Friendship Homophily 1 Running head: FRIENDSHIP HOMOPHILY
Friendship Homophily Among Ethnic German Immigrants. A Longitudinal Comparison Between Recent and More Experienced Immigrant Adolescents
Peter F. Titzmann and Rainer K. Silbereisen Center for Applied Developmental Science, Friedrich-Schiller-University, Jena, Germany
Final version of this manuscript was published here:
Titzmann, P. F., & Silbereisen, R. K. (2009). Friendship homophily among ethnic German immigrants. A longitudinal comparison between recent and more experienced immigrant adolescents. Journal of Family Psychology, 23, 301-310. doi: 10.1037/a0015493 http://www.apa.org/pubs/journals/fam/
Correspondence concerning this article should be addressed to Peter F. Titzmann Center for Applied Developmental Science Friedrich-Schiller-University Am Steiger 3/1 07743 Jena Germany
[email protected] Tel: +49 (0) 3641 945220 Fax: +49 (0) 3641 945202
Friendship Homophily 2 Abstract Immigrant adolescents have been found to show high levels of friendship homophily (the tendency to prefer intra-ethnic over inter-ethnic friends). Based on acculturation theories, we predicted longitudinal changes and inter-individual differences in level and rate of change in friendship homophily by immigrant composition of school, parental objections to the cultural adaptation of their offspring, discrimination, and language use. Altogether 218 newcomer and 272 experienced adolescents who had emigrated from the former Soviet Union to Germany participated in four annual assessments. Friendship homophily was assessed by percentage of intra-ethnic friends. Results revealed higher levels of friendship homophily among newcomer (75%) than among experienced (65%) adolescents at T1, and a significant decrease over time was found only among newcomers. Higher age, higher percentage of immigrants at school, and lower levels of new language use all related to higher friendship homophily. Rate of change in friendship homophily was strongly related to change in new language use. Findings suggest that friendship formation with native adolescents is related to social learning processes, in which new language use plays a crucial role.
Keywords: friendship homophily, adolescent immigrants, acculturation, longitudinal
Friendship Homophily 3 Friendship Homophily Among Ethnic German Immigrants. A Longitudinal Comparison Between Recent and More Experienced Immigrant Adolescents As friendships across racial or ethnic groups can reduce prejudice, decrease implicit and explicit racial bias, reduce discrimination, improve interactions between members of two groups, and enhance immigrants’ access to resources and information (Aberson, Shoemaker & Tomolillo, 2004; Antonio, 2001; Pettigrew & Tropp, 2000), inter-ethnic friendships have become an important benchmark in efforts to reduce racial segregation (Aboud, Mendelson & Purdy, 2003). However, despite the fact that inter-ethnic friendships are desirable, studies repeatedly found a preference for intra-ethnic over inter-ethnic friendships (e.g., Wade & Okesola, 2002). This preference for intra-ethnic friends has been assumed to be related to similarities within an ethnic group, such as having a similar cultural background, comparable acculturation-related challenges, easier communication in the same language, or a similar social status (Maharaj & Connolly, 1994). In the literature, this phenomenon is referred to as friendship homophily: Homophily in general “is the principle that contact between similar people occurs at a higher rate than among dissimilar people” (McPherson, Smith-Lovin & Cook, 2001, p. 416). It can be applied to specific relationships, such as marriage, and is applied here to friendship. The aim of this study was to examine mean level and rate of change in friendship homophily among newcomer and experienced immigrants, and to predict inter-individual differences in level and rate of change in friendship homophily by several factors, such as the ethnic composition of the school context, parents’ objections to their children’s adaptation, the use of the new language, and perceived discrimination. As several studies found differential effects of predictors on outcomes depending on the length of residence (Beiser, Hou, Hyman & Tousignant, 2002; Wall, Power & Arbona, 1993), we distinguished two groups of immigrants, newcomer and experienced adolescents. For example, one study found
