Consumer Behavior V Scholarly Activity

Situational Ethnicity and Consumer Behavior

DOUGLAS M. STAYMAN ROHITDESHPANDE*

The role of situalional ethnicity in consumption behavior is examined, and the rela- tionship between ethnicity and consumption is argued to be affected by the situa- tional contexts in which choices are made. Situational effects are proposed to oper- ate through changes in the level of felt ethnicity and in the relationship between felt ethnicity and behavior. An empirical study demonstrates these effects by showing the impact of two situational dimensions—social surroundings and antecedent conditions^-on ethnic food choices.

I mmigrants are a growing presence in the UnitedStates, especially from countries in Latin America and Asia (U.S. Bureau of the Census 1988). These growth patterns in ethnic subcultures have significant implications for the consumption aspects of Ameri- can life. In response, a body of research on the influ- ence of ethnicity on consumption has begun to emerge (Hirschman 1981,1983; Pitts, Sheth, and Va- lencia 1986).

Recent research has built on earlier consumer eth- nicity work by grounding it more fully in theory in sociology and cultural anthropology. Such studies in- clude the pioneering work of Hirschman (1981) on Jewish-American innovativeness and several studies of Hispanic Americans, including that by Wallendorf and Reilly (1983) on the impact of assimilation on food consumption; by O'Guinn and Meyer (1984) on media usage; Saegert, Hoover, and Hilger (1985) on brand loyalty; Deshpande, Hoyer, and Donthu (1986) on the impact of perceived ethnic affiliation; and Faber, O'Guinn, and McCarty (1987) on accul- turation and the importance of product attributes.

Most of this work has examined ethnicity as an in- terindividual, group membership characteristic. That is, ethnicity has been viewed as a nominally codable demographic classification. Subjects have been classi- fied as Hispanics, for instance, based on last name

•Douglas M. Stayman is Assistant Professor of Marketing, Col- lege of Business Administration, University of Texas, Austin, TX 78712. Rohit Deshpande is Professor of Marketing, Amos Tuck School of Business Administration, Dartmouth College, Hanover, NH 03755. Order of authorship was determined randomly, as both authors contributed equally. Financial assistance was provided by grants from the University Research Institute at the University of Texas, Austin, and the Tuck Associates Program at Dartmouth College. The authors would like to thank Mark Alpert, Steve Brown, Ajay Kohli, Ron Faber, the Northeast Marketing Consor- tium participants, and especially the JCR reviewers for construc- tive suggestions on earlier versions of the article, as well as Dave Aaker and Kevin Keller for assistance with the research at the Uni- versity of California, Berkeley.

361

(Saegert et al. 1985) and country of origin (Wallen- dorf and Reilly 1983) in cross-sectional studies.

Recent work, however, suggests that ethnic identi- fication and behavior is at least partly situationally determined. That is, ethnicity is not just who one is, but how one feels in and about a particular situation. Thus, one can think of ethnicity and related behavior not only as a stable sociological trait of individuals that is manifested in the same way at all times, but also as a transitory psychological state manifested in different ways in different situations. O'Guinn and Faber (1985) and Faber et al. (1987) hint at such an influence within the consumption context by measur- ing the effect of acculturation roles on consumption. However, no work in consumer research has focused on the social psychological role of transitory ethnic states in moderating the relationship between choice and individual variables such as ethnicity, as well as its role in affecting general consumption patterns (the focus of much situational research; see Kakkar and Lutz 1981).

The objectives of this article are (1) to introduce work on situational ethnicity from the social sciences and relate it to consumer research, (2) to test hypothe- ses about situational effects on ethnic consumer choice in a controlled laboratory study of food con- sumption choices, and (3) to suggest directions for fu- ture research.

SITUATIONAL ETHNICITY IN SOCIAL SCIENCE

Most earlier research approaches ethnicity from the etic perspective. Here, the researchers' perceptions of ethnicity are based on last names of individuals or languages spoken at home. However, this ignores in- dividual perceptions and mental states. Following the anthropological tradition of emic measurement, re- searchers have begun to use a more subjective defini-

© JOURNAL OF CONSUMER RESEARCH • Vol. 16 • December 1989

362 THE JOURNAL OF CONSUMER RESEARCH

tion of ethnicity (Cohen 1978). In this tradition, an individual first identifies which ethnic group(s) s/he belongs to (what we will refer to as self-designated eth- nicity) and then indicates how strongly s/he identifies with that group (what we will refer to as felt ethnicity).

Work in sociology echoes the anthropological per- spective. For example, Yancey, Eriksen, and Juliani (1976, p. 399) suggest that "ethnicity should not be regarded as an ascribed attribute with only two dis- crete categories, but as a continuous variable. . . . Indeed, the small amount of explained variance con- tributed by one's specific ethnic category (as mea- sured in earlier work) could partly be due to the fact that ethnicity was treated as an ascribed trait." This sociological view is displayed in the work of Desh- pande et al. (1986) on strength of ethnic identifica- tion.

