The Development and Validation of a Colorism Scale

Abstract

Two studies were conducted to develop and then validate a scale to measure the construct of Colorism. While colorism is a long discussed phenomena within the Black community in the United States, there have been virtually no attempts to measure the degree to which individuals embrace it. The In-Group Colorism Scale (ICS) was developed to assess the degree to which skin tone variation is important across five essential domains: Self-Concept, Affiliation, Attraction, Impression Formation, and Upward Mobility. The scale was empirically tested and then replicated using two distinct national samples of Black Americans (total sample of 783 participants). The ICS proved to have both good reliability and good structural validity. Moreover, the ICS proved to be significantly related to other important constructs such as parental socialization, skin tone, self-esteem, stereotypes, racial identity, and socioeconomic status. The implications of the ICS for both research and practical applications are discussed.

Keywords

colorism racial identity skin tone bias shadism

Colorism generally refers to the fact that variations between skin tone gradients are used as criterion for within-group differentiation and stratification among people of color. According to Burke (2008), “Colorism is the allocation of privilege and disadvantage according to the lightness or darkness of one’s skin” (p. 17). The implications of colorism extend across multiple levels of analyses for Black Americans, from the intrapersonal (e.g., self-esteem; Harvey, Labeach, Pridgen, & Gocial, 2005), to the interpersonal (e.g., romantic pairings; Hill, 2002), to the intragroup (e.g., familial socialization; Wilder & Cain, 2011), and ultimately to the intergroup (e.g., social inequality; Burton et al., 2010). It is important to note that at the root of these implications is the notion that individuals (both within and outside communities of color) do, in fact, embrace the ideology of colorism. That is, these implications hinge on the degree to which people actually assign significance and meaning to differences in skin tone gradients. To date, most of the evidence that people do assign significance and meaning to differences in skin tone comes from experimental studies (see Maddox & Chase, 2004; Maddox & Gray, 2002). These studies focus on manipulations of skin tone salience rather than direct assessments of the significance and meaning of colorism.

The purpose of this current research is to present a cross-validation of a scale designed to assess the degree to which people assign significance and meaning to skin tone gradients (i.e., colorism). This scale could be used to assess such significance and meaning on the part of those inside a community of color. This endeavor reflects a shift in the predominant approach to the way that colorism is often analyzed. For the most part, the analysis of colorism has been heavily focused around three primary poles: inequality and socioeconomic mobility, attractiveness, and racial socialization of children (cf. Burke, 2008; Burton et al., 2010). These topics reflect a macro-level of analysis in which they point to the role of colorism in the larger social context. Within this larger context, colorism serves as a social stratification mechanism that typically pits lighter skinned group members on the top and darker skinned group members on the bottom across a number of outcomes. What appears to be notably absent from this analysis is an approach that recognizes that the significance or meaning associated with skin tone gradients might vary from person to person. Thus, the degree to which colorism is likely to be evoked in a given situation will depend on the degree to which people in the situation, whether explicitly or implicitly, believe that skin color variations matter. We suggest that this approach reflects a more psychosocial level of analysis.

The Nature of Colorism

A psychosocial approach to the study of colorism reflects an individual level of analysis of the perceived significance and meaning associated with skin tone variation. Across the literature on colorism, there are four primary areas in which colorism has been perceived to hold significance and meaning: Self-Concept, Impression Formation, Attraction/Affiliation, and Upward Mobility.

Self-Concept

Perhaps the most paramount implications of colorism are its impact on how Black Americans define and evaluate themselves. The preference for light skin that was a product of slavery has over time embedded itself into the definitive core values of Black American culture (Coard, Breland, & Raskin, 2001). With respect to self-evaluations, the affective component of the self-concept, early research found that Black children’s self-evaluations were based a good deal on their skin tone, with higher self-evaluations being linked to lighter skin (Jordan & Hernandez-Reif, 2009). Furthermore, Robinson and Ward (1995) found a relationship between self-esteem and self-reported skin tone among adults as well. Thus, the idea that a person’s self-concept is likely to be influenced by the significance placed on skin tone is well supported across age levels in the extant literature.

Impression Formation

The idea that the skin tone of a Black American might convey important information about the character of the person is rooted in slavery. Within slavery, Black Americans were differentiated by their perceived intelligence, skillfulness, and monetary worth based on their skin tone in favor of light over dark skin tone (Coard, Breland, & Raskin, 2001; Maddox & Gray, 2002). Research has found that these early distinctions between persons of light and dark skin tone have maintained their course such that the impressions that people form of Black Americans are still heavily dependent on their skin tone (Chin-Quee, 1992; Maddox & Chase, 2004; Maddox & Gray, 2002). These findings suggest that skin tone may be an important factor when forming impressions of people of color.

Attraction/Affiliation

A great deal of the literature on Colorism has focused on the attractiveness of skin tones. Historically, there has been a tendency for Black Americans of lighter skin tone to be favored as social companions (Hughes & Hertel, 1990). Furthermore, beyond mere friendship or association, preferences for romantic mates with a lighter skin tone has also been found in a number of studies (Harvey et al., 2005; Wilder, 2009; Wilder & Cain, 2011). With regard to Black Americans, lighter skin tone is generally rated as more attractive than darker skin tone by both Black Americans and White Americans (Hill, 2002). For example, Maddox and Gray (2002) found that the term “attractive“ was more likely to be attributed to pictures of people with light skin tone, and “unattractive“ was more likely to be assigned to people with dark skin tone.

Upward Mobility

The driving force behind much of the research on colorism has been the notion that skin tone plays a significant role in shaping life chances and opportunities (Wilder, 2010). For example, studies have explored the wealth inequality that results from differing social privileges and opportunities afforded to some skin tones over others. Several studies (Goldsmith, Hamilton, & Darity, 2007; Hill, 2002; Hughes & Hertel, 1990; Keith & Herring, 1991) have suggested that skin tone variations have important implications for occupational status, income, and level of educational attainment. Thus, there seems to be a widespread explicit and/or implicit belief that skin tone could be used as a gateway to upward mobility on these traditional indicators of socioeconomic success.

