Effects of Strength of Ethnic Identity and Product Presenter Race on Black Consumer Attitudes: A Multiple-Group Model Approach
Troy Elias
University of Florida
Osei Appiah
The Ohio State University
Li Gong
SAP
Abstract
This study examines the relationship between ethnic identity and same- or different-race sources on blacks' consumer attitudes. Social identity theory provides a theoretical framework to explain black consumers' decision-making processes. Using structural equation modeling, the authors show that the race of a character used in an ad transversely moderates the relationship between ethnic identity and consumer attitudes. Specifically, they find that as blacks' strength of ethnic identity increases, they tend to display more favorable consumer attitudes if the product presenter of an e-commerce website is also black and weaker but negative attitudes if the product presenter is white.
Keywords: ethnic identity, race, structural equation modeling, e-commerce
Introduction
To help their clients differentiate themselves from all the clutter online, advertisers strive to make their clients' websites more interesting and engaging. Their goal is to increase the interactive capabilities of their clients' sites so that Internet users not only visit the sites but also spend more time there, continue to come back, and recommend the site to other consumers. In doing so, advertisers and marketers frequently use computer-generated characters, such as avatars (user-controlled) and computer agents (computer-controlled), as digital communicators both on websites and in computer games and applications (Gong and Nass 2007).
For example, IKEA offers the interactive agent "Ask Anna" on its website. Anna engages website visitors in conversation, helps them locate products, and addresses any customer service issues they might have. Other companies also use computer-generated anthropomorphic characters as part of their advertising and marketing strategies. For example, AccuWeather, L'Oreal, Global Gillette, and DaimlerChrysler all use avatars to handle consumer questions online (Ahrens 2006), and other companies use avatars to provide entertainment gratifications for their online visitors. For example, ESPN offers an avatar-based program titled "Voice of the Fan" that allows fans to record their own analyses, synopses, and opinions of various aspects of different sports or sports figures. The premise of this study is based on the concept of the "Ask Anna" computer agent. The study is an exploratory step into understanding how such an interactive and engaging feature on a commercial website affects consumer attitudes, particularly those of multiracial consumers.
On a purely intuitive level, we do not expect that all consumers who fall under a specific racial category will respond the same way to the same stimuli all the time. Several studies focusing on ethnic minorities' social identities have found that individual members vary greatly in their identification with their racial in-group (see Berry 1997; Phinney 1990, 1992). Phinney (2005, p. 189) defines ethnic identity as "a self-constructed understanding of oneself in terms of one's cultural and ethnic background and the attitudes and feelings associated with that background." Consequently, the current study examines the concurrent effects of blacks' ethnic identity and the use of black and white product presenters on an e-commerce website on blacks' evaluative responses.
This study contributes to the advertising and strategic communication literature by exploring the impact of ethnic identification on consumer attitudes in an e-commerce setting. Although often overlooked, ethnic identification could be a major construct in the prediction of ethnic minorities' behavior, attitudes, and consumption practices. In contrast with previous studies (e.g., Appiah 2001, 2004), we use structural equation modeling (SEM) to analyze the relationships among ethnic identification, consumer attitudes, and a source's race and also use a generic e-commerce website as opposed to a racially targeted website. Racially targeted websites contain a level of personalization based on the ethnicity and interests of the ethnic group in that context; in contrast, the site we use pertains primarily to the impact of a source's race on consumer attitudes in a context that is not explicitly racially relevant to any one ethnic group. In his study, Appiah (2004) finds that blacks responded more favorably across a broad range of consumer attitudes; however, his stimuli dealt with issues that specifically pertained to blacks. In contrast, we do not attempt to evoke blacks' social identities and, as a result, reinforce the impact of social identity using a minimum of ethnic cues. Overall, these differences are subtle, but they are not without merit.
Appiah (2001) argues that a source's race is often an important and salient communication characteristic for ethnic minorities, particularly for those whose ethnic identity is central to their self-worth. Therefore, though advertisers and marketers try to segment markets using abstract social categories, such as race or nationality, it is up to the individual person to determine the social category with which he or she identifies best (Brewer 1991). Ultimately, this self-categorization dictates the symbols associated with the person's relevant social membership to which he or she responds.
Accordingly, Phinney and Onwughalu (1996) argue that in general, the social identity of ethnic minorities involves two elements: (1) race, which refers to phenotypical differences between groups (e.g., skin color), and (2) ethnicity, which incorporates the cultural intergroup differences in attitudes, beliefs, and norms. To date, research exploring the parameters of race and ethnicity has tended to treat members of ethnic minority racial groups as homogeneous collectives who uniformly and consistently exhibit same-race in-group preferences (Gong, Appiah, and Elias 2007). Although this treatment is understandable, given that race is a prominent social identity in U.S. history for ethnic minority groups (Smedley 1999), it is flawed. Historically, race has proved to be especially relevant to one particular U.S.-based ethnic group, namely, blacks.
Being black in the United States constitutes a salient identity distinction. Thus, as a racial and ethnic minority, blacks are largely more saliently aware of their race than whites (Jaret and Reitzes 1999), place more importance on race than whites (Jaret and Reitzes 1999), and consider their race a prominent factor in their interpersonal communications with whites (Hecht, Larkey, and Johnson 1992). Race also tends to be a prominent factor in blacks' consumption of the media (Appiah 2004; Dates 1980). However, prior research has mostly overlooked the notion that all persons belonging to an ethnic minority group may not attach the same magnitude or valence to their racial or ethnic membership (Gong, Appiah, and Elias 2007).
Whereas many blacks might consider race an important social identity, others may believe that their ethnic membership carries less significance. Accordingly, several empirical studies have delineated ethnic identity's propensity to predict variation in ethnic minorities' behavior, attitudes, social interactions, and consumption practices (Appiah 2001, 2004). A discernable pattern of these studies is that people who display higher levels of ethnic identification and racial affirmation tend to display a greater degree of same-race in-group preference than people who display lower levels of identification.