Recent work in social psychology, anthropology, and' ethnic studies views ethnicity in a compatible, but different, way. Rather than being viewed as an ac- culturating influence of environment, level of felt eth- nicity is seen as more situation specific. For example, McGuire et al. (1978) apply the spontaneous self-con- cept construct to ethnicity in showing that the sa- lience of ethnicity to self-concept differs according to the amount of ethnic heterogeneity in the environ- ment. First mention of situational ethnicity has been traced to anthropologist John Paden in discussing ethnic categorization in urban Africa. In 1967, Paden noted that "situational ethnicity is premised on the observation that particular contexts may determine which ofa person's communal identities or loyalties are appropriate at a point in time" (cited in Okamura 1981, p. 452). Like McGuire's research on self-con- cept, Paden's research describes the influence of eth- nic identity on an individual as depending upon the social situation and the individual's perception of that situation. Thus, social science research supports the view that situations likely influence the effects of ethnicity on behavior and that the relevant construct to focus on is not self-designated ethnicity as an indi- vidual trait but felt ethnicity as a situationally deter- mined state.

SITUATIONAL ETHNICITY IN CONSUMER BEHAVIOR

In marked contrast to early consumer research that focused on either person or product-object character- istics, the work on situational influences on consump- tion found strong relationships with purchase and us- age behavior (see Kakkar and Lutz 1981 for a review). The work of Belk (1974, 1975a) and of Lutz and Kak- kar (1975) is of special importance in this regard, and a brief summary of that work will indicate its rele- vance for situational ethnicity and consumption.

Belk (1974) defines a "situation" as something out- side the basic tendencies and characteristics of the in-

dividual and beyond the characteristics of the stimu- lus object to be acted upon. He identifies five objec- tive dimensions of situations (1975a): the physical surroundings (e.g., geographic location), social sur- roundings (e.g., the presence or absence of others), temporal perspective (e.g., time of day), task defini- tion (e.g., information search or product purchase), and antecedent states (e.g., momentary moods or conditions immediately preceding choice). There has not, however, been uniform agreement on whether it is more valid to define a situation in primarily objec- tive terms, such as those used by Belk, or in primarily subjective terms, i.e., by focusing on a consumer's in- ternal perceptions, interpretations, and responses. Kakkar and Lutz (1981) prefer the latter approach, first defining a situation in objective terms and then examining the meaning that an individual ascribes to the situation (in a sense providing a validity check for the objective characteristics).

Although the majority of empirical work on Belk's five dimensions has been on the task definition (cf. Belk 1975a), the antecedent state and the social situa- tion dimensions are most relevant to the study of situ- ational ethnicity. In the case of the former, one ante- cedent condition is how ethnic a consumer feels prior to a decision. Consider, for example, the situation in which an Irish-American celebrant of St. Patrick's Day is deciding where to eat out. The choice, given the probable heightened state of felt ethnicity, may be more likely an Irish restaurant than if the same indi- vidual were choosing a restaurant on a different, non- Irish, holiday. This suggests that a consumer's felt ethnicity at the time of a behavior is influenced both by his/her ongoing self-designated ethnicity as well as by the antecedent conditions prior to the behavior.

The second of Belk's five dimensions that appears to be important to situational ethnicity is the social situation. As McGuire et al. (1978) suggest, the sa- lience of ethnicity in a given social situation is likely to increase or decrease depending on the extent to which one's ethnicity is similar to or different from that of others in a given environment or situation. In addition, many different social situations have different norms of ethnic behavior (e.g., type and amount of food and drink considered appropriate at Chinese versus Jewish weddings), suggesting that the relationship between ethnicity and behavior is affected by the type of product being considered.'

When the influences of antecedent state and self- designated ethnicity on felt ethnicity are combined with the moderating effects of social situation and product type on the relationship between felt ethnic- ity and behavior, we arrive at the model illustrated by Figure A.

'Food consumption situations are used here and in the experi- ment because they are ethnically cued (Patai 1977).

SITUATIONAL ETHNICITY 363

FIGURE A

PROPOSED MODEL OF SITUATIONAL ETHNICITY AND CONSUMPTION

Self-Designated Ethnicity

Antecedent State

Felt Ethnicity

Pro( Ty

\ /

duct pe

Social Surroundings

Behavior

HYPOTHESES The argument advanced in this article is that situa-

tional factors (i.e., antecedent state and social sur- roundings) and type of product influence the relation- ship between etJinicity and behavior. Thus, we hy- pothesize:

H I : A combination of self-designated ethnicity, antecedent state, social surroundings, and product type affect behavior.

We also propose:

H2: Self-designated ethnicity and antecedent state combine to influence felt ethnicity.

Further, as Figure A indicates, felt ethnicity should be a better predictor of behavior than the traditional measure of self-designated ethnicity. Also, since our argument is that how ethnic one feels in a given situa- tion is what should influence behavior, such situa- tion-specific felt ethnicity should be a better predictor of behavior than the level of ethnicity felt at other times (non-situation-specific felt ethnicity). This leads to:

H3: Situation-specific felt ethnicity is a better predictor of behavior than either self-desig- nated ethnicity or non-situation-specific felt ethnicity.

Finally, as also indicated in Figure A:

H4: The felt ethnicity-behavior relationship is moderated by the social surroundings and the product type being considered.

METHOD Design

The hypotheses were tested using a four-factor, mixed effects design (Keppel 1982, pp. 452-463). Be-

tween-subject factors included self-designated ethnic- ity (Anglo-, Chinese-, and Mexican-American) and antecedent state (non-ethnic/relevant, ethnic/rele- vant, and ethnic/non-relevant). Within-subject fac- tors included social surroundings (with parents or with business associates) and food type (Anglo, Chi- nese, or Mexican).