Given that skin tone variation appears to hold significance across these five areas, and perhaps others, colorism can be said to be multidimensional in nature. Thus, any approach to the analysis of it must also be multidimensional. In so doing, an allowance is made for the complexity that is likely to exist in the relationships both among the various dimensions of it and between those dimensions and other important constructs.

Previous Measures

In our review of the extant literature on colorism, we could not find a single self-report scale that explicitly measures the significance and meaning that an individual places on skin tone variations. This gap is not totally unexpected due to the level at which colorism has most typically been conceptualized. As discussed earlier, for the most part, colorism has been considered as a macro-level structural-political phenomena. Thus, the unit of analysis has been at the societal level and the indicators for its existence have also been societal indicators (e.g., discrimination rates, wealth inequality estimates, marital rates, etc.). At this level, whole societies are judged on the degree to which colorism is present. Attempts to examine colorism at the relatively more micro-individual level are sparse. This level allows for individual variation in the degree to which colorism is embraced and thereby locates colorism within the psyche of the individual more so than within the culture and climate of the society.

Most of the research on colorism at the psychosocial level has focused on assessing the skin tone of participants and then examining the association between their skin tones and important outcomes such as indicators of subjective well-being (e.g., self-esteem; Harvey et al., 2005; Thompson & Keith, 2001). These measures of participant skin tones are, at best, indirect measures of the significance and meaning of skin tone in which the significance of skin tone is only established to the degree to which correlations are found between skin tone and the particular outcome under study. Moreover, because they focus more on the participants’ own skin tone rather than their perceptions of others’ skin tones, they are limited to establishing the significance on only one of the four dimensions identified earlier, namely self-concept.

There is, however, an implicit measure of skin color bias that could be used to assess the significance that one might place on skin tone variations among Black Americans in particular. Project Implicit represents a collaborative research effort between researchers at Harvard University, the University of Virginia, and University of Washington. The website of Project Implicit features a Skin-tone Implicit Association Test (IAT; “light skin-dark skin”) among 11 other IAT tests. The accumulated results of participants who have been visiting the website and taking the test has revealed a general automatic preference for light skin relative to dark skin. This skin tone test, as with all IAT tests, measures implicit or unconscious processes believed to be associated with skin tone preferences. Thus, theoretically, such a measure could be used to reveal the salience of colorism for participants when making stereotypic judgments. However, it does not allow for probing the complex ideology of colorism due to the fact that the IAT appears to be measuring colorism as unidimensional rather than multidimensional. Thus, it could be indicative of significance but not necessarily meaning. A purposeful attempt to explore colorism must take into consideration the complex multidimensional nature of colorism in order to best understand both its structure and relationships with other important phenomena.

Colorism Versus Racism

Given the fact that both terms hold implications for people of color, it is nevertheless important to distinguish colorism from racism. While the definitions of racism vary considerably, the common denominator between the various definitions is that racism is the belief that some race or races are superior/inferior to others. Thus, racism may be considered an intergroup phenomenon in which it involves a social comparison between two or more racial groups. While colorism does, in fact, have its roots in the intergroup distinction between free Whites and enslaved Black Americans in the United States, it is much more of an intragroup phenomenon. That is, it reflects biases toward certain variations (e.g., lighter skin) within racial groups rather than between them. Of course, these biases might be held by either those outside or those inside the focal group. Hence, while connected in their origin, racism and colorism are not necessarily dependent on each other for their modern existence. Thus, a decrease or even annihilation of racism does not automatically preclude the existence of a colorism problem since colorism can exist independent of an intergroup context.

Scale Development

We utilized Simms’s (2007) guidelines on scale construction to develop and validate an In-Group Colorism Scale (ICS). The ICS was developed to assess the significance and meaning of skin color variations on the part of those inside a community of color. While this particular version was developed on Black Americans in the United States, it is our contention that this scale could be modified and used for virtually any people of color minority group.

The goal for the ICS was to ultimately create a scale that was (a) comprehensive of the dimensions of colorism discussed earlier but nevertheless (b) was short in length. The rationale for the latter is to make the scale economical to use in both research and applied contexts. In both contexts, the relationships between the ICS scale and other constructs are likely to be of interest over and above mere descriptive information on the ICS itself. Short scales better allow for the examination of such relationships. However, this objective was to be balanced with the first and primary objective of having a comprehensive and multidimensional measure.

Simms’ (2007) model is based on Loevinger’s (1957) seminal model of scale construction and it integrates the three mutually exclusive approaches to scale construction: rational-theoretical, empirical criterion keying, and factor-analytic/internal consistency. Thus, Simms’s model reflects an eclectic and comprehensive approach to scale development and validation.

As a first step, subject matter experts (SMEs) were enlisted to develop and then subsequently review/edit potential items for the ICS scale. Based on the literature reviewed earlier on colorism, the researchers a priori chose to measure five important facets of colorism: Self-Concept, Impression Formation, Attraction, Affiliation, and Upward Mobility. Attraction appears to include many of the same facts as affiliation. However, attraction does include aspects (i.e., beauty, romance) that are not necessarily components of mere affiliation (i.e., friendship). Thus, we chose to treat these as related but distinct dimensions of colorism. SMEs were initially asked to draft items for each of the five dimensions of colorism. The pool of items was then reviewed by a different panel of SMEs independently of the first panel. Some items were then further revised and/or dropped.

