• Keine Ergebnisse gefunden

Instruments, Cross-Cultural Equivalence, and Reliabilities

3. Method

3.4 Instruments, Cross-Cultural Equivalence, and Reliabilities

The questionnaire that was developed in a research cooperation of the VOC-IR teams from the Universities of Konstanz and Chemnitz contained several questions relating to adolescents’ personal background and socio-economic situation, religious orientation, general and family-specific values, values of children, self-construal, attachment security, various aspects of the parent-child relationship, and to the social support between parents and adolescents (Trommsdorff et al., 2002). Furthermore, there were questions with regard to life satisfaction, optimism and peer popularity. Finally, several questions were asked with respect to adolescents’ future orientation, primarily with regard to plans for a future family. For the present study, only parts of the questionnaire were used: values (general, family-specific, value of children), self-construal, and family-oriented future orientation. The parts of the VOC-IR questionnaire for adolescents (in English language) that were used in this study are included in Appendix D.

The language of the original version of the questionnaire was English. All language versions were translated from English into the respective target language by a native speaker of the target language who was also fluent in English. The translated versions were then back translated into English by another translator fluent in the target language as well as in English.

Finally, the English back translations were compared to the original versions by members of the Konstanz team, primarily by native English speakers, and any inconsistencies were corrected in cooperation with the translators from the respective countries. With this procedure the cross-cultural equivalence of the different language versions with respect to item-content was ensured. In the few cases were a literal translation was difficult or impossible the translators tried to render the meaning of the item as closely as possible. Van de Vijver (2007) suggests that a literal translation is no longer the first choice for the development of cross-culturally valid instruments, but that instruments should be “adapted” to avoid inadequate items in a target language. By applying the traditionally recommended back translation technique and simultaneously trying to be sensitive with respect to the cultural adequacy of the items the comparability and culture-sensitivity of the instruments were enhanced.

In the following, the instruments used in the present study, their cross-cultural structural equivalence, and their reliabilities are reported. With the exception of work related to the construction of the questionnaire (e.g., selection of instruments and their pretest in a pilot-study) which was done by the VOC-IR teams from Konstanz and Chemnitz prior to this study, all following analyses were done as part of this dissertation.

All instruments were tested with respect to their structural or construct equivalence across cultures except for the single-item questions regarding background information and future orientation where tests of cross-cultural construct equivalence are not applicable.

Structural or construct equivalence across cultures refers to the cross-cultural equality or close similarity of the dimensionality of a construct (and the respective instrument) (van de Vijver

& Leung, 1997a, 1997b). For example, an instrument that is intended to measure individualism and collectivism as two separate dimensions (and proved to do so in the culture where the instrument was developed) may fail to be cross-culturally equivalent when there are cultures where only one dimension results or when there are other cultures where more than two dimensions are necessary to account for the variability of the items. Obviously, a valid cross-cultural comparison of an instrument can only be accomplished when the instrument proves to be cross-culturally equivalent, otherwise “apples” are compared with “oranges”

(van de Vijver, 2007).

The procedure used to check for cross-cultural construct equivalence starts with a principal component analysis of an instrument across all participants and cultures (with per culture z-standardized values to account for cross-cultural mean differences that may inflate or deflate the overall correlations among items, see (Leung & Bond, 1989). The resulting structure is called the pooled solution. In a second step, for each culture a separate PCA is conducted, using the same number of dimensions that resulted in the pooled solution. In the last step, each of these culture-specific single solutions is target-rotated on the pooled solution. The loadings of the target-rotated single solutions are then compared to the loadings of the pooled solution using the proportionality coefficient Tucker’s Phi. A Tucker’s Phi value of .95 and higher indicates high similarity between the culture-specific solution and the pooled solution, and a value between .90 and .95 indicates acceptable similarity (van de Vijver & Leung, 1997a). Others consider even a cut-off value of .85 indicating sufficient equivalence (Ten Berge, 1986). If all (or nearly all) culture-specific solutions show acceptable or high similarity to the overall or pooled solution one can conclude that the instrument is structurally equivalent across cultures (van de Vijver & Leung, 1997a).