The design is similar to that used in previous situa- tion research (e.g., Belk 1974), with two differences. First, the ethnicity and antecedent state manipula- tions have been added as between-subject factors. Thus, the different groups rated all foods for both sit- uations (order was rotated). Second, the foods chosen within the situations relate to the different ethnic groups, creating an expected person by food type in- teraction not present in a simple split-plot design.^

Subjects Subjects were recruited from two sources. One hun-

dred fourteen subjects were members of the market- ing department undergraduate subject pool at a large West Coast university, where the predominant mi- nority population is Chinese-American. The remain- ing 281 subjects were recruited from undergraduate business organizations at a large Southwestern uni- versity, where Mexican-Americans are the predomi- nant minority population. (No differences in results were found due to location.) The former received sub- ject pool credit for participation, while the latter and their organizations each received $3 per subject.

^Lutz and Kakkar (1975) and Bonner (1985) suggest that this de- sign may inflate the product by situation term due to sensitizing subjects to the situational context, although Belk (1975b) refutes this due to split-plot controls. However, the attempt here is not to determine the "amount" of the effect, but to use the effect to test for between-subject differences. Thus, an effect of situation and product would not support the hypotheses without an interacting effect of ethnicity.

364 THE JOURNAL OF CONSUMER RESEARCH

Although clearly a convenience sample, student subjects were considered appropriate, since this re- search is a theory test based on the concerns of inter- nal validity and need for a homogeneous population. For these reasons, the sample need not be representa- tive of larger populations. The nature of the sample used allowed extensive pretests to choose appropriate situations, products, and measures specific to the population of interest.

Independent Variables The independent variables were the subjects' self-

designated ethnicity, antecedent state, social situa- tion, and food type. Following Deshpande et al. (1986), subjects' self-designated ethnicity was mea- sured using the emic self-report method. This mea- sure included Asian and Hispanic categories, but because this research specifically focused on Chinese- and Mexican-Americans, a country-of-ancestry ques- tion was included. Finally, since the study focus is subculture rather than nationality, foreign students and Americans other than Anglos, Chinese, or Mexi- cans were excluded. The resulting sample iticluded 195 Anglo-Americans, 54 Chinese-Americans, and 57 Mexican-Americans from the 395 original sub- jects. For simplicity, the hyphenated "American" ap- pendage is dropped in the following sections.

As noted earlier, the two situational dimensions varied were antecedent state and the social surround- ings. The other three situational dimensions—loca- tion, task, and temporal perspective—were held con- stant. This was accomplished by including the depen- dent variables in a survey on food choice that varied the situation by the persons present (social surround- ings), but not by physical location (a restaurant),^ task (ordering for oneself), or temporal perspective (din- ner). In pretests, the social surroundings dimension was rated significantly less similar {p < 0.01; ^tests) than the three other dimensions (3.1 versus 5.8, 5.7, and 6.0) on a seven-point scale of similarity of the sit- uations.

Antecedent state was manipulated by having sub- jects read a page entitled "A Study of Recruitment Bias," which summarized statistics on alleged dis- crimination in corporate hiring of undergraduate stu- dents and in promoting those hired to the ranks of top management. This issue was chosen based on pretest- ing that showed recruitment and promotion bias to be a relevant issue to the business undergraduate pop- ulation of the study. The questionnaire asked for sub- jects' opinions of the results of the study.

meal choice constrains somewhat the type of restaurant, since ethnic food is not served at all restaurants. However, our pur- pose in holding location constant was primarily to control varia- tion in terms of eating at home versus eating out, a major distinc- tion in earlier situational research (e.g., Belk 1975a).

Three levels of induction of antecedent state were obtained through altering the focus of the bias in the page read by the subjects. In the non-ethnic/relevant condition, the reported bias was toward undergradu- ate business students in general and was thus ex- pected to be relevant to all subjects in the experiment independent of their ethnicity. The ethnic/relevant condition contained reported bias in corporate hiring of either Asian or Hispanic students and was thus ex- pected to be relevant only to the specific ethnic group targeted. (Asian and Hispanic subjects were exposed only to bias toward their relevant ethnicity; half of the Anglos were exposed to Asian and half to Hispanic bias). The ethnic/non-relevant condition reported bias against Arab-Americans, who were fewer than half of one percent of the population at each school. (Pretesting had indicated that such a bias was not an important issue to any of the target groups.)

The social surroundings dimension was manipu- lated through the description of the persons accompa- nying the subject in the situations presented in the food choice questionnaire. Based on focus groups and pretesting of ethnic relevance of different social sur- roundings, two were chosen: eating with business as- sociates (non-ethnic) or with one's parents (very eth- nic). The pretests included ratings of the extent to which different situations varied on such dimensions as personal relevance and importance of choices made, as well as ratings of ethnic differences expected across a number of social surroundings. Presence of parents and business associates constituted the only situations for which the sole significant difference was ethnicity.

Finally, food ethnicity was manipulated through having subjects rate foods of three ethnic types in each situation. Based on focus groups with students of each ethnic group, eight foods were pretested for ethnicity. The foods chosen were rated by all ethnic groups as either American (grilled chicken), Mexican (chicken mole), or Chinese (steamed yellowfish). All three foods chosen were rated in the same general price range, and moderately expensive foods were chosen, since pretests suggested that very inexpensive or expensive foods may be differentially appropriate across situations or ethnic groups (e.g., with parents who will pay for the meal).

