The Ethnicity and Consumption Relationship

Johanna Zmud, University of Southern California
Carlos Arce, NuStats, Inc.
ABSTRACT - This paper investigates the influence of situational factors on the relationship between ethnicity and consumer behavior. Findings based on analyses using data from a 1989 Survey of Hispanics in the Los Angeles metropolitan area conducted for The Los Angeles Times, suggest that behavior is a function of ethnicity, social surroundings, and product type. Results also indicate that ethnicity is a dynamic, complex construct that is comprised of both inherited and acquired characteristics. While the available data are limited, this analysis raises the possibility that the situational ethnicity model proposed by Stayman and Deshpande (1989) can be improved with the addition of the variable, cultural identity.
[ to cite ]:
Johanna Zmud and Carlos Arce (1992) ,"The Ethnicity and Consumption Relationship", in NA - Advances in Consumer Research Volume 19, eds. John F. Sherry, Jr. and Brian Sternthal, Provo, UT : Association for Consumer Research, Pages: 443-449.

Advances in Consumer Research Volume 19, 1992      Pages 443-449

THE ETHNICITY AND CONSUMPTION RELATIONSHIP

Johanna Zmud, University of Southern California

Carlos Arce, NuStats, Inc.

ABSTRACT -

This paper investigates the influence of situational factors on the relationship between ethnicity and consumer behavior. Findings based on analyses using data from a 1989 Survey of Hispanics in the Los Angeles metropolitan area conducted for The Los Angeles Times, suggest that behavior is a function of ethnicity, social surroundings, and product type. Results also indicate that ethnicity is a dynamic, complex construct that is comprised of both inherited and acquired characteristics. While the available data are limited, this analysis raises the possibility that the situational ethnicity model proposed by Stayman and Deshpande (1989) can be improved with the addition of the variable, cultural identity.

The dramatic rise in immigrants to the U.S., especially from countries in Latin America and Asia, has prompted researchers to investigate the hypothesis that one's ethnicity influences consumption patterns (Stayman and Deshpande, 1989; Deshpande, Hoyer, Donthu, 1986; Pitts, Sheth, and Valencia, 1986; Hirschman, 1981; Belk, 1974, and Lutz and Kakkar, 1975). Most recent work in this area has focused upon refining the concept of ethnicity. As Deshpande, Hoyer, Donthu (1986) assert, the lack of discipline that has been followed in defining ethnicity (i.e., surname, country of origin, paternal ancestry, self-identification, language spoken at home) has lead to inconsistent and contradictory findings which hinder theory building. The discrepancy in definitions is problematic also in that it implies differing assumptions about the nature of ethnicity. Contributing much to both the conceptualization of ethnicity and to the literature on consumer subcultures is the work of Stayman and Deshpande (1989), who have specified a model of situational ethnicity. (See Figure 1.)

According to Stayman and Deshpande, situational factors (i.e., social surroundings and type of product) influence the relationship between ethnicity and consumer behavior. In other words, such behavior is a function of ethnicity, social surroundings, and type of product. Social surroundings and type of product are rather straight-forward variables. However, ethnicity continues to be a troublesome concept. "Felt ethnicity" is a new concept identified by Stayman and Deshpande; it is defined as a transitory psychological state of individuals that is manifested in different ways in different situations. It is distinguished from the idea of ethnicity as a stable, sociological trait that is manifested in the same way at all times. The theoretical rationale underlying this idea of dynamic ethnicity is the contention by sociologists (Yancey et al., 1976) and psychologists (McGuire et al., 1978) who argue in different literatures that ethnicity is not just who one is, but how one feels in different situations. In specifying ethnicity as a dynamic concept, Stayman and Deshpande have found consistent support for the effect of situational factors on consumer behavior. For example, in their most complex test of the model (a four-factor, mixed effects design), Stayman and Deshpande (1989) found that Hispanic, Asian, and Anglo subjects switched the type of food consumed (among Mexican food, Chinese food, and American food) based on their social context. Their use of situation-specific, felt ethnicity increased explained variance in likelihood of choice over that explained by either self-designated ethnicity alone or use of non-situation specific, felt ethnicity. The present study seeks to more rigorously test this finding through the use of structured-equation modeling with the LISREL program (Joreskog and Sorbom, 1988). Thus, the first hypothesis to be tested in this study is as follows:

H1: Behavior is a function of felt ethnicity, social surroundings, and product type.

