An Examination of Ethnicity Measures: Convergent Validity and Cross-Cultural Equivalence

Michel Laroche, Concordia University
Annamma Joy, Concordia University
Michael Hui, Concordia University
Chankon Kim, Concordia University
ABSTRACT - With its focus on English-Canadian and French-Canadian ethnicity, this study examines five measures of ethnicity with respect to their convergent validity and cross-cultural equivalence. Results show that the five measures including four unidimensional measures and self-identification do converge in capturing both French-Canadian and English-Canadian ethnicity. The assumption of cross-cultural equivalence of the five measures was unsubstantiated. However, three of the measures showed equivalent estimates of measurement parameters as indicators of English-Canadian and French-Canadian ethnicity.
[ to cite ]:
Michel Laroche, Annamma Joy, Michael Hui, and Chankon Kim (1991) ,"An Examination of Ethnicity Measures: Convergent Validity and Cross-Cultural Equivalence", in NA - Advances in Consumer Research Volume 18, eds. Rebecca H. Holman and Michael R. Solomon, Provo, UT : Association for Consumer Research, Pages: 150-157.

Advances in Consumer Research Volume 18, 1991      Pages 150-157

AN EXAMINATION OF ETHNICITY MEASURES: CONVERGENT VALIDITY AND CROSS-CULTURAL EQUIVALENCE

Michel Laroche, Concordia University

Annamma Joy, Concordia University

Michael Hui, Concordia University

Chankon Kim, Concordia University

[The authors gratefully acknowledge the financial support of the Social Sciences and Humanities Research Council of Canada.]

ABSTRACT -

With its focus on English-Canadian and French-Canadian ethnicity, this study examines five measures of ethnicity with respect to their convergent validity and cross-cultural equivalence. Results show that the five measures including four unidimensional measures and self-identification do converge in capturing both French-Canadian and English-Canadian ethnicity. The assumption of cross-cultural equivalence of the five measures was unsubstantiated. However, three of the measures showed equivalent estimates of measurement parameters as indicators of English-Canadian and French-Canadian ethnicity.

INTRODUCTION

Studies of ethnicity as a correlate or cause of consumption patterns have emerged as a major stream of consumer research. Studies to date have focused on such diverse ethnic markets as black consumers (Sexton 1972; Sturdivant 1973), Hispanics (Deshpande, Hoyer, and Donthu 1986; Wallendorf and Reilly 1983), Jewish Americans (Hirschman 1981), Asian Americans (Lee 1989), and French Canadians (Schaninger, Bourgeois, and Buss 1985). Despite the proliferation of findings in the area, the fundamental issues concerning the conceptualization and operationalization of the concept of ethnicity are in need of clarification and refinement. Criticisms have been directed at the overly simplistic treatment of the concept and the operationalization inadequately reflecting the complex domain of ethnicity (O'Guinn and Faber 1985; Valencia 1985). This study has the following purposes: the first is to examine more closely the concept of ethnicity itself and the diverse operational practices as shown in the current consumer behavior literature; the second is to examine the convergent validity of several measures of ethnicity that are found in the past studies of consumer behavior; and finally, the cross-cultural equivalence of these measures of ethnicity is investigated in the bi-cultural environment of Quebec.

LITERATURE REVIEW

Definition of Ethnicity

The concept of ethnicity, as depicted in the anthropological and sociological literature, in a loose sense, refers to the character or quality encompassing several cultural identifiers which is used to assign people to groupings. According to Cohen (1978), ethnicity is defined as a series of "nesting dichotomizations of inclusiveness and exclusiveness." The process of assigning persons to groups is both subjective and objective, carried out by self and others, and depends on what diacritics are used to define membership. Yinger's (1985) definition of an ethnic group encompasses a similar conception. He defines it as "a segment of a larger society whose members are thought, by themselves and/or others, to have a common origin and to share important segments of a common culture and who, in addition, participate in shared activities in which the common origin and culture are significant." (p. 159) Four elements often cited in the existing definitions of ethnicity are: 1. a self-perpetuating population; 2. sharing of cultural values; 3. a field of communication and interaction; and 4. a membership which identifies itself, and is identified by others as constituting a distinguishable category (Barth 1969). The important point is that the concept characterizes solidarity and loyalty among the group members generated by sharing of common cultural traits. Such traits, according to Weber (1961), may be as diverse as to include common customs, language, religion, values, morality, and etiquette.