Dependent Variables Felt ethnicity was assessed through subjects' self-

rated strength of ethnic identification (Deshpande et al. 1986) at two points, before and after the bias sec- tion was read. A nine- rather than five-point scale (very strong to very weak) was used to increase mea- sure sensitivity. Thus, the ethnic/relevant antecedent condition was expected to increase the reported felt ethnicity of Chinese and Mexican subjects, while the non-ethnic/relevant and ethnic/non-relevant condi- tions were not expected to affect felt ethnicity.

SITUATIONAL ETHNICITY 365

Involvement in the bias study was assessed to con- trol for a possible alternative explanation for the bias induction. Involvement was measured in the recruit- ment study through three questions on how impor- tant subjects thought the issue was and how relevant it was to them. The three items (alpha = 0.63) were averaged for the analyses.

Following Belk (1974), a variation of Triandis's (1964) "behavioral differential" approach was used to assess the dependent variables oi" interest, i.e., food appropriateness and likelihood of choice. Specifi- cally, subjects responded to a number of questions for a given dinner situation. The same foods and ques- tions were then used across situations. Two scales were constructed for measuring food choice behavior across situations. Based on a factor analysis of pretest questions, seven-point scales for appropriateness (ap- propriate/inappropriate and suitable/unsuitable) and likelihood (likely to eat/unlikely to eat and desirable/ undesirable) were constructed. The two measures in each scale (alpha = 0.86 and 0.94) were averaged for analysis, and alphas were similar for each group.

Procedure A procedure similar to that used by Gardner (1986)

and others was employed, whereby ostensibly differ- ent studies are used in one session to separate differ- ent manipulations. Subjects were told that a number of business school faculty were working together within one experimental setting for efficiency rea- sons. In an initial, unrelated study, the prior felt eth- nicity, generation, and country-of-ancestry questions were asked. This was followed by the recruitment bias study, which included the post-bias felt ethnicity measure. Finally, subjects completed the food choice behavior study before being paid (if appropriate) and debriefed. Informal questioning at the end indicated that no subjects guessed the true purpose of the study.

RESULTS Only the results for the likelihood scale are reported

here, since results for appropriateness are similar and since likelihood was considered both more in keeping with previous research (e.g., Belk 1974) and more di- rectly related to consumption. Mean values for likeli- hood are in Table 1."

'Lutz and Kakkar (1975) note the problems with the use of likeli- hood as a dependent variable, since ratings across products may not be independent. Therefore, an additional ANOVA was run with likelihood for the other foods as a covariate. In that analysis, only the ethnicity by food type and four-way interaction were affected to any extent. In both cases, significant effects (p < 0.05) became moderately significant (p < 0.10). Thus, it was concluded that like- lihood was an appropriate measure.

A longer version of the paper containing appropriateness and ANOVA results is available from the authors. It should be noted that in the ANOVA, the results for the effect of social situation are

Test of Hypothesis 1 The first hypothesis states that behavior (likeli-

hood) is affected by a combination of four factors: self-designated ethnicity, antecedent state, social sit- uation, and type of product. The hypothesis thus sug- gests that the relationship between each variable and likelihood is affected by all three other variables, without specifying the exact mediational model pro- posed. Thus, a four-way interaction was suggested and tested by analysis of variance using Keppel's (1982) general mixed effect procedure.

The significance of the four-way interaction (F = 2.65,;; < 0.05) supports the hypothesis. This result gives evidence on two levels. First, it suggests the im- portance of the combination ofdifferent factors—sit- uational and personal—in determining ethnically re- lated behavior. Second, it provides initial evidetice for the validity of the proposed model, since the inter- action is consistent with (although it does not spe- cifically test) the model.

Although a four-way interaction is difficult to inter- pret. Figure B provides some insight into why the effect emerged. The figure shows a number of differ- ences across the four factors in the study. For exam- ple, steamed yellowfish was considered fnore likely when with parents rather than business associates only by the Chinese subjects. As might be expected, chicken mole was generally considered more likely when with parents only by Mexican subjects, and grilled chicken was considered more likely when with parents only by Anglos.

Likelihood of congruent ethnic choice increased in the ethnic/relevant bias condition for the with-par- ents situation, but decreased in the business situation (middle graph of top row, left graph of middle row). One interpretation of this interesting effect of the bias manipulation is that subjects in a heightened state of ethnic awareness in an ethnically diverse group (e.g., with business associates) try not to choose an ethni- cally relevant food. However, added ethnic awareness leads to more likely choice of an ethnic food in an ethnic situation (e.g., with parents). While the spe- cifics may be of interest, the larger point of Figure B is that the numerous differences across both the social surroundings and antecedent state conditions for the different ethnic groups and food types support a role for situational ethnicity in choice on both dimen- sions.

Test of the Mediational Model Three hypotheses were proposed concerning the

mediational model in Figure A: that self-designated

consistently significant, while those for antecedent situation are mixed, being significant for both variables only for the four-way interaction.