After assessing cross-cultural equivalence, scale reliabilities were checked according to the dimensions that appeared in the pooled solutions PCA. When the reliabilities were acceptable scales were constructed by averaging across those items with clear loadings on one of the factors in the pooled solution. Since exact criteria for a simple structure (Thurstone, 1947) are not available, the criterion for a clear loading of an item on a specific factor was set to a difference of more than .20 between the loading on this factor and loadings on other factors and to a minimum loading of .50.

3.4.1 Individualism/Collectivism

To assess individualism and collectivism a short version of the Schwartz and Bilsky (1987) instrument developed by Chan (1994) as part of the COLINDEX was used (Schwarz, Chakkarath, Trommsdorff, Schwenk, & Nauck, 2001). The instrument consisted of 13 items with seven items measuring individualistic values (items 1, 3, 5, 7, 9, 11, 13) and six items measuring collectivistic values (items 2, 4, 6, 8, 10, 12). Chan (1994) reported high reliabilities for the short version of individualism and collectivism in a US sample (Cronbach’s Alpha = .84 and .74, respectively) as well as in a Hong Kong sample (Cronbach’s Alpha = .78 and .73, respectively). The participants were asked to indicate to what extent these values constitute a guiding principle in their life and rate their importance from 1 = ‘Not important at all’ to 5 = ‘Very important’. The wording and sequence of the items was as follows:

1. An exciting life (stimulating experiences)

2. Honor of your parents and elders (showing respect)

3. Pleasure (enjoyment, amusement, or gratification of desires)

4. Social order (stability of society) 5. Creativity (uniqueness, imagination) 6. National security (protection of your own nation from enemies)

7. A varied life (filled with challenge, novelty, and change) 8. Self-discipline (self-restraint, resistance to temptation)

9. Being daring (seeking adventure, risk) 10. Politeness (courtesy, good manners) 11. Freedom (freedom of action and thought)

12. Obedience (fulfilling duties, meeting obligations)

13. Independence (self-reliance, choosing your own goals & interests

Principal Component Analysis (Pooled Solution)

The Varimax-rotated PCA across all participants resulted in a pooled solution with a pattern of eigenvalues (scree-plot) that clearly suggested two components. The eigenvalues of the first and second component were 3.86 and 1.64, respectively, and the two components explained 42.30 % of the total variance. The pattern of loadings showed the expected dimensions of Individualism and Collectivism (see Table 4). Items loaded clearly on the respective factor with loadings on the other factor relatively low, fulfilling the criteria for a simple structure according to Thurstone (1947).

Table 4

Factor Loadings for Individualism/Collectivism

Item Individualism Collectivism

07 A varied life (filled with challenge, novelty, and change) .65 .10 03 Pleasure (enjoyment, amusement, or gratification of desires) .64 .09 01 An exciting life (stimulating experiences) .63 -.03 05 Creativity (uniqueness, imagination) .62 .11 11 Freedom (freedom of action and thought) .57 .27 13 Independence (self-reliance, choice of own goals and interests) .55 .35 09 Being daring (seeking adventure, risk) .54 .18 10 Politeness (courtesy, good manners) .05 .72 02 Honor of your parents and elders (showing respect) .03 .71 12 Obedience (fulfilling duties, meeting obligations) .17 .70 08 Self-discipline (self-restraint, resistance to temptation) .16 .62 04 Social order (stability of society) .16 .59 06 National security (protection of my own nation from enemies) .23 .55 Note. Principal component analysis with Varimax-rotation. Bold printed factor loadings mark items that were included in the respective scales. Pooled Solution, n = 2566.

Cross-Cultural Equivalence and Reliabilities

The target rotations of the culture-specific PCAs on the pooled solution resulted in satisfactory Tucker’s Phi values with the exception of South Africa where a coefficient <.90 resulted for both factors (see Table 5). Thus, in all cultures except South Africa the culture-specific factor solution was very similar to the overall pooled solution, indicating cross-cultural equivalence of the Individualism/Collectivism construct. A closer inspection of the South African culture-specific factor solution showed that in this culture a uni-dimensional solution resulted with most of individualistic and collectivistic items loading substantially on the first factor.

