Racial Appearance Bias: Improving Evidence-Based Policies to Address Racial Disparities

Racial Versus Racial Phenotypicality Biases

Social psychologists tend to focus their energies exploring the antecedents and consequences of racial bias. Research shows that a person’s racial category membership can be inferred from a variety of cues such as physical appearance (skin color, clothing style, body movement, and morphology), language/dialect, nationality, and name (Brewer & Feinstein, 1999; Fiske, Lin, & Neuberg, 1999; Freeman & Ambady, 2011; Maddox, 2004; Zebrowitz, 1996). Once activated, racial stereotypes and prejudices can guide judgment and behavior toward individual group members. Armed with this and other evidence, researchers seek to contribute to the development of interventions designed to mitigate racial bias.

That said, one overlooked aspect of racial bias may undermine efforts to address racial disparities: Within-race variation on some cues also links to several important outcomes. When considering physical appearance, racial phenotypicality bias reflects a tendency to use within-race variation in facial appearance to guide judgments and behavior (Maddox, 2004). For example, Blacks are stereotypically thought to have dark skin tone, broad noses, full lips, and tightly curled hair. Research exploring within-race variation on these features contends that individuals with more stereotypical appearance have poorer outcomes compared with their less stereotypical counterparts (Adams, Kurtz-Costes, & Hoffman, 2016; Maddox, 2004). The current selective review focuses on research with relevance for the development of policy and practice. Afterward, we offer some considerations for those seeking to develop policy and programs to address the effects of phenotypicality bias.

Criminal Justice

Research in this domain is rooted in the association between phenotypical appearance and negative stereotypes. Higher phenotypicality has been consistently associated with more stereotypically Black crimes, worse interactions with police, and worse criminal justice outcomes. High phenotypical Black men are generally associated with more negative characteristics than their low phenotypical counterparts. As such, research has found that Black men are perceived as more physically bigger and threatening than White men, and this difference is related to phenotypicality (Wilson, Hugenberg, & Rule, 2017). These perceptions predict support for physical force used against Black suspects.

Racial phenotypicality has also been generally associated with the concept of immorality and crime. The notion that “Black is bad” may be a commonly used heuristic in social judgment. In archival newspaper data, photographs accompanying articles about Black and White celebrities and politicians were printed to depict subjects’ skin tones as darker if the article was critical versus complimentary (Alter, Stern, Granot, & Balcetis, 2016). Furthermore, participants indicated that photographs showing darker skin tones were more representative of a person who had committed an immoral versus moral act.

The association between Blackness and “bad” extends into populations working within the criminal justice system. Police officers primed with crime were more likely to falsely identify highly stereotypic Black faces in a recognition task and were more likely to rate these faces as criminal (Eberhardt, Goff, Purdie, & Davies, 2004; Study 4).

Subsequent investigations suggest that racial phenotypicality may be associated with certain categories of criminal activity. For example, perceivers mistakenly recalled the appearance of the suspects accused of stereotypicality Black crimes (e.g., robberies, drive-by shootings) as being higher in racial phenotypicality compared with those accused of stereotypically White crimes (e.g. Internet hacking, serial killing; Osborne & Davies, 2012). Furthermore, victim characteristics can interact with the stereotypicality of a crime. Specifically, when eyewitnesses believed the victim of a stereotypically Black crime was White or female, they falsely exaggerated the perpetrator’s Black phenotypic appearance (Davies, Hutchinson, Osborne, & Eberhardt, 2016).

The impact of suspect race on the nature of police encounters also shows phenotypicality bias. In one study, participants showed a higher propensity to shoot high versus low phenotypical Black targets holding weapons (Kahn & Davies, 2011). Police are more likely to shoot unarmed Blacks than unarmed Whites (e.g., Correll, Park, Judd, & Wittenbrink, 2002). In another, White suspect phenotypicality seemed to influence decisions to shoot suspects among a sample of police officers and community members (Ma & Correll, 2011).

Finally, existing data can help to evaluate the impact of racial phenotypicality on sentencing decisions (Blair, Judd, & Chapleau, 2004): Although sentence severity was relatively the same between Black and White inmates with equivalent criminal backgrounds, Afrocentric features within both Black and White inmate populations predicted harsher criminal sentences (see also Eberhardt, Davies, Purdie-Vaughns, & Johnson, 2006).

Education

Higher phenotypicality has been associated with poor educational success, higher disciplinary outcomes, and poor interactions with fellow students. Darker skinned Blacks complete fewer years of schooling (Hughes & Hertel, 1990; Monk, 2014) and are less likely to pursue education beyond high school (Ryabov, 2013). Skin tone also relates to school suspension, particularly for female students (Blake, Keith, Luo, Le, & Salter, 2017; Hannon, DeFina, & Bruch, 2013). Even after controlling for frequency of delinquent behavior, as well as a host of individual- and school-level characteristics associated with school discipline, dark-skinned Black students were at a higher risk for school suspension than were light-skinned Black students. African American females with the darkest skin complexion exhibit the highest risk for school suspension. Compared with their light-skinned and White female peers, dark-skinned Black girls were up to 3 times as likely to receive an out-of-school suspension.