With the understanding that an individual's ethnicity is a complex state of being, Stayman and Deshpande specify it with by use of two theoretical concepts: self-designated ethnicity and felt ethnicity. In terms of the relationship between these two variables, felt ethnicity is a function of self-designated ethnicity, and antecedent state. According to Stayman and Deshpande, antecedent state refers to how ethnic one feels in the face of bias towards the target group. Stayman and Deshpande supported the earlier work of Deshpande, Hoyer, and Donthu (1986) in finding that self-designated ethnicity and antecedent state influence felt ethnicity. However, a relationship not specified by Stayman and Deshpande was the notion that social surroundings may influence felt ethnicity as well. For example, research in language acquisition and sociolinguistics has uncovered the phenomenon of "code switching". This is a practice of bilinguals who switch languages according to the social context in which the language use takes place. While early in the field of linguistics code-switching was thought to connote the idea that the individual was in a transitional state C going from one language system to another one. More recent work has shown that code-switching is actually a function of changes in an individual's strength of ethnic identity which may vary according to social context (Gumperz, 1978, 1976; Valdes Fallis, 1980). Thus, it may be that the Stayman and Deshpande model could be improved with the conceptualization of social surroundings as an antecedent condition to felt ethnicity and not simply as a moderator of the felt ethnicity and behavior relationship.

Another potentially rich modification in the model involves the addition of a new variable, cultural identity. While Stayman and Deshpande found that situation-specific felt ethnicity is a better predictor of behavior than non-situation specific or self-designated ethnicity, the differences in R-square among the three conditions are not highly significant. This implies that the construct, felt ethnicity, can be improved. In going back to earlier work of Deshpande (Deshpande, Hoyer, and Donthu, 1986), it appears as though Weber's definition of "ethnicity" was used in early theorizing. According to Weber (1986), ethnicity is a common inherited and inheritable trait that actually derives from a common descent. Viewed in this way, ethnicity is seen as a stable trait. One way to view ethnicity as having both stable and varying properties is to suggest (as does anthropological theory) that ethnicity functions in two distinct ways: a cultural identity which is stable and a social identity which varies (Fitzgerald, 1974). Social identities, then, vary to suit the social surroundings. The individual shifts behavior with each adaptation to new situational demands. Social identification facilitates change; and opens the individual up to experiences that produce new identities or new social stances. In anthropological terms, social identity refers to the phenomenon of "acculturation." This is the integration of ethics at the primary level (i.e., kinship, neighbors, and close friends) (Alba, 1976, Gallo, 1974).

FIGURE 1

CASUAL MODEL OF SITUATIONAL ETHNICITY AND CONSUMPTION

Cultural identity, on the other hand, refers to one's basic group identity. It refers to identity at the group level vis-a-vis some other group. A sense of cultural identity supplies a unifying "identity" thread as people strive toward consistency from one situation to another. McGuire et al. (1978) suggests that individuals in a multi-ethnic society such as the United States are likely to have a set of ethnic and other identities that may be differentially salient. Thus, there appears to be theoretical support for the addition of both social surroundings and cultural identity to the list of variables that predetermine felt ethnicity. The second hypothesis to be tested is:

H2: Felt ethnicity is a function of self-designated ethnicity, antecedent state, social surroundings, and cultural identity.

In essence, I am suggesting that the more elaborate situational ethnicity model (see Figure 2) is a more promising representation of reality than the Stayman and Deshpande model presented in Figure 1. In using structural-equation modeling to test this proposition, I submit the final hypothesis:

H3: The modified situational ethnicity model is a good fit to the observed data.