With regard to the size of the overall scale and its subdimensions, Comrey (1988) suggests that it is seldom necessary to use more than 20 items to obtain adequate reliability for a scale. Furthermore, homogeneous subdimension scales have consistently obtained acceptable reliabilities with four items. Thus, the target for the total number of items for the ICS scale was 20, with 4 items for each of the 5 subdimensions of the scale. To move toward that goal, our initial pool of items was trimmed down. However, following the principle of overinclusiveness, we trimmed the scale down to 50 items with 10 items per dimension. The 50-item version of the scale was then used in the psychometric evaluation step of the Structural Validity phase of Simms’ model (2007).

ICS Validation Studies

Once the initial set of items was developed, the next two phases of Simms’s model (2007; Structural Validity and External Validity) required distributing the items to samples of respondents from the target population(s). To facilitate these phases, we conducted a cross-validation study requiring two separate samples. The purpose of the first study was to focus on establishing a set of items for each of the subdimensions of the ICS. In this study, all initial 50 items were constructed into a scale whereby they were placed on a 7-point Likert-type scale format (strongly agree to strongly disagree). The purpose of the second study was to focus on confirming the structural validity and criterion validity of the ICS. Because we used a different sample of respondents, the second study provided us with the opportunity to confirm the proposed structure of the ICS. Moreover, the criterion validity test required an assessment of how the ICS scale relates to other measures that are said to lie within its nomological network (i.e., other measures that theoretically should be related or unrelated to varying degrees if the scale truly and accurately measures the latent construct of colorism).

The measures for both studies were administered using online surveys. The links for the surveys were distributed across a number of mediums. For both studies, an initial group of Black American respondents were recruited within a Midwestern university population. This group of respondents was then asked to forward the link to their network of friends. Also, the link was sent to a national network of Black American professionals who were also asked to forward the link to their network of family and friends. Finally, an online survey service called Amazon’s Mechanical Turk (Mturk) was utilized to further recruit Black American participants. Mturk has been demonstrated to yield highly representative samples, both geographically and socioeconomically (see Buhrmester, Kwang, & Gosling, 2011).

Study 1

Method and Results

Sample

Approximately 383 Black Americans completed the online survey containing all 50 items and demographic information that included sex, age, education, and income. The mean age of this sample was 41.59 years. Seventy-seven percent of the sample were female, while 23% were male. Both the education and socioeconomic status (SES) of our sample were skewed toward the higher end with a median income of $50,000 and median education of 4 or more years of college.

Exploratory Factor Analyses (EFA)

Consistent with Simms’s (2007) model, Exploratory Factor Analyses (EFA) were used as a method of item selection. We initially used EFA as an item reduction technique within factors and then as an initial attempt to explore the dimensionality of the scale. EFAs were conducted in order to reduce the initial set of 50 items down to 20 (4 for each factor). Separate EFAs were conducted within each of the proposed factors to select items that best represented the intended construct. An attempt was made to balance out homogeneity with comprehensiveness. That is, while our primary criterion for selection was the items factor loading on the primary factor, we also tried to avoid selecting items that were too redundant in content with other items. All EFAs were conducted using Principle Components Analyses (PCAs) with Varimax rotation.

Item Reduction

Self-Concept

These items were designed to assess the degree to which respondents based their self-concepts on their skin tone (e.g., “My skin tone is an important part of my self-concept”). The initial 10 items were entered into a PCA. The top four loading items that were nonredundant were then isolated and reentered into a separate PCA. This analysis yielded a single unidimensional factor that accounted for 71% of the variance, with all four items loading above .77. These four items combined yielded good reliability as indicated by a coefficient alpha of .87.

Impression Formation

These items were designed to assess the degree to which respondents form impressions of other Black Americans based on their skin tone (e.g., “You can tell a lot about a person by their skin tone”). The initial 10 items were entered into a PCA. The top four loading items that were nonredundant were then isolated and reentered into a separate PCA. This analysis yielded a single unidimensional factor that accounted for 56% of the variance, with all four items loading above .68. These four items, combined yielded acceptable reliability as indicated by a coefficient alpha of .73.

Upward Mobility

These items were designed to assess the degree to which respondents believe that the upward mobility of Black Americans depends on their skin tone (e.g., “Skin tone plays a big part in determining how far you can make it”). The initial 10 items were entered into a PCA. The top four loading items that were nonredundant were then isolated and reentered into a separate PCA. This analysis yielded a single unidimensional factor that accounted for 77% of the variance, with all four loadings above .78. These four items, combined yielded excellent reliability as indicated by a coefficient alpha of .90.

Attraction

These items were designed to assess the degree to which respondents find certain skin tones more romantically attractive than others (e.g., “I prefer light skin over dark complexion skin when choosing romantic interests”). The initial 10 items were entered into a PCA. The top four loading items that were nonredundant were then isolated and reentered into a separate PCA. This analysis yield a single unidimensional factor that accounted for 67% of the variance, with all four loadings above .76. The combination of the four items yielded good reliability as indicated by a coefficient alpha of .81.

Affiliation

These items were designed to assess the degree to which respondents prefer to have friendships and associations with certain skin tones (e.g., “I usually choose who I’m going to be friends with by their skin tone”). The initial 10 items were entered into a PCA. The top four loading items that were nonredundant were then isolated and reentered into a separate PCA. This analysis yielded a single unidimensional factor that accounted for 63% of the variance, with all four loadings above .71. These four items combined yielded good reliability as indicated by a coefficient alpha of .80.

Dimensionality

All 20 items were entered into a PCA. An examination of the Scree Plot suggested six components before the curve flattened out. However, only the first five factors provided eigenvalues above 1, accounting for 67% of the variance. Furthermore, a Parallel Analysis was conducted using Vivek, Singh, Mishra, & Donovan’s (2007) Parallel Analysis Engine using the recommend random correlation matrices (100) and percentile of eigenvalues (95; Vivek, Singh, Mishra, & Donovan, 2008). Comparisons between the generated eigenvalues and extracted eigenvalues revealed that the first four extracted components yielded eigenvalues larger than the generated eigenvalues. However, the fifth extracted component eigenvalue was just slightly below the generated eigenvalue (1.14 vs. 1.19). With all three criteria for selecting factors taken into consideration, five factors were retained to be further tested in the confirmatory phase of the scale development.

