Cross-Cultural Research Methodology As a Case of Construct Validity

W. Fred van Raaij, Tilburg University
ABSTRACT - Cross-cultural consumer research witnesses an increasing interest of researchers, both for the managerial strategy of multinational companies and in order to establish the universality or specificity of theoretical construct and findings.
[ to cite ]:
W. Fred van Raaij (1978) ,"Cross-Cultural Research Methodology As a Case of Construct Validity", in NA - Advances in Consumer Research Volume 05, eds. Kent Hunt, Ann Abor, MI : Association for Consumer Research, Pages: 693-701.

Advances in Consumer Research Volume 5, 1978      Pages 693-701

CROSS-CULTURAL RESEARCH METHODOLOGY AS A CASE OF CONSTRUCT VALIDITY

W. Fred van Raaij, Tilburg University

ABSTRACT -

Cross-cultural consumer research witnesses an increasing interest of researchers, both for the managerial strategy of multinational companies and in order to establish the universality or specificity of theoretical construct and findings.

In this paper the validity of the hypothetical constructs is discussed in order to create functionally equivalent constructs across cultures and equivalence of samples from the cultures.

A theoretical structure in which a hypothetical construct is embedded serves as a manner for attaining the functional equivalence of instruments and constructs in order to interpret the findings and to implement managerial strategy.

INTRODUCTION

In 1889 Sir Francis Galton presided over a meeting of the Royal Anthropological Institute of Great Britain in which a paper was read using cross-cultural data. At the meeting Galton criticized this paper, raising the problem of independence of cases. Elder (1976) describes this as follows:

"If two societies exhibit the same juxtaposition of institutions or traits, how does one know whether these two societies represent two independent cases of that juxtaposition? What if society A "borrowed" the juxtaposition from society B, in which case the juxtaposition is historical rather than functional? What if A and B are merely variants of some common overarching society C? In either case, what appear to be two cases may actually be two illustrations of the same case, in which event the cross-cultural comparison has not produced any further evidence for the generalization of the juxtaposition than did the single-culture observation."

This illustrates the theoretical value of cross-cultural research: the generalization of findings and constructs across cultures, or better, following Popper (1959), the attempts to invalidate a theory or "behavioral law" as culture-specific.

From a practitioner's viewpoint, there exists the need for information on consumer behavior in other cultures and nations in order to develop international marketing operations. It's not the validation or invalidation of theories that is important here, but rather the answer on the question of applying a global or culture-specific marketing strategy.

Triandis, Malpass, and Davidson (1973) mention the following reasons for conducting cross-cultural research:

"To check the generality of behavioral laws; to increase the range of our observations on variables of interest; to determine the variations found in subjective culture variables in different settings; to take advantage of natural experiments involving combinations of variables (conditions) that cannot be obtained in the laboratory; to study the manifestation of psychological variables in different cultural contexts; and to study cultures for their own sake."

Such research can serendipitously lead to a better insight into the communalities and specificities of consumer behavior in different cultural settings and the universality of behavioral "laws" and concepts across cultures (Poortinga, 1975).

Some introductory remarks make clear the ground for the remainder of this paper:

First, cross-cultural research as comparative research is not unique. All behavioral research is comparative; that is, it involves a comparison of experimental and a control group, e.g., samples of working and non-working wives, husband and wife roles, middle and lower social class consumers, users and non-users of a service, etc. Cross-cultural research is unique in that it considers different cultural settings or "geographical environments'' (Koffka, 1935) that shape consumer behavior and are shaped by consumer attitudes and behaviors. Therefore, it means that we cannot control for the "disturbing'' factors that may create unwanted differences between the samples under study, and it means that we cannot use the same instrument (questionnaire) for both samples in different cultures.

Secondly, consumer research is largely "made in the USA." Concepts and instruments have been developed in the United States and not in the "other" culture of the cross-cultural study. This may introduce "ethnocentrism" in the type of questions we address, the concepts we employ, and the explanations we give of the results.

For instance, the study of "consumer satisfaction" is relevant in a western mass-consumption society but not, or in a different way, in a developing country. Another example is a book by Nieuwenhuijze (1963) reporting the results of an analysis of Islam, that was banned in Indonesia on the grounds that a non-Muslim could not properly examine either the history or the tenets of Islam. A last example is a questionnaire statement: "A good citizen is responsible for shoveling the sidewalk in front of his home" for measuring "social responsibility.'' This statement assumes private ownership of houses, one-family housing, and a climate with snow in winter, and is clearly not applicable in an African country. In that case, we have to measure "social responsibility" with another statement or set of statements.

Thirdly, the terms "cross-cultural" and "cross-national" are used in the literature. I prefer "cross-cultural" because this term reflects more possible differences in consumer behavior than "cross-national." In cross-cultural research we compare consumer behavior in different cultures, including subcultures, e.g. differences between French- and English-speaking Canadians; black and white subcultures; and minority ethnic groups with the majority ethnic group. Many of the problems and caveats that apply for cross-cultural research are also valid for market segmentation studies using the segmentation variables of race, ethnic background, and social class.

Cross-cultural studies on consumer behavior include: shopping behavior (Green and Langeard, 1975; Douglas, 1976), innovator characteristics (Green and Langeard, 1975), female role perception (Douglas, 1976), media usage (Douglas, 1977), family buying decisions (Hempel, 1974), life-style research (Plummer, 1977), and information seekers (Thorelli, Becker, and Engledow, 1975; Anderson and Engledow, 1977; Becker, 1976). These studies compare a US sample of consumers with samples in France, England, Canada, Mexico, and Germany, all nations of the western world. Sheth and Sethi (1973) developed a theory of cross-cultural buyer behavior for the diffusion of innovations, and Dubois (1972) developed a framework for the study of the cultural factors that affect the rate of adoption of an innovation. Green and White (1976) give methodological considerations in cross-cultural consumer research. De Vos and Hippler (1969) and Triandis, Malpass, and Davidson (1973) review cross-cultural research in psychology.