METHODS

Design

This study is a sub-study, using data collected in a 1989 Telephone Survey of Hispanics in the Los Angeles Metropolitan area, conducted for the Los Angeles Times newspaper. This researcher was a member of the team that conducted the study in July/August 1989. The researchers followed two criteria in drawing a stratified sample representative of the area: (1) random digit sample production of households in all zip codes which contained at least 20% Hispanic population; (2) Spanish surname random sampling in all zip codes with less than 20% Hispanic population; and (3) purposive sampling of Hispanics without telephones. A total of 1,733 Hispanic adults, including 165 in non-telephone households were interviewed in summer 1989. A total of 420 non-Hispanic interviews were conducted as comparative data, using a random digit sample. From this larger sample, the subsample for the present study was selected. A primary goal of the study was to develop a theory on the dynamics of ethnicity in a multi-cultural society and effects on consumer behavior. Thus, Hispanic subcultures were used as the subsample: Mexican-Americans, Mexicans, and Central Americans (N=1337).

Three survey instruments were developed: (1) English language; (2) Spanish language; and (3) non-Hispanic. The Spanish language questionnaire was concurrently developed with the English, and then back-translated to English to identify and resolve any problems with disequivalency of language. Bilingual/bicultural interviewers conducted the interviews.

FIGURE 2

PROPOSED SITUATIONAL ETHNICITY MODEL

Measurement

The Los Angeles Times questionnaire contained a number of questions from which measures of the variables of the study were operationalized. The measures for the exogenous variables are as follows. Self-designated ethnicity was measured using the emic self-report method asking, "What is your ethnic background?" Antecedent state, as conceptualized by Stayman and Deshpande, is measure of "how ethnic one feels". Their assumption is that perception of discrimination would affect this state. For purposes of my analysis, the variable was measured on a 4-pt. scale of agree-disagree with the statement, "being too Hispanic gets you more problems than benefits." The new variable, cultural identity, was measured on a 4-pt. frequency scale in answer to the question: "Would you say you go to social gatherings with non-Hispanics?" (frequently, sometimes, seldom, never). The assumption underlying this measure is that Hispanics would frequent such gatherings as they assimilate. Thus, the stronger the assimilation score, the weaker the cultural identity. Social surroundings is a dichotomous variable measuring the presence of "change" in behavior from one type of social situation to another. This operationalization is conceptually equivalent to the Stayman and Deshpande manipulation of social surroundings in their experimental design. The items from the Los Angeles Times survey used to measure this concept was: "What language do you use when you talk to your parents?" and "When you talk to your co-workers?"--only English, mostly English, both, mostly Spanish, only Spanish. A dummy variable was constructed of change, no change. Product type was also a measure of the presence of "change", using the dummy variable technique. The items chosen concerned the change in use of language in the context of two different product types--a media product (newspaper) and a banking product (loans).

Endogenous variables were measured in the following manner. Felt ethnicity was a measure of strength of ethnic identification. A 5-point scale of frequency of attendance at local Hispanic community festivals was used. The assumption was that strong ethnic identifiers would attend such festivals more frequently than weak ethnic identifiers. Behavior was a measure of a consumer choice between a Spanish-language newspaper and an English-language newspaper.

TABLE 1

CORRELATION MATRIX

Analysis

The original data file was used in the re-analysis of The Los Angeles Times survey data.

Descriptive statistics and a correlation matrix were estimated using SYSTAT. The structural-equation parameters were estimated using the LISREL VII program. The correlation matrix (see Table 1) was input to the LISREL VII program. Each correlation coefficient is based on 1336 cases.

RESULTS

Macro Results

Hypothesis 3 tested the overall fit of the model. The data did not provide well-founded support for this hypothesis. The initial iteration produced a model with X2=245.37, df = 4, p = .000. The large value of X2, relative to the degrees of freedom, indicates that the proposed model did not fit the data well. The difference between the observed correlation matrix and that resulting from the hypothesized model was statistically significant. The Goodness of Fit Index (GFI) was .953; however, the Adjusted Goodness of Fit (AGFI) was .674. The AGFI takes into account degrees of freedom; therefore, it is a more accurate indication of the fit of the model. This lower AGFI indicates a problem model. A better fit might have been found if LISREL had been used to test the interaction between social surroundings and felt ethnicity or behavior. As the model was specified, the test becomes one of situational change in behavior but not change "specific" to a particular situation.