366 THE JOURNAL OF CONSUMER RESEARCH

Self-designated ethnicity and

antecedent state

Anglo NER ER ENR

Mexican NER ER ENR

Chinese NER ER ENR

Grilled chicken

7.28 7.50 7.55

5.12 4.63 4.95

5.53 5.32 5.42

TABLE 1

MEAN LIKELIHOOD RATINGS FOR FOOD CHOICES

Social situation

Dinner with parents

Chicken mole

4.60 5.41 4.62

5.90 7.97 6.03

4.00 4.65 5.12

Steamed yellowfish

4.41 3.95 4.57

4.43 4.14 3.12

4.84 5.65 5.44

Dinner with business associates

Grilled chicken

7.15 7.32 7.34

5.74 6.08 5.44

6.00 6.24 5.56

Chicken mole

5.16 5.84 5.54

4.81 3.17 4.82

4.27 6.77 4.42

Steamed yellowfish

5.21 4.70 5.73

5.14 5.56 4.62

4.50 3.33 5.19

NOTE: NER = non-ethnic ijut relevant bias manipulation in report of corporate discrimination against undergraduate business students; ER = ettinic and relevant bias against Hispanics, Asians; ENR = ethnic but non-relevant bias against Arab-Americans.

ethnicity and antecedent conditions combine to in- fluence felt ethnicity (Hypothesis 2), that situation- specific felt ethnicity is a better predictor of behavior than either self-designated ethnicity or non-situation- specific felt ethnicity (Hypothesis 3), and that the felt ethnicity-behavior relationship is moderated by the social situation and product type (Hypothesis 4). The following sections discuss each hypothesis.

Influences on Fell Ethnicity. Evidence supporting the combined influence of self-designated ethnicity and antecedent conditions on felt ethnicity (Hypoth- esis 2) is given by the self-reported level of ethnic identification before and after the bias manipulation. The mean change in this measure (pre- minus post- manipulation measure) for each group is given in Ta- ble 2. A two-way ANOVA (bias manipulation by eth- nic group) showed the expected interaction between the bias manipulation and ethnicity (F = 4.67, p < 0.001). In addition, there was also a main effect of ethnicity (F = 8.41, ^ < 0.001). The involvement means, also reported in Table 2, provide additional evidence for the effect of the bias manipulation and ethnicity. For involvement, the main effects of both the bias manipulation and ethnicity are significant {F = 4.65 and 27.81, p < 0.01 and 0.001, respectively), as was the interaction term {F = 2.22, p < 0.05).

As can be seen in Table 2, the interaction effects are due to the expected effect of ethnicity combined with the personal relevance of the bias manipulation to in- fluence a directional change in felt ethnicity. Both the Chinese and Mexican subjects showed a strengthen- ing of ethnic identification for the ethnic/relevant

condition, but a weakening in all other conditions. Further, although all subjects reported the non-eth- nic/relevant business bias study to be more important than the ethnic/non-relevant Arab bias study, the Chinese and Mexican subjects reported the ethnic/ relevant bias study to be slightly more important than the business bias study. Hence, the ethnic/relevant condition created the expected heightened level offelt ethnicity in the Chinese and Mexican groups.

Mediational Role of Felt Fthnicity. Hypothesis 3 suggested that situation-specific felt ethnicity would predict choice better than either self-designated eth- nicity or non-situation-specific felt ethnicity. To test this effect, a series of regressions was run. Since social surroundings and food type are expected to affect the ethnicity-behavior relationship, separate regressions were run for each food in each social situation. Thus, three food types by two social surroundings yielded six regressions for each set of independent variables assessed.

Performance of three sets of independent variables was assessed to compare the predictive ability of situ- ation-specific felt ethnicity to other ethnicity mea- sures: (1) only self-designated ethnicity, (2) self-desig- nated ethnicity and non-situation-specific felt ethnic- ity (from the pre-measure), and (3) self-designated ethnicity and situation-specific felt ethnicity (after the bias manipulation). Since the relationship be- tween level of felt ethnicity and behavior may differ by ethnic group, the latter two equations included dummy variables for self-designated ethnicity as a main effect and interaction terms for the effect of self-

SITUATIONAL ETHNICITY 367

FIGURE B

INTERACTION OF FOOD TYPE BY SOCIAL SITUATION, ANTECEDENT STATE, AND SELF-DESIGNATED ETHNICITY

Steamed Yellowfish

Mexicans Likelihood

6,0

5,5

5,0

4,5

4,0

3.5

Non-Ethnic Ethnic Bias

Chinese Likelihood

6.0

5.5

5,0

4.5

4,0

3.5

Non-Ethnic Ethnic Bias

Anglos Likelihood

6,0

5,5

5,0

4,5

4,0

3,5

-

-

-

-

Non-Ethnic Ethnic Bias

Chicken Mole

Mexicans Likelihood

8,0

7,0

6,0

5,0

4,0

3,0

-

-

Non-Ethnic Ethnic Bias

Chinese Likelihood

8,0

7,0

6,0

5,0

4,0

3,0

Non-Ethnic Ethnic Bias

Anglos Likelihood

8,0

7.0

6,0

5.0

4.0

3.0

-

-

-

, •

-

-

^on-Ethnic Ethnic Bias

Mexicans Likelihood

7.5

7.0

6,5

6.0

5,5

5,0

4,5

Non-Ethnic Ethnic Bias

Grilled Chicken

Chinese Likelihood

7,5

7.0

6.5

6,0

5.5

5,0

4.5

Non-Ethnic Ethnic Bias

NOTE: All means are based on a 1-9 scale.