HISTORY

Cross-cultural research is not new. A long cross-cultural tradition exists in anthropology. Benedict (1934) suggests that cultures - especially of small, isolated groups - are often integrated wholes wherein the parts coalesce around certain basic values. To extract parts from the wholes, compare them (out of context) with parts extracted from other cultures, and then conclude that one has found cross-cultural similarities or differences does gross injustice to social reality. Osgood (1967), using the semantic differential technique, demonstrated that persons speaking different languages have different subjective cultures. Arguments still continue over the degree to which language determines thought processes and thus subjective culture: The Whorfian Hypothesis (Whorf, 1956). The argument in anthropology is that cross-cultural data can be used primarily for the identification of cultural uniqueness or the establishment of cross-cultural contrasts, describing the contrasting cultures in their own context. However, in consumer research, our interest is not primarily the Eskimo of Northern Canada or the Temne of Sierra Leone (Berry, 1967). We study more-or-less similar cultures in Northern America or Western Europe. We may expect more culturally unique behavior patterns in the study of Japanese consumers or consumers in developing countries.

Borrowing the psycholinguistic terminology, a distinction will be made between an emic (culturally specific) and an etic (culturally universal) approach. In an emic approach (from phonemics) the behavior is described in terms and concepts of that specific culture and with internal criteria; hence cross-cultural comparison is difficult or impossible. In an etic approach (from phonetics) the behavior is described in universal categories and with external criteria. Cross-cultural comparison is feasible in the etic approach. Davidson (1977) criticizes the "emic-etic dilemma" and advocates the use of etic concepts (with which comparisons among cultures can be made) and emic ways of measurement (in the context of the culture under study). We will see that this is the central issue in cross-cultural methodology, also referred to as the functional equivalence of measuring instruments.

Another historical development is the evolutionary or revolutionary theory of socio-economic change. Cross-cultural contrasts are used to identify, in a quasi-experimental way, why socio-economic changes in one culture differ from those in another culture. Weber (1958) discusses why industrial capitalism developed in western Europe but not in China or India (Protestant ethic). Certain cultural conditions were different or absent in China or India, and industrial capitalism did not develop there. Hsu (1963) explained why clan, caste, and club emerged respectively in China, India, and the United States. Implicit in Hsu's analysis is the counterfactual premise that if family structure had been different in the USA, clans or castes rather than clubs might have emerged.

Marx (1972) argued for a contextual view of social phenomena; the basic mode of production in a culture or nation determines social and cultural phenomena. Engels (1949) identified a sequence of politico-economic stages based on the means of production and the class relationship to the production means. Knowing these stages (feudal, bourgeois-capitalist, socialist) one can develop generalizations for societies in the same stage. Parsons (1966) distinguishes primitive, advanced primitive, archaic, advanced intermediate, and modern societies, identifying certain prerequisites as necessary but not sufficient conditions for a society's transition from one stage to another. Meaningful cross-cultural comparisons can only be made between societies at the same level of development (differentiation). Consumer behavior differs depending on whether it occurs in a capitalist nation, a dependent nation (colony) or a socialist nation.

SAMPLING

Cross-cultural research involves the comparison of samples from different populations (cultures). What type of sample we select depends on our research objective.

1. For a comparison of income and age distributions in different cultures one needs a random sample to employ statistical tests for the significance of differences.

2. For descriptive studies on attitude-value structures, attitude-behavior relationships, life-style, or opinion research sample representativeness is essential. This means that the sample has to be representative for the culture from which it is drawn. In this case, we study the relationships between variables in different cultures and we are not primarily interested in the absolute scores (distributions) of the variables as such.

3. In some cross-cultural studies we need functional equivalence of samples, e.g., comparing salesmen in an American and a Japanese department store. Here we try to find "similar" organizations and samples of the personnel of the organizations. Generalization of the results to the general culture, however, becomes hazardous.

4. For causal studies across cultures, we may employ "matched" samples with matching variables such as age, income, education, etc. One increases the power of the statistical test for the significance of differences using paired observations. But the danger exits that one eliminates real cross-cultural differences through the matching procedure. One has to have a theory to distinguish between the matching variables and the independent variables in a research design.

Except for the random samples, sampling procedures involve an a priori distinction between variables "that are controlled for" in sampling and "real" independent variables that are important from a theoretical viewpoint. To control for sampling inadequacies and to partial out covarying factors that could not be controlled for in the sampling procedure, the technique of analysis of covariance (ANCOVA) may be used. But one has to be aware of the dangers of these control procedures. Controlling for obvious variables such as age, income, and education in the sampling procedure may create samples unrepresentative of the population. A sample from a culture in India with the same age, income, and education distribution as a US sample may be largely unrepresentative of the Indian culture.

Pretest data will often help to pinpoint the best sample. Pearlin and Kohn (1966) were interested in sampling people from equivalent class positions in the USA and Italy. Their preliminary data showed that income and education were not suitable for manifesting similar class positions, but that occupational prestige was. They approached the sampling problem by minimizing the extremeness of the groups chosen to be interviewed, finally selecting members of the middle and working classes. They concluded that "some intra-class variation is obscured by using only those two broad social class categories but what is lost in precision is gained in increased cross-national comparability" (p. 468).