Micro Level

Hypothesis 1 tested the proposition that behavior is a function of felt ethnicity, social surroundings and product type. The R2 for this equation was .088. Much of the variance was left unexplained. In terms of the individual links, the maximum likelihood estimates were: .138 for the link between felt identity and behavior; -.191 for the link between social surroundings and behavior; and .266 for the link between product type and behavior. All of these were significant at the .01 level.

Hypothesis 2 tested the proposition that felt ethnicity is a function of self-designated ethnicity, antecedent state, social surroundings, and cultural identity. The R2 for this equation was .036, indicating that much variance is left unexplained. The maximum likelihood estimates were: -.034 for the link with ethnicity; -.040 for the link with antecedent state; .001 for the link with social surroundings; and .122 for the link with cultural identity. Of these only the link with cultural identity was significant at the .01 level. The total coefficient of determination for structural equations was .104, indicating the presence of much unexplained variance in the whole model.

Post-Hoc Analyses

The modification indices, presented in the LISREL VII printout, indicated that the model could be significantly improved with the addition of a link between cultural identity and behavior. This link makes theoretical sense because of anthropological studies which indicate that cultural identity transcends situational adjustments and in doing so gives common meaning, stability, and predictability to the individual's behavior (Fitzgerald, 1974). With this rationale, another iteration of the LISREL analysis was conducted. In this second iteration, all the previous links were left "as is."

Macro Results

The second iteration produced a model with X2=11.43, df = 3, p = .000. The smaller value of X2 indicated a much improved fit of the new model with the observed data. The GFI was .998, and the AGFI was .977. The close match between the GFI and the AGFI indicates a good fitting model. Another indication of an improved model is found in comparison of the root mean square residuals between the first and second LISREL iterations. An estimate under .05 is considered acceptable. The estimate for iteration #1 was .076; the estimate for iteration #2 was .016. This parameter supports the good fit of the post-hoc model.

It must be noted that the chi square statistic tends to be sensitive to sample size, implying that the low chi square estimate obtained may be an artifact of the large sample size (N=1337). To test the robustness of the chi square obtained, the post-hoc model was tested against a null model. The null model hypothesized no links among variables, and this model was run in LISREL. Using a chi square difference test, it was found that a X2 of 297.80 with 8 degrees of freedom is significant. It appears more likely that the post-hoc model is a good fit of the data.

Micro Level

The R2 for hypothesis 1 was .218; it was much improved. In terms of the individual links, the maximum likelihood estimates were: -.363 for cultural identity; -.14 for felt identity; -.154 for social surroundings; and .285 for product type. All of these were significant at the .01 level.

The R2 for this equation was .024, which is slightly lower than the first iteration. The maximum likelihood estimates were: -.006 for the link with ethnicity; -.040 for the link with antecedent state; -.001 for the link with social surroundings; and .121 for the link with cultural identity. Again, only the link with cultural identity was significant at the .01 level. The total coefficient of determination for structural equations was improved to .229.

DISCUSSION

This study lends support to the growing body of research on the link between ethnicity and consumer behavior. Most importantly, this study supports the notion of ethnicity as a dynamic, complex concept. The new variable, cultural identity, proved to be a significant addition to the Stayman and Deshpande model. Figure 3 illustrates the post-hoc model and reports the LISREL VII estimates of the parameters in this model. The model, as estimated, supports the proposition that behavior is a function of felt ethnicity, cultural identity, social surroundings, and product type.

However, the proposition that felt ethnicity is a function of self-designated ethnicity, antecedent state, product type, social surroundings, and cultural identity is only weakly supported. There is good reason to believe that problems may lie in the operationalization of the variable, felt ethnicity. The diagnostic information contained in the LISREL VII printout indicates that the standardized residuals for this variable are problematic (i.e., larger than 2.48 for the links with behavior, cultural identity, and product type.) The present study relied on data collected for a different purpose; and thus, the researchers were constrained by limitations in the pre-existing items. For this reason, several leaps in logic were made in operationalizing the variables. For example, the measure of felt ethnicity was behavioral rather than perceptual. Social situation was measured as situational change in general, rather than as a specific change in the situation under which a particular behavior was performed. While the study suggests an improved model of the ethnicity and consumption relationship, the results are limited.

