes?’
Ethnic background was measured by the mother’s and father’s countries of origin. Children are defined as ‘German’ and ‘Turkish’ if both biological parents are either of German or of Turkish origin. With a share of 9 percent, the Turkish group is by far the largest immigrant group in our sample. All other immigrant groups, as well as children of ethnically mixed couples, are defined as ‘other migrants’, which resulted in 154 native Germans, 23 Turks and 80 other migrants. Due to the small sample size, only such a rough measure was available for this article.
To control for cultural capital (Bourdieu, 1986) in ego’s family, homophily: number of books was included in the model. Therefore, absolute differences in the number of books between ego and alter were multiplied by −1. The wording of the item was ‘How many books do you have at home?’ and response categories were ‘1. none, or just a few (0–10)’, ‘2. one shelf (11–25)’, ‘3. one rack (26–100)’, ‘4. two racks (101–200)’, and ‘5. three or more racks (201 and more)’. Although this is a categorical measurement, it was considered as a metric variable because of its approximate normal distribution. Without such a control, ethnic homophily could be a spurious effect of status homophily, especially since many Turkish and other immigrant children have low-educated parents. Whether ego has his or her own room was controlled, which is also an indicator of living conditions.
The percentage of Turkish and other immigrant children at the class level is of importance because it indicates the context of inter-ethnic friendships. Table 1 shows the descriptive statistics.
Descriptive statistics, directed graphs, N = 4382
Notes: * both actors nominate each other as friends
In social network data, observations are not independent of each other. The adage ‘friends of my friends are my friends’ expresses the dependence on ties in transitive triads (Snijders et al., 2006). In the following analysis, so called p* models (Robins et al., 2007) were estimated to predict the likelihood of the presence of a tie conditional upon all other ties in a network. In simple terms, p* models are logistic regressions, controlling for the embeddedness of a dyad in the whole network. First, a network characteristic, like the number of transitive triads, is calculated, conditional upon the forced presence of a specific tie xij. Secondly, this characteristic is calculated again, conditional upon the forced absence of this tie. Intuitively, the difference in both values can be perceived as the contribution of each tie xij to the respective network characteristic, and it accounts for the statistical dependence of ties. By using PREPSTAR software (Crouch and Wasserman, 1997), these values were computed, added to the data, and controlled in the logistic regression models. Currently, p* models are still under development. As Lubbers and Snijders (2007: 506) conclude from their comparison of models, ‘. . . for testing covariates, it is important to control for structural effects, but the precise specification of the structural part ( . . . ) seems to matter less’. In the following analyses, mutuality, transitive triads, out-stars and in-stars are controlled in order to capture the statistical non-independence. The latter represent changes in expansiveness and popularity if a respective tie is present or absent (Robins et al., 2007: 182).
Dyads of potential ties will be regarded as directed graphs (digraphs), meaning that ego’s nomination of a friendship must not be reciprocated by alter. This occurs quite often in the data: for instance, the average percentage of accordance in friendship nomination or non-nomination within classes is 83.2 percent. Moreover, attendance of other children’s birthday parties was predicted by estimating an additional error component at the dyad level. This error component is assumed to capture the dependence of actors in networks in terms of mutuality and other unobserved characteristics (Snijders and Bosker, 1999).
Results
Table 2 shows the odds ratios of a link in each dyad. For each dependent variable, the first model is a simple logistic regression. The second model also considers the share of immigrants and the p* parameters as structural properties of the class networks. For each model, the McKelvey and Zavoina R2 (Snijders and Bosker, 1999: 225) was computed. Below, the interpretation of effects is focused on the p* models. Since these are conditional logistic models, results are interpreted as odds ratios. Models 1 and 2 estimate the effects on children’s friendship ties, and Models 3 and 4 show the effects on acquaintance ties among their parents. Models 5 and 6 show the effects on children’s birthday party invitations. Finally, Models 7 and 8 replicate Models 4 and 6 for the subsample of friendship dyads, completed by the effect of intergenerational closure on birthday invitations in Model 9. 1
Determinants of directed graphs. Logistic models for friendships, intergenerational closure, and visits to birthday parties
Notes: N(observations) = 4382, N(dyads) = 2191, N(students) = 257, N(classes) = 15
friendships only: N(observations) = 1326, N(dyads) = 848, N(students) = 251, N(classes) = 15 + p<=.10; *p<=.05; **p<=.01; ***p<=.001
In all p* models presented in Table 2, the spatial proximity of places of residence has a positive effect on network integration. If alter lives close to ego, the odds of friendship increase by a factor of 3.5 and the odds of contact between parents increase by a factor of 6. Also quite strong is the effect on the odds of visits to birthday parties, which increase by a factor of 4.2 if alter lives close to ego. Moreover, with respect to intergenerational closure, network integration of ego increases with higher homophily of cultural capital in the ‘objectified state’ (Bourdieu, 1986) – measured as the similarity in the numbers of books in ego’s and alter’s households. Regarding children’s gender, it is not surprising that, in a sample of fourth graders, mixed-gender dyads have lower odds of being friends. Regarding ethnic background, we find significant differences in friendship choice between German-German dyads and other combinations of ethnic background. Hence, there is some degree of friendship segregation, but friendship choice does not seem to be strongly related to ethnicity or migration background (H1).