Measurement of Ethnicity

While most researchers tend to share a similar conception of ethnicity, with respect to its operationalization, there are two schools of thought whose perspectives are in disagreement. They are labeled as "subjective" and "objective" approaches. The difference between the two is significant for the reason that the groupings resulting from each approach do not coincide (Nagata 1974; Anderson and Frideres 1981). In the subjective approach, ethnicity is viewed as a matter of personal belief and its operational definition reflects the individual's own psychological identification based on internal beliefs relating to his/her cultural attributes that are perceived to be relevant (Barth 1969; Shibutani and Kwan 1965).

In contrast, the objective approach dictates that researchers measure ethnicity with a set of objective cultural attributes such as religion, language, and cultural tradition which they perceive are relevant. While the subjective approach is sometimes criticized for its extreme subjectivism, the objective approach has drawn criticisms for the imprecision inherent in determining the "objective criteria". Some researchers (c.f., Van den Berghe 1975) propose a combination of the both approaches in conceptual and operational definitions to overcome the limitations in each.

Much of the traditional marketing research of consumer ethnicity relied on a single objective indicator in identifying ethnic membership. Hirschman (1981) points out that this practice may have been largely due to the lack of well-conceived a priori conceptual and operational scheme. Rather, classification of subjects into the ethnic group of interest was frequently post hoc in design.

A more elaborate multi-dimensional operationalization is found in the study of Hispanic ethnicity by Valencia (1985). The author in developing an "Hispanicness" index, combines six indicators: self-identification, English language ability, the extent of Spanish language use at home, language preference, relative length of residence in the U.S., and miscegenation. Similarly, the general acculturation scale proposed by O'Guinn and Faber (1985) for Hispanics incorporate multiple dimensions tapping national origin, a general language preference, and demographics.

With respect to the issue of subjective/objective operationalization of ethnicity, the tendency in the current consumer behavior literature has been to emphasize the subject's self-perception in the ethnic identification process. Except for a few studies which incorporate both subjective as well as objective measures (Valencia 1985; Bergier 1986), self-identification has become a prevalent approach in operationalizing ethnicity (Deshpande, Hoya, and Donthu 1986; Hoyer and Deshpande 1982; Hirschman 1981).

Multidimensionality of Ethnicity

Ethnicity is commonly conceived as a multidimensional concept. Without reference to one specific ethnic identity, the existing literature reveals, among others, the following facets of ethnicity: language use (Tzu 1984; Kim 1985); religion (Yinger 1985; Segalman 1967); social interaction (Driedger 1975); endogamy (Sanua 1965; Driedger 1975); media communication (Kim 1985; Shibutani and Kwan 1965).

However, the various cultural traits of individuals take on different magnitude of importance as ethnicity indicators, depending upon which ethnic group is under consideration. For instance, in studying Jewish ethnicity, the language dimension plays a less important role than in the case of studying Hispanic ethnicity in the U.S. or the French-Canadian ethnicity in Canada. While there is little disagreement regarding the multidimensionality of the concept, delineation of essential elements in measuring a particular ethnic identity remains a challenge primarily faced by those researchers who rely on the objective operational approach. Their objective assessment of one's ethnic identity requires that one or more of important dimensions be identified and used in the measurement process. On the other hand, in the subjective measurement approach, the individual's self-identification of ethnicity is presumed to be based on a set of ethnicity attributes that the individual perceives to be important.

It follows then, the diversity in the current operational definitions of ethnicity is largely a result of differences among the various objective measures in the selected set of ethnicity attributes. One can suspect, however, that various dimensions of ethnicity are not independent of one another. For instance, an individual's language preference (or use) is highly likely to be correlated with his/her ethnic mode of social interaction as well as media communication. Similarly, a certain degree of overlap should exist between self-perception of ethnic identity and the various objective criteria of ethnicity.