Therefore, we posit that blacks with strong ethnic identities (i.e., who have a greater sense of connection to their ethnic group and are willing to express traditional black values) should exhibit a greater sense of awareness of and appreciation for black characters. This appreciation should ultimately result in more favorable responses to in-group characters (e.g., blacks) than out-group characters (e.g., whites) in both mediated and interpersonal contexts. In contrast, blacks with weaker ethnic identities may harbor attitudes and behaviors that are not strongly tied to black culture and that more closely reflect attitudes and behaviors of the dominant mainstream white culture (Appiah 2004; Green 1999). Consequently, these group members may appreciate black media characters less than their strong ethnic identifier counterparts. Among weak ethnic identifiers, their "blackness" may not be a particularly salient or meaningful characteristic (Appiah 2004), leading them to respond no differently to media with white characters than to media with black characters. Therefore, an e-commerce website featuring racially similar and dissimilar characters may only be important for blacks whose ethnicity is both salient and an essential part of their self-concept.
Empirical work demonstrates that blacks with strong ethnic identification perceive themselves as more similar to, identify more strongly with, and have greater liking for black media and characters (Appiah 2001, 2004) than blacks with weak ethnic identities. Research also shows that blacks with weak ethnic identities typically respond either indifferently or more favorably to white media and models than to black media and models (Appiah 2004; Green 1999). Thus, ethnic identity seems to be a key factor that determines the degree of in-group preference among ethnic minorities, particularly blacks. This information should be of interest to entrepreneurs, established businesses, and advertising and marketing specialists as the U.S. consumer market becomes increasingly diverse. Accordingly, these parties should tailor their advertising, products, and media to each market segment.
The Selig Center's (Humphreys 2008) estimates and projections of buying power for 1990-2013 further highlight the importance of the black demographic: In 2007, U.S. blacks' buying power ($913 billion) was larger than the economies of all but 13 countries in the world. According to Hallman (2008), U.S. black buying power rose from $318 billion in 1990 to $590 billion in 2000 and will rise to $1.1 trillion in 2011, a 237% increase. In addition, Hallman predicts a rise in black buying power in 47 states, with the largest market share increases in Mississippi (24.3%), Maryland (22.2%), Georgia (20.8%), and the District of Columbia (30.6%). According to U.S. Census data, the number of black-owned businesses is also on the rise. In 2002, their number totaled nearly 1.2 million, and revenues for these businesses amounted to $88.6 billion (African Americans by the Numbers 2010). Also in 2002, black-owned businesses accounted for 5% of all nonfarm businesses in the United States. According to Hallman (2008), these businesses increased 45% from 1997 to 2002, exceeding the overall 10% increase in U.S. businesses, and should continue to increase.
In general, by 2011, the combined buying power of blacks, Asians, and Native Americans in the United States will be $1.8 trillion, nearly quadruple its 1990 level of $454 billion, a gain of $1.3 trillion or 289% (Hallman 2008). These three ethnic groups also will account for 14.3% of the U.S. total buying power in four years. Consequently, the Selig Center (Humphreys 2008) posits that as the U.S. consumer market continues to grow more diverse, advertising, products, and media must be tailored to meet each ethnic consumer market. In addition, the gains in black buying power over the years reflect more than population growth and inflation; they are credited to improvements in employment opportunities, higher education levels, and the growth of the black population compared with the total population (Humphreys 2008). For example, the projected U.S. black population (including those with a combined race) for July 1, 2050, is 65.7 million (African Americans by the Numbers 2010). According to these projections, in 2050 blacks will constitute 15% of the nation's total population. The nation's black population will grow almost 34% by 2011, compared with 18.5% for whites and 25.4% for the total U.S. population. Thus, social identity theory, which examines social psychological processes of intergroup behavior (Tajfel 1974; Tajfel and Turner 1986), could be a powerful vehicle for advertisers and marketers in understanding how to communicate more effectively with this important demographic.
Social Identity Theory of Intergroup Behavior
Tajfel (1974, p. 69) defines social identity as "that part of an individual's self-concept that is derived from his knowledge of his membership of a social group (or groups) together with the emotional significance attached to that membership." In their seminal article, Tajfel and Turner (1986) posit that social identity is facilitated when a person distinguishes between his or her in-group and some relevant out-group, with the member's in-group being evaluated more favorably. However, an unsatisfactory social identity will lead people either to strive to leave their existing group and join some other, more positively distinct group or to try to make their existing group more positively distinct (Tajfel 1974; Tajfel and Turner 1986).
Social identity theorists (e.g., Hogg 2004; Tajfel and Turner 1986) argue that the inherent human motivation to maintain positive social identity is the enhancement and maintenance of self-esteem. People have a fundamental need to achieve and maintain positive social identities in their social groups as a means to enhance their sense of worth. The implications are especially relevant to ethnic minorities, especially because minority size tends to be highly correlated with disadvantages in power and status (Brewer 1991). Thus, we argue that ethnic minorities should be more motivated than members of an ethnic majority to improve perceptions of their self-worth, which in turn should lead to higher levels of ethnic identification with ethnic groups because the outcome and the reputation of these groups are indelibly linked to members' self-worth (Brewer 1991).
Many ethnic minority groups have noticeable phenotypical differences (e.g., skin color, facial bone structure) from mainstream whites, which limit their ability to move from one ethnic group to another. For example, moving from the black to white ethnic social group is never a viable option, except in the rare cases when a person has overwhelmingly white mainstream features and characteristics. Therefore, ethnic minorities are able to achieve positive social or ethnic identity only by renegotiating the positive distinctiveness of their own ethnic group (Phinney 1990; Tajfel 1974, 1981; Tajfel and Turner 1986). Consequently, a common behavioral outcome of positive social identity is in-group favoritism (Jackson et al. 1996; Tajfel 1981; Tajfel and Turner 1986).
Hypothesis
Historically, blacks have endured conscious and deliberate acts of discrimination in the United States that have institutionalized their group identity (Entman and Rojecki 2000; Mills 1997). These group identities have formed not only through the dissemination of cultural stories but also through systematic efforts from colonial times and more recent legalized racial segregation, which has created social, political, and economic advantages for whites (Entman and Rojecki 2000). These advantages have led many blacks to perceive racial biases against them in critical areas, such as the criminal justice system, politics, housing, and health care.