Items loaded on their respective subscales with the exception that two items belonging to the affiliation factor (i.e., “I’m usually uncomfortable being around people who are a certain skin tone” and “I usually choose who I’m going to be friends with by their skin tone”) split-loaded on both the affiliation factor and the impression formation factor.

Preliminary Discussion

Following Simms’s (2007) guidelines for scale development and validation, we were able to develop a multidimensional scale to measure the significance and meaning placed on skin tone with regard to one’s self-concept, how impressions are formed of others, beliefs about upward mobility, what is seen as attractive, and the people with whom one desires associations and/or friendships. The use of SMEs helped establish the scales’ substantive validity (aka content validity). Furthermore, the structural validity of the scale was established by a series EFAs in which these analyses demonstrated a unidimensional structure within each of the subscales.

Study 2

Confirmatory Factor Analysis (CFA)

It is important to conduct confirmatory tests of models in order to affirm the structures that are initially discovered using exploratory methods. Because our initial list of factors were derived a priori from the literature, we did not use exploratory methods to derive the dimensionality of the ICS, but rather as a method of item selection. Nevertheless, confirmatory analyses are required to test the fit of items to the proposed multidimensional scale. Thus, one purpose of this second study was to validate the structure of the ICS. Another purpose was to provide evidence of the criterion validity of the scale by demonstrating its association with measures of constructs believed to be associated with colorism.

We recruited a different set of Black participants using the same recruiting methods as in study one. Furthermore, because colorism has been rumored to correlate with SES, we attempted to stratify our participants along our measure of income to ensure representation at all SES levels. This approach would allow us to examine the structure and relevance of our colorism measure at all levels of SES.

In addition to completing the ICS measure, participants also completed measures of constructs that have been previously associated with colorism. Thus, we were able to examine the relations of the ICS measure with several variables previously posited as antecedents and/or consequences of colorism. These measures included scales to assess parental socialization, endorsement of positive and negative stereotypes, racial identity, skin tone, self-esteem, and several demographic variables (e.g., sex, age, education).

There were two main hypotheses for the current study. The first hypothesis involves confirming the hypothesized five-part dimensionality proposed in the ICS. This model will substantiate the proposed structure of colorism as consisting of five distinct dimensions (i.e., Self-Concept, Impression Formation, Attraction, Affiliation, and Upward Mobility) that are nonetheless united as indicators of the more global construct of colorism.

Hypothesis 1: CFA will demonstrate that a five-factor structure is a good fit and the best fit among alternative measurement models.

The second hypothesis concerns the relationship between Colorism and phenomena believed to serve as either antecedents or consequences of colorism. The extant literature on colorism focuses primarily on the association of these variables with either the skin tone of the participant or the participant’s preference for one shade (e.g., light skin) over another (e.g., dark skin). Our focus is more on the significance and meaning that is placed on skin tone overall and not on particular skin tone preferences. Thus, the previous literature on colorism is of limited value in generating hypotheses for this study. Nevertheless, we argue that skin tone preferences imply that skin tone holds significance and meaning for those that endorse them. Thus, we make the following hypotheses.

Hypothesis 2: Higher colorism scale ratings should be significantly associated with measures of skin tone parental socialization, skin tone stereotype endorsement, racial identity, participant skin tone, and self-esteem.

We intentionally left open the direction of these associations as arguments can be made in either direction (e.g., parental socialization may either heighten the significance of skin tones or diminish it). However, it might be argued that heightened skin tone significance is more likely to be associated with higher skin tone stereotype endorsement rather than lower.

Method

Participants

Approximately 355 Black Americans completed the online survey. The mean age of this sample was 37.17 years. Seventy-seven percent of the sample was female, while 23% were male. The sample median income was $50,000 and median education was “4 or more” years of college.

Procedure

Data for the current study was collected via the Qualtrics online surveying platform. On retrieving the link to the online survey participants viewed a screen presenting a recruitment statement. Participants were then prompted to agree to participate or if they chose not to participate they were instructed to logout of the survey and close the browser window. Following the recruitment statement, participants were presented with the ICS and instructed to respond openly to each item. After participants completed the measure of interest, they responded to the remaining battery of scales included in the survey (e.g., Parental Socialization, Racial Identity, and Skin Tone). The order of the measures included in the survey, with the exception of the ICS, was randomized. Once all measures were completed, the responses were saved and stored. Each participant was thanked for their participation and presented with an onscreen explanation of research.

Measures

Parental Socialization

To determine the impact of an individual’s upbringing on their perception of skin tone, the researchers developed a scale to measure parental socialization toward skin tones. This five item scale employs a 5-point Likert-type format with response options ranging from not at all (1) to extremely (5). A sample item from this scale is “As you were growing up, how much was SKIN TONE or topics related to it talked about by the people or in the settings listed below?” (e.g., your parents, at your place of worship). This scale had Cronbach’s alpha of .79 for the current study.

Stereotype Endorsement

Stereotype endorsement was measured using the Maddox Trait Scale (Maddox & Gray, 2002). The trait scale has been previously validated and found to be a psychometrically sound questionnaire measuring individual endorsement of stereotypes about minority groups (e.g., Black Americans). This 22-item scale asks participants to rate each stereotyped trait presented on a 10-point Likert-type format with response options ranging from more descriptive of dark skinned Black Americans (10) to more descriptive of light skinned Black Americans (1). A total scale score is computed for this scale with negatively worded items being reverse scored prior to data analysis. The following are sample items from this scale, “Athletic,” “Criminal,” and “Superior” (Maddox & Gray, 2002).