In examining the relationship between ethnicity and consumption, this study suggests that cultural identity may have a greater influence on behavior than felt ethnicity. In order to fully explore the implications of this finding, it is necessary to examine more closely the conceptual differences between a social identity and a cultural identity. As was mentioned previously, social identities are assumed to vary situationally, while cultural identities have been conceived as stable. Evidence has been mounting since the 1970s that the phenomenon of situational variance in ethnic identity exists. Theorists, such as Stein and Hill (1977), Gans (1979), and Padilla (1985), discuss the situational aspects of ethnicity as strategic or symbolic manipulations. According to Websters Ninth New Collegiate Dictionary, situation can be defined as "relative position at a certain moment." When people shed, resurrect, or adopt ethnicity at certain moments only, it can be assumed that they do so out of "choice." If so, then race would act as a constraint on the range of choices. Therefore, Whites would enjoy a great deal of freedom in terms of situationally choosing ethnic identities; those defined in racial terms as "non-White" would enjoy much less freedom. Then in terms of situational ethnicities, the important research questions should be: how and why people choose a particular ethnic identity from a range of possible choices. The recognition of such questions has led to a recent increase of work investigating the conditions for choosing identities among White ethnics (Waters, 1990; Alba, 1990). It is suggested that further research investigate the differences in situational effect between Whites and non-Whites, with the hypothesis that situational effects will be stronger among Whites. As Coleman and Rainwater (1978) argue, Whites do not perceive ethnic discrimination to be a threat to their individual life chances, and their ethnic identify can be used at will and discarded when its psychological or social purpose is fulfilled.

This view that Whites are more likely to be "situational ethnics" should not be taken to mean that non-Whites do not exhibit situational variance in behavior. In fact, the present study documented the situational variance in language use among Hispanics. Bilingual Hispanics are likely to speak Spanish with their parents and English with their children. However, the range of possible choices for situational variance is much less among Hispanics than among Whites.

The more fundamental way in which ethnic identity influences consumption could be through cultural identity. Cultural identity is comprised of what people have inherited (e.g., race, origin, history, religion, language) and of what they have acquired (e.g., language, nationality) (Isaacs, 1975). As was mentioned previously, cultural identity could be the "stable" dimension of an individual's ethnic identity as it refers to one's basic group identity. Of importance, differences in consumption patterns have been evidenced by culture. In the U.S., for example, evidence exists that Hispanics are brand-loyal (Saegert, Hoover, Hilger, 1985; Segal and Sosa, 1983). While cultural identity does influence behavior, its strength of influence is dependent upon the importance attached to a particular identity. It is likely that the level of importance one attaches to a particular cultural identity will vary. Presumably, some individuals will attach a great deal of importance to that identity. For others, a cultural identity will be intermittently important, and still others will assume the label and little else.

Theory supports the notion that one's cultural identity is not stable but dynamic. Because what people acquire can be constantly new and changing, their cultural identity under certain conditions (e.g., immigrant status) is dynamic C almost in a continuing state of becoming. Thus, the strength of influence of cultural identity on consumption behavior is dependent upon the level of importance placed on that cultural identity. For any given member of an ethnic group, the level could vary. For this reason, it is suggested that further research explore the varying strength of ethnic identification among individuals of a particular ethnic or minority group. In a follow-up study currently in progress, the researchers are testing the differential effects of a cultural identity on behavior on a large sample of Hispanics in three cities: New York, Miami, and Houston. The study was specifically designed to test the associations between cultural identity and behavior as well as the dynamic nature of culture identity given the process of assimilation. It uses a full LISREL model that combines both a measurement model and a theoretical model. Such information is useful to the prediction of consumer behavior which relies both on knowledge about individuals within a particular cultural context and on knowledge about individuals in their social relations.

FIGURE 3

POST-HOC SITUATIONAL ETHNICITY MODEL

REFERENCES

Alba, Richard (1990). Ethnic Identity: The Transformation of White America. New Haven: Yale University Press.

Alba, Richard (1976). "Assimilation among American Catholic national origin groups." American Sociological Review, 41, 1020-1046.