The perceived establishment of racial discrimination and disparities have ultimately shaped the way blacks approach not only whites in interpersonal contexts but also white mainstream media, including white characters and models. As a result these factors, as well as the consistently negative portrayals of blacks in the media (Abraham 2003), in general blacks may discern competition from whites, which might threaten their positive group identity and lead them to display more in-group favoritism (Mastro 2003).
In contrast, Appiah (2001) posits that blacks tend to appreciate being associated with a target market and thus are more likely to respond to targeting cues directed toward their race. In turn, this appreciation leads blacks to foster more favorable attitudes toward black-targeted media than white-targeted media (Aaker, Brumbaugh, and Grier 2000). It follows then that blacks with higher levels of ethnic identification should be more likely than those with lower levels to identify with ethnically targeted media and link the targeted media and themselves.
In addition, the media's current representation of blacks may lead blacks to seek out or respond more favorably to positive media messages and characters that support or enhance their need for a positive social identity and distinctiveness. Therefore, we argue that seeing a positive black spokesperson in a mediated or face-to-face advertising setting may enhance black consumers' identity, which in turn may lead them to identify with and develop more positive attitudes toward a black spokesperson (in-group member) than they would a white spokesperson (out-group member). Moreover, because blacks may not identify with many white models, they may perceive content from white sources as personally irrelevant and not directed at them, causing them to tune out and even disidentify with white sources (Appiah 2002). We argue that this dis-identification should be particularly true for blacks with strong ethnic identities, who may only consider black characters relevant and worthy of attention (Appiah 2004). Scholars have argued that in-group favoritism and out-group derogation occurs most often among people with greater in-group identification (Appiah 2001; Fujioka 2005). Therefore, racially similar product presenters should be particularly salient to blacks and lead them to not only identify with but also respond more favorably to black endorsers. In light of these arguments, we propose the following:
H1. A product presenter's race acts as a transverse moderator, such that ethnic identity positively influences blacks' consumer attitudes if the presenter's race is black but negatively influences their consumer attitudes if the presenter's race is white.
Although studies have examined the role of ethnic identity on blacks' consumer behaviors and attitudes (see Appiah 2004; Appiah and Elias 2009; Ethier and Deaux 1994), few studies have used an analytical approach that enables testing of group mean differences on higher-order latent variables. Therefore, we use SEM because it enables a holistic analysis of the relationships involved in blacks' consumer processing and decision making based on their exposure to ethnically targeted and nontargeted advertising messages. In addition, SEM offers a means to test hypotheses at a higher level of abstraction. In contrast with analysis of variance and multiple regression analysis, SEM can differentiate between observed and latent variables (Kline 2005). Finally, SEM enables a more appropriate statistical treatment of ethnic identity than prior studies have used.
Prior research has often measured ethnic identity as a continuous variable using Phinney's (1992) Multigroup Ethnic Identity Measure. Reponses to the items on the scales are usually averaged and interpreted such that higher scores represent strong ethnic identity. Experimental studies on ethnic identity typically apply a median split to create two groups, one representing relatively strong and one representing relatively weak ethnic identity (e.g., Appiah 2004; Forehand and Deshpandé 2001; Operario and Fiske 2001). Researchers have indicated though that dichotomizing continuous data is not a good strategy because it may lead to the loss of information, lower statistical power, and lower reliability (Aguinis and Stone-Romero 1997; Chen, Cohen, and Chen 2007); however, SEM can circumvent some of these problems.
Figure 1 presents the overall hypothesized model for the relationships among ethnic identity, consumer attitudes, and black and white product presenters. In the subsequent section, we explain the study.
Figure 1. Hypothesized Ethnic Identity-Consumer Attitude Model
Method
This study examines whether the relationship between blacks' ethnic identity and consumer attitudes remains equivalent for blacks who view a black versus a white product presenter on a simulated e-commerce website. In contrast with Appiah (2004), in the sites we used as stimuli, we did not incorporate levels of embedded racial cues that extend beyond the race of the product presenter. That is, we provided no additional racial cues to indicate that the products being advertised or presented were culturally relevant to the black community or that the website on which they were presented expressly related to the black culture. Instead, the site examined the impact of a source's race on consumer attitudes in a generic e-commerce environment. To test this relationship, we developed a second-order latent variable, consumer attitude, from four first-order latent variables, including identification with and perception of homophily with the product presenter, social presence rating of the product presenter, and attitude toward the website.
Participants
We recruited 115 black undergraduate and graduate students (51 men, 64 women) from a large U.S. Midwestern public university to participate in the study. Of the sample, 56% were women, and 82% were between 18 and 22 years of age. Participants were recruited from both communication classes, with the incentive of earning extra course credit, and university organizations that targeted black students, with the incentive of entrance into a cash prize drawing. Because selecting only black students is difficult, we included all ethnic groups during data collection but chose only the black participants for study analyses. Participants were randomly assigned to either a black or a white product presenter.
Stimuli Materials
We selected a black male and a white male image from a pool of 73 images. Before image selection, a different sample of students rated the 73 images on a series of attributes that included physical attractiveness and race, on a seven-point semantic differential scale (e.g., "ugly-beautiful"). From the ratings 22 black students provided, the two selected facial images did not differ in perceived physical attractiveness (t(21) = .18, p = .86). Racial categorization data also confirmed perceptions of the images' intended race at approximately 100%. We selected only male images to control for the gender of the visual representation; the images depicted each face looking forward with relatively neutral expressions.
The black and white male images appeared as interactive product presenters on a prototype e-commerce website. The product presenter was animated with Adobe Flash software and presented information on 12 different products. We chose a broad range of products, including a car, a pair of jeans, a couch, ice cream, and a personal digital assistant, to avoid any unusual effects occurring because of the use of any one specific product. All product brands were fictitious. The product presenter began the presentation of each product with an introductory screen. To introduce each product, the presenter's image would appear in the center of the screen, and a text bubble adjacent to the presenter would display an introductory message (e.g., "The following product is a Sonaka Laptop"). After the participant clicked on the "continue" button, the next screen would show a detailed product page. At this point, the presenter's image appeared on the left-hand side of the product detail page. The rest of the page displayed an image of the product and a larger text bubble with more detailed product information.