Racial Identity

Racial Identity was measured using the Multidimensional Inventory of Black Identity-Short form (MIBI-SF; Sellers et al., 1997). The MIBI is a measure comprising seven subscales that are representative of three dimensions: centrality, regard, and ideology. The seven subscales included are Racial Centrality, Private Regard, Public Regard, Nationalist, Assimilationist, Humanist, and Oppressed Minority. The 27-item short form of the MIBI was used in order to maintain a reasonable length for the study survey as a whole. Each of the subscales used a 7-point Likert-type format with response options ranging from strongly disagree (1) to strongly agree (7). For this measure, participants were instructed to indicate the degree to which they agreed with each of the statements presented within the subscales of the MIBI. A sample item from the centrality subscale is, “In general, being Black is an important part of my self-image” (Sellers et al., 1997; Sellers et al., 1998).

Skin Tone

Participant skin tone was measured via the Measurement of Skin Tone from the National Survey of Black Americans (1979-1980). This scale asks Black individuals to estimate their own skin tone based on a scale of 1 (very dark brown) to 5 (very light brown/very light skinned).

Self-Esteem

Self-Esteem was measured via the Rosenberg Self-Esteem Scale (RosSES). The RosSES is a previously validated and psychometrically sound questionnaire measuring self-esteem (Rosenberg, 1965). This 10-item scale uses a 4-point Likert-type format with response options ranging from strongly agree (1) to strongly disagree (4). RosSES is an affect and personality trait scale). This scale had Cronbach’s alpha of .91 for the current study. A total scale score is computed for RosSES with negatively worded items being reverse scored prior to data analysis. The following is a sample item from this scale, “I take a positive view of myself” (Rosenberg, 1965).

Demographics

Participant information regarding their sex, ethnicity, age, family income, education level, and percentage of Black American friends and associates with the same skin tone as the participant were collected for descriptive purposes.

Results

Confirmatory Factory Analyses

CFAs using maximum likelihood estimation methods were conducted to determine if the five-dimension structure developed in the first study could be replicated using a confirmatory approach. We compared our proposed Five-Factor model against a One-Factor model. The One-Factor model assumes all items to be indicative of a single One-Dimensional model. The Five-Factor model was hypothesized as the best fit for the data.

We tested two different versions of the Five-Factor model. We tested a Higher Order factors model against a Correlated Factors model. The Higher Order model allows for the examination of the degree to which the five factors reflect a higher order factor (e.g., an Overall Colorism factor), whereas a Correlated Factors model would treat the five factors of colorism as correlated, but distinct factors.

As seen in Table 1, the model fit indexes suggest that both of the Five-Factor models provided a better fit to the data than the One-Factor model. Whereas the One-Factor model failed to meet the standard requirements of good model fit, the proposed Five-Factor models proved to be significantly better fits to the data. The scaled chi square difference test revealed that there was no significant difference (chi square difference = 10.5, degrees of freedom [df] difference = 5, p > .05) between the Correlated Factors model (comparative fit index [CFI] = .989) and the Higher Order model (CFI = .986). However, both models displayed significant reductions in chi square relative to the One-Factor model. Thus, the data equally fit a Higher Order factors model as well as a Correlated Factors model. This suggests that whereas the multidimensional nature of the model was confirmed, the data were also adequately represented by a hierarchical factorial structure. The five factors were explained by a higher order second factor of general colorism—the overall degree to which skin tone is seen as significant and meaningful.

Table 1.

Confirmatory Factor Analyses of the In-Group Colorism Scale.

			Fit indexes
Models	χ²	df	χ²/df ratio	CFI	RMSEA
Null	26012	190	136.91
One-Factor	3399	170	19.99	.875	.235
Five Correlated Factors	442	160	2.766	.989	.070
Higher Order Factor	537	165	3.255	.986	.081

Note: df = degrees of freedom; CFI = comparative fit index; RMSEA = root mean square error of approximation. All χ²s p < .001.

Thus, the five factors of colorism were reconfirmed on a different sample of participants. Notably, all items loaded highly and significantly onto its respective factor. The standardized loadings for the final CFA model are presented in Table 2. As with the first study, each of the subscales proved to be reliable indicators of their respective constructs (see Table 2 for subscale reliabilities). The subscales were scaled to consistently reflect the degree to which the participants placed significance and meaning on each of these dimensions. Thus, higher Self-Concept subscale scores indicated higher significance and meaning placed on skin tone as a determinant of one’s self-concept (see Table 2 for content). Higher Impression Formation scores indicated higher significance and meaning placed on skin tone as determinant of how one forms impressions of others. Higher Affiliation scores indicated higher significance and meaning placed on skin tone as a determinant of affiliations and friendships. Higher Attraction scores indicated higher significance and meaning placed on skin tone as a determinant of attraction and beauty. Finally, higher Upward Mobility scores indicated higher beliefs in the notion that skin tone is a significant and meaningful determinant of a person’s potential for upward mobility.

Table 2.

Confirmatory Factor Analysis for In-Group Colorism Scale (ICS) Subscales.