RESEARCH OBJECTIVES

As can be summarized from the above review, the past measures of ethnicity found in the consumer behavior literature fall in to one of the following categories: those with a single or multiple objective indicators; those with a subjective indicator using self-identification; and those with objective indicators and self-identification. The past operational practices raise several issues worth examining. Among these are the extent of convergence between the various unidimensional, single objective item measures of ethnicity, the extent of convergence between self-identification and single or multidimensional objective measures of ethnicity, and the stability of various unidimensional ethnicity indicators in a measurement situation involving more than one ethnic identity.

The focus of this study is on the convergent validity of five measures of English-Canadian and French-Canadian ethnic identification in the context of Quebec's bi-cultural=l environment. The five selected measures include self-identification and four unidimensional indicators, namely, language use in various social communication settings, religious beliefs, social interaction, and upbringing. Also investigated is the stability of these measures in the identification the two different ethnic identities of Quebec, i.e., English-Canadian and French-Canadian ethnicity.

METHODS

Data

Data for this come from a survey of residents in various districts of the Greater Montreal area conducted in 1988. The population consisted of only those people who identified themselves as French or English Canadian. The area sampling method employed in this study involved first numbering of the 1986 census tracts of the Greater Montreal area and drawing fifteen tracts randomly with the use of a table of random numbers. As the census tracts were drawn, some judgments were exercised to eliminate those with large concentrations of industrial/commercial activities and/or other ethnic minorities. Within each of the chosen census tracts, a number of streets were further picked at random and efforts were made to survey as many households on these streets as possible. Interviewers, after the initial introduction, used a filter question to screen out those who identified themselves as belonging to neither group. Those who qualified were asked whether they preferred a French or English questionnaire. The self-administered questionnaire was then left with the consenting individual to be picked up at a later time. The sample analyzed in this study consists of 810 respondents.

Measurement

The questionnaire first measured respondents' language use in eleven interpersonal and mass-communication contexts. They were asked to estimate the percentage of times they used French, English or other languages (adding up to 100) in the following contexts: 1. with spouse; 2. with children; 3. with relatives; 4. at work; 5. when watching television; 6. when listening to radio; 7. when reading newspapers; 8. when reading magazines or books; 9. when shopping; 10. with close friends; 11. when in school.

The ethnic mode of social interaction, as all other measures to follow, was measured in two parts; French-Canadian social interaction and English-Canadian social interaction. The primary reason for this was that the degree of French-Canadian interaction may not mirror that of English-Canadian social interaction. i.e., a respondent may interact with other ethnic cultures, just as s/he may use languages other than French or English in various communication contexts. The following questions using 10-point Likert scales (l=Strongly Disagree; 10=Strongly Agree) were used to measure the degree of English-Canadian (French-Canadian) social interaction:

1. All my closest friends are Anglophones (Francophones).

2. All my neighbors are Anglophones (Francophones).

3. I am very comfortable dealing with Anglophones (Francophones).

4. I go to places where I can be with Anglophones (Francophones).

5. I often participate in the activities of Anglophone (Francophone) community or political organizations.

For the measure of self-identification, respondents were asked to indicate the degree of agreement (1 =Strongly Disagree; 10=Strongly Agree) with the statement, "I consider myself to be Anglophone (Francophone)." Again, instead of assuming the unidimensionality which would have required one measurement scale, individuals' self-identification with English-Canadian ethnicity as well as that with French-Canadian ethnicity were measured separately.