Procedure
For the study, the participants visited a lab where they sat at a computer terminal. We informed them that the purpose of the study was to test a prototype e-commerce website that provides product information. On the lab computer, participants interacted with either the black or the white product presenter as they navigated through all 12 products. They completed the study by filling out an online questionnaire.
Measures
The postexperiment questionnaire contained measures of identification and homophily with the product presenter, social presence rating of the product presenter, and attitude toward the website. The questionnaire also included demographic questions and a measure of ethnic identity. Note that the use of the variables identification, homophily, and attitude toward the site are particularly important. A large body of research suggests that audiences are influenced by feelings of congruence with elements within the message, such as the product endorser. When a person sees a communication source that he or she perceives as similar to him or her, the communication is more effective and/or influential in shaping or changing attitudes. This finding suggests that audiences would respond more positively to endorsers who are more similar to them than endorsers who are less similar. Specifically, audiences should respond more favorably to racially similar than racially dissimilar endorsers. Homophily refers to the phenomenon of perceived similarity between two people; Rogers and Bhowmik (1970, p. 526) define it as "the degree to which pairs of individuals who interact are similar with respect to attributes, such as beliefs, values, education, social status, etc." However, communication and consumer researchers have conducted little, if any, work on the relationship between identification and homophily or on how ethnic identification may moderate this relationship.
We measured identification with the product presenter with a scale that averaged two items: "How strongly did you identify with the product presenter?" and "How strongly did you relate to the product presenter?" (Aaker, Brumbaugh, and Grier 2000). Participants answered these two questions on a seven-point scale (1 = "not strongly at all," 7 = "very strongly"). The two items had high reliability, Cronbach's alpha was .96, and the average response was 3.23 (SD = 1.77).
We measured homophily with the product presenter with a scale that comprised three seven-point semantic differential items, adapted from McCroskey, Richmond, and Daly's (1975) homophily scale: "is like me-is unlike me" (reverse coded), "is different from me-is similar to me," and "thinks like me-does not think like me" (reverse coded) (α = .82). We averaged the items to create the scales; higher scores indicated greater homophily. The average response on this scale was 3.16 (SD = 1.49).
We measured social presence with the four-item scale Short, Williams, and Christie (1976) developed. The four seven-point semantic differential items were "impersonal-personal," "unsociable-sociable," "insensitive-sensitive," and "cold-warm" (α = .90). These items are highly relevant to computer-mediated communication and have been used in previous research (Lee and Nass 2002; Rice and Love 1987). We also averaged these items, and the average response was 4.39 (SD = 1.30). Higher scores indicated more social presence.
We measured attitude toward the website with a scale used in previous studies (e.g., Appiah 2001, 2004; Desphandé and Stayman 1994). The scale included 11 seven-point semantic differential items that we averaged: "boring-interesting," "bad-good," "negative-positive," "useless-useful," "worthless-valuable," "poor-outstanding," "not for me-for me," "weak-strong," "not appealing-appealing," "not attractive-attractive," and "not likable-likable" (α = .96). Again, higher numbers meant a more positive evaluation of the website. The average response was 4.25 (SD = 1.32).
We measured ethnic identity with Phinney's (1992) Multigroup Ethnic Identity Measure. High numbers indicate strong ethnic identity. The overall ethnic identity scale had a good reliability level (α = .80), and the average response was 5.43 (SD = .74). Demographic variables included gender, age, and ethnicity. All 115 participants in the study selected "Black/African American" as the only answer for their ethnicity.
Analysis
We analyzed the study using AMOS 19. The method of estimation used throughout the study was maximum likelihood. (Zero-order correlations and tables are available on request.) To assess the relative fit of the nested models, we employed the chi-square difference test and used the fit statistics comparative fit index (CFI) and root mean square error of approximation (RMSEA) to compare alternative models. Prior research maintains that CFI is a good index for smaller sample sizes (Bentler 1990) and posits that RMSEA is among the most informative criteria in covariance structure modeling (Byrne 2001). Consistent with extant literature, a well-fitting model should meet the requirement of a RMSEA lower than .06 and a CFI greater than .95 (for summaries of recommended fit statistics and their cutoff values, see Holbert and Stephenson 2002; Hu and Bentler 1999).
The SEM technique we used for analysis is a hybrid model (Holbert and Stephenson 2002) and represents a full model that consists of measurement and structural parameters (Stephenson and Holbert 2003). According to Holbert and Stephenson (2002), the hybrid model allows all observable variables to load individually on their respective latent variables, in line with the work of Joreskog (1973a).
We focused primarily on the hypothesized areas that could potentially contain measurement problems, namely, the measurements of consumer attitudes and ethnic identity. Accordingly, we constructed a hybrid measurement model. This model was not fully latent because the measurement for ethnic identity had 20 items and the measures for attitude toward the website had 11 items. Because of the relatively small number of participants in the study (N = 115), we item-parceled these constructs into five and three indexes, respectively.
Item parceling helps reduce the number of indicators to a more manageable number by taking a set of observable variable items associated with a single latent variable and constructing an additive index from those indicators (Little et al. 2002). Researchers often employ item parceling when the number of observable items associated with any one latent variable is large (i.e., more than seven), as is the case with both the ethnic identity and attitude toward the website measures in this study. In general, studies with relatively smaller sample sizes should use parceling (Little et al. 2002).
For the initial model, which does not account for differences between the race of the product presenter, the results indicate an adequate fit: (χ2 = 141.17, d.f. = 99, p < . 05; CFI = .967; RMSEA = .061; and 90% confidence interval = .036-.083). None of the observable variables we used to construct consumer attitudes cross-loaded on any of the second-order latent constructs. Thus the measures we used to develop this higher-order construct were valid. In light of the acceptable fit estimates, we did not respecify the initial model. Thus, this analysis is based on the overall sample.
We used isolated equality constraints, which is common when testing hypotheses related to group invariance (see Byrne 2001). Thus, we conducted the analyses in accordance with two groups-blacks who viewed the black presenter, and blacks who viewed the white presenter-with the relationship between ethnic identity and consumer attitude first unconstrained and then held constant. Consistent with literature (Byrne 2001), we estimated a baseline model for each group separately to test for group invariance (i.e., equivalence across conditions).