Factors and items	Loadings
Self-Concept	(.85)
• My skin tone is an important part of my self-concept.	.75
• My skin tone is an important component of who I am.	.84
• My skin tone affects my self-esteem.	.80
• My skin tone is a big part of my identity.	.85
Impression Formation	(.77)
• You can tell a lot about a person by their skin tone.	.76
• Black Americans with lighter skin tone tend to be more pleasant people to deal with.	.79
• Dark-skinned people are more difficult to work with.	.80
• There are real differences between light skin and dark skinned people.	.76
Affiliation	(.82)
• I’m usually uncomfortable being around people who are a certain skin tone.	.81
• Most of my friends tend to be the same skin tone.	.84
• I usually choose who I’m going to be friends with by their skin tone.	.86
• The majority of my current friends are the same skin tone as me.	.77
Attractiveness	(.81)
• I’m primarily attracted to people of a certain skin tone.	.70
• I prefer light skin over dark complexion skin when choosing romantic interests.	.78
• I prefer a romantic partner who has the same skin tone as me.	.76
• Lighter skin tone makes others more attractive.	.86
Upward Mobility	(.91)
• Even if you work really hard, your skin tone matters most.	.80
• Skin tone plays a big part in determining how far you can make it.	.91
• Skin tone effects how much money you can make.	.90
• If you want to get ahead, you have to be the right skin tone.	.91

Note: Alphas for each subscale are in parentheses.

Consistent with the fact that the data adequately fit a higher order model, we also found that all of items combined into a highly reliable (α = .91) total score. Thus, while it might be beneficial to take advantage of the multidimensional nature of the ICS and primarily utilize the subscales, we also examined the utility of using a total score. Higher total scores reflect higher significance and meaning assigned to skin tone variations overall.

Potential Antecedents and Consequences of Colorism

Beyond establishing structural validity, the criterion validity of a measure should be established by demonstrating its associations with measures of constructs previously established in the literature as antecedents and/or consequences of the focal construct. As our concern was with the criterion validity of the ICS more so than establishing directionality of the related constructs as either antecedents or consequences, we chose correlational analyses. Table 3 contains a matrix of correlations between the colorism overall and subscale scores with potential antecedents/consequences. Consistent with our second hypothesis, several significant and meaningful correlations were found:

Table 3.

Correlations Between Overall and Subscales of Colorism and Other Variables.

	Parental Socialization	Skin Tone	Positive Traits	Negative Traits	Racial Centrality	Private Regard	Public Regard	Racial Nationalism	Racial Assimilationism	Self-Esteem
Colorism
Overall	.19	−.13*	−.03	.33**	−.03	−.28**	.08	.23**	−.27**	−.37**
Self-Concept	.20	−.23**	−.03	.17**	.17	.01	.02	.08	−.12*	−.18**
Impression Formation	.07	−.02	−.02	.32**	−.17	−.35**	.19**	.21**	−.26**	−.30**
Affiliation	.07	−.07	.02	.29**	−.13*	−.27**	.14**	.22**	−.24**	−.27**
Attraction	.12	−.02	−.05	.28**	−.16**	−.30**	.09	.21**	−.16**	−.34**
Upward Mobility	.15	−.05	−.09	.24**	.05	−.08	−.14**	.09	−.13*	−.23**
N	355

Note: *p <.05, ** p<.01

Colorism and Socialization

A small but significant positive correlation was found between Overall Colorism endorsement ratings and racial parental socialization. This effect was best accounted for by the Self-Concept, Attraction, and Upward Mobility subscales, as these were all significantly positively correlated with socialization.

Colorism and Skin Tone

A small but significant negative correlation was found between Overall Colorism endorsement ratings and the Skin Tone hue reported by the participants. This effect was best accounted for by the Self-Concept subscale. Higher Self-Concept subscale ratings were associated with darker skin tones.

Colorism and Stereotypes

Interestingly, there were no significant correlations between the colorism scales and the positive trait assignment. However, the Overall Colorism scale and all of the subscales were positively correlated with the assignment of negative traits to darker skinned people.

Colorism and Racial Identity

There were several significant correlations between the MIBI Racial Identity subscales and the colorism subscales. Racial Centrality ratings were positively correlated with the Self-Concept subscale, but negatively correlated with the Impression Formation, Affiliation, and Attraction subscales. Similarly, Racial Private Regard ratings were negatively correlated with the Impression Formation, Affiliation, and Attraction subscales. Racial Public Regard was positively correlated with Impression Formation and Affiliation subscales, but negatively correlated with the Upward Mobility subscales. The Racial Nationalism subscale was positively correlated with the Impression Formation, Affiliation, and Attraction subscales. Finally, the Racial Assimilation subscale was significantly negatively correlated with all of the subscales.

Colorism and Self-Esteem

There was a significant negative correlation between the Overall Colorism scale and Self-Esteem such that higher endorsement ratings of colorism were associated with lower self-esteem scores. Each of the Colorism subscales were also significantly negatively correlated with Self-Esteem. However, there were sex differences. Whereas women reported a negative correlation between the Self-Concept subscale and Self-Esteem scale (r = −.23, p < .001, n = 264), the correlation was not significant for males (r = −.028, p < .791, n = 90).

Socioeconomic Status and Colorism

With one small exception, neither of our two indicators (i.e., Income and Education) of SES were significantly correlated with either the Overall Colorism scale or the subscales. The one exception is that there was a small significant positive correlation between Education and the Impression Formation subscale, such that more highly educated individuals placed slightly more importance on skin tone when forming impressions of other Black Americans, r(355) = .11, p = .045.

Beyond direct effects of SES on colorism variables is the possibility that the impact of colorism on other key outcomes might be conditioned on and/or be moderated by SES. We explored this possibility by running several analyses whereby Overall Colorism was regressed onto the variables listed above with Income serving as the conditional variable and moderator. We utilized Hayes’s (2013) PROCESS Bootstrapping method of conditional effects and moderation testing. This method has become the gold standard for testing conditional effects and moderated regression analyses. There were four notable results. First of all, a conditional effects analysis revealed that the effect of higher Overall Colorism on Negative Trait Assignment to darker skin tones was significantly higher among those reporting higher incomes¹ (b = .71) relative to those reporting middle incomes (b = .47) and lower incomes (b = .23). All effects had p < .015. This finding was clearly evident that the effects of Overall Colorism ratings were linearly dependent on reported Income ratings, as the interaction effect was significant, b = .21, t(355) = 2.94, p = .004. Thus, the association between higher colorism importance ratings and higher negative trait assignment to darker skin hues was significantly stronger among those reporting higher incomes.