Brislin, Lonner, and Thorndike (1973) advocate the plausible rival hypotheses approach. The differences or similarities that are found may be attributed to different sampling methods or to different qualities of the samples (age, socioeconomic status) rather than to "real" cross-cultural differences or similarities. A rival hypothesis may explain the obtained differences/ similarities as well as the hypothesis one is interested in. In the research design one has to eliminate all plausible rival hypotheses that may explain the results from sampling inadequacies. It may be clear that "identical'' samples in cross-cultural research are actually impossible; that the costs of random samples are often too high to justify their additional benefits. Essentially, equivalence of samples, based on the control of variables in a design that eliminates plausible rival hypotheses, must be the objective of sampling in cross-cultural research.

Rival hypotheses cannot always be ruled out in a design with only one sample from each culture. If we study the differences and similarities of French and American wives, it may be illuminating to include control samples of French and American husbands. How do the wives differ from the husbands (not necessarily their husbands!) in the two cultures, and an interaction (in the statistical sense) may exist between sex and culture. A good example is the 2 x 2 design by Douglas (1976), where she compares working and non-working wives in France and in the USA, using multivariate ANCOVA. Cross-cultural differences in Douglas' study prove to explain more variance than the differences between working and non-working wives.

Thorelli, Becker, and Engledow (1975), on the other hand, conclude that information-seeking consumers (higher education and higher income) are more similar across cultures (the USA and Germany) than the control group of non-information-seekers. Katona, Strumpel, and Zahn (1971) find more similarity across nations for white-collar than for blue-collar respondents.

FORMAL AND FUNCTIONAL EQUIVALENCE

Formal equivalence of a questionnaire or a questionnaire item means that an identical questionnaire and respondent instruction are employed for all respondents involved in the study. Formal equivalence is the ideal of the survey technician but is impossible in cross-cultural research in nearly all cases. Translation problems and procedural problems such as whether to use a mail questionnaire or a personal interview, may arise. Formal equivalence can be attained to some extent through back-translation (translating the questionnaire and instructions into the other language and then back into the original language by independent translators, and then comparing the two versions), the use of bilingual respondents, expert judges, and writing translatable English. See Brislin, Lonner, and Thorndike (1973, pp. 32-81). Formal equivalence is treated as an instrument reliability problem: How reliably does the instrument measure the variables in different cultural settings? It will be clear that we can easily attain formal equivalence in measuring behavioral variables (e.g., possession of durables; number of hours of TV watching; frequency of shopping trips). In these cases, we are interested in measuring and comparing behavioral variables as such. But as soon as these behavioral variables become an indicator or operationalization of a hypothetical construct, it becomes another case. "Frequency of church attendance" may be selected as an indicator of the hypothetical construct "religiosity." In this case, church attendance may be a good indicator of religiosity in one culture but not in another.

Functional or conceptual equivalence of instruments for measuring a certain construct is a validity problem. The survey methodologist strives for functional equivalence of instruments, i.e., the instrument must have a functionally equivalent meaning for the respondents in both cultures. Although questionnaire items may be different for different cultures the intention is that they measure the same hypothetical construct. "Religiosity'' can be operationalized by church attendance in one culture while meditation may be an indicator of religiosity in another culture. In such cases, functional equivalence of instruments can only be attained through a theoretical a priori framework and pilot studies how a hypothetical construct becomes manifested in actual behavior or attitudes in the culture under study.

Straus (1969) distinguishes formal and functional equivalence of stimulus material (questionnaire, test, task) and of mode of quantifying. Instead of the terms "formally and functionally equivalent" he uses the terms "phenomenally identical" and "conceptually equivalent." Straus (1969) obtains the four possibilities of Figure 1, using our terminology.

FIGURE 1

A TAXONOMY OF MEASUREMENT EQUIVALENCE

The mode of quantifying involves the development of norms for coding and scoring. A psychological test can only be used for a new population provided new norms are developed based on standardizing the test among samples of that new population (culture). The four quadrants have the following interpretation:

1. Culturally universal measurement refers to research methods that use identical stimulus materials in each culture, except for translation, and that record and quantify the responses in identical ways. This means the formal equivalence of instrument and quantification of responses. This is the "etic" approach to cross-cultural research. It is likely that there are only a limited number of instances. The meaning and intention of this observable behavior, however, may be different across cultures.

2. Culturally ipsatized measurement refers to instances in which the identical instrument is used but the recording and/or interpretation of the responses are judged relative to others in that culture (internal criteria) rather than relative to external standards of scoring or interpretation. An attitude scale measuring "ecological responsibility," for example, may be identical across cultures but the norms for high or low scores may differ across cultures.

3. Culturally modified measurement refers to measurements in which the indicators for the hypothetical construct are altered in order to make them culturally appropriate but the original scoring is maintained. Przeworski and Teune (1966/67) provide an example in their cross-cultural study in the U.S.A. and Poland. The hypothetical construct "political activity" is measured in the U.S. with the indicators "contribute money to parties or candidates" and "place sticker on car," but in Poland they used the equivalent indicators "fight for the execution of economic plans" and "attempt to influence economic decisions." Przeworski and Teune (1966/67) also had a set of five identical indicators for both cultures.