We use the fit of this simultaneously estimated model to provide the baseline value against which we compared all subsequently generated nested models. We also treated significant changes in the model fit estimates (i.e., χ2 and degrees of freedom) between the fit estimates of the unconstrained model and the constrained model as evidence of a moderating effect. Therefore, any significant differences between the model fit of the unconstrained and the constrained two-group models indicate a moderating effect of the product presenter's race on blacks' consumer attitudes.
Results
Our general hypothesis predicts that as blacks' ethnic identity increases, their consumer attitudes will be positive if the product presenter's race is black but negative if the product presenter's race is white. Overall, this hypothesis was supported. The following sections break down the relevant models according to their chi-square and global fit assessments and their subsequent interpretations.
Model 1: Unconstrained Two-Group Hybrid Model
The fit of this simultaneously estimated model provides the baseline against which we evaluate all subsequently specified models (Figures 2 and 3). The results indicate a good fit (χ2 = 255.46, d.f. = 198, p < .05; CFI = .950; RMSEA = .051; and 90% confidence interval = .030-.068).
Figure 2. Model 1: Unconstrained Two-Group Hybrid Model (black product presenter, unstandardized coefficients)
Notes: Error terms were included in the estimation but are not shown for ease of interpretation.
*p < .05.
**p < .01.
***p < .001.
Figure 3. Model 2: Unconstrained Two-Group Hybrid Model (white product presenter, unstandardized coefficients)
Notes: Error terms were included in the estimation but are not shown for ease of interpretation.
*p < .05.
**p < .01.
***p < .001.
Model 2: Constrained Two-Group Model (Black Product Presenter Versus White Product Presenter)
The fit estimates for Model 2, in which 62 participants saw only a black product presenter and 53 participants saw only a white product presenter, demonstrated an adequate fit. Specifically, the chi-square and goodness-of-fit estimates from the model generated after the equality constraint between the two groups was introduced were as follows: χ2 = 261.26, d.f. = 199, p < .05; CFI = .946; RMSEA = .053; and 90% confidence interval = .033-.069 (see Figures 4 and 5).
Figure 4. Model 3: Constrained Two-Group Hybrid Model (black product presenter, unstandardized coefficients)
Notes: Error terms were included in the estimation but are not shown for ease of interpretation.
*p < .05.
**p < .01.
***p < .001.
Figure 5. Model 4: Constrained Two-Group Hybrid Model (white product presenter, unstandardized coefficients)
Notes: Error terms were included in the estimation but are not shown for ease of interpretation.
*p < .05.
**p < .01.
***p < .001.
The model comparison analysis indicates that the chi-square difference between the unconstrained model (χ2 = 255.46, d.f. = 198) and the isolated constrained equivalences of the multiple-group model (product presenter = black or white; χ2 = 261.26, d.f. = 199) of 5.8, with a corresponding single degree of freedom, is significant (p = .016) (Table 1). This finding indicates that the race of the presenter moderates the relationship between ethnic identity and consumer attitudes. In addition, the unconstrained multiple-group models show the effects of the manipulation of the product presenter's race. The coefficient goes from .14 for the black product presenter to -.11 for the white product presenter. This transverse interaction suggests that as ethnic identity increases, the relationship between ethnic identity and consumer attitude becomes weaker when the product presenter is white than when the product presenter is black. This finding adds support to social identity theory and also provides potentially powerful implications.
Table 1. Goodness-of-Fit Statistics for Tests of Invariance
| Model Description | Groups | Comparative Model | χ2 | df | ∆χ2 | ∆df | Statistical Significance |
| Hypothesized model (Model 1) |
Black presenter/white presenter | - | 255.46 | 198 | - | - | - |
| Factor loading constrained | Black presenter/white presenter | Model 1 | 261.26 | 199 | 5.8 | 1 | p < .05 |
Notes: ∆χ2 indicates the difference in χ2 values between models; ∆df refers to the difference in degrees of freedom between models.
General Discussion
The results of the study indicate that even on a generic e-commerce website, as ethnic identity increases, the race of the source in a persuasive e-commerce message moderates blacks' consumer attitudes. The study demonstrates that for blacks, the relationship between ethnic identity and consumer attitude is negative when the product presenter is white and also tends to be weaker than when the product presenter is black. That is, the data suggest that as blacks' ethnic identity increases, so do their attitudes toward their own group, in support of social identity theory. Specifically, social identity theory suggests that the more strongly members identify with their in-group, the less favorable attitudes they hold toward out-group members (Negy et al. 2003). Prior research has argued that a person's desire to maintain a positive social identity through social comparisons results in higher levels of both in-group favoritism and out-group derogation (Fujioka 2005). This view is supported by other researchers, who maintain that in-groups favor messages that negatively characterize out-groups (Jackson et al. 1996). Few studies to date have empirically tested in an e-commerce environment whether in-group members are likely to perceive out-group members unfavorably (Hewstone, Rubin, and Willis 2002). The findings from this study suggest that as blacks' ethnic identity increases, their attitudes toward white product endorsers decrease or become more negative.
Implications
The results of this study offer several implications for businesses with an online component. Marketers and advertisers who employ computer agents on their clients' websites should ensure that the sites provide many options in terms of the agents' race and gender, because consumers interact more easily with virtual characters to which they feel similar. Black consumers in particular may perceive a greater degree of inclusiveness on commercial websites that provide information from or interaction opportunities with racially diverse agents. Blacks may also appreciate the time, energy, and thought firms took to include a racially similar representative in a nontargeted commercial website.
This study also has implications for black-owned sites. According to Noyes (2010), African American entrepreneurs face fewer opportunities and more challenges than members of other ethnic groups. In the United States, black-owned businesses tend to start out with less capital, are four times more likely to be denied credit than white-owned firms, and are 20% more likely to fail within their first four years than white-owned businesses (Noyes 2010). In addition, African Americans are less likely to benefit from the multigenerational family and social ties that so often lead to business partnerships among white-owned firms. African American firms hire minority employees in much higher proportions than other companies do, and fewer successful African American companies means fewer jobs and financial opportunities for multicultural neighborhoods and the country as a whole (Noyes 2010).