In contrast, both intergenerational closure and birthday invitations are strongly reduced in ethnically mixed dyads. Compared with the reference group of all-German dyads at the children’s level, the odds of parental contact are much lower in Turkish-German dyads (Model 4). We observed in Models 1 and 2 that children’s friendship choice is rather moderately affected by ethnic origin, and, at the same time, we find a strong effect of ethnicity on intergenerational closure in Model 4. Hence, intergenerational openness occurs mainly in ethnically heterogeneous dyads (H2). In addition, the positive effect of Turkish-Turkish dyads is strong (factor of 3.6) and significant, relative to the reference group of all-German parents. This result hints at a highly cohesive Turkish community: parents of Turkish children are better connected among each other than parents of native German children (H3). Following Coleman’s argument, the decisive point is that most immigrant children seem to have normal ties to their peers, while immigrants in the parental generation are rarely part of host country networks. Even though the effect of role models cannot be directly measured in this study, there are strong arguments for how important these role models are – especially when immigrant children do not have the opportunity to learn specific habitus forms within their own families (Jencks and Mayer, 1990: 114f; Leventhal and Brooks-Gunn, 2000; Pong and Hao, 2007; Wilson, 1987). Under the assumption that cultural differences between home and host society actually exist, visiting their native friends at home can surely serve as a formative experience for immigrant children. At the very least, immigrant children might notice differences from their families, which can be a stimulus for assimilative efforts in later life.
Model 6 presents the effects on ego’s participation at alter’s birthday party. Compared with the reference group, we find significantly lower odds in German=>Turkish, Turkish=>German and ‘other mig.’=>Turkish dyads. Results from a larger sample would possibly indicate that Turkish children generally have lower odds of receiving invitations to birthday parties, simply because such invitations are less common in the Turkish community. Nonetheless, there is a large and positive effect of intergenerational closure on the odds of joining alter’s birthday (increase by a factor of 7.5). Intergenerational closure, in turn, strongly depends on the ethnic composition of dyads, as we have seen in Model 4. It has been argued in the theoretical section that, conditional upon being friends, the quality of friendship corresponds with intergenerational closure of networks, since birthday parties and similar activities usually require parental commitment – which is less likely in the case of intergenerational openness.
The last three Models, 7 to 9, are based on a subsample of friendship dyads. Model 7 shows the effects on intergenerational closure and the two remaining models show the effects on ego having attended alter’s birthday party. Again, we find strong, significant and negative effects of ethnically heterogeneous friendships on intergenerational closure in Model 7. Even if students are multi-ethnic friends, it is rather unlikely for their parents to have contact with each other. This result is in line with the notion of an intergenerational gap in social assimilation. Furthermore, Model 8 shows strong, significant and negative effects of ethnically heterogeneous dyads on the odds of birthday party invitations. Even if the likelihood of friendship choice depends only moderately on ethnic origin, the quality of friendships depends strongly on it – at least with respect to birthday party invitations. Many interethnic friends do not share the important experience of inviting each other to birthday parties. Since these events are important issues in conversations among students during school breaks, for instance, they are obviously a relevant component of social integration. From the children’s point of view, it makes a difference whether or not their parents have contact with each other because participating at birthday parties strongly coincides with intergenerational closure (H4). According to Model 9, intergenerational closure also increases the odds of birthday invitations in the subgroup of friends by a factor of five. Indeed, there is a correspondence between intergenerational closure and invitations to birthday parties, even if birthday parties occur less frequently, e.g. in the group of Turkish children.