4. Culturally specific measurement in the fourth quadrant represents the minimum formal equivalence in order to achieve the maximal functional equivalence. This is the "emic" approach to cross-cultural research, although it does not state the absolute uniqueness of cultures (Benedict, 1934). To measure the extent to which a sample of youth participate in an adolescent subculture, items of dress reflecting adolescent fads of the particular subculture might be used as a part of the index. In one subculture such indicators may be tight pants and in another subculture bell-bottom pants.

Straus (1969) provides four techniques for establishing functional equivalence of measures:

1. Use of expert judges to evaluate the appropriateness of the items, questions, and instructions. This procedure may help provide for the content validity of the instruments.

2. Revalidation and restandardization of the instrument in the culture in which it will be employed (ipsatizing).

3. Semantic differential procedure to obtain "psycholotical translation" of indicators. Blood and Takeshita (1964), as cited by Straus (1969), found that the word deito, a translation of the English "date", was not widely enough known in Japan to use it in the questionnaire. They therefore substituted the work "otsusia" with connotations similar to "date."

4. Construct validation. Construct validity is determined by establishing that certain hypothetical constructs account to some degree for performance on a psychological test, or that the test accounts to some degree for performance on variables for which the test is taken as explanatory (APA standards). Zaltman, Pinson, and Angelmar (1973) define construct validity as "the extent to which an operationalization measures the concept (hypothetical construct) which is purports to measure." For instance, Blood and Takeshita (1964) measured the relative power of husband and wife in the Japanese family. The item "Who decides about what type of car to get?" had to be eliminated and had to be substituted by some major purchase. Formal equivalence is given up for the sake of functional equivalence. The construct validation of such an instrument would consist of determining if the scores are related to other variables which theory or prior knowledge indicated should be related (Straus, 1969).

Another approach to functional equivalence of instruments is provided by Przeworski and Teune (1966/67) distinguishing between identical and equivalent indicators. An identical indicator is universal (etic) across cultures and an equivalent indicator is specific (emic) for a culture. We have seen the example of equivalent indicators in their study on political activity in the U.S.A. and Poland. By combining identical and equivalent indicators reliable and valid comparisons can be made. By this procedure a longer and more culturally relevant, and perhaps more valid, set of items is acquired. A set of items for a culture consists of universal and specific items. Inter-item correlation reveals that the specific items add to the universal items.

Green and White (1976) make a distinction between the functional equivalence of instruments, of phenomena ("shopping"), and of constructs ("Cognitive consistency''). However, an instrument is functionally equivalent across cultures if it measures the "same" construct.

FIGURE 2

INDICATORS, INTERVENING VARIABLES, AND HYPOTHETICAL CONSTRUCTS - SHETH (1967)

THEORETICAL STRUCTURE

Construct validity is not only the extent to which an operationalization (set of indicators) measures the hypothetical construct which it purports to measure. It is also the relationship of the hypothetical construct with other hypothetical constructs and intervening variables in a theory. A theory is a set of propositions relating hypothetical constructs, formulated in such a way that testable hypotheses can be derived (de Groot, 1969). Maehr (1974) gives an example of the relationships of the hypothetical construct "achievement motivation" and other hypothetical constructs. Cross-cultural studies of "achievement motivation" suggest that the psychological dynamics underlying achievement behavior (e.g. entrepreneurial action) might not be the same across cultures. In other words, variables which motivate individuals to be economically productive in one culture may differ from those motivating productive work in another culture. These variables may be the need for affiliation, concern for the reaction of others, and strong obligations and identifications to one's group. In Japan the basic cultural value is loyalty to one's group rather than the advancement of one's personal status. The hypothetical construct "achievement motivation" is conceptually different across cultures and even irrelevant in some cultures. We have to specify when and under what conditions achievement motivation will occur. Maehr (1974) proposes three theoretical structures in the study and prediction of achievement motivation.

1. In the first structure, social structure (SS) or, more specifically, the social learning experience provided by the milieu in which the person develops, influences the subjective culture (SC) or the assumed predispositions (attitudes, values, norms) and the personality traits (P) to respond in a given manner. SC determines achievement motivation (M) and achievement behavior (B). In a diagrammatical expression this may be indicated as:

SS --------> SC/P --------> M --------> B.

This is the traditional approach by McClelland (1961) where the personal history of the individual determines his achievement motivation. Only through changes in SC/P achievement motivation may be changed.

2. A second structure may be diagrammed as follows:

S --------> (SC/P --------> M --------> B

where a situation or context (S) has demonstrable effects on motivation (M). Subjective culture (SC) and personality (P) are placed in parentheses indicating the relative unimportance of the attitudes, norms, values and personality variables. Here situational variables play a major role in determining achievement motivation. Maehr (1974) identified four situational dimensions that appear to be critical in determining achievement motivation and behavior. An individual achieves as a member of a social group and is influenced by the norms of his group. Another dimension is the locus of control. How much (internal) control has the individual over the outcomes (success or failure) of his behavior? Feedback about performance has to be communicated to the achiever through interpersonal contacts. A task dimension involves task difficulty and intrinsic task interest. The situational model implies that the above situational variables greatly affect achievement motivation and behavior. It is better changing the situation rather than the personality of the individual to enhance achievement motivation.

3. The third theoretical structure represents the most sophisticated analysis of achievement motivation:

C --------> SC/P <--------> S = M --------> B.

This structure identifies both subjective culture/personality (SC/P) and situation (S) as simultaneously critical variables. This SC/P x S model goes back to Lewin; one of the prominent and highly developed of such models is the motivation model proposed by Atkinson (1964, Chapter 10). See also Van Raaij and Wandwossen (1978). The question becomes to find what situations match what motives in order to elicit a specific behavior such as achievement motivation. This theoretical structure requires considerable theoretical and methodological sophistication.