Skin color discrimination also has a history within Historically Black Colleges and Universities (HBCUs). In general, HBCUs served as either a barricade or a conduit toward achieving social acceptance and influence within the Black community (Gasman & Abiola, 2016). Whether Black colleges were a barrier to success depended upon one’s skin color, family status, and family education (Taylor, 2009). Lighter skinned Blacks frequented the top three HBCU institutions, above the others, creating clusters of light-skinned Blacks and making their presence the norm at the most prestigious institutions (Gatewood, 2000). The segregation by skin tone within HBCUs was particularly pronounced in Black sororities. Although most Black Greek organizations started as a way of bringing Blacks together on college campuses and creating a sense of solidarity, potential sorority sisters at many institutions were subject to blue vein and brown paper bag tests to see whether they were too dark to be admitted to the organization (Giddings, 2007; Sheffer-Parrott, 2009). These findings are representative of a general association between racial phenotypicality and social rejection (e.g., Hebl, Williams, Sundermann, Kell, & Davies, 2012).

One component of the relationship between phenotypicality and poor educational outcomes may partially be rooted in the relationship between SES and phenotypicality. In the United States, parental SES is a major influence on educational success (Kao & Thompson, 2003). Therefore, because more phenotypically Black individuals tend to have lower SES, their children may lack exposure to the best schools and resources.

Employment and Income

Higher phenotypicality is associated with lower socioeconomic success due to lower perceived intrinsic ability and higher perceived threat. Compared to lower phenotypical Blacks, higher phenotypical Blacks obtain lower levels of employment and economic success, lower levels of job quality, and lower salaries within less prominent or well-paying jobs (Hughes & Hertel, 1990; Ryabov, 2013; Seltzer & Smith, 1991; Visser, 2017). Furthermore, years in the labor force is negatively associated with successful self-employment for dark-skinned Blacks (Devaraj & Patel, 2017).

These disparities may stem, in part, from a perceived association between phenotypicality and innate competence. Participants falsely remembered Black faces as lighter than they were when subliminally primed with “educated” compared with the words “ignorant” or “athletic” (Ben-Zeev, Dennehy, Goodrich, Kolarik, & Geisler, 2014). Black–White biracial individuals are judged as more competent when they have low (vs. moderate or high) levels of Black phenotypicality (Skinner & Nicolas, 2015).

Accordingly, higher phenotypic Blacks are generally perceived to be less desirable job candidates than are lower phenotypic Blacks (Harrison & Thomas, 2009): Dark-skinned job candidates were more harshly evaluated and less likely to be recommended for the job compared with lighter candidates, even when the resumes and qualifications of the two candidates were identical. Furthermore, participants evaluated highly educated darker candidates on par with light-skinned candidates with less education and experience. In another study, darker job candidates were judged as less suitable than lighter candidates, particularly for jobs involving higher client contact or status (Derous, Pepermans, & Ryan, 2016).

Finally, perceived racial phenotypicality may also depend on the nature of the current economic environment. When participants felt their economic resources were scarce, they perceived mixed-raced faces as more stereotypically Black than they objectively were, visualized faces that were more stereotypically Black, and allocated less money to these Black faces (Krosch & Amodio, 2014). This perceptual distortion may have reflected a motivation to conserve economic resources for one’s own racial group, as perceiving people as more phenotypically Black may facilitate discrimination.

Mental and Physical Health

The evidence on the relationship between racial phenotypical appearance and health outcomes is scant and inconsistent, suggesting a need for more empirical investigation. Research has long showed a link between discrimination and health outcomes (Clark, Anderson, Clark, & Williams, 1999). Accordingly, because higher phenotypical Blacks experience more discrimination than lower phenotypical Blacks (Keith & Herring, 1991), higher phenotypical Blacks may also endure worse mental and physical health. Indeed, darker skinned Blacks were more likely to perceive the racism they experienced as stressful (Klonoff & Landrine, 2000). Consistent with this idea, darker skin tone is associated with indicators of heart disease, such as higher diastolic and systolic blood pressure (e.g., Gleiberman, Harburg, Frone, Russell, & Cooper, 1995). In a study exploring this link, greater degree of Afrocentric facial appearance predicted poor health outcomes, and this relationship was mediated by participants’ perceived experience of unfair treatment within several social domains (e.g., education, jobs, criminal justice; Hagiwara, Penner, Gonzalez, & Albrecht, 2013).