Friendship Homophily 4 that inter-ethnic friendships were positively related to depressive symptoms in early stages of the acculturation process, whereas they were negatively related to depressive symptoms later in the immigration process (Silbereisen & Schmitt-Rodermund, 2000). Such studies also show that friendships may serve different functions along the immigration process and that the selection of friends may be related to different variables in early vs. later stages of the immigration process. Using two groups of length of residence allowed us to examine discrepancies in strength of associations between predictors and friendship homophily between both groups. The study was conducted with a sample of ethnic German immigrants from the former Soviet Union, a particular group of diaspora migrants who returned to Germany, the country of their ancestors, after generations of living in countries of the former Soviet Union. Although this group is quite privileged and receives instant citizenship and financial support upon entry to Germany, the adolescents were found to experience similar problems to other immigrant groups, such as difficulties in language acquisition and in social integration (Dietz, 2003). Group Level Changes in Friendship Homophily Although ethnic German immigrants (as well as other immigrant groups) were found to engage mainly with intra-ethnic peers (Titzmann, Silbereisen, & Schmitt-Rodermund, 2007), there are good reasons to expect change over time. Immigrant adolescents have been found to adjust socio-culturally in terms of language (Fuchs, Schwietring & Weiss, 1999), autonomy expectations (Schmitt-Rodermund & Silbereisen, 1999), and identity (Birman & Trickett, 2001). This socio-cultural adaptation, which is assumed to be more pronounced shortly after immigration than later (e.g., Masgoret & Ward, 2006), should diminish the cultural gap between immigrants and non-immigrants and should thus increase the likelihood of inter-ethnic friendships through increased similarity. In line with this assumption,
Friendship Homophily 5 Silbereisen and Titzmann (2007) found friendships to be almost 100% intra-ethnic shortly after arrival in the host country and to be between 60% and 70% intra-ethnic among adolescents who had been in the country for about 10 years. These results are, however, based on cross-sectional designs and may be attributed to differences in immigrating cohorts. Given the assumptions mentioned, we expected friendship homophily to be higher among newcomers than among experienced adolescent immigrants (Hypothesis 1), because newcomers should be less well-adjusted in comparison to experienced adolescents. Second, since socio-cultural adjustment (and related friendship homophily) is supposed to be more pronounced early after immigration, we expected a more pronounced longitudinal decrease in friendship homophily in the group of newcomers than among experienced immigrants (Hypothesis 2). Predicting Inter-individual Differences in Friendship Homophily The changes in friendship homophily described so far have concerned the means in the newcomer and experienced groups. Adaptation processes are, however, not identical across individuals. Adolescents in both groups are likely to differ with regard to level and rate of change in friendship homophily. Some adolescents in the experienced group, for instance, may tend to withdraw into intra-ethnic relations, whereas others may establish friendships with natives resulting in an overall group change of zero. As described earlier, the predictors for these individual shifts may differ between early and later years of the acculturation process. Thus, the second aim of this study was to test whether we can explain interindividual differences in level and rate of change in friendship homophily by a set of predictors in the two groups. These predictors were the ethnic composition of the school, the families’ objections to their offspring’s adaptation to the new culture, the individual’s use of the new language, and level of perceived discrimination. These predictors represent core developmental factors for inter-ethnic friendship selection in different contexts.