4. Not proposed by Maehr (1974) but as a counterconceptualization, Inkeles (1971) states that personality structure does not necessarily precede modernization and economic development. The influences may go in the opposite direction. That is, institutional changes may change the outlook of individuals exposed to them. According to Inkeles' theory modern social and cultural structures and systems may cause individuals' attitudes, values, behaviors, and need dispositions to change in such a way as to make effective adaptation to these institutions by becoming increasingly "modern." The greater the degree of exposure and the longer the exposure time to the various modern institutions, the more the individual will shift toward the modern direction on the psychological continuum, Inkeles' theory offers an alternative way for emphasizing the role of the individual in social institutions. The diagram becomes as follows:

PE --------> SS --------> S --------> M = B --------> SC/P.

The physical environment (PE) and/or the social structure (SS) expose an individual to situations (S) of modernization where achievement motivation (M) and behavior (B) are the best adaptive responses. M and B change the subjective culture (SC) and personality (P) of the individual afterwards making it consonant with behavior.

These four theoretical structures have consequences for the measurement and change of achievement motivation. In the first structure, projective personality tests have been used, and we may change M and B through changes in SC/P (personality change and attitude change). In the second theoretical structure, M and B are determined by situational dimensions. Group norms, locus of control, feedback, and task dimensions are indicators of M and B, and we may influence M and B by changing the situational dimensions. The interaction of SC/P and S determines M and B in the third structure. In the fourth structure, achievement motivation is the result of environmental factors (PE and SS), and is in fact adaptive behavior to the changing (modernized) environment. Personality and attitude (SC/P) measures are not a good predictor of M and B in a changing environment but the situation (S) itself may become an indicator of achievement motivation (M) and behavior (B). Maehr (1974) advocates a methodology of identifying settings in which achievement motivation occurs or does not occur in various cultures and then proceeds to characterize the nature of these settings. Much of the research that attempts to understand motivational patterns of (sub)cultural groups involves placing individuals in a "middle-class biased" or "western-biased" performance setting and then observing behavior. But instead, we have to devise a theoretical structure of achievement motivation for the culture under investigation and from this theoretical structure functionally equivalents may be derived for measuring the hypothetical construct of achievement motivation.

CROSS-CULTURAL CONSUMER RESEARCH

Triandis, Malpass, and Davidson (1973) distinguish classes of distal variables (PE, SS, O) which act as independent variables in studies in which classes of proximal variables (A, SC, P, B) are the dependent variables. These classes of variables are for consumer research:

PE: physical environment: type of economy, distribution structure, climate, population density, GNP, transportation system, consumption level, health and education system. Sethi (1971) uses PE variables for clustering nations into eight groups.

SS: social structure: social class, family organization, social and sex roles, social norms, language, socialization and education patterns. Sheth and Sethi (1973) call the PE and SS variables exogenous and treat them as givens or constraints of their model. These variables are not explained in terms of their structure or any changes over time.

O: other persons: influence attempts, opinion leadership, social interaction, word of mouth, imitation, conformism/deviancy, cooperation/competition.

A: persons' abilities: intelligence, cognitive style, cognitive development, memory, information-process-ing capabilities.

SC: person's subjective culture: attitudes, values, personal norms, aesthetic judgments, opinions, life style, beliefs and stereotypes, meaning, satisfaction.

PD: personality dispositions: self-concept, anxiety, aggression, alienation, achievement and affiliation motives, adjustment, cosmopolitanism, information-sensitivity.

B: behavior: shopping, communication, consumption, and disposition behavior.

In cross-cultural consumer research, as indicated earlier, we cannot simply compare concepts from a class across cultures. We have to study the functional equivalence of these concepts (hypothetical constructs) across cultures through both their operationalization (indicators) and their relationship to other concepts in a theoretical structure. Maehr (1974) and Inkeles (1971) already provided four possible theoretical structures for the study of achievement motivation and economic behavior.

I will try to give some examples in consumer research of possible theoretical structures.

1. The cross-cultural study of consumer satisfaction has to take into account the PE variables: type of economy, distribution structure, consumption level, transportation system, and income distribution, and the SS variables: social class, family organization, social and sex roles. If the cross-cultural differences of the PE and SS variables are great it becomes meaningless to compare consumer satisfaction scores. Consumer satisfaction is a function of the PE, SS variables and probably also of O and A variables.

PE --------> SS --------> A/O --------> SC

Consumer satisfaction as such may become functionally equivalent across cultures that differ on PE/SS variables but we have to employ a different (emic) set of indicators for each culture. Comparing the consumer satisfaction of a mass-consumption society with a developing country becomes a "tour de force" in finding sets of equivalent indicators. In a case of extreme differences, we better make cross-cultural comparisons of PE/SS variables only. If only slight differences on PE/ SS variables of two cultures exist we may find sets of equivalent indicators in both cultures for the hypothetical construct "consumer satisfaction.'' The procedures may be largely emic in that they reflect the PE/SS structure in a culture but may become "pseudo-etic" in that a comparison is made between similar evaluative consumer reactions to the consumption system.

2. Douglas (1976), Green and Langeard (1975) compare among other variables grocery shopping behavior in France and the U.S.A. Shopping behavior (B) can be thought of as a function of the PE variables: transportation system, private ownership of cars, distribution (retailing) structure, consumption level, and the SS variables: family organization, social and sex roles, working or nonworking house- wives. Intervening variables from the O, SC, and PD classes may influence shopping behavior: social interaction, word of mouth, life style, attitudes, self-concept, and opinion leadership. Two theoretical structures can be hypothesized:

I:  PE --------> SS --------> (O/SC/PD) --------> B.