with Business Associates

— — with Parents

Anglos Likelihood

7.5

7.0

6,5

6,0

5,5

5.0

4.5

-

-

-

-

-

-

Non-Ethnic Ethnic Bias

368 THE JOURNAL OF CONSUMER RESEARCH

TABLE 2

FELT ETHNICITY CHANGE AND INVOLVEMENT FOR THE

TABLE 3

PREDICTION OF BEHAVIOR BY MEASURES OF ETHNICITY ANTECEDENT STATE MANIPULATION

Self-designated ethnicity and

antecedent state

Anolo NER ER ENR

Mexican NER ER ENR

Chinese NER ER ENR

Change in felt

ethnicity

-0.62 -0,50 -0,37

-0,24 0,95

-0.18

-0,21 1.18

-0,33

Involvement

3.51 4,20 4,50

2.51 2,32 2,98

3,11 2,10 3,87

n

57 89 49

21 19 17

19 17 18

Food and social situation

Chicken mole With parents With business

Steamed yeilowfish With parents With business

Grilled chicken With parents With business

Unadiusted fl' values

Calf Oclt

designated ethnicity

only

,08 .11

,04 ,07

,12 ,11

Self- designated

onrl nno-ai IU 1 i\jf 1 situation- specific ethnicity

.13

.16

,07 ,08

.23 ,13

Self- designated

anfi al IU

situation- specific ethnicity

,19 ,20

,10 ,13

,21 ,18

NOTE: NER = non-ethnic but relevant bias manipulation in report of corporate discrimination against undergraduate business students; ER = ethnic and rele- vant bias against Hispanics, Asians; ENR = ethnic but non-relevant bias against Arab-Americans,

designated ethnicity on the felt ethnicity variable (Cohen and Cohen 1983). The unadjusted R^ values are given in Table 3.

As can be seen in Table 3, there was a uniform increase in explained variance when felt ethnicity was added to self-designated ethnicity. The Fadd statistic (Kerlinger and Pedhazur 1973) can be used to com- pare the values that include both felt and self-desig- nated ethnicity with those that reflect self-designated ethnicity only. For four of six comparisons, addition of non-situation-specific felt ethnicity increased ex- plained variance a t p < 0.05 (Faddxioo > 3.23), and it was marginally significant for a fifth (Fadd = 2.31, p < 0.10). The addition of situation-specific felt ethnic- ity increased explained variance for all six compari- sons at p < 0.001 {Fadd > 6.92).

Table 3 also shows that using situation-specific felt ethnicity adds to explained variance over that ex- plained by non-situation-specific felt ethnicity. For five of the six comparisons, the i?-square is greater with situation-specific felt ethnicity than with non- situation-specific felt ethnicity. The mean explained variance increases from 13 percent using non-situa- tion-specific felt ethnicity to 17 percent when using situation-specific ethnicity. The Fadd statistic could not be used for this non-nested comparison; there- fore, Akaike's (1973) criterion was employed. For four of the six tests, the model with situation-specific felt ethnicity outperformed that with non-situation- specific felt ethnicity.

In summary, these results suggest that level of felt ethnicity, rather than just ethnic classification, influ- ences behavior, supporting the results of Deshpande et al; (1986). However, these results go further to sup-

port Hypothesis 3, that situation-specific felt ethnic- ity is a better predictor of behavior than non-situa- tion-specific felt ethnicity.

Moderating Fffect of Social Situation and Type of Food. A moderating effect of social situation and food type on the felt ethnicity-likelihood relationship (Hypothesis 4) was assessed through analyzing the B- values for the felt ethnicity variable across the six in- stances in which situation-specific felt ethnicity was used. That is, if social surroundings and food type moderate the relationship between felt ethnicity and likelihood, we would expect that the slope of the rela- tionship (the B-values) would vary across social sur- roundings and food types. As can be seen in Table 4, there are a number of such differences supporting the hypothesis.

Of particular note are the sign reversals both within food type across social surroundings and within social surroundings across food types. Sign reversals are noteworthy because they indicate that it is not only the level of felt ethnicity that influences behavior but, depending on the situation, increases in felt ethnicity may either increase or decrease certain behaviors.

For example, the results for Hypothesis 1 can be in- terpreted within the regression results. The interpre- tation from the ANOVA effects—that subjects try not to choose an ethnically relevant food when in an eth- nically integrated situation, but are more likely to choose an ethnic food in an ethnic situation—is sup- ported by the positive weights for chicken mole for Mexicans and steamed yellowfish for Chinese when in the with-parents situation and negative weights for the same combinations in the with-business associ- ates situation. For chicken mole for Chinese subjects and steamed yellowfish for Mexican subjects, the re- sults are reversed (positive in the with-business and

SITUATIONAL ETHNICITY 369

TABLE 4

FELT ETHNICITY-BEHAVIOR RELATIONSHIP BY SITUATION AND ETHNIC GROUP

B-values, felt ethnicity by level of self-designated ethnicity

Food and social situation

Chicken mole With parents With business

Steeinned yellowfish With parents With business

Grilled chicken With parents With business

Anglos

-.34 -.29

-.26 -,31

,01 .22

Mexicans

,37 -,41

-.40 ,07

-,27 ,29

Chinese

-.36 .26

,40 -,43

-,23 .70

NOTE: Following Cohen and Cohen (1983), each equation had a variable for the interval felt ethnicity measure, tvi/o dummy variables for the nominal seif- designated ethnicaty measure, and two variables for the interaction. At least one of the two interaction terms was significant at p < 0,05. indicating that the appropriate interpretation of the felt ethnicity B-vaiues was for each level of self-designated ethnicity separately—Anglos, f^exicans, Chinese,

negative in the with-parents situations), indicating that the results are due to specific ethnic relevance of the foods, rather than general food ethnicity.