Friendship Homophily 6 Friendships are among the most important social relations in adolescence and are often established at school (e.g., Kiesner, Kerr & Stattin, 2004). According to the opportunity hypothesis (Hallinan & Teixeira, 1987), a lower percentage of same-ethnic peers in the school context should reduce levels of friendship homophily, because it increases the likelihood of successful inter-ethnic exchange and subsequent friendships. Cross-sectionally, this assumption was supported in a sample of ethnic Germans (Silbereisen & Titzmann, 2007). Thus, we expected the percentage of ethnic German immigrants at school to be related to higher levels of friendship homophily and to a less pronounced decrease of friendship homophily over time. Outside of the school context, parents need to be considered as a source of interindividual variation in friendship homophily. In a study by Mounts (2000), parents reported a substantial number of indirect and direct strategies to regulate their offspring’s peer relations, for instance, by giving their opinion about chosen friends or by limiting time spent with certain peers. Only 10% of parents reported doing nothing about their child’s peer associations. We examined the parents’ objections to their offspring’s adaptation to German peers as perceived by the adolescent. We expected that parents who do not want their child to adapt too much to the new culture will reinforce the intra-ethnic peer contacts of their children. Nevertheless, parents can also be expected to acculturate (Birman & Trickett, 2001) and are assumed to reduce their opposition to the cultural adaptation of their children over time. Such changes in parental objections to their offspring’s adaptation should therefore be positively related to changes in friendship homophily, i.e., a decrease in the parents’ objections should be related to a decrease in friendship homophily and vice versa. A third source of inter-individual variation in friendship homophily is the use of the new language as a central facet of socio-cultural adjustment (Masgoret & Ward, 2006) and as a prerequisite for successful social integration (Esser, 1980). Indeed, this variable has been
Friendship Homophily 7 found to be strongly related to adaptation to the new context (Kang, 2006). We therefore expected immigrant adolescents who use the new language more often to report lower initial levels of friendship homophily and a more pronounced decrease in it over time. This is due to their obvious ability to communicate better with native peers, to having better opportunities for inter-ethnic contact, and to a smaller cultural gap with their native peers. However, language use itself can be expected to change over time (Fuchs, Schwietring & Weiss, 1999), and we expected an additional negative association between changes in language use and changes in friendship homophily, i.e., adolescents who increase their spoken German should decrease their friendship homophily more than average and vice versa. Perceived discrimination, our final predictor, refers to negative experiences or barriers encountered in the new context (Esser, 1980). Experiences of feeling rejected because of one’s ethnicity may lead to a withdrawal into one’s own ethnic community and may reinforce friendship homophily. We expected higher levels of discrimination to be related to higher levels of and a less pronounced decrease in friendship homophily. Furthermore, changes in perceived discrimination should be positively related to changes in friendship homophily. In other words, we expected a more pronounced decrease in perceived discrimination to be related to a more pronounced decrease in friendship homophily and vice versa. Thus, in order to predict friendship homophily in both groups, two further hypotheses were tested. Hypothesis 3 regarded the associations of first wave of assessment predictors with levels and rate of change of friendship homophily. We predicted that higher percentages of ethnic German immigrants at school, higher levels of parental objections to their offspring’s cultural adaptation, lower levels of new language use, and higher levels of perceived discrimination would be related to higher levels of friendship homophily at first assessment. We further predicted that these variables would be related to a less pronounced decrease of friendship homophily over time as compared to average level of change.