II: PE --------> SS --------> B --------> (O/SC/PD).

PE/SS variables influence the O/SC/PD group, and O/SC/PD determines behavior, or PE/SS determines behavior and the O/SC/PD variables change into a direction to become consistent with behavior. Green and Langeard (1975) conclude that "most of the differences that we found (between samples of French and U.S. women) could be attributed to social and environmental that characterize the two nations. These finding emphasize the need for further theoretical and empirical cross-national work to achieve a better understanding of the relationships between buyer behavior and environmental forces." Douglas (1976) concludes in a similar fashion: "Retail environmental factors seem to play a key role in shaping and conditioning behavior patterns between countries may in many respects merely reflect current market conditions such as the availability of different products and services or the number of supermarkets or small, traditional retailers." She recommends further investigation of the role of retail environmental factors in influencing customer response patterns. Shopping behavior does not reflect "real" cross-cultural differences but reflects only the different distribution structures in France and in the U.S.A. Green and White (1976) add that shopping is not functionally equivalent in France and in the U.S.A. "In France shopping is reputed to be an integral part of the housewife's social life, whereas in the United States shopping tends to be considered a chore." Even a third theoretical structure is not ruled out, namely that the distribution structure reflects the social interaction patterns, the subjective culture, and personality dispositions of the customers:

III. (O/SC/PD) --------> (SS) --------> PE --------> B.

Does the "adaptive consumer" adapt his behavior to changes in the distribution structure (model II), or do we have to change the O/SC/PD cluster too before new shopping patterns develop? (model I). Or, do "adaptive retailers" change the distribution structure in line with changes in the O/SC/PD cluster? Most retailers want us to believe that model III is true because it reflects the marketing concept.

3. Hempel (1974) found "a surprisingly high degree of cross-cultural similarity in household decision-making processes. In most instances, the differences between the roles perceived by husbands and wives within the same cultural setting were greater than the differences between cultures for either sex." Hempel compared samples in Hartford, Connecticut, and Preston-Lancaster, England, with nearly identical data collection methods, except for some minor adaptations to English word usage. Husband-wife interaction is a function of the SS variables: social class, family organization, social and sex roles, socialization and education patterns. A simple theoretical structure can be postulated:

(PE) --------> SS --------> O.

The United States and the United Kingdom are similar cultures in many respects. A valid cross-cultural comparison can be made and functionally equivalent instruments (no translation problem) are easily developed.

4. Beginning with the pioneering study of Horton (1943), social scientists have attempted to find out why people drink alcohol. A correlation has been found between a PE variable, type of economy (hunting, fishing vs. agricultural), and a B variable, alcohol consumption. In general, hunting tribes drink more heavily than tribes that depend on agriculture:

PE --------> SS --------> (PD) --------> B.

But a number of variables may intervene the PE qB relationship. The degree of societal organization has received the strongest support as an intervening variable. In more highly organized cultures there are more social controls that inhibit drinking. The reflection of social organization on the individual disposition (PD) has been called alienation. The greater the alienation the more alcohol is consumed.

A recommendation for research designs is that if two hypothetical constructs are not comparable across cultures, i.e. are not functionally equivalent, the researcher needs to go to the next-higher order construct in the theoretical structure. If consumer satisfaction is an incomparable construct, compare the distribution system and/or the social system across cultures.

CONSTRUCT VALIDITY

Construct validity is the extent to which an operationalization (indicator) measures the construct which it purports to measure (Zaltman, Pinson, and Angelmar, 1973). Or alternatively, in order to show that an indicator applies to a construct, it is necessary to derive hypotheses from the theory related to the construct, that can be tested in reality (indicators) (Cronbach, 1964, p. 105). Thus, we have to include in the operationalization of a construct the theoretical structure in which the construct is embedded.

The validity of constructs across cultures can be attained through four essential procedures:

1. The operationalizations (measuring instruments, indicators) should measure the construct which they purport to measure. If identical indicators (etic measures) are not available, a set of identical and equivalent indicators (Przeworski and Teune, 1966/67), or equivalent indicators (emic measures) should measure the hypothetical construct.

2. The validity of a construct can be established in some cases through systematic variation with a criterion. Criterion-related validity is concerned with prediction (predictive validity) or with simultaneous performance of some other variable (concurrent validity). For instance, we measure the concept of "innovation-proneness'' in several cultures using functionally equivalent indicators. The degree of adoption of an innovation may be the criterion measure. This means that we can validate our measuring instrument for the hypothetical construct only afterwards.

3. A hypothetical construct has relations to other constructs in a theoretical structure. Examples are provided in an earlier paragraph. Causal explanations for the obtained cross-cultural differences and similarities are proposed in a theoretical structure.

4. A last essential procedure of construct validity is the similarity of components of the construct across cultures. Frijda and Jahoda (1966) call this "dimensional identity." "Intelligence", for instance, consists of several components (verbal, numerical, etc.) obtained after factor-analyzing the concept. "Intelligence'' is functionally equivalent across cultures if the same components are found in different cultures. "Perceived risk" may differ in its components (social, physical, financial risk, etc.) across cultures but that means that the construct validity of the hypothetical construct "perceived risk" is low. Especially in cross-cultural assessment (test) psychology, the componential similarity of a construct is stressed. However, we may study the components of "intelligence" in another culture with an emic approach, i.e. investigating the unique componential structure of a construct in that culture.