Supporting the relative effects of felt ethnicity on food choice consistent or inconsistent with the social surroundings—e.g., Mexican foods for Mexicans when with parents (consistent) versus Mexican foods for Mexicans when with business associates (inconsis- tent)—are the effects of the non-ethnic food, grilled chicken. The positive weights for grilled chicken for the Mexican and Chinese subjects in the with-busi- ness-associates situation, but negative weights in the with-parents situation, indicate that high levels offelt ethnicity appeared to drive subjects toward ethnically related f̂ ood in an ethnically consistent situation and toward non-ethnic foods in an ethnically inconsistent situation.

DISCUSSION AND CONCLUSION To summarize, we found consistent support for the

effect of social surroundings on choice by ethnic groups. Chinese, Mexican, and Anglo subjects had different perceptions of the appropriate food for con- sumption in situations where business associates were present versus those where parents were present. In addition, there was a four-way interaction between ethnic classification, food type, and the two situation dimensions studied, i.e., social surroundings and an- tecedent conditions. This result extends the discus- sion of consumer ethnicity into the realm of the con- sumption situation literature initiated by Belk (1974), Kakkar and Lutz (1981), and others.

We also found evidence for situationally deter- mined felt ethnicity. Chinese and Mexican subjects

in relevant bias conditions increased their strength of ethnic identification, providing empirical support for contentions by sociologists (Yancey et al. 1976) and psychologists (McGuire et al. 1978) that ethnicity is not just who one is, but how one feels also. To our knowledge, this is the first time such an empirical re- sult has been reported in any of the literatures studied (i.e., consumer behavior, anthropology, sociology, and social psychology).

In addition, regression analysis supported the pro- posed mediational model of fielt ethnicity effects on behavior. Use of situation-specific felt ethnicity in- creased explained variance in likelihood of choice over that explained by either self-designated ethnicity alone or use of non-situation-specific felt ethnicity. In addition, the difference in the weights for felt ethnic- ity across social surroundings and food types sup- ported a moderating role for those variables.

The ability of situation-specific felt ethnicity to bet- ter explain choice than non-situation-specific felt eth- nicity extends the research of Deshpande et al. (1986), who measured felt ethnicity cross-sectionally rather than in specific choice conditions, in assessing the effect of felt ethnicity on behavior. Their ability to find significant results may owe to the fact that their measure of felt ethnicity was contained in the same questionnaire as the choice measures even though it was not choice specific. Thus, the measure may have been able to predict intentions, but may not have been able to predict actual choices in different situa- tions as well.

Beyond remedying the obvious study limitations of generalizability due to subject population, single product category, and two social surroundings, this experiment points to potentially fruitful avenues for future research. Two areas of particular importance stand out. The first is to assess not only whether, but also why and when ethnicity is important in con- sumption situations. The present study has suggested the role of the spontaneous self-concept (McGuire et al. 1978) and social norms in affecting the level and effect of ethnic identification. These areas point to in- teresting additional questions. For example, work on the spontaneous self-concept suggests that individu- als 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. Hence, we may ex- pect people with a strong single ethnic background to have greater stability in choice related to ethnicity across situations, although they may be more likely influenced by other trait variables such as gender. More work remains to be done to identify whether and when the self-concept notion applies to con- sumption situations. Additionally, different ethnic groups may bring different subcultural perspectives to bear regarding sensitivity to group norms. Hence, we may expect differences in the effect of situational factors on members ofdifferent ethnic groups. For ex-

370 THE JOURNAL OF CONSUMER RESEARCH

ample, Asian cultures have been found to be more sensitive to group norms than Western cultures are (cf. Lee and Green 1988) and thus may be relatively more affected by the within- versus between-culture situation represented by our with-parents versus with-business-associates condition. Other areas that may help explain conditions under which situational effects will occur include perceived power in a situa- tion, which may affect the importance of group norms, and social visibility of choice. The present re- search has indicated the importance of assessing situ- ational factors. Future research must compare alter- native explanations behind such effects.

Second, more research is needed to build upon the proposed model through which situational ethnicity is expected to affect behavior. For example, our re- search assessed only a mediational role of social situa- tions between felt ethnicity and behavior. However, social situations may also affect the level of felt eth- nicity and thus play a dual role. Such a hypothesis suggests that further research, using more types of sit- uations and measures of effects, is necessary to de- velop a comprehensive model.

Future research should also expand the inquiry of situational effects into areas outside the limitations of the present study. A primary objective should be to assess effects of situations on the relationship between choice and individual differences other than ethnicity (such as age or sex). A second fruitful area may be assessing the other three dimensions of situations: physical surroundings (eating out versus eating at home), temporal perspective (breakfast versus din- ner), and task definition (choice for oneself versus choice for others). The latter dimension may be par- ticularly important for other expressive product cate- gories (e.g., clothing or homes) where subcultural norms might play a greater role.