Friendship Homophily 8 Hypothesis 4, our final hypothesis, predicted significant associations between changes in predictors and changes in friendship homophily. We expected that more pronounced decreases in parental objections to their child’s cultural adaptation, more pronounced decreases in discrimination, and more pronounced increases in language use would be related to a more pronounced decrease in friendship homophily. Method Sample The data are part of a project on adolescent immigrants from the former Soviet Union to Germany and Israel1. The initial sample comprised 1,437 ethnic German adolescents, which represents approximately 44% of the ethnic German students in the participating schools. Given the fact that according to Mohler, Koch and Gabler (2003) surveys in Germany rarely reach a response rate higher than 50%, this percentage is acceptable. We concentrated our analyses on first generation ethnic German immigrants from the former Soviet Union. Although ethnic German immigrants also come from other countries, such as Poland, these were found to show higher levels of adjustment problems and few ethnic Germans emigrated from these countries in recent years (Schmitt-Rodermund & Silbereisen, 2004). In order to avoid any confounds we excluded 270 immigrants who came from other countries than the former Soviet Union. We only included adolescents who took part in at least three out of four waves of measurement in order to limit the amount of imputed data (using the AMOS regression algorithm). Thus, the final sample comprised 490 ethnic German adolescents from 53 schools in nine cities from four federal states of Germany (34% of the whole sample). Most of the participants (about 45%) attended the lowest (Hauptschule) or intermediate (Realschule) school tracks, 22% attended the highest educational track (Gymnasium), 20% came from comprehensive schools (Gesamtschule), and remaining adolescents came from vocational schools (Berufsschule). These numbers are comparable to
Friendship Homophily 9 other studies (e.g., Baumert & Schümer, 2002). Immigrants in Germany find themselves in lower school tracks more often than native adolescents. As expected, the schools in this study varied substantially with regard to the concentration of immigrants. Schools in neighborhoods with a high share of immigrants had a higher concentration of ethnic Germans. Schools in eastern federal states of Germany had a lower concentration, because these former communist federal states received significantly fewer immigrants due to their lower economic capabilities in accommodating and supporting new immigrants (Bade & Oltmer, 1999). For the first wave, the questionnaires were completed at school, while the follow up questionnaires were answered by mail with 12 months intervals between the assessments. For all waves the questionnaires included both languages (German and Russian) to enable the participants to choose the language in which they were most fluent. Of the 490 ethnic German adolescents, 327 (67%) participated at all four waves, 104 (21%) in the first three waves, and the remaining 59 (12%) missed participating in either the second or third wave. In order to split the sample into a newcomer and an experienced group, we decided to use 7 years after migration as the cut-off criterion. After 6 to 7 years, parents of adolescent immigrants were found to have a more balanced orientation between their home and their new country, rather than being mainly focused on the country of origin (Birman & Trickett, 2001). This resulted in a sample of 218 newcomer ethnic German immigrants with seven years of residence or less and 272 experienced ethnic German immigrants with more than 7 years of residence2. To solve the common problem of confounding length of residence and age in acculturation research (Fuligni, 2001), a specific design was chosen in which length of residence and age were not correlated (see Appendix). Thus, the two groups only differ in length of stay, but not in age, so that differences found between newcomers and experienced adolescents cannot be attributed to different life-stages. Measures
Friendship Homophily 10 Friendship homophily. We defined friendship homophily as the percentage of intraethnic friends out of all friends. Thus individuals could vary in friendship homophily between zero and 100%. Participants reported the number of friends that were a) native Germans and b) ethnic Germans. On average, participants reported a rather large number of friendships (mean between 23.9 and 25.7, median between 17 and 20 depending on the wave of measurement). These numbers indicate that the friendships mentioned included an extended peer network. Friendship homophily was calculated by dividing the number of intra-ethnic friends by the total number of intra- and inter-ethnic friends multiplied by 100. Demographic variables. These variables were part of the questionnaire. Participants reported their year of birth, year of arrival in their respective country, and their gender. The percentage of ethnic German immigrants at each school was derived from official