The culture-specific reliabilities of the resulting Individualism (7 items) and Collectivism (6 items) scales are shown in Table 5. Most Cronbach’s alphas were > .70, indicating a satisfactory reliability. Only one coefficient was < .60 (Individualism in Switzerland, Cronbach’s Alpha = .57). The measures Individualism and Collectivism were generated by averaging across the respective items.

Table 5

Cross-Cultural Equivalence and Reliability of Individualism/Collectivism

Tucker‘s Phi Cronbach’s Alpha

Culture Individualism Collectivism Individualism Collectivism

China .97 .97 .76 .77

France .96 .95 .63 .72

Germany .98 .96 .70 .77

India .90 .94 .72 .81

Indonesia .98 .96 .70 .74

Israel .97 .93 .70 .75

Japan .97 .98 .76 .72

South Africa .84 .84 .78 .69

Switzerland .93 .93 .57 .75

Turkey .97 .96 .73 .78

3.4.2 Family Values

An instrument developed by Georgas (1989) was used to measure adolescents’ family values (Schwarz et al., 2001). Two sub-dimensions from Georgas’ (1991) Family Values Scale were chosen for the present study: Responsibilities of children toward family and relatives and Responsibilities of parents toward children. Short versions of the scales consisting of 5 and 7 items were tested in a pilot study of the VOC-IR project. Due to the length of the main study questionnaire it was decided to shorten the instrument again. Of the remaining 7 items, 4 items belonged to the subscale Responsibilities of children toward family and relatives (items 1, 3, 5, 7) and 3 items belonged to the subscale Responsibilities of parents toward children (items 2, 4, 6). Adolescents were asked to indicate how much they agree with the presented statements from 1 = ‘strongly disagree’ to 5 = ‘strongly agree.’ The wording and sequence of the items was the following:

1. One should maintain good relationships with one’s relatives.

2. Parents shouldn’t get involved in the private lives of their married children.

3. Children have an obligation to care for their parents when their parents are old.

4. A family’s problems should be solved within the family.

5. We should honor and protect our family’s reputation.

6. Parents should help their children financially whenever the children need it (even when the children are adults).

7. Children should obey their parents.

Principal Component Analysis (Pooled Solution)

Instead of the expected two-factorial solution the pattern of eigenvalues suggested a uni-dimensional structure. However, the first component explained only 31.55 % of the total variance with an eigenvalue of 2.21. The pattern of loadings showed that five items (1, 3, 4, 5, 7) loaded substantially on the factor and that two items (2 and 6) did not load substantially (see Table 6).

Table 6

Factor Loadings for Family Values

Item Family Values

05 We should honor and protect our family’s reputation. .72 03 Children have an obligation to care for their parents when their parents are old. .65 04 A family’s problems should be solved within the family. .65 07 Children should obey their parents. .58 01 One should maintain good relationships with one’s relatives. .58 06 Parents should help their children financially whenever the children need it

(even when the children are adults). .41 02 Parents shouldn’t get involved in the private lives of their married children. .04 Note. Principal component analysis. Bold printed factor loadings mark items that were included in the scale.

Pooled Solution, n = 2566.

Cross-Cultural Equivalence and Reliabilities

When there is only one factor no target rotation can be done (and is necessary) before assessing the similarity of the factor loadings between the culture-specific solution and the pooled solution (van de Vijver & Leung, 1997a). The Tucker’s Phi coefficients showed acceptable to high similarities of the country-specific solutions with the pooled solution indicating cross-cultural equivalence (see Table 7).

Only the five items with substantial loadings in the pooled solution were used in the computation of the reliabilities. Cronbach’s alphas were rather low across cultures with four cultures < .60 and only three cultures > .70 (see Table 7). This suggests a substantial hetero-geneity in the items which is also reflected in the rather low explained variance in the PCA.