The respondent's religious beliefs were measured using three 10-point Likert statements:

1. I consider myself to be a strong Protestant (Catholic) believer.

2. I had a strong Protestant (Catholic) childhood upbringing.

3. Protestant (Catholic) beliefs are important part of my life.

Finally, the measure of upbringing contained two Likert statements:

1. My parents were Anglophones (Francophones).

2. I grew up in mostly Anglophone (Francophone- neighborhood.

ANALYSIS AND RESULTS

Prior to conducting the analysis to examine the convergence of these measures in capturing English-Canadian and French-Canadian ethnicity as well as their cross-cultural equivalence, items in each of the ten measures (five English-Canadian and five French-Canadian ethnicity measures) were analyzed for their internal consistency. This analysis was aimed at testing the feasibility of combining the items in each measure into a simple index. Toward this end, ten correlation matrices and Cronbach's reliability coefficients were computed and examined. The average correlation among the items in a measure ranged from .40 for French-Canadian social interaction to .83 for English-Canadian religious beliefs. The Cronbach's reliability coefficients, which ranged between .77 and .97, similarly indicate a high level of internal consistency among the items in all of the ten measures. Subsequently, the items in each measure were averaged for the ensuing analysis.

The investigation of convergence among the five measures (for each ethnic identity) and their cross-cultural equivalence employed the multi-sample LISREL analysis (Joreskog and Sorbom 1983). The original sample was divided randomly into two split-half subsamples. In one, only the items which measure E-C ethnicity were considered while in the other subsample, only the items which measure F-C ethnicity were considered. The examination of convergence involved testing a confirmatory factor model which specifies that the five index measures are indicators of E-C ethnicity for the first split-half subsample and F-C ethnicity for the other subsample (see Figure A). The goodness of fit of the model will thus indicate if each set of five indicators converge in measuring their respective ethnicity (EC ethnicity and F-C ethnicity). The multi-sample LISREL also allows for testing the equivalence of these measures in capturing E-C ethnicity and F-C ethnicity, to be discussed later.

FIGURE A

MULTI-SAMPLE LISREL MODEL

TABLE 1

GOODNESS-OF-FIT OF VARIOUS COMPETING MODELS

Convergent Validity of the Five Measures

If these five measures have convergent validity, they should all load on the same common factor. The model testing the convergent validity of the measures as shown in Figure A (M-1) produces a chi-square value of 20.05 with 10 degrees of freedom. The probability level of this chi-square value (p=.029) indicates that the goodness-of-fit of the model is below the common .05 acceptable level.

The undesirable goodness-of-fit level can be attributed to a number of factors related to LISREL itself such as large sample size (N=810) and violation of the multi-normal distribution assumption. A review of the modification indices obtained from the LISREL analysis also suggests that for the E-C ethnicity sample, a correlation may exist between the error terms of two of the indicators, language and social interactions. The possibility of a pair of correlated errors provides an alternative explanation to why M-1 does not produce a good fit. This latter argument is confirmed when the parameter (642) representing the correlation

between the two error terms is set free in another multi-sample LISREL analysis. The respecified model (M-2) gives a chi square of 11.17 with 9 degrees of freedom (p=. 07) implies an excellent fit. Two other goodness-of-fit indicators, Goodness-of-Fit Index (GFI1 and GFI2) and Root Mean Square Residual (RMR1 and RMR2), also suggest that the respecified model (M-2) has a better fit than the original model (M-1, see Table 1).

Conceptually speaking, a correlation between the two error terms in the E-C sample is deemed possible as the language indicator actually refers to the respondents' language use in various types of social interactions (e.g. at work, with friends, ...). On the other hand, the two error terms are found to be independent in the F-C ethnicity sample. The finding may be due to the fact that in Quebec, French is the dominant language and therefore, the use of French does not necessarily imply positive attitudes in dealing with francophones.

Factorial Congruence: between E-C Ethnicity and F-C Ethnicity

The findings obtained from the first part of our analysis provide strong evidence for convergence validity for the five measures. Moreover, the one-factor structure is found to hold when the five indicators are employed to operationalize either E-C ethnicity or F-C ethnicity. The finding also provides preliminary evidence to the cross-cultural equivalence of the five variables as measures of ethnicity or what Olmedo (1979) has labelled as derived etics - "the development of explanatory constructs that are applicable to all cultures" (P. 1064). -A more stringent test of cross- cultural equivalence, however, involves both the factor structure as well as the estimated parameters (i.e. the l's and the d's). The equivalence of the factor structure has already been confirmed, the second part of our analysis therefore deals with the equivalence of the estimated parameters.

