• Keine Ergebnisse gefunden

Georgas, Berry, van de Vijver, Kagitcibasi, and Poortinga (2006)

2. Theoretical Outline

2.2 Family Models and Selves

2.2.2 Empirical Studies Based on the Theory of Family Change

2.2.2.1 Georgas, Berry, van de Vijver, Kagitcibasi, and Poortinga (2006)

Few studies have been carried out that explicitly tested or validated Kagitcibasi‘s theory of family models. One of these is the recent large-scale cross-cultural study by Georgas, Berry, van de Vijver, Kagitcibasi, and Poortinga (2006). The methodology, results, and conclusions of this study are reviewed here in detail since it is the only study so far relating diverse family variables to Kagitcibasi’s theory in a larger number of cultures. The study aimed to explore cross-cultural differences and similarities across 30 nations with respect to family networks, family roles, emotional distance from family members, independent and interdependent self-construal, personality, family values, and Schwartz’

general values. The study is an expansion (both culture- and concept-wise) of the study by Georgas et al. (2001). Hypotheses with regard to the pattern of cross-cultural differences and similarities were derived from the eco-cultural framework (e.g., Berry, 1976; Berry et al., 2002) and from Kagitcibasi’s model of family change (e.g., Kagitcibasi, 2002).

The study – reported in a volume of 550 pages – contains a quantitative and a qualitative part. In the quantitative part, overall 5482 university students (mean age 21 years) from 27 countries were compared with respect to the above mentioned characteristics.5 In the qualitative part, collaborators from the participating countries summarized country-specific aspects with respect to families and family change in separate chapters. This review will focus on the quantitative part of the study only, and here specifically on the results related to Kagitcibasi’s theory. The two most important hypotheses relevant to the theory of family change can be summarized as follows6.

Hypothesis 1: Regarding socioeconomic development, increasing affluence and increasing education will predict (a) greater distance of residence and less frequent contact; (b) decreased salience of emotional roles and less emotional closeness (less emotional interdependence); (c) decreased salience of material/ instrumental roles (less material interdependence); (d) less instrumental (material/ economic) role of the child; (e) less interdependent self; (f) less hierarchy and less salience of kin relationships; and (g) less personal embeddedness. Inverse associations with the above variables are predicted for the percentage of population engaged in agriculture, which is inversely related to affluence at the cultural level.

5 The study started out with 30 countries, but data from South Africa, Botswana, and Mongolia could not be used for the quantitative analyses due to technical problems. The other 27 countries were Algeria, Brazil, Bulgaria, Canada, Chile, China (Hong Kong), Cyprus, France, Georgia, Germany, Ghana, Greece, India, Indonesia, Iran, Japan, Mexico, The Netherlands, Nigeria, Pakistan, Saudi Arabia, South Korea, Spain, Turkey, Ukraine, United Kingdom, United States.

6 Hypotheses are not numbered in the original text, but are numbered for reference purposes in this review.

Hypothesis 2: Regarding the role of the culture of relatedness vs. culture of separateness, three (a-priori) country clusters were formed that should be associated with the corresponding three family patterns as formulated in Kagitcibasi’s Model of Family Change:

Cluster 1: Low affluence/High relatedness: Algeria, Ghana, India, Indonesia, Nigeria, Pakistan.

Higher emotional and material interdependencies/bonds (Family Model of Interdependence)

Cluster 2: High (increasing) affluence/High relatedness: Brazil, Bulgaria, Chile, Cyprus, Georgia, Greece, Hong Kong, Iran, Japan, Mexico, Saudi Arabia, South Korea, Spain, Turkey, Ukraine. Relatively higher emotional, weaker material interdependencies/bonds (Family Model of Emotional Interdependence)

Cluster 3: High affluence/Low relatedness: Canada, France, Germany, The Netherlands, UK, USA. Relatively weaker emotional and material interdependencies/bonds (Family Model of Independence)

Despite the differences between clusters, however, emotional bonds may be expected to be important in all family types. (Berry, Kagitcibasi, Georgas, Poortinga, & van de Vijver, 2006, pp.

108-110; Georgas, van de Vijver, Berry, Poortinga, & Kagitcibasi, 2006, p. 117f)

Hypotheses 1a – 1g are also in accordance with the eco-cultural framework. In short, they deal with the influence of modernization and industrialization (i.e., increased affluence, education, and less people in agriculture) on the importance of the family and relationships.

Overall, these linear relationships are in line with Kagitcibasi’s model of family change.