A conditional effects analysis revealed that the effect of higher Overall Colorism ratings on higher Public Regard ratings was significant among those reporting lower Income, b = .29, t(355) = 2.15, p = .03, but not among those in the middle and higher income ranges. The interaction effect was marginally significant, b = −.16, t(355) = −1.79, p = .07. Thus, the association between higher colorism importance ratings and higher Public Regard ratings was somewhat stronger among those reporting lower incomes.

Finally, a conditional effects analysis revealed that the effect of higher Overall Colorism on higher Racial Nationalism scores was significant among those reporting higher incomes (b = .30) and those reporting middle incomes (b = .18). Both effects had p < .007. However, the effect was not significant among those reporting lower incomes, b = .06, t(3255) = 1.13, p = .26. The interaction effect was significant, b = .10, t(355) = 1.98, p = .049. Thus, the association between higher colorism importance ratings and higher Racial Nationalism scores was significantly stronger among those reporting higher and middle incomes relative to those reporting lower incomes.

Discussion

The first task of the current research was to create a scale that would accurately capture the construct of colorism on a psychosocial level. Previous research has conceptualized colorism primarily as a macro-societal variable and lacked the ability to measure the extent to which individuals endorsed the construct on a personal and interpersonal level. A cross-validation study using two distinct samples suggested a strong fit for a five-dimension structure united by a higher order factor. This reconfirmation along with proper factor loadings suggests that the scale adequately captures the subscales Self-Concept, Impression Formation, Attraction, Affiliation, and Upward Mobility. Moreover, the confirmatory analysis provided evidence that these five factors are also indicators of a global colorism construct (i.e., the overall perceived significance and meaningfulness of skin tone).

The second task of this research was to examine the associations between colorism and other measures that may be considered as antecedents and/or consequences of colorism. All of the variables measured were correlated to some extent with either overall colorism or with a specific subscale. For example, there was a significant correlation between colorism and racial parental socialization. Not surprisingly, this finding suggests that the way in which the participants were taught to think about skin tones was associated with the extent to which they endorsed colorism. Examining the specific subscales provides greater understanding of the nature of these associations. The more participants were given messages about skin tone being important, the more they endorsed the idea of skin tone being an important factor in their own self-concept, their attraction to other Black Americans, and their beliefs about the upward mobility of Black Americans. This finding is consistent with previous discussions on colorism that has consistently highlighted the importance of family and other socializing agents as sources of learning and reinforcing ideas about skin color (Wilder & Cain, 2011).

With regard to the issue of the participants’ skin tones, participants in this sample who reported having darker skin reported that skin tone played a more significant role in their self-concept relative to those with lighter skin. Identity is in part a function of what is reflected back to an individual about whom they are (Tatum, 2003). Within social contexts in which skin tone might render someone unique or deviant, it seems reasonable that people would incorporate it as an important aspect of self. Thus, it is not surprising that those with darker skin would feel that their skin tone is more germane to their identity. Yet it is important to note that self-esteem was not lower in those participants with darker skin tone, rather self-esteem was lower in those participants who endorsed attitudes of colorism. Hence, it is likely that colorism might be an important mediator to the presumed relationship between darker skin tones and lower self-esteem.

Overall, greater endorsement of colorism was correlated with lower self-esteem for men and women equally. However, there were sex differences on the self-concept subscale. Whereas there was a negative correlation between heightened importance of skin tone for the participants’ self-concept and self-esteem for women, there was no significant correlation between these variables for men. This finding somewhat fits with previous research, which found that increased colorism was associated with decreased self-esteem for Black American women but not for Black American men (Thompson & Keith, 2001).

We found significant relationships between colorism and skin tone stereotypes. Interestingly, there were no significant relationships between colorism and skin tone stereotypes for positive traits. Instead, there were consistent correlations between colorism and skin tone stereotypes for negative traits. In every case, the more significance that participants placed on colorism the more likely they were to assign negative traits to darker skinned Black Americans. While the ICS was not designed to detect preferences for either light or dark skin as much as to detect the importance placed on colorism in general, it does appear that higher colorism scores imply less bias toward lighter skinned Black Americans. This finding is consistent with previous research on the relationship between negative stereotypes and skin tone (Maddox & Gray, 2002).

With regard to racial identity, the more central race was to the participants’ self-identity, the more they reported that skin tone was important for their sense of self-concept but less important for forming impressions of, and their attraction and affiliation with other Black Americans. Furthermore, high personal regard for Black Americans was correlated with skin tone being less important in forming impressions, affiliation, and attraction. Similar to private regard and centrality, participants who endorsed nationalism racial ideology also reported lower scores on skin tone being important in forming impressions, affiliations, and attraction. Thus, substantial relations were found between racial identity and colorism.

SES appears to be an important variable in understanding how colorism operates among Black Americans. Those participants with greater levels of education placed more importance on skin tone when forming impressions. Moreover, higher colorism scores were substantially more indicative of negative trait assignment to darker skinned Black Americans among those in the highest income levels. These dynamics could be seen as reflective of a historical preference being given to those with lighter skin tone in regard to educational and occupational attainment (Hughes & Hertel, 1990) as well as access to social clubs and organizations (Frazier, 1957). However, the intersection of colorism and racial identity tells a more nuanced story. Colorism importance seemed to be attached to outsider views of Black Americans for lower income individuals but not for middle or higher income Black Americans. That is, lower income individuals were more likely to associate the importance they placed on skin tone with their perceptions of non-Black Americans attitudes toward their racial group. More positive perceptions were associated with higher colorism ratings. Furthermore, for individuals who were middle or high income, there was a stronger positive association between nationalism racial ideology and the importance of skin tone compared with individuals from the lower socioeconomic level. On the surface, these findings seem somewhat perplexing. For example, we would have expected a negative connection between nationalism racial ideology and skin tone importance among those with higher incomes. Nevertheless, these findings point to a more complex interaction between SES, racial identity, and colorism. Future research should make attempts to replicate and further parse these findings.