Three multi-sample LISREL analyses are conducted to test (a) the equivalence of the l's (the l elements are set to be invariant between the two samples), (b) the equivalence of the d's (the d elements are set to be invariant between the two samples), and (c) the equivalence of both the l's and the d's (both the l and the d elements are set to be invariant between the two samples). Chi-square values of 28.27 (d.f.=14, p=.013), 35.44 (d.f.=14, p=.001), and 47.19 (d.f.=19, p=.000) are obtained from the three models, labelled as M3, M4, and M5 respectively (Table 1). Since all the three constrained models are nested with M-2, the chi-square difference between M-2 and M-3 (or M-4 or M-5) also has a chi-square distribution. All the three incremental chi-square values, 13.81 (d.f.=5) for M-3, 20.98 (d.f.=5) for M4, and 32.73 (d.f.=10) for M-5 are significant.

TABLE 2

LISREL ESTIMATES FOR M-7

Accordingly, the hypothesis that the l or the d matrix (or both) is equivalent between the two samples is therefore rejected. However, the findingdoes not deny the possibility that some of the parameters may be identical between the two samples. In fact, the modification indices obtained from the most constrained model (M-5) suggest that different estimates are required only for some of the parameters. Multi-sample LISREL analysis gives (a) a chi-square value of 33.48 (d.f.=17, p=.010) when the constraints on the l and the d of social interaction are relaxed (labelled as M-6 in Table 1), and (b) a chi-square value of 16.77 (d.f.=15, p=.333) when the constraints on the l's and the d's of social interaction and religion are relaxed (labelled as M-7 in Table 1).

The three models, M-5, M-6 and M-7, are nested and therefore the chi-square difference can be used as an indicator of relative fit between M-5, M-6 and M-7. Compared between M-5 and M-6, the latter model has 2 degrees of freedom less but its chi-square value is also 13.71 smaller. This shows that M-6 has a significantly better fit than M-5 (p<.001). By the same token, M-7 can also be proved to have a significantly better fit than either of the two more constrained models, M-5 and M-6 (p<.001 in either case). On the other hand, the chi-square difference between M-2 (the least constrained model) and M-7 (2.31, d.f =6, p>.10) is nonsignificant. This last finding suggests that further relaxation of constraints on M-7 will not produce any significant improvement in terms of goodness-of-fit. It is therefore concluded that M-7 is the r model which exhibits the optimal fit. The following discussion will therefore be based on this optimal model.

An examination of the LISREL estimates for M-7 reveals that among the five indicators included in this study, "religion" is the least valid measure of either E-C ethnicity or F-C ethnicity. Compared with the other four indicators, "religion" produces the smallest (though significant) l value and the greatest d value in both the E-C and F-C ethnicity subsamples (Table 2). Nonetheless, the estimated value of l is greater and the estimated value of d is smaller in the E-C sample than in the F-C sample. The finding suggests that Protestantism as anindicator of E-C ethnicity shows greater validity than Catholicism as an indicator of F-C ethnicity.

"Self-identification" and "language" have a similar factor loading (k) and measurement error (a) and their estimated values show that as a single indicator of ethnicity, either variable is considerably better than the other three indicators (Table 2). Moreover, the estimated values are found to be identical across the E-C ethnicity and the F-C ethnicity subsample and this suggests that either "self-identification" or "language use" has satisfied the most stringent requirement as an "etic" (vs "emic") measure of ethnicity. The above findings in part corroborate Hirschman's (1981) argument that the subjective self-labeling method of measuring ethnicity is the most appropriate approach. More recently, language use has also been employed to operationalize ethnicity in the study of cultural impact on consumption patterns (Kim, Laroche, and Joy 1990). The findings obtained from this study provide strong support to the employment of language use as an alternative operationalization of ethnicity.