Although the use of an interactive computer agent on black-owned companies' website will not solve all their problems, it can potentially stimulate return visits from consumers and provide an engaging and interactive experience that makes their time spent on the site both enjoyable and informative. We argue that black consumers who visit these sites online are already strong ethnic identifiers and, as such, should respond to same-race agents with more favorable attitudes toward the website, higher levels of social presence, and greater degrees of identification and homophily with the agent, as our model indicates.
Blacks in the United States spend only five cents of every dollar with black-owned businesses (Brown 2010). If blacks directed more dollars toward black-owned businesses, these enterprises would strengthen, which in turn would aid the development of black communities. According to Thomas Boston, a Black Enterprise Board of Economists member and chief executive officer of an Atlanta-based consulting firm, "On average about 70% of the workforce of a Black-owned business is Black, and for every $100,000 of revenue that a Black-owned business generates, that accounts for one additional job in that city" (Brown 2010). In general, by supporting black businesses, consumers can help provide employment within the black community.
Further Research
The implications of this study indicate that marketers should develop ways to differentiate strong black ethnic identifiers from weaker identifiers. Also beneficial would be further exploration of the mediated characters that elicit favorable consumer attitudes among both strong and weak black ethnic identifiers. Appiah and Elias (2009) expand on this point by examining the impact of ethnically ambiguous female characters versus single-race black and white characters. They find that blacks with strong ethnic identities had more positive attitudes toward the black female agent and blacks with weak ethnic identities had more positive attitudes toward the white female agent. They also find that blacks with strong ethnic identities had more positive attitudes toward the ethnically ambiguous female agent. Finally, their analysis showed that blacks with weak ethnic identities perceived themselves as more similar to the white female agent. Understanding how blacks with weak ethnic identities responded in that study is important because it suggests that advertisers wanting to reach a broad-based audience with a mainstream message and character can be successful in appealing to blacks, but in reality this promise only pertains to those with weak ethnic identities. Using mainstream messages and characters may suffice for advertisers trying to attract those in the middle in terms of their identities, values, opinions, and cultural practices, but doing so will not attract populations with strong ethnic identities. Understanding weak ethnic identifiers and their responses to persuasive messages is also important given the growing population of biracial and multiracial people in the United States. Seven million Americans identified themselves as biracial or multiracial in 2000 (La Ferla 2003), and thus an increasingly large number of people likely possess weak ethnic identities because they may lack strong ties to any one particular ethnic group. Therefore, the findings of this study may serve as a catalyst for those wanting to understand audiences with both stronger and weaker ethnic identities and how they might respond to persuasive messages that vary in cultural relevance.
In addition, future studies could examine the impact of character's synthetic or real human features on consumer attitudes. As mentioned previously, the characters we used in this study were "natural" or real human-looking agents. Future studies could examine whether black and white consumers respond more favorably to black and white human agents or black and white virtual agents on e-commerce websites. Specifically, does the race of a computer-generated synthetic character generate the same racial responses in black and white consumers as natural-looking, human product presenters? Virtual human entities offer a growing opportunity for research to test how far the effects of ethnic identity spread. Does ethnic identity predict racial preference in the same way with respect to virtual humans? If so, does it operate at the same or lesser strength? These questions have important theoretical implications beyond relevance to the area of digital communication.
Conclusion
This study is important for several reasons. First, studies on race and ethnicity tend to use a broad brush to paint a picture of the psychological underpinnings of the communication and behavioral practices of blacks in the United States. As a result, the complexities and multidimensional impact of race on blacks' and other ethnic minorities' decision-making processes are underemphasized. Perhaps the most recent and salient example of the intricacy of the issues facing blacks was during Barack Obama's presidential campaign. In 2007, Joe Biden, then-U.S. senator of Delaware, created controversy when he described Obama as articulate and clean. In response, in an article in Time magazine, Coates (2007) argues that though Obama's biracial identity helped him develop a sizable following in Middle America, it also provided an opportunity to question his authenticity as a black man. Coates also argues that in calling Obama articulate and clean, Biden was implying that ordinary black people and even those who aspire to the highest office in the land are none of these things. The implication of Coates's argument extends beyond the political landscape and is significant for advertisers and marketers, as well as communicators in the health and public service sphere. Because blacks are accustomed to interacting with myriad other black people, using a black image may not be enough to elicit favorable responses in black and minority audiences if that image does not sufficiently embody racial sincerity and similarity to its audience. However, the strength of ethic identity could override this response. Although an understanding of racial sincerity is beyond the scope of this study, research could explore it in the future.
Second, blacks' social identity remains a complex entity when examined from the perspective of social identity theory. Omi and Winant (1994) distinguish between two languages of race-one in which members of racial minorities, particularly blacks, view race as a fundamental aspect of history and everyday experience and another in which whites view race as a tangential, unimportant reality. Many studies have reinforced this observation and have shown that a source's race and ethnicity in a given context is a salient communication feature for ethnic minorities (Appiah 2002; Deshpandé and Stayman 1994). However, marketers continue to attempt to reach consumers of all ethnicities with mainstream media and characters. The current literature, along with the results of this study, shows that marketers would be better off using black characters in their online ads, especially when trying to reach that particular demographic. As this study indicates, black viewers are more likely to experience social presence when interacting with these characters, express a more favorable evaluation of the website, identify better with the source, and believe that they are more similar to the source if the characters are black. Black participants with stronger ethnic identities were negatively affected by white characters presenting the products. Thus, further research might explore the ways communication to ethnic groups could be improved and made more effective not only for advertising and political campaigners but also for the distribution of health-related information.
Limitations
A main limitation of the study was the small sample size. We used maximum likelihood, a large-sample estimator that assumes multivariate normality, to test the structural equation model, consistent with the literature (Holbert and Stephenson 2002; Joreskog 1973b). Recommendations of the appropriate sample size for conducting analyses using SEM are mixed (for a review, see Holbert and Stephenson 2002). However, because of the relative simplicity of the model, the reliability of the measures, and the absence of a hard-and-fast rule, we deemed the sample size appropriate. Note that our use of a college student sample, which featured students with a high education level, may have led to more liberal responses than that of the general population. Additional limitations of the study include the artificial nature or viewing environment of the laboratory, which forced participants to follow the commands of a computer agent. This environment is in contrast with how people normally interact with most e-commerce websites. Finally, the stimuli for the study included men only, which could potentially result in gender-specific effects.