Phenotypicality may also relate to mental health. For example, dark-skinned Chicano men born in the United States show higher levels of depression than their light-skinned counterparts (Codina & Montalvo, 1994). In focus groups, Black women of differing skin colors reported distinct experiences related to psychological duress (e.g., bullying, devalued social capital, abuse), but psychological duress was not measured directly (Hall, 2017).

Although a handful of studies have suggested a link between racial phenotypicality and health outcomes, not all the evidence is supportive, some finding no association between skin color, self-reported discrimination, and self-reported mental or physical health (Borrell, Kiefe, Williams, Diez-Roux, & Gordon-Larsen, 2006; Keith, Lincoln, Taylor, & Jackson, 2010).

Work to Mitigate the Impact, Rather Than the Expression, of Racial Phenotypicality Bias

On HBCU campuses, several solutions might combat discrimination as a function of skin tone (Gasman & Abiola, 2016): providing regularly structured opportunities for students to engage in discussion through instruction and campus-wide events, creating professional development opportunities explicitly focused on skin color discrimination, and conducting self-studies to determine whether any internal patterns exhibit signs of colorism within the institution. Efforts of this type could have several goals. One is to attempt to develop individuals’ understanding of the nature of bias to prevent bias in their interactions. Short-term efforts to control an individual’s use of racial stereotypes can be effective with sufficient institutional support (Kalev, Dobbin, & Kelly, 2006; Penner et al., 2014).

That said, even the short-term benefits may not extend to racial phenotypicality. Participants given explicit instructions to suppress racial stereotypes were able to do so; however, racial phenotypicality continued to affect their judgments of others (Blair, Judd, & Fallman, 2004). Although it has not been explored empirically, this instructional benefit could extend to racial phenotypicality if people were more aware of that dimension of bias. As several authors have noted, antibias education alone is not sufficient to mitigate bias expression (e.g., Glaser et al., 2014; Penner et al., 2014). In addition, strategies to limit the expression of bias in the long-term may require too much conscious effort to be effective (Lai et al., 2014).

Accordingly, many authors recommend moving away from approaches that attempt to change the hearts and minds of potential evaluators and instead focus on contextual factors that guide individuals’ judgments. Here, education efforts may be crucial as policy makers craft interventions that would address race and racial phenotypicality. Although not reviewed in detail above, marketing practices and media consumption can have a direct impact on perceptions and evaluations of skin tone differences. When participants read a tabloid-like news story about dark-skinned criminals, they judged subsequently presented dark-skinned faces as more threatening than light-skinned faces (Arendt, Steindl, & Vitouch, 2015). Similarly, individuals with higher levels of news viewing reported more emotional discomfort in response to a crime story about a dark-skinned Blacks perpetrator compared with less frequent news viewers (Dixon & Maddox, 2005). This suggests that stereotypical depictions of Blacks heighten viewers’ sensitivity to phenotypic variation. News organizations have choice in the stories that they select to air; movie and television producers have discretion in terms of who they cast for certain roles. Awareness of racial and racial phenotypicality biases could lead to informed decisions that provide viewers with challenges to the status quo.

The impact of stereotypical depictions extends to the classroom and organizational perceptions. In structural and symbolic interventions to address racial bias in classroom design, potential considerations should inform selecting classroom artifacts (Cheryan et al., 2014). One is to strive for diverse representations and avoid stereotypical depictions of people, such as historical figures. Efforts like these can work to make underrepresented students feel more included, benefitting performance. A racial phenotypicality “tweak” to this policy might urge educators to go a step further and consider the facial characteristics as a component of the stereotype. For example, a classroom filled with images of lighter skinned Black scholars and darker skinned Black athletes feeds into the status quo. Other evidence suggests that these representations impact perceptions of organizations. Black participants evaluated organizations with highly phenotypically Black employees as more trustworthy and a more desirable workplace than organizations that included only low phenotypically Black or White employees (Kahn, Unzueta, Davies, Alston, & Lee, 2015).

In many contexts, one might cringe at the thought of using a selection policy that considers racial phenotype in determining whom to hire for a position or whom to grant a loan. Likewise, an affirmative action policy that considers phenotype for college admission seems heavy handed. Rather than including phenotype in the selection process, it might be better to make sure that decision makers are aware of its potential impact on judgments. Research on jury decision making suggests that individuals who discuss the possibility that race may be a factor in judgments of culpability make less biased decisions (Sommers & Ellsworth, 2001). Making phenotype salient may encourage discussion among jurors and selection committee members that will serve to mitigate phenotypicality biases.