Nevertheless, the Family Values scale was included for the following analyses to incorporate this important and prominent construct in the present study. The measure Family Values was generated by averaging across the five items showing substantial loadings in the pooled solution.

Table 7

Cross-Cultural Equivalence and Reliability of Family Values

Family Values

Culture Tucker‘s Phi Cronbach’s Alpha

China .99 .58

France .94 .70

Germany .99 .57

India .99 .81

Indonesia .98 .63

Israel .95 .60

Japan .98 .62

South Africa .92 .45

Switzerland .98 .76

Turkey .99 .67

3.4.3 Independence/Interdependence

To assess adolescents’ interdependent and independent self-construal a shortened version of the instrument developed by Singelis (1994) was used (Schwarz et al., 2001). The original instrument uses 12-items to measure each Independence and Interdependence. To shorten the instrument the five highest loading items for each scale from a factor analysis reported by Singelis (1994) were chosen. Furthermore, the reference group indicated in the items was changed from “persons around me” or “my group” to “family” and “family members” where applicable. This was done to tap a domain-specific family self-construal because the family is the main focus of the study. The final instrument consisted of 10 items with items 3, 4, 5, 6, and 7 representing Independence, and items 1, 2, 8, 9, and 10 representing Interdependence. Adolescents were asked to indicate their agreement with the following 10 statements on a 5-point scale from 1 = ‘Strongly disagree’ to 5 = ‘Strongly agree.’

1. It is important to me to respect decisions made by my family.

2. I often have the feeling that my relation with my family is more important than my own accomplishments.

3. I prefer to be direct and forthright when dealing with family members.

4. I enjoy being unique and different from family members in many respects.

5. My personal identity, independent of my family, is very important to me.

6. I value being in good health above everything else.

7. Being able to take care of myself is a primary concern for me.

8. My happiness depends on the happiness of my family.

9. It is important for me to maintain harmony within my family.

10. I would sacrifice my self-interest for the benefit of my family.

Table 8

Factor Loadings for Independence/Interdependence

Item Independence Interdependence

05 My personal identity, independent of my family, is very

important to me. .74 -.08

04 I enjoy being unique and different from family members in

many respects. .68 -.07

07 Being able to take care of myself is a primary concern for me. .57 .18 06 I value being in good health above everything else. .43 .29 09 It is important for me to maintain harmony within my family. .05 .71 02 I often have the feeling that my relation with my family is more

important than my own accomplishments. .00 .65 01 It is important to me to respect decisions made by my family. -.03 .65 10 I would sacrifice my self-interest for the benefit of my family. .09 .64 08 My happiness depends on the happiness of my family. .08 .61 03 I prefer to be direct and forthright when dealing with family

members. .11 .48

Note. Principal component analysis with Varimax-rotation. Bold printed factor loadings mark items that were included in the respective scales. Pooled Solution, n = 2566.

Principal Component Analysis (Pooled Solution)

An inspection of the scree-plot suggested a two-factorial structure with eigenvalues of 2.57 and 1.47 for the first and second factor, respectively, and an explained variance of 40.40 %. The pattern of factor loadings showed that all Interdependence items loaded substantially on the second factor without cross-loadings on the first factor (see Table 8).

Regarding Independence, only three of five items showed clear loadings on the first factor without cross-loadings on the second factor. Items 3 and 6 did not load substantially on any of the two factors. Therefore, there were 5 items making up the scale of Interdependence and 3 items making up the scale of Independence.

Cross-Cultural Equivalence and Reliabilities

The Tucker’s Phi values after the target rotations showed that the country-specific factor solutions were very similar to the pooled solution except for South Africa (both factors) and for Israel (only Independence, see Table 9). Thus, the Independence/Interdependence construct showed substantial cross-cultural equivalence. For the South African adolescents, the scree-plot also suggested two factors in the culture-specific factor analysis, but the pattern of factor loadings was not in line with the expected item-assignment of Independence/

Interdependence. In Israel, only one of the Independence items showed substantial loadings.