References
Aaker, Jennifer, Anne Brumbaugh, and Sonya Grier (2000), "Non-Target Market Effects and Viewer Distinctiveness: The Impact of Target Marketing on Attitudes," Journal of Consumer Psychology, 9 (3), 127-40.
Abraham, Linus (2003), "Media Stereotypes of African-Americans," in Images That Injure, Paul Martin Lester and Susan Dente Ross, eds. Westport, CT: Praeger, 87-92.
African Americans by the Numbers (2010), Infoplease, http://www.infoplease.com/spot/bhmcensus1.html (accessed November 28, 2010).
Aguinis, Herman and Eugene Stone-Romero (1997), "Methodological Artifacts in Moderated Multiple Regression and Their Effects on Statistical Power," Journal of Applied Psychology, 82 (1), 192-206.
Ahrens, Frank (2006), "The Nearly Personal Touch: Marketers Use Avatars to Put an Animated Face with the Name," The Washington Post, July 15, http://www.washingtonpost.com/wp-dyn/content/article/2006/07/14/AR2006071401587.html (accessed December 7, 2007).
Appiah, Osei (2001), "Black, White, Hispanic and Asian American Adolescents' Responses to Culturally Embedded Ads," Howard Journal of Communications, 12, 29-48.
--- (2002), "Black and White Viewers' Perception and Recall of Occupational Characters on Television," Journal of Communication, 52 (4), 776-93.
--- (2004), "Effects of Ethnic Identification on Web Browsers' Attitudes Toward and Navigational Patterns on Race-Targeted Sites," Communication Research, 31, 312-37.
--- and Troy Elias (2009), "Ethnic Identity and the Effects of Ethnically-Targeted and Ethnically-Ambiguous Computer-Generated Agents on Browsers Evaluations of a Commercial Website," in Virtual Social Identity and Consumer Behavior, Natalie T. Wood and Michael R. Solomon, eds. Mahwah, NJ: Lawrence Erlbaum Associates, 159-180.
Bentler, Peter, M. (1990), "Comparative Fit Indexes in Structural Models," Psychological Bulletin, 107, 238-46.
Berry, John W. (1997), "Immigration, Acculturation, and Adaptation," Applied Psychology: An International Review, 46 (1), 5-34.
Brewer, Marilyn. B. (1991), "The Social Self: On Being the Same and Different at the Same Time," Personality and Social Psychology Bulletin, 17, 475-82.
Brown, Carolyn M. (2010), "I Will Support the Creation and Growth of Minority-Owned Businesses," Black Enterprise, http://www.blackenterprise.com/personal-finance/wealth-for-life-principles/2009/03/13/i-will-support-the-creation-and-growth-of-minority-owned-businesses/ (accessed November 26, 2010).
Byrne, Barbara M. (2001), Structural Equation Modeling With AMOS: Basic Concepts, Applications, and Programming. Mahwah, NJ: Lawrence Erlbaum Associates.
Chen, Henian, Patricia Cohen, and Sophie Chen (2007), "Biased Odds Ratios from Dichotomization of Age," Statistics in Medicine, 26 (18), 3487-97.
Coates, Ta-Nehisi Paul (2007), "Is Obama Black Enough?" Time, February 1, http://www.time.com/time/nation/article/0,8599,1584736,00.html (accessed November 28, 2010).
Dates, Jannette (1980), "Race, Racial Attitudes and Adolescent Perceptions of Black Television Characters," Journal of Broadcasting, 24, 549-60.
Desphandé, Rohit and Douglas M. Stayman (1994), "A Tale of Two Cities: Distinctiveness Theory and Advertising Effectiveness," Journal of Marketing Research, 31 (February), 57-64.
Entman, Robert M. and Andrew Rojecki (2000), The Black Image in the White Mind: Media and Race in America. Chicago: University of Chicago Press.
Ethier, Kathleen A. and Kay Deaux (1994), "Negotiating Social Identity when Contexts Change: Maintaining Identification and Responding to Threat," Journal of Personality and Social Psychology, 67 (2), 243-51.
Forehand, Mark R. and Rohit Deshpandé (2001), "What We See Makes Us Who We Are: Priming Ethnic Self-Awareness and Advertising Response," Journal of Marketing Research, 38 (August), 336-48.
Fujioka, Yuki (2005), "Emotional TV Viewing and Minority Audience: How Mexican Americans Process and Evaluate TV News About In-Group Members," Communication Research, 32 (October), 566-93.
Gong, Li, Osei Appiah, and Troy Elias (2007), "See Minorities Through the Lens of Ethnic Identity: Reflected unto Racial Representations of Real Humans and Virtual Humans," paper presented at the International and Intercultural Communication Division of the 93rd Annual Convention of the National Communication Association, Chicago (November).
--- and Clifford Nass (2007), "When a Talking-Face Computer Agent Is Half-Human and Half-Humanoid: Human Identity and Consistency Preference," Human Communication Research, 33, 163-93.
Green, Corliss L. (1999), "Ethnic Evaluations of Advertising: Interaction Effects of Strength of Ethnic Identification, Media Placement, and Degree of Racial Composition," Journal of Advertising, 28 (1), 49-64.
Hallman, Charles (2008), "By 2011 Black Buying Power to Hit Trillion-Dollar Mark," New America Media, January 10, http://news.newamericamedia.org/news/view_article.html?article_id=990c03344a0ec68e9edcfcaaa70e7563 (accessed November 28, 2010).
Hecht, Michael L., Linda Kathryn Larkey, and Jill N. Johnson (1992), "African American and European American Perceptions of Problematic Issues in Interethnic Communication Effectiveness," Human Communication Research, 19 (March), 209-36.
Hewstone, Miles, Mark Rubin, and Hazel Willis (2002), "Intergroup Bias," Annual Review of Psychology, 53, 575-604.