Finally, as in the case of "religion", different estimated values of l and d were obtained for "social interaction" between the E-C ethnicity and the F-C ethnicity subsample. Our findings indicate that "social interaction" is a better measure when it is used to operationalize F-C ethnicity (a higher estimated l and a lower estimated d are obtained) than when it is used to operationalize E-C ethnicity (see Table 2). Since francophones consist of the majority of the Quebec population, it is not surprising that even for anglophones, interactions with francophones may be inevitable in many occasions. An alternative explanation is that when one is uncertain about the ethnic origin of the people one interacts with, one will likely regard them as francophones who may be in fact English, Italians, Greeks or others. These arguments may explain why "interactions with francophones" as an indicator of F-C ethnicity shows greater validity than "interactions with anglophones" as an indicator of E-C ethnicity.

Reliability of Findings

Since all the above findings are derived from two random split-half samples, there is a possibility that the findings are an outcome of capitalization on chance. To examine the reliability of the findings, we ran all the tests again with five other sets of random split-half samples. In general, the findings are consistent with that obtained from the original analysis.

CONCLUSIONS

With its focus on English-Canadian and French-Canadian ethnicity in Quebec, this study examined five measures of ethnicity in terms of their convergent validity and cross-cultural equivalence. The five measures including four unidimensional objective measures and self-identification did show convergence in capturing both French-Canadian and English-Canadian ethnicity. These measures, however, displayed varying degrees of validity. In both cases involving E-C and F-C ethnicity, measures of language use and self-identification proved to be the most valid indicators. Language and self-identification separately or in conjunction have recently been the two most frequently used elements of Hispanic and English- or French-Canadian ethnicity measures seen in consumer behavior research. Findings of this study provide support for those measures.

Findings from the test of cross-cultural equivalence of the five measures are also worth reiterating. The assumption of equivalence of the overall variable structure was unsubstantiated in this research. However, it was shown that three of the measures (self-identification, language use, and upbringing) have equivalent values of measurement parameters as indicators of E-C and F-C ethnicity. Thus, using the "emic-etic distinction" (Price-Williams 1975; Olmedo 1979), these three measures meet the requirement of etic measures of ethnicity, i.e., they are equally valid measures of two ethnic identities. The search for equivalent measurement for many consumer behavior constructs including ethnicity in cross-cultural settings is a difficult but important task for researchers. Most of the traditional measurement instruments were developed within narrow cultural perspectives. The increasing level of importance attached to the cross-cultural and cross-national investigations of consumer behavior may subsequently require a reexamination and reorientation o, measurement practices.

REFERENCES

Anderson, Alan B. and James Frideres (1981), Ethnicity in Canada:- Theoretical Perspectives, Toronto: Butterworths.

Barth, Frederik (1969), Ethnic Groups and Boundaries: The Social Organization of Culture Difference, London: Allen and Unwin.

Bergier, Michel J. (1986), "Predictive Validity of Ethnic Identification Measures: An Illustration of the English/French Classification Dilemma in Canada," Journal of the Academy of Marketing Science, 14 (Summer), 3742.

Cohen, Ronald (1978), "Ethnicity: Problems and Focus in Anthropology" Annual Review of Anthropology, VQ1 7, 379403.

Deshpande, Rohit, Wayne D. Hoyer and Naveen Donthu (1986), 'The Intensity of Ethnic Affiliation: A Study of the Sociology of Hispanic Consumption," Journal of Consumer Research, 13 (September), 214-220.

Driedger, Leo (1975), "In Search of Cultural Identity Factors: A Comparison of Ethnic Students," Canadian Review of Sociology and Anthropology, 12 (2), 150-162.

Hirschman, Elizabeth C. (1981), "American Jewish Ethnicity: Its Relationship to Some Selected Aspects of Consumer Behavior," Journal of Marketing, 45 (Summer), 102-110.

Hoyer, Wayne D. and Rohit Deshpande (1982), "Cross-Cultural Influences on Buyer Behavior: The Impact of Hispanic Ethnicity," in Assessment of Marketing Thought and Practice, eds., Orville C. Walker et al., Chicago: American Marketing Association, 89-92.