statistics. Perceived discrimination. Discrimination was assessed using the mean of three items taken from Steinbach and Nauck (2000). Adolescents were asked: “Sometimes native Germans, ethnic Germans and other immigrants are treated differently. How often did you have to face disadvantages in the following situations during the last twelve months, because you are an immigrant? “…at school/ work”; “…when visiting administrative officials”, “…in shops”. The answering format was “never”, “once or twice”, “three to five times”, “six to ten times”, and “more than ten times” (T1: α = .63; T4: α = .73). Language use. The mean of two items (“language spoken with friends” and “language used for reading books or journals”) was used to assess the level of new language use. Participants rated their language use on a 5-point Likert scale ranging from “never in German” to “always in German” (correlation of items: T1: r =.52 T4: r =.48). Two other items originally included in the questionnaire were excluded, namely “language used with parents” and “language of TV shows and radio programs”. These were found to be poor indicators, since the use of the new language with parents and of TV shows depends less on
Friendship Homophily 11 the decision of the adolescents, but rather on the parent’s decision or on the availability of Russian programs (not similar across Germany). Parents’ objections to their offspring’s adaptation. These objections as perceived by the adolescents were measured using three self-developed items (Titzmann, SchmittRodermund & Silbereisen, 2005): “My parents did not understand why I wanted to be like local adolescents”; “My parents did not want me to be too oriented towards local adolescents”; and “My parents did not want me to dress like local adolescents” (T1: α = .64, T4: α = .66). As the mean of these items was highly skewed we used an index of 0 to 3 indicating whether an adolescent agreed that none, one, two or all three of these situations happened over the last twelve months. Results Before testing the hypotheses, we compared the two groups of immigrants descriptively (see Table 1). Newcomers reported greater parental objections, less use of the new language, and more perceived discrimination. Furthermore, the change in these variables across the time of the study was only significant among newcomers. In this group, parental objections to their offspring’s adaptation decreased, language use increased, and discrimination decreased. These differences between newcomers and experienced adolescents could be expected. Note that the percentage of ethnic German adolescents at school differed between the two groups, which may be a result of the German policy of settling ethnic German immigrants. In order to distribute immigrants evenly across the country according to the economic capabilities of regions, immigrants are only free to decide where to live after several years of residence in Germany. Any relocation before this time would result in losing welfare benefits. Thus, after a few years, ethnic Germans often move to areas where relatives and other ethnic Germans live, and thus may end up in more segregated environments.
Friendship Homophily 12 However, the range in percentage of immigrants at school is comparable (newcomers: range = 0.63% to 38.4%; experienced: range = 0.14% to 38.4%). Table 1 about here Group Level Changes in Friendship Homophily In order to test Hypotheses 1 and 2, we created a latent linear growth curve model (e.g., Ferrer & McArdle, 2003) using the statistical package AMOS. A two group (newcomer vs. experienced) model was set up with T1 measures as intercepts and a linear slope in both groups. Slope and intercept were allowed to covary, and variances in the manifest measures of friendship homophily were constrained to be equal across the different measurement points in each group. The two group structural equation model showed an acceptable fit (chi-square = 24.1 [16df], p = .08; NFI = .96; CFI = .99, RMSEA = .03; e.g., Hu & Bentler, 1999). In order to test whether the linear model describes our data sufficiently, we tested this model against an unconditional model in which the growth parameters were estimated freely. However, the unconditional model did not significantly improve model fit (chi-square = 20.7 [12 df], p = .06; CFI = .99; NFI = .97; RMSEA = .04). Thus the linear model was the best and most parsimonious in representing our data. Among newcomers, the results showed a significant intercept and slope. The intercept was 75% ethnic German friends out of all friends, with a significant longitudinal decrease of about 2% per year in the study. For experienced adolescents, only the intercept was significant, at about 65% of ethnic German friends with no significant decrease. Furthermore, intercept and slope were only significantly correlated among newcomers (r = -.50, p < .01), but not among experienced immigrants (r = -.25, p = .33). Thus, adolescents in the newcomer group with a high percentage of ethnic German friends at T1 tended to have a more pronounced decrease than was the average in this group.