Conclusion
It has been shown in this article that inter-ethnic friendship ties of 10-year-old children have a high risk of being ‘intergenerationally open’. There is, in other words, a gap in social assimilation between generations of children and their parents. As a result of intergenerational openness, invitations to birthday parties are much less likely, and thereby the opportunities for immigrant children to interact with their native friends’ parents – who represent host country role models – decrease. In anthropological studies, circles of birthday invitations were considered as integrated segments of local communities. Being part of these circles has highly integrative effects on immigrant children. However, due to intergenerational openness – especially in Turkish-German dyads – friendships exist without the involvement of parents, which may lead to these friendships being of lower quality. These results should sensitize researchers to the fact that processes of the social assimilation of immigrant children and their parents do not operate independently of each other. This result is in line with the selective acculturation argument (Quillian and Campbell, 2003: 543) and points to an interesting explanation of it – namely the intergenerational ‘gap’ in social assimilation. Results of further analyses (not shown) suggest that cultural capital – measured by the number of books at home – is the main resource to be used in bridging this gap. Although the data set is rather small for a detailed analysis, it can be shown that, for immigrant parents, the chance of having ties to native parents increases by a factor of 4.9 if they have at least two bookshelves at home. This is in line with the idea that social assimilation requires cultural and educational resources, which are devalued for many immigrants in the wake of their migration.
This study’s results suffer from some shortcomings. First, the data set is comparatively small, even though the overall number of dyads is high. In the logistic regression p* models, some of the estimated coefficients are large without showing significance. In most cases, this results from small data sets (Joyner and Kao, 2000: 820) where certain combinations of characteristics show low case numbers within each cell. An analysis based on a larger sample will provide more precise and reliable estimates.
Moreover, in some cases, parents might have contact to each other because of their children’s birthday invitations. Hence, causality is not always clear-cut. On the other hand, the strong effect of intergenerational closure on birthday party invitations at least shows that there is a close correspondence between segregation at the parental level and the quality of children’s friendships, meaning that friends also invite each other to birthday parties. Regardless of the causal directionality, the strong association of these variables indicates that ethnic segregation in networks still exists: either invitations of Turkish and other immigrant children to natives’ parties (and vice versa) rarely occur, and immigrant parents thus do not take part in circles of exchange and do not become acquainted with native parents; or immigrant children do not attend natives’ birthdays (and vice versa) because there is no contact between their parents. In both cases, the result indicates that ethnic boundaries separate both groups. Accordingly, social assimilation at the parental level corresponds with social assimilation and acculturation at the children’s level, at least with respect to social ties which go beyond friendship nomination. This is in line with the idea that the segregation of parental networks is obstructive to inter-ethnic transmission of role models and upper service-class habitus forms.
A further problem in this article is that some important assumptions are based solely on theoretical arguments. For example, the assumption that the circulation of gifts and children has an integrative effect is derived from theoretical arguments and justified by referring to existing literature. An empirical clarification of this effect should be based on longitudinal data, as this would allow for a decomposition of cause and effect on a time scale.
Nevertheless, the overall results seem to be plausible with regard to the theoretical considerations. First, there is an intergenerational gap in social assimilation. While children’s ties are only moderately determined by ethnic origin or migration background, strong segregation occurs in networks at the parental level. Segregation of native and immigrant parents corresponds closely with fewer invitations to birthday parties. Second, this study showed the fruitfulness of the social network approach in migration and integration research. It has been demonstrated that the suggested network generator works well and provides reliable information on class-networks of fourth graders. This is a very promising result for future research, which should be based on a large-scale survey, especially if it is conducted as a panel study. A larger survey would also allow for a detailed analysis of the small group of immigrant parents who establish ties with natives, regardless of the overall tendency towards segregation in networks.
Footnotes
Acknowledgements
Many thanks to the Deutsche Forschungsgemeinschaft (DFG) for support (grant no. 574 807), Sahra Oldenburg and Priya Fielding-Singh for her very professional assistance, and special thanks to the reviewers of Sociology and Stefanie Kley for their helpful comments, and Dirk Baier and Christian Pfeiffer for pointing my attention to birthday parties.