However, the specific pattern of values with the hypothesized emergence of the family model of emotional interdependence is only reflected in Hypothesis 2 that is based on the a-priori clustering of cultures according to their “cultural relatedness” crossed with affluence.

Thus, according to the model of family change, linear negative associations (correlations) between affluence indicators and material as well as non-material interdependencies in the family can be expected since both kinds of interdependencies should decline across the full range of cultures’ economic development (i.e., from poor countries to rich countries) and from cultures with a family model of (total) interdependence to those with a family model of independence. When it comes to comparisons of clusters (relatedness crossed with affluence, thus referring to the three proto-typical family models), however, expectations based on the eco-cultural framework are no longer equivalent to those based on the family change model. According to the eco-cultural framework – at least as it is presented in the study by Georgas et al. (2006) – all kinds of family interdependencies should decline from low affluence over medium affluence to high affluence cultures while the family change model predicts that the decline is differential: material interdependencies should decline gradually from low affluence to high affluence, but emotional interdependencies should stay as high in emotional-interdependence-cultures (medium/high affluence – high relatedness) as in total-interdependence-cultures (low affluence – high relatedness), and should be lower only in independence-cultures (high affluence – low relatedness). Unfortunately, these differences

in the predictions of the eco-cultural framework and the model of family change are not made explicit in Georgas et al.’s (2006) study. This lack of clarity leads to more ambiguities and inconsistencies with respect to the report and interpretation of results in the course of the study, as will be laid out in the following.

With respect to hypothesis tests regarding the eco-cultural framework (not reported here in detail), a country-clustering based on the level of affluence was employed (low – medium – high affluence) that was very similar to the clustering reported above. The clusterings only differed with respect to the cluster assignment of Hong Kong, Japan, South Korea, and Spain:

while these countries were in Cluster 3 (high affluence) in the clustering that was solely based on the level of affluence, they were assigned to Cluster 2 in the clustering according to cultural relatedness crossed with affluence, because of their presumed cultures of relatedness.

The clusters based on the crossing of relatedness and affluence are referred to differently in the course of the reported analyses: while in the hypotheses the clusters are characterized according to Kagitcibasi’s original conceptualization (Cluster 1: low affluence/high relatedness; Cluster 2: high (increasing) affluence/high relatedness; Cluster 3: high affluence/low relatedness), Cluster 2 is referred to later (in the chapter reporting the results of the study) as medium affluence/medium relatedness (van de Vijver, Mylonas, Pavlopoulos, &

Georgas, 2006, p. 148). This inconsistency is relevant since it is crucial for Kagitcibasi’s theory according to which the family model of emotional interdependence is thought to develop in industrializing cultures with high relatedness. While medium affluence is unproblematic (with the exception of Japan) since it reflects the economic development taking place in these countries, the label medium relatedness is questionable, since it insinuates that these countries’ cultures of relatedness have begun to develop in the direction of separateness, a prediction that would be contrary to expectations based on Kagitcibasi’s theory.

Unfortunately, this difficulty is not discussed by the authors.

The instruments employed in the study by Georgas et al. (2006) were tested with regard to cross-cultural equivalence. Though the authors conclude that all instruments showed sufficient construct equivalence to be compared across cultures, the criteria for the establishment of equivalence were not very strict. For instance, for the constructs of self-construal, personality, and general values it was first tried to establish cross-cultural equivalence via factor-analyses and following target rotations for the full instruments. After this procedure did not yield stable cross-cultural factor solutions, the instruments were split up into their proposed single scales and subsequently equivalence was tested for each of the

scales separately. This is methodologically problematic since it questions the very nature of the test of cross-cultural construct equivalence (same or similar patterning of a constructs’

components across cultures) and ends up with little more than a cross-cultural comparison of the reliabilities of the single scales. The authors argue that it may have been the use of short versions of the original instruments that lead to the problems in the establishment of equivalence. Overall, they conclude that all instruments showed sufficient equivalence to be included in the cross-cultural comparisons, but caution to interpret findings of scales and samples that did not show satisfactory equivalence.

Hypothesis 1 was tested through culture-level analyses. Culture-level characteristics reflecting socioeconomic development like affluence, proportion of people working in agriculture, and level of education were correlated with the cultural mean values of the dependent variables, or the dependent variables were regressed on the socioeconomic indicators (N = 27). The correlation coefficients of affluence, education, and proportion of people working in agriculture with the relevant dependent variables showed a parallel pattern of results, with those regarding the proportion of people working in agriculture pointing in the opposite direction, as expected. Since the relations of affluence with the dependent variables were stronger than those of education and people in agriculture with the dependent variables, only the results regarding affluence will be reviewed here in detail.