Friendship Homophily 13 For testing our hypotheses regarding group differences in level and rate of change, the parameters for intercept and slope were constrained to be equal across both samples and the change in model fit was tested for significance via change in chi-square. For the intercept, this test revealed a significant difference (chi-square = 15.8, p < .05) indicating that newcomers indeed reported significantly higher levels of friendship homophily than experienced immigrant adolescents, which supports Hypothesis 1. Constraining the slope parameter to be equal across the two groups did not change the model fit significantly (chi-square = 1.5, p = .22) so that the decrease found for newcomers was not significantly different from the average change in the experienced sample: Hypothesis 2 was therefore not supported by our data. Predicting Inter-individual Differences in Friendship Homophily In order to predict differences in intercept and change in friendship homophily, we added the predictors to the latent growth model. This model included measures for the percentage of intra-ethnic friends among all friends as manifest variables as well as the latent variables defining the growth curves of friendship homophily. The manifest predictors in the model were age, ethnic school composition, parental objections, language use, and perceived discrimination at T1. These T1 variables were direct predictors of intercept and slope in friendship homophily. The T4 measures of parental objections, language use, and perceived discrimination were used to estimate change between T1 and T4. These changes were modeled as latent change parameters (e.g., Ferrer & McArdle, 2003) predicting the slope in friendship homophily. This model allowed all hypothesized associations to be estimated simultaneously. First, it can estimate whether inter-individual differences in the predictors at T1 are related to inter-individual differences in intercept (Level at T1) and slope (rate of change) of friendship homophily. These paths test the expected associations of Hypothesis 3. Second, associations between changes in parental objections, language use, and perceived discrimination with changes in friendship homophily, as predicted in Hypothesis 4, were
Friendship Homophily 14 assessed. To control for the possibility that change in predictors is related to friendship homophily at first assessment, change in predictor and intercept of friendship homophily were allowed to covary freely. The model fit (see Table 2) was acceptable (e.g., Hu & Bentler, 1999). Table 2 shows the estimated parameters in the model. Age as control variable was related to higher values of friendship homophily in the intercept in both groups3. Older adolescents in both groups reported a higher percentage of ethnic German immigrants in their network. Among newcomers, however, an additional effect on the slope was found, indicating that older adolescent newcomers had a more pronounced decrease in their friendship homophily than younger adolescent newcomers. This association can probably be explained by regression to the mean. Those newcomers who were 19 years or older at T1 reported about 94% ethnic German friends among all friends and consequently can only shift towards the lower mean of the whole newcomer group over time. The school composition showed strong effects in both groups. As expected in Hypothesis 3, newcomers and experienced adolescents reported higher levels of friendship homophily if they attended schools with a high percentage of ethnic German immigrants. The standardized regression coefficients (β) were .31 for newcomers and .47 for experienced immigrant adolescents. School composition had, however, no effect on the rate of change in friendship homophily in either group; this part of Hypothesis 3 was thereby not supported. The strongest predictor in our analysis was language use. In both groups, higher levels of new language use at first wave of assessment predicted lower levels in initial friendship homophily (β = -.52 for newcomers, β = -.34 for experienced adolescents) as was expected in Hypothesis 3. Also in line with this hypothesis was the association between language use at T1 and a more pronounced decrease in friendship homophily in the newcomer group (β = .37). This association was, however, not significant in the experienced group. One part of
Friendship Homophily 15 Hypothesis 4 expected associations between changes in language use and changes in friendship homophily. This expectation was indeed supported for language use by our data. In both groups, a more than average increase in language use was significantly related to a more than average decrease in friendship homophily (β = -.63 for newcomers, β = -.67 for experienced adolescents). With regard to higher levels of perceived discrimination and more parental objections to their adolescents’ adaptation, Hypothesis 3 predicted higher levels of friendship homophily and a less pronounced decrease than average. These associations were not confirmed. Also in contrast to our expectations, newcomers who felt more discriminated against at first wave of assessment decreased more than average in their friendship homophily (β = -.38). None of the associations between changes in parental objections to their children’s adaptation or perceived discrimination and changes in friendship homophily, as posited as part of Hypothesis 4, was found. Table 2 about here Taken together, the third and fourth hypothesis was only supported with regard to language use, which was found to be the most prominent predictor of individual differences in level and rate of change in friendship homophily. In order to find out whether the apparent differences in strength of association between predictors and friendship homophily differ significantly between both groups, each of these parameters was constrained to be equal across the samples. The resulting change in model fit was tested for significance via the chi square test (e.g., Jöreskog & Sörbom, 1989; Lennox & Lennox, 1995). Results showed that only the parameter of discrimination to the slope of friendship homophily was significantly different between newcomers and experienced adolescents (chi-square change = 4.6, p