Joreskog, Karl G. and Dag Sorbom (1983), LISREL V1, Chicago: National Educational Resources, Inc.

Kim, Chankon, Michel Laroche, and Annamma Joy (1990), "An Empirical Study of Ethnicity on Consumption Patterns in a Bi-Cultural Environment," Advances in Consumer Research, Vol. 17, eds., Marvin E. Goldberg, Gerald Gorn and Richard W. Pollay, Provo, UT: Association for Consumer Research, 839-846.

Kim, Young Yun (1985), "Communication and Acculturation," in Intercultural Communication: A Reader, eds., Samovar, Larry A. and Richard E. Porter, Belmont, California: Wadsworth Publishing Company, 379-386.

Lee, Wei-Na (1989), "Acculturation and Consumption-Related Adjustments: The Chinese Subculture," in Proceedings of the Society for Consumer Psychology, ed. D.W. Schumann, American Psychological Association, 127-134.

Nagata, Judith A. (1974), "What is a Malay? Situational Selection of Ethnic Identity in a Plural Society," American Ethnologist, 1(2): 331 -350.

O'Guinn, Thomas C. and Ronald J. Faber (1985), "New Perspectives on Acculturation: The Relationship of General and Role Specific Acculturation with Hispanics' Consumer Attitudes," in Advances in Consumer Research, Vol. 12, eds., Elizabeth C. Hirschman and Morris B. Holbrook, Provo, UT: Association for Consumer Research, 113-117.

Olmedo, Esteban L., "Acculturation: A Psychometric Perspective," American Psychologist, Vol. 34, No. 11 (November), 1061-1070.

Price-Williams, D.R. (1975), Explorations in Cross-Cultural Psychology, San Francisco: Chandler and Sharp.

Sanua, Victor D. (1965), "A Study of Adolescents Attending Jewish Community Centers in New York," Journal of Jewish Communal Service, 41, 401 -424.

Schaninger, Charles M., Jacques B. Bourgeois and W. Christian Buss (1985), "French-English Canadian Subcultural Consumption Differences," Journal of Marketing, 49 (Spring), 82-92.

Segalman, Ralph (1967), "Jewish Identity Scale: A Report." Jewish Social Studies, 29.

Sexton, Donald E. (1972), "Black Buyer Behavior," Journal of Marketing, 36 (October), 36-39.

Shibutani, T. and K.M. K wan (1965), Ethnic cation: A Comparative Approach, New York: MacMillan Co.

Sturdivant, F.D. (1973), ' Subculture Theory: Poverty, Minorities, and Marketing," in Consumer Behavior: Theoretical Sources, eds., S.F. Ward and T.S. Robertson, Englewood Cliffs, NJ: Prentice Hall, 469-520.

Tzu, Lao (1984), "Strangers' Adaptation to New Cultures," in Communicating with Strangers: An Approach to Intercultural Communication, eds., Gudykunst, W.B. and Y.Y. Kim, Addison-Wesley Publishing Company, Inc., 205-222.

Valencia, Humberto (1985), "Developing an Index to Measure Hispanicness," in Advances in Consumer Research, Vol. 12, eds., E.C. Hirschman and Morris B. Holbrook, Provo, UT: Association for Consumer Research, 118-121.

Van den Berghe, Pierre L. (1975), "Ethnicity and Class in Highland Peru," in Ethnicity and Resource Competition in a Plural Society, ed., Leo A. Despres, The Hague: Mouton Publishers, 71-85.

Wallendorf, Melanie and Michael Reilly (1983), "Ethnic Migration, Assimilation, and Consumption," Journal of Consumer Research, 10 (December), 293-302.

Weber, Max (1961), "Ethnic Groups," in Theory of Society, eds., Talcott Parsons et al., New York: Free Press, 301-309.

Yinger, Milton J. (1985), "Ethnicity," Annual Review of Sociology, 11, 151-180.

----------------------------------------