Belk, R. (1974). "An exploratory assessment of situational effects in buyer behavior." Journal of Marketing Research, 1, 156-163.

Coleman, R. and Rainwater, L. (1978). Social Standing in America: New Dimensions of Class. New York: Basic Books.

Deshpande, R., Hoyer, W., and Donthu, N. (1986). "The Intensity of Ethnic Affiliation: A study of the Sociology of Hispanic Consumption." Journal of Consumer Research, 13 (September), 214-220.

Fitzgerald, T. (1974). Social and cultural identity: Problems of persistence and change. Athens, GA: Southern Anthropological Association.

Gallo, P.J. (1974). Ethnic alienation: The Italian-Americans. Cranbury, NJ: Fairleigh Dickenson University Press.

Gans, H. (1979). "Symbolic ethnicity: The future of ethnic groups and cultures in America." Ethnic and Racial Studies 2: 1-20.

Gumperz, John. (1976) "The sociolinguistic significance of conversational codeswitching." Papers on Language and Context, University of California, Berkeley

(1978) "The conversational analysis of interethnic communication." Interethnic Communication. E.L. Ross (ed) Proceedings of the Southern Anthropological Society. Auburn: University of Georgia Press.

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

Isaacs, H. (1975). Idols of the Tribe: Group Identity and Political Change. New York: Harper & Row.

Joreskog, K., and Sorbom, D. (1988). "LISREL 7: A Guide to the Program and Applications." Chicago SPSS.

Kakkar, P., and Lutz, R. (1981). "Situational Influence on Consumer Behavior: A Review." Perspectives in Consumer Behavior. H. Kassarjian and T. Robertson (eds.). Glenview IL: Scott, Foresman, 204-214.

Lutz, R., and Kakkar, P. (1975). "The Psychological Situation as a Determinant of Consumer Behavior." Advances in Consumer Research, Vol. 2, M.J. Schlinger (ed). Ann Arbor, MI: Association for Consumer Research, 439-454.

McGuire, W., and McGuire, C. (19787). "Salience of Ethnicity in the Spontaneous Self-Concept as a Function of One's Ethnic Distinctiveness in the Social Environment." Journal of Personality and Social Psychology, 36, 5, 511-520.

O'Guinn, T., and Faber, R. (1985). "New Perspectives on Acculturation: The Relationship of General and Role Specific Acculturation with Hispanics' Consumer Attitudes." Advances in Consumer Research, Vol. 12, E. Hirschman and M. Holbrook (eds). Provo, UT: Association for Consumer Research, 113-117.

Padilla, F. (1985). "On the nature of Latino ethnicity." In The Mexican American Experience, de la Garza, Bean, Bonjean, Romo, and Alvarez (eds). Austin: University of Texas Press.

Pitts, R., Sheth, J., and Valencia, H. (1986). Proceedings of the conference on cultural and subcultural influences in consumer behavior and marketing. Chicago: American Marketing Association.

Saegert, J., Hoover, R., and Hilger, M. (1985). "Characteristics of Mexican American consumers." Journal of Consumer Research 12, 1: 104-09.

Segal, N.M. and Sosa,L. (1983). "Marketing to the Hispanic community." California Management Review 26: 120-34.

Stayman, D., and Deshpande, R. (1989). "Situational Ethnicity and Consumer Behavior." Journal of Consumer Research, 16, 361-371.

Stein, H. and Hill, R. (1977). The Ethnic Imperative: Examining the New White Ethnic Movement. University Park: Pennsylvania Sate University Press.

ValdTs Fallis, G. (1980). "Code switching as a deliberate verbal strategy." In Discourse Processes: advances in research and theory. R. Freedle (ed), Norwood, NJ: Ablex.

Yancey, W., Eriksen , E., and Juliani, R. (1976). "Emergent Ethnicity: A Review and Reformulation." American Sociological Review, 41, 391-403.

Waters, Mary. (1990). Ethnic Options: Choosing Identities in America. Berkeley: University of California Press.

Weber, M. (1968). "Economy and Society: An Outline of Interpretive Sociology." G. Roth and C. Wittich (eds). Berkeley: University of California Press.

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