Hogg, Michael A. (2004), "Social Categorization, Depersonalization, and Group Behavior," in Self and Social Identity, Marilynn B. Brewer and Miles Hewstone, eds. Malden, MA: Blackwell Publishing, 203-31.
Holbert, Robert Lance and Micael T. Stephenson (2002), "Structural Equation Modeling in the Communication Sciences, 1995-2000," Human Communication Research, 28 (October), 531-51.
Hu, Li-tze and Peter M. Bentler (1999), "Cutoff Criteria for Fit Indexes in Covariance Structure Analysis: Conventional Criteria Versus New Alternatives," Structural Equation Modeling, 6 (1), 1-55.
Humphreys, Jeffrey M. (2008), "The Multicultural Economy 2008," The Selig Center for Economic Growth, http://www.terry.uga.edu/selig/docs/buying_power_2008.pdf (accessed December 12, 2009).
Jackson, Linda A., Linda A. Sullivan, Richard Harnish, and Carole N. Hodge (1996), "Achieving Positive Social Identity: Social Mobility, Social Creativity, and Permeability of Group Boundaries," Journal of Personality and Social Psychology, 70 (2), 241-54.
Jaret, Charles and Donald C. Reitzes (1999), "The Importance of Racial-Ethnic Identity and Social Setting For Blacks, Whites, and Multiracials," Sociological Perspectives, 42 (4), 711-37.
Joreskog, Karl Gustav (1973a), "A General Method for Estimating a Linear Structural Equation System," in Structural Equation Models in the Social Sciences, Arthur S. Goldberger and Otis Dudley Duncan, eds. New York: Seminar Press, 85-112.
--- (1973b), "Testing Structural Equation Models," in Testing Structural Equation Models, Kenneth A. Bollen and J. Scott Long, eds. Newbury Park, CA: Sage Publications, 294-316.
Kline, Rex B. (2005), Principles and Practice of Structural Equation Modeling, 2d ed. New York: The Guilford Press.
La Ferla, Ruth (2003). "Generation E.A: Ethnically Ambiguous," The New York Times, December 28, http://www.nytimes.com/2003/12/28/fashion/28ETHN.html (accessed December 30, 2003).
Lee, Eun-Ju and Clifford Nass (2002), "Experimental Tests of Normative Group Influence and Representation Effects in Computer-Mediated Communication: When Interacting Via Computers Differs from Interacting with Computers," Human Communication Research, 28 (3), 349-81.
Little, Todd D., William A. Cunningham, Golan Shahar, and Keith F. Widaman (2002), "To Parcel or Not to Parcel: Exploring the Question, Weighing the Merits," Structural Equation Modeling, 9 (2), 151-73.
Mastro, Dana E. (2003), "A Social Identity Approach to Understanding the Impact of Television Messages," Communication Monographs, 70 (2), 98-113.
McCroskey, James C., Virginia P. Richmond, and John A. Daly (1975), "The Development of a Measure of Perceived Homophily in Interpersonal Communication," Human Communication Research, 1, 323-32.
Mills, Charles W. (1997), The Racial Contract. Ithaca, NY: Cornell University Press.
Negy, Charles, Tara L. Shreve, Bernard J. Jensen, and Nizam Uddin (2003), "Ethnic Identity, Self-Esteem, and Ethnocentrism: A Study of Social Identity Versus Multicultural Theory of Development,"Cultural Diversity & Ethnic Minority Psychology, 9 (4), 333-44.
Noyes, Katherine (2010), "Support African-American Entrepreneurs: Buy from Black-Owned Businesses," http://charityguide.org/diversity (accessed November 19, 2010).
Omi, Michael and Howard Winant (1994), Racial Formation in the United States: From the 1960s to the 1980s. New York: Routledge.
Operario, Don and Susan T. Fiske (2001), "Ethnic Identity Moderates Perceptions of Prejudice: Judgments of Personal Versus Group Discrimination and Subtle Versus Blatant Bias," Personality and Social Psychology Bulletin, 27 (5), 550-61.
Phinney, Jean S. (1990), "Ethnic Identity in Adolescents and Adults: Review of Research," Psychological Bulletin, 108, 499-514.
--- (1992), "The Multigroup Ethnic Identity Measure: A New Scale for Use with Diverse Groups," Journal of Adolescent Research, 7 (2), 156-76.
--- (2005), "Ethnic Identity in Late Modern Times: A Response to Rattansi and Phoenix," Identity, 5 (2), 187-94.
--- and Mukosolu Onwughalu (1996), "Racial Identity and Perception of American Ideals among African American and African Students in the United States," International Journal of Intercultural Relations, 20 (2), 127-40.
Rice, Ronald E. and Gail Love (1987), "Electronic Emotion: Socioemotional Content in a Computer-Mediated Communication Network," Communication Research, 14 (1), 85-108.
Rogers, Everett M. and Dilip K. Bhowmik (1970), "Homophily-Heterophily: Relational Concepts for Communication Research," Public Opinion Quarterly, 34 (4), 523-38.
Short, John, Ederyn Williams, and Bruce Christie (1976), The Social Psychology of Telecommunications. New York: John Wiley & Sons.
Smedley, Audrey (1999), Race in North America: Origin and Evolution of a Worldview. Boulder, CO: Westview Press.
Stephenson, Michael T. and R Lance Holbert (2003), "A Monte Carlo Simulation of Observable Versus Latent Variable Structural Equation Modeling Techniques," Communication Research, 30, 332-54.
Tajfel, Henri (1974), "Social Identity and Intergroup Behaviour," Social Science Information, 13 (2), 65-93.
--- (1981), Human Groups and Social Categories. New York: Cambridge University Press.
--- and John C. Turner (1986), "The Social Identity Theory of Intergroup Behavior," in Psychology of Intergroup Relations, Stephen Worchel and William G. Austin, eds. Chicago: Nelson-Hall, 7-24.
About the Authors
Troy Elias (Ph.D., The Ohio State University) is an assistant professor at the College of Journalism and Communications at the University of Florida. Email: telias@jou.ufl.edu.
Osei Appiah (Ph.D., Stanford University) is an associate professor at the School of Communication at The Ohio State University.
Li Gong (Ph.D., Stanford University) is a principal customer researcher and strategist at SAP.