Some race-based policies may be less susceptible to the residual influence of racial phenotypicality. To reduce the impact of racial bias on policing, strategies focused on decision makers might include limiting the degree of discretion they have in judgments (Glaser et al., 2014). Doing so could limit the uncertainty in decision-making contexts where biases tend to have greater impact. Similarly, strategies could encourage physicians to individuate their clients and thus decrease the likelihood of seeing them through a racial lens (Penner et al., 2014). Each of these strategies focuses on structuring the interaction in ways that can limit the impact of racial bias (e.g., Brewer & Feinstein, 1999; Fiske, Lin, & Neuberg, 1999). In medical and educational contexts, stereotype threat interventions focused on patients and students may have similar benefits toward reducing racial biases (Glaser et al., 2014; Penner et al., 2014).

Although it is unclear as to whether these strategies can effectively limit the impact of racial and racial phenotypicality biases, we cannot know this if policy makers do not know to include assessments of racial phenotypicality when exploring potential disparities (see below).

Education About the Potential Impact of Racial Phenotypicality Bias

Although skin tone–based discrimination is discussed within the African American community, it has been a taboo topic in interracial contexts (Russell, Wilson, & Hall, 1992). This is born in part out of concern over in-group hypocrisy. Although once explicit, in modern times, Whites may be largely unaware of the influence that racial phenotypicality can have on their judgments of Blacks (Hagiwara, Kashy, & Cesario, 2012). Thus, while the benefits of bias education can be mixed in terms of limiting individuals’ expressions of racial bias (e.g., Kalev et al., 2006), educating individuals about the existence of racial phenotypicality bias would seem to be crucial for policy makers to be mindful in their design of programs. Thus, we would recommend efforts to educate policy maker and other individuals about the nature of racial phenotypicality biases.

Assess Race and Racial Phenotypicality to Document Potential Disparities

Before implementing a phenotype-related policy, decision makers should first and foremost assess the extent to which racial phenotypicality bias may impact outcomes in their domain. As stated above, the literature exploring the role of phenotypicality is considerably smaller than that exploring race. As a result, we have less information about where phenotypicality might contribute to racial disparities. If researchers fail to assess phenotypicality in addition to race, we lack the ability to explore ways in which phenotypical appearance may influence the outcomes of others. If not, disparities will slip through the holes in the race-based filter. How to do this is a nontrivial question.

One consideration is when to focus on several or a single feature. Although several facial characteristics can contribute to racial phenotypicality, most empirical efforts focus either solely on skin tone variation or more global assessments that include skin tone. Thus, skin tone is likely a good proxy for racial phenotypicality. However, other research has suggested that skin tone and facial features can have unique effects on perception and judgment (Hagiwara et al., 2012). Relatedly, skin tone may be a stronger proxy for phenotypicality in assessments of some groups versus others where skin tone variation is limited, such as some East Asian groups in the United States. (Strom, Zebrowitz, Zhang, Bronstad, & Lee, 2012).

Another consideration is whether one should rely on self- or observer-reported racial phenotypicality. Typically, racial identity assessments are collected through self-reports in survey and experimental research. From a practical standpoint, self-report would be easiest approach when collecting information on racial phenotypicality. However, the observer reports of racial phenotypicality may differ from self-reports. People’s self-reported skin tone may reflect a bias as those with extremely light or dark skin tone based on observer assessments show self-assessments closer to the middle of the skin tone scale (Coard, Breland, & Raskin, 2001).

The use of self-assessment versus observer assessment may also depend on the domain of inquiry. To the extent that racial phenotypicality biases can only operate when an individual is visible to a perceiver, they are more likely to affect outcomes in contexts involving face-to-face interaction or visual information (photographs or video). In these contexts, outcomes may be driven more by the targets’ appearance as seen by others than by how the targets see themselves, suggesting that observer assessments would be more predictive. However, the degree of consistency between the self-assessment and observer assessment of racial group member can predict health outcomes. Individuals who considered themselves White but were seen as non-White by others had relatively high risk for negative health outcomes (Veenstra, 2011). Thus, it may be important to include both self-assessment and observer assessments of race and racial phenotypicality.

Caveat

The recommendations in this analysis are based on a literature focused heavily, almost exclusively, on perceptions and outcomes associated with men. Although one might assume the implications for women are similar, this assumption is unwarranted (Eagly & Kite, 1987). Similar to the current thesis, scholars exploring intersectionality, or judgments of individuals as simultaneously members of multiple social categories (e.g., race and gender), stress that models focused on a single dimension of identity will fail to capture nuances in the treatment and experience of group members. Social psychological research is coming around to this realization, evidenced through an increase of research exploring the perceptions and experiences of multiracials (Gaither et al., 2014; Young, Sanchez, & Wilton, 2017) and the implications of intersectionality for judgment and outcomes (Remedios & Snyder, 2015a). Those seeking to design and improve effective social policy would be wise to heed this call for research and documentation (Remedios & Snyder, 2015b).