ETHNOCENTRISM

One of the biases in cross-cultural research has been called "ethnocentrism", i.e. bias arising from blindness for the unique and different characteristics of another culture. Consumer research is largely "made in the U.S.A." with all the risks that western-American- or middle-class biases pervade this type of research in the research questions we address, the concepts and theories we use, and the interpretations we give. We should encourage researchers in other cultures to study their own reality rather than to replicate American studies. Faucheux (1976) recommends a separate development of social psychology on the European continent. We may recommend the same for consumer research in order to become aware of the ethnocentrism and the blindness to other cultural values. A rationalistic tradition in Europe encourages theoretical speculation, sometimes at the cost of a proper care for empirical validation. This tradition produced scientists such as Freud, Lorentz, Piaget, Levi-Strauss, and Chomski, and does not have the illusion that a theory will spring from a gathering of "facts" and measurements collects with a preconceived idea. Some suggestions for avoiding ethnocentric biases are:

1. Double investigation: The researchers in a cross-cultural study perform their research or experiment in both cultures and compare their results.

2. Decentering of indicators: Questionnaire items are developed in both cultures are then an identical set and two equivalent sets are formed. See Brislin, Lonner, and Thorndike (1973, pp. 37-39).

3. Select and/or train the staff members in the other culture to introduce their own questions, concepts, and interpretations.

GALTON'S PROBLEM AND EVOLUTIONARY THEORY

How do we get subclasses of cultures in which a valid cross-cultural comparison is possible? And if we want to generalize a finding, how do we get two or more cultures that are not simply two or more illustrations of the same case? This is Galton's problem. Evolutionary theory stated that only cultures in the same stage of differentiation can be meaningfully compared. In any case, we have to classify cultures or nations into diverse subclasses or types. Sethi's (1971) cluster-analytical approach may serve this purpose on a global scale. More refined classification is needed within the western cultures. To a certain extent, we may assume that concepts are functionally equivalent within such a subclass. Gal-ton's problem, however, is to find mutually "independent'' cultures (in different subclasses) to make universal generalization of a phenomenon possible, if one's hypothesis proves to be true for these independent cultures. We cite Elder (1976, p. 217):

"Galton's problem bears an interesting relationship to the problem raised by critics of universal cross-cultural generalizations, to the effect that one can meaningfully compare only cultures rather similar to each other in terms of size, structural complexity, etc. In both cases, the methodological goal is to identify cultural subsets to which specific cultures can be assigned. In the case of Galton's problem, the purpose of assigning cultures to subsets is to try to guarantee subsequently that cultures being compared are drawn from different subsets, thereby permitting universal cross-cultural generalizations. In the case of the critics of such generalizations, the purpose of assigning cultures to subsets is to try to guarantee subsequently that cultures being compared are drawn from the same subset, thereby permitting limited but meaningful cross-cultural generalization."

CONCLUSION

From a theoretical viewpoint mere comparison of consumer responses to questionnaire items in different cultures does not guarantee meaningful and useful information for theoretical purposes and for managerial action. Functional equivalence of constructs and instruments has to be established in order to guarantee "real" cross-cultural comparison. Plausible rival hypotheses that also explain the obtained differences/similarities have to be ruled out by using equivalent samples, and functionally equivalent measures and quantification modes.

Cross-cultural research as comparative research creates some additional problems because of the non-equivalence of instruments that have to be employed. Only through (1) equivalence of indicators, (2) criterion-related validity of constructs, (3) a theoretical structure, and (4) componential similarity of constructs can the validity of hypothetical constructs be determined.

REFERENCES

Ronald Anderson and Jack Engledow, "A Factor Analytic Comparison of U.S. and German Information Seekers," Journal of Consumer Research, 3(March, 1977), 185-196.

John W. Atkinson, An Introduction to Motivation, New York: Wiley, 1964.

Helmut Becker, "Is there a Cosmopolitan Information Seeker?" Journal of International Business Studies, 7(Spring, 1976), 77-89.

Ruth Benedict, Patterns of Culture, Boston: Houghton Mifflin, 1934.

F. Kenneth Berrien, "Methodological and Related Problems in Cross-Cultural Research," International Journal of Psychology, 2(1, 1967), 33-43.

John W. Berry, "Independence and Conformity in Subsistence Level Societies," Journal of Personality and Social Psychology, 7, 1967, 415-418.

R. O. Blood and Y. J. Takeshita, "Development of Cross-Cultural Equivalence of Measures of Marital Interaction for U.S.A. and Japan," Transactions of the Fifth World Congress of Sociology, 4(1964), 333-334.

Richard W. Brislin, Walter J. Lonner, and Robert M Thorndike, Cross-Cultural Research Methods, New York: Wiley, 1973.

Lee J. Cronbach, Essentials of Psychological Testing, New York: Harper & Row, 1964.

Andrew R. Davidson, "The Etic-Emic Dilemma: Can Methodology provide a Solution in the Absence of Theory?" in Y. H. Poortinga (Ed.), Basic Problems in Cross-Cultur-al Psychology, Amsterdam: Swets & Zeitlinger, 1977, 49-54.

Adriaan D. de Groot, Methodology: Foundations of Inference and Research in the Behavioral Sciences, The Hague: Mouton, 1969.

George A. de Vos and Arthur A Hippler, "Cultural Psychology: Comparative Studies of Human Behavior," in G. Lindzey and E. Aronson (Eds.), The Handbook of Social Psychology, Reading: Addison-Wesley, 1969, Volume 4, 323-417.