With regard to family networks, affluence was negatively related to geographic proximity and visits, but positively to telephone calls (Hypothesis 1a). With respect to family roles, affluence was negatively related to father’s expressive and financial role, mother’s and grandparents’ expressive roles, and to 20-year old males’ and females’ expressive as well as instrumental roles (Hypothesis 1b to 1d). Affluence was negatively related to emotional closeness to both nuclear and extended family members (Hypothesis 1b), as well as to independent and interdependent self-construal (Hypothesis 1e). It was strongly negatively related to the family values scales of hierarchy and kin relationships (Hypothesis 1f), as well as to Schwartz’ embeddedness and harmony scales (Hypothesis 1g). These culture-level correlations were largely in line with hypotheses.

Hypothesis 2 was tested employing individual-level analyses. ANOVAS were carried out for all dependent variables with country as a factor, and additionally with the different country-clusterings as factors in separate analyses. The focus was not on the significance of the analyses (since very small differences become significant in analyses with N = 5469) but on the effect sizes (proportion of variance accounted for by the factors). According to Cohen

(1988) proportions of explained variance (η2) of .01, .06, and .14 were used to distinguish between small, medium, and large effects. Only the results with respect to the country-factor and the factor of the crossed relatedness/affluence clusters will be reported in the following.

By definition, the variance accounted for by the country factor sets the upper limit of the variance that can be possibly explained by the three clusters since these are nested in the country factor. The pattern of differences among the three relatedness/affluence clusters is reported on the basis of results displayed in Tables 7.13 (p. 154f) and 7.14 (p.157). The authors do not describe the results in such detail.

With respect to family networks, country explained 17 % of geographic proximity as well as telephone calls (large effects), and 11 % of visits (medium effect), while the relatedness clusters explained 6 % – 7 % of these variables (medium effects). The pattern was such that in low relatedness cultures family members lived farther away from each other than in “medium”7 and high relatedness cultures. Visits and telephone calls were highest in medium relatedness cultures. These results are in line with the hypotheses.

With regard to family roles, medium to large country effects were found for the emotional/expressive roles (11 % – 20 % explained variance), as well as for the instrumental/

financial roles (8 % – 19 % explained variance). The effect sizes of the factor relatedness/

affluence cluster were small to medium for the expressive roles (1 % – 10 %) as well as for the instrumental roles (0 % – 6 %). The hypotheses with respect to the cultures of relatedness were such that “medium-relatedness cultures” (characterized earlier by increasing affluence and high relatedness) should show expressive roles as high as high-relatedness cultures while at the same showing relatively low instrumental roles, resulting in a pattern according to the family model of emotional interdependence. The results with regard to the pattern across the three clusters showed that for nearly all expressive roles high relatedness cultures showed higher values in expressive roles than medium and low relatedness cultures (which did not differ from each other). With respect to instrumental roles, a similar pattern was shown for the fathers and 20-year old sons and daughters instrumental roles, but no differences occurred for the other family positions.

In my view, the results were not in accordance with the expected pattern. With respect to expressive roles, medium relatedness cultures were more similar to low relatedness cultures (contrary to expectations); with regard to instrumental roles the expected differences were

7 The term “medium relatedness cluster” is used here since the authors use it themselves throughout the results section of the book. This does not qualify my reservations with respect to this term (see above).

found for some roles, but not for others. Interestingly, when looking at the results of the expressive roles according to the (pure) “affluence clusters,” the results were more in line with the predictions made by family model theory: here, for some roles medium affluence cultures were as high as low affluence cultures or in between low and high affluence cultures for other roles. This suggests that the clustering according to relatedness with Hong Kong, Japan, South Korea, and Spain assigned to the medium relatedness cluster in addition to the other medium affluence countries (see above) is problematic. The values of these four high affluence countries seem to make the cluster means of expressive roles more similar to those of the high affluence/low relatedness cultures. Thus, medium affluence may be compatible with high emotionality/expressiveness while high affluence may not, even if these highly affluent cultures can supposedly be characterized by high relatedness. Unfortunately, the incompatibility of the results regarding expressive and instrumental family roles with the hypotheses according to the family model theory is not clearly stated neither in the result-chapter (van de Vijver et al., 2006) nor in the discussion-result-chapter (Georgas, Berry, &

Kagitcibasi, 2006).