This led to Israeli adolescents’ low equivalence to the pooled solution with respect to Independence.

Table 9

Cross-Cultural Equivalence and Reliability of Independence/Interdependence

Tucker‘s Phi Cronbach’s Alpha

Culture Independence Interdependence Independence Interdependence

China .96 .99 .47 .75

France .93 .96 .39 .78

Germany .98 .99 .40 .74

India .94 .98 .47 .69

Indonesia .96 .98 .36 .60

Israel .74 .93 .45 .75

Japan .95 .99 .63 .74

South Africa .69 .73 .39 .36

Switzerland .97 .98 .48 .82

Turkey .91 .98 .48 .72

Satisfactory to high reliabilities resulted for the five Interdependence items with the exception of South Africa (see Table 9). In contrast, the three Independence items showed very low reliabilities. These low reliabilities lead to the exclusion of the Independence scale from further analyses. Therefore, only for the dimension of Interdependence a measure was generated by averaging across the respective items.

3.4.1 Value of Children

This is an instrument from the original VOC study (Arnold et al., 1975) (Schwarz et al., 2001). The items refer to economic-utilitarian, social, and psychological/emotional values of children. The adolescents had to rate reasons for wanting a child and reasons for not

wanting a child. Only a selection of the original items was used in the present study.

Additionally, newly developed items and some items from the Family and Fertility Survey (FFS) (Pohl, 1995) were included. Altogether, there were 18 VOC-items referring to reasons for wanting a child. Adolescents had to indicate on a 5-point scale from 1 = ‘Not important at all’ to 5 = ‘Very important’ how important for them personally were the following reasons people may give for wanting to have children in general:

1. Because a child helps around the house.

2. Because any new family member makes your family more important.

3. Because having children increases your sense of responsibility and helps you to develop.

4. Because it is a joy to have a small baby.

5. Because it is fun to have young children around the house.

6. Because of the pleasure you get from watching your children grow.

7. Because of the special feeling of love that develops between a parent and a child.

8. Because parenthood improves your standing and betters your reputation among your kin.

9. Because people with children are less likely to be lonely in old age.

10. Because raising children helps you to learn about life and yourself.

11. Because some of your older relatives feel that you should have more children.

12. Because your life will be continued through your children.

13. To be sure that enough children will survive to adulthood.

14. To carry on the family name.

15. To have one more person to help your family economically.

16. To have someone to love and care for.

17. When it is a duty to have children according to your belief.

18. Your children can help you when you’re old.

Principal Component Analysis (Pooled Solution)

The pattern of eigenvalues (scree-plot) of the pooled solution suggested extracting two factors with eigenvalues of 5.94 and 2.08, and with a combined explained variance of 44.53 %. The first factor combined economic/utilitarian and social/normative reasons for having children and was labeled Utilitarian/Normative VOC. The second factor was made up of items reflecting emotional values of children (see Table 10). Items 02, 09, and 12 showed cross-loadings and no substantial loadings on one of the factors. They had to be excluded to reach a factor solution with clearly separable dimensions. There were eight items representing the Utilitarian/Normative VOC (01, 08, 11, 13, 14, 15, 17, 18) and seven items representing the Emotional VOC (03, 04, 05, 06, 07, 10, 16).

Table 10

Factor Loadings for the Value of Children Construct

Item Utilitarian/

Normative VOC

Emotional VOC 15 To have one more person to help your family economically. .71 .08 14 To carry on the family name. .67 .15 11 Because some of your older relatives feel that you should have

more children. .63 .02

01 Because a child helps around the house. .62 .16 18 Your children can help you when you’re old. .62 .19 08 Because parenthood improves your standing and betters your

reputation among your kin. .60 .28

13 To be sure that enough children will survive to adulthood. .58 .08 17 When it is a duty to have children according to your belief. .55 .18 09 Because people with children are less likely to be lonely in old age. .49 .34

12 Because your life will be continued through your children. .46 .39

12 Because your life will be continued through your children. .46 .39