Susan P. Douglas, "Do Working Wives read different Magazines from Non-Working Wives?" Journal of Advertising, 6(Winter, 1977), 40-43 and 48.

Bernard Dubois, "A Cultural Approach to the Study of Diffusion and Adoption of Innovations," in M. Venkatesan (Ed.), Proceedings of the 3rd Annual Conference of the Association for Consumer Research, 1972, 840-850.

Joseph W. Elder, "Comparative Cross-National Methodology," Annual Review of Sociology, 2(1976), 209-230.

Friedrich Engels, "The Origin of the Family, Private Property, and the State," in: Karl Marx and Friedrich Engels: Selected Works, 2(1949), Moscow: Foreign Languages Publishing House, 155-296.

Claude Faucheux, "Cross-Cultural Research in Experimental Social Psychology," European Journal of Social Psychology, 6(3, 1976), 269-322.

Nico Frijda and Gustav Jahoda, "On the Scope and Methods of Cross-Cultural Research," International Journal of Psychology, 1(2, 1966), 109-127.

Robert T. Green and Philip D. White, "Methodological Considerations in Cross-National Consumer Research," Journal of International Business Studies, 7(Fall-Winter, 1976), 81-87.

Donald J. Hempel, "Family Buying Decisions: A Cross-Cultural Perspective," Journal of Marketing Research, 11(August, 1974), 295-302.

D. Horton, "The Functions of Alcohol in Primitive Societies: A Cross-Cultural Study," Quarterly Journal on the Study of Alcoholism, 4(1943), 199-220.

F. L. K. Hsu, Clan, Caste and Club, Princeton: Van Nostrand. 1963.

A. Inkeles, "Continuity and Change in the Interaction of the Personal and the Socio-Cultural Systems," in: B. Barber and A. Inkeles (Eds.), Stability and Social Change, Boston: Little Brown, 1971.

George Katona, Burkhard Strumpel, and Ernest Zahn, Aspirations and Affluence, Comparative Studies in the United States and Western Europe, New York: McGraw-Hill, 1971.

Kurt Koffka, Principles of Gestalt Psychology, New York: Harcourt Brace, 1935.

Martin L. Maehr, "Culture and Achievement Motivation," American Psychologist, 29(December, 1974), 887-896.

Karl Marx, Grundrisse: Foundations of the Critique of Political Economy, New York: Random House, 1973.

C. A. O. Nieuwenhuijze, Cross-Cultural Studies, The Hague: Mouton, 1963, Chapter 8.

Charles E. Osgood, "On the Strategy of Cross-National Research into Subjective Culture," Social Science Information, 6(1, 1967), 5-37.

Charles E. Osgood, William H. May, and Murray S. Miron, Cross-Cultural Universals of Affective Meaning, Urbana, IL: University of Illinois Press, 1975.

Yalcott Parsons, Societies: Evolutionary and Comparative Perspectives, Englewood Cliffs: Prentice-Hall, 1966.

L. Pearlin and M. Kohn, "Social Class, Occupation, and Parental Values: A Cross-National Study," American Sociological Review, 31(1966), 466-479.

Joseph T. Plummer, "Consumer Focus in Cross-National Research," Journal of Advertising, 6(Spring, 1977), 5-15.

Ype H. Poortinga, "Some Implications of three Different Approaches to Intercultural Comparison," in: J. W. Berry and W. . Lonner (Eds.), Applied Cross-Cultural Psychology, Amsterdam: Swets & Zeitlinger, 1975, 327-332.

Karl R. Popper, The Logic of Scientific Discovery, New York: Dover, 1959.

Adam Przeworski and Henry Teune, "Equivalence in Cross-National Research," Public Opinion Quarterly, 30(4, 1966/67), 551-568.

S. Prakash Sethi, "Comparative Cluster Analysis for World Markets," Journal of Marketing Research, 8 (August, 1971), 348-354.

Jagdish N. Sheth, "A Review of Buyer Behavior," Management Science, 13(August, 1967), B-718-756.

Jagdish N. Sheth and S. Prakash Sethi, "An Approach to Developing a Theory of Cross-Cultural Buyer Behavior," Paper presented at the Annual Meeting of the American Psychological Association, Montreal, August 27-31, 1973.

Murray A. Straus, "Phenomenal Identity and Conceptual Equivalence of Measurement in Cross-National Comparative Research," Journal of Marriage and the Family, 31(May, 1969), 233-241.

Hans B. Thorelli, Helmut Becker, and Jack Engledow, The Information-Seekers -- An International Study of Consumer Information and Advertising Image, Cambridge, Mass; Ballinger Press, 1975.

Harry C. Triandis, Roy S. Malpass and Andrew R. Davidson "Psychology and Culture,' Annual Review of Psychology, 24(1973), 355-378.

W. Fred van Raaij and Kassaye Wandwossen, "Motivation-Need Theories and Consumer Behavior," in H. K. Hunt (Ed.) Advances in Consumer Research, Vol. 5 Association for Consumer Research, 1978.

Max Weber, The Protestant Ethic and the Spirit of Capitalism, New York: Scribner's, 1958.

B .L. Whorf, "A Linguistic Consideration of Thinking in Primitive Communities," in: J. B. Carroll (Ed.) Language, Thought, and Reality: Selected Writing of of Benjamin Lee Whorf, Cambridge, Mass: MIT Press, 1956.

Gerald Zaltman, Christian R. A. Pinson, and Reinhard Angelmar, Metatheory and Consumer Research, Hinsdale, IL Dryden Press, 1973.

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