With regard to emotional distance, the factor country explained 4 % (nuclear family) and 9 % (extended family) of the variance. The relatedness clusters did not differentiate with respect to emotional distance (1 % and 0 % explained variance, respectively). Overall, emotional distance was lower with respect to the nuclear family (more closeness) than with respect to the extended family. With regard to self-construal, the factor country explained 17 % and 29 % of the variance of independent and interdependent self-construal, respectively (large effects). The relatedness clusters explained only 1 % and 5 %, respectively. While no significant differences were found for independent self-construal, high relatedness cultures showed a higher interdependent self-construal than low relatedness cultures (with medium relatedness cultures in between, while not significantly different from the other two). With respect to family values, country explained 54 % of the variance of the subscale hierarchy (roles of mother and father), and 29 % of the subscale kin relationships. Here, the relatedness clusters also explained a large part of the variance with 37 % and 15 %, respectively. High relatedness cultures showed markedly higher hierarchy than medium relatedness cultures which in turn were higher than low relatedness cultures. With regard to kin relationships, high relatedness cultures were higher than both medium and low relatedness cultures (which did not differ from each other), but the differences were rather weak compared to the hierarchy scale. With regard to Schwartz’ values, the country effect was large for embeddedness and hierarchy (29 % and 16 %, respectively), and medium for the other four scales. Since

hypotheses with regard to the model of family change were only formulated for embeddedness and since this was the only scale with a large effect of the relatedness clusters (14 %), only the differences with respect to embeddedness are reported here (hierarchy and harmony showed similar effects): High relatedness cultures showed the highest embeddedness, followed by medium relatedness and low relatedness cultures.

With respect to the patterns expected according to the theory of family change, the variables emotional distance, interdependent self-construal, and embeddedness represented the emotional interdependencies. Independent self-construal and hierarchy were assumed to be related to the material interdependencies (since the latter are assumed to be strongly related to hierarchical family relationships and obedience orientation). The kin relationships scale is not easy to categorize: the authors locate it in the emotional domain since it has to do with good relationships, but it also includes items regarding obedience and other traditional values like family reputation. Therefore, it might be also characterized as belonging to the material/hierarchy domain. The results of the above named variables were partly in accordance with the expectations according to the family change model: the emotional domain variables interdependent self-construal and embeddedness showed higher values in high relatedness cultures than in low relatedness cultures with medium relatedness cultures in between. According to the expectations, however, medium relatedness cultures should have been closer to high relatedness cultures than to low relatedness cultures with respect to these emotional variables. Emotional distance was virtually unrelated to the relatedness clusters.

The material domain variable hierarchy showed strongly declining values from high relatedness cultures to low relatedness cultures, which was in line with expectations. The variable kin relationships was higher in high-relatedness cultures, and lower in both medium and low relatedness cultures, suggesting that it might have more material/hierarchical as compared to emotional components.

In their discussion of the results with regard to the hypotheses based on the model of family change, van de Vijver et al. (2006) conclude that “various hypotheses derived from Kagitcibasi’s model were supported” (p. 170). In their argumentation the authors also include the absolute levels of endorsement of some of the variables in addition to cross-cultural differences and culture-level associations: “The emotional/expressive role of the family was important in all relatedness clusters. Even in the independence cluster, which showed the lowest expressiveness scores, the global means were still relatively high and they are higher than the means of the instrumental role for each position.” (p. 170). Furthermore, they argue

that especially the fact that the instrumental role of the child decreased with rising affluence (while other family positions did not) speaks in favour of the hypotheses derived from the model of family change. A similar conclusion is drawn by Georgas, Berry, and Kagitcibasi (2006) in their chapter synthesizing the results of the overall study. They affirm that “to a large extent the predictions of the Family Change Model are borne out by the results” (p.

229). In addition to repeating conclusions drawn in the results chapter, they conclude that “the sharp decreases in family hierarchy values with higher socioeconomic levels while relationships with family/kin remained high, point to the apparent prevalence of the Family Model of Emotional Interdependence […]. This pattern involves autonomy of the individuals together with relatedness, thus the ‘autonomous-related self’ (Kagitcibasi, 1996b, 2005) is particularly salient for our young and educated sample.” (p. 229).

From my perspective, these conclusions are only partially justified. First of all, in the results section the pattern of differences in the dependent variables according to the relatedness clusters is not described systematically with reference to the patterns that would

From my perspective, these conclusions are only partially justified. First of all, in the results section the pattern of differences in the dependent variables according to the relatedness clusters is not described systematically with reference to the patterns that would