[Received June 1988. Revised June 1989.]

REFERENCES Akaike, H. (1973), "Information Theory and an Extension

of the Maximum Likelihood Model," in Proceedings of the Second International Symposium of Information Theory, eds. N. Petrov and F. Csadki, Budapest, Hun- gary: Akademiai Kiado, 267-281.

Belk, Russell W. (1974), "An Exploratory Assessment of Situational Effects in Buyer Behavior," Journal of Mar- keting Research, 1 (May), 156-163.

(1975a), "Situational Variables and Consumer Be- wioT," Journal of Consumer Research, 2 (December),

157-167. (1975b), "The Objective Situation as a Determinant

of Consumer Behavior," inAdvances in Consumer Re- search, Vol. 2, ed. Mary Jane Schlinger, Ann Arbor, MI: Association for Consumer Research, 429-437.

Bonner, P. Gregg (1985), "Considerations for Situational Research," in Advances in Consumer Research, Vol.

12, eds. Elizabeth C. Hirschman and Morris B. Hol- brook, Provo, UT: Association for Consumer Re- search, 368-373.

Cohen, Jacob and Patricia Cohen (1983), Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences, Hillsdale, NJ: Lawrence Eribaum Associates.

Cohen, Ronald (1978), "Ethnicity: Problems and Focus in Anthropology," in Annual Review of Anthropology, Vol. 7, eds. Bernard J. Siegel et al., Palo Alto, CA: An- nual Reviews, Inc., 379-403.

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.

Faber, Ronald J., Thomas C. O'Guinn, and John A. Mc- Carty (1987), "Ethnicity, Acculturation, and the Im- portance of Product Attributes," Psychology and Mar- keting, 4 (Summer), 121-134.

Gardner, Meryl P. (1986), "Responses to Emotional and In- formational Appeals: The Moderating Role of Con- text-Induced Mood States," working paper. Graduate School of Business, New York University, New York, NY 10003.

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

(1983), "Cognitive Structure Across Consumer Eth- nic Subcultures: A Comparative Analysis," in Ad- vances in Consumer Research, Vol. 10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Arbor, MI: Associa- tion for Consumer Research, 197-202.

Kakkar, Pradeep and Richard J. Lutz (1981), "Situational Influence on Consumer Behavior: A Review," in Per- spectives in Consumer Behavior, eds. Harold H. Kas- sarjian and Thomas S. Robertson, Glenview, IL: Scott, Foresman, 204-214.

Keppel, Geoffrey (1982), Design and Analysis: A Research- er's Handbook, Engiewood Cliffs, NJ: Prentice-Hall.

Kerlinger, Fred N. and Elazar J. Pedhazur (1973), Multiple Regression in Behavioral Research, New York: Holt, Rinehart & Winston.

Lee, Choi and Robert T. Green (1988), "Cross-Cultural Ex- amination of the Fishbein Behavioral Intention Model," Working Paper No. 87/88-5-7, Department of Marketing, College of Business Administration, UniversityofTexas, Austin, TX 78712.

Lutz, Richard J. and Pradeep Kakkar (1975), "The Psycho- logical Situation as a Determinant of Consumer Behav- ior," in Advances in Consumer Research, Vol. 2, ed. Mary Jane Schlinger, Ann Arbor, MI: Association for Consumer Research, 439-454.

McGuire, William J., Claire V. McGuire, Pamela Child, and Terry Fujiota (1978), "Salience of Ethnicity in the Spontaneous Self-Concept as a Function of One's Eth- nic Distinctiveness in the Social Environment," Jour- nal of Personality and Social Psychology, 36 (5), 511 - 520.

Okamura, Jonathan Y. (1981), "Situational Ethnicity," Ethnic and Racial Studies, 4 (October), 452-465.

O'Guinn, Thomas C. and Ronald J. Faber (1985), "New Perspectives on Acculturation: The Relationship of General and Role Specific Acculturation with Hispan- ics' Consumer Attitudes," in Advances in Consumer

SITUATIONAL ETHNICITY 371

Research, Vol. 12, eds. Elizabeth C. Hirschman and Morris B. Holbrook, Provo, UT: Association for Con- sumer Research, 113-117.

- and Timothy P. Meyer (1984), "Segmenting the His- panic Market: The Use of Spanish-Language Radio," Journal of Advertising Research. 23 (6), 9-16.

Patai, Raphael (1977), The Jewish Mind, New York: Charles Scribner's Sons.

Pitts, Robert E., Jagdish N. Sheth, and Humberto Valencia, eds. (1986), Proceedings of the Conference on Cultural and Subcultural Influences in Consumer Behavior and Marketing, Chicago: American Marketing Associa- tion.

Saegert, Joel, Robert .F. Hoover, and Marye T. Hilger (1985), "Characteristics of Mexican American Con-

sumers," Journal of Consumer Research, 12 (June), 104-109.

Triandis, Harold (1964), "Exploratory Analyses of the Be- havioral Components of Social Attitudes," Journal of Abnormal and Social Psychology, 68 (April), 420-430.

U.S. Bureau of the Census (1988), Statistical Abstract of the United States, Washington, D.C: U.S. Government Printing Office.

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

Yancey, William L., Eugene P. Eriksen, and Richard N. Ju- liani (1976), "Emergent Ethnicity: A Review and Re- formulation," American Sociological Review, 41 (June), 391-403.