Limitations

A potential limitation to our studies has to do with the degree to which our samples can be said to represent the whole Black U.S. population. First, it was the desire of the researchers to develop a measure of colorism that could be used across communities of color and not necessarily just within the Black U.S. population. Thus, the emphasis of this study was focused more on establishing the reliability and validity of the ICS than on representing the focal population. Larger sample sizes allow for greater specification and thus generalization of sample parameters (i.e., means and standard deviations). In the case of this study, generalizing the mean colorism scores from our sample to Black Americans in general is inadvisable. Instead, the value added of this current research is not so much in the absolute parameter values of the variables, but rather on the hypothesized covariances between them. A much broader attempt to recruit a geographically stratified Black American sample would be needed to establish true population parameters.

There are other potential limitations that are relevant to the usage of self-report survey methods. Common method variance (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003), which occurs when the same method is used to measure different constructs can be shown to induce correlations between constructs. While common method variance has been shown to be quite pervasive in self-report methods, its true impact on the validity of research has been debated. A more recent review of the various methods of assessing common method variance has suggested that while common variance is pervasive, its impact is relatively small, and tends to vary depending on the methods used to assess it (see Meade, Watson, & Kroustalis, 2007). Furthermore, while common method variance might be problematic for some of our analyses (i.e., correlations between colorism and racial identity), it would be a null issue for our main hypotheses (CFAs on the factor structure of the colorism scale) because these analyses involve competing correlations (for which methods bias is held constant).

Another related problem with self-report methods has to do with temporal restrictiveness, which means that the constructs are only measured at one point in time, when certain constructs (e.g., attitudes) have been found to fluctuate considerably over time (Sonnentag, 2003). Thus, it is not necessarily clear whether the importance of colorism fluctuates over short or long periods of time. Of course, the only way to assess this would be conduct repeated measures over time.

Implications and Conclusion

The ICS, as a valid measure of colorism, can add significant value for future research on colorism. For research focused on the degree to which individuals might vary in the importance that they place on skin tone, the ICS should prove quite helpful. While we did find high interitem reliability and high validity using the overall score, we believe that it would be advantageous to take advantage of the multidimensional nature of the ICS and utilize the subscales. Furthermore, for research interested in examining potential covariates of colorism, the ICS scale has, herein, demonstrated considerable overlap with constructs that have been previously identified as important domains among Black Americans. Thus, the ICS can facilitate future research that might attempt to examine potential antecedents and/or consequences of the significance and meaning that communities of color might place on skin tone variations.

While the ICS was initially developed and validated using Black Americans, we suggest that the scale can be tailored for use with any community of color. As recent publications (see Norwood, 2014) and recent conferences (2015, Global Perspectives on Colorism conference in St. Louis, Missouri) have attested, colorism is a universal feature of human existence. Colorism has been reported to be just as significant in places like Israel, China, and Columbia as it is in the United States (Norwood, 2014). Only one item on the ICS makes reference to “Blacks” This item could easily be revised to refer to a different racial or ethnic group. The remainder of the 19 items on the scale are generic with respect to any particular racial or ethnic group.

In sum, the current study suggests that colorism can be conceptualized and measured as an individual psychosocial level variable. Furthermore, this study suggests that there is considerable merit in thinking of colorism as a multidimensional construct in which different dimensions may hold different implications for other constructs of relevance to communities of color, Black Americans in particular. It is clear that the issue of colorism, while one of the oldest constructs of Black American folklore and history, is still very globally prevalent and psychologically relevant today.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Buhrmester

Kwang

Gosling

S. D.

(2011). Amazon’s Mechanical Turk: A new source of inexpensive, yet high-quality, data? Perspectives on Psychological Science, 6, 3-5.

Burke

(2008). Colorism. In Darity Jr.

(Ed.), International encyclopedia of the social sciences (Vol. 2, pp. 17-18). Detroit, MI: Thomson Gale.

Burton

L. M.

Bonilla-Silva

Ray

Buckelew

Hordge-Freeman

(2010). Critical race theories, colorism, and the decade’s research on families of color. Journal of Marriage and Family, 72, 440-459.

Coard

S. I.

Breland

A. M.

Raskin

(2001). Perceptions of and preferences for skin color, Black racial identity, and self-esteem among African Americans. Journal of Applied Social Psychology, 31, 2256-2274.

Comrey

A. L.

(1988). Factor-analytic methods of scale development in personality and clinical psychology. Journal of Consulting and Clinical Psychology, 56, 754-761.

Frazier

F. E.

(1957). Black bourgeoisie: The rise of the new middle class. New York, NY: Free Press.

Goldsmith

Hamilton

Darity

W. A.

(2007). From dark to light: Skin color and wages among Black Americans. Journal of Human Resources, 4, 701-738.

Harvey

R. D.

Labeach

Pridgen

Gocial

T. M.

(2005). The intragroup stigmatization of skin tone among Black Americans. Journal of Black Psychology, 31, 237-253.

Hayes

A. F.

(2013). Introduction to mediation, moderation, and conditional process analysis. New York: The Guilford Press.

10.

Hill

(2002). Skin color and the perceptions of attractiveness among Black Americans: Does gender make a difference? Social Psychology Quarterly, 65, 77-91.

11.

Hughes

Hertel

(1990). The significance of color remains: A study of life chances, mate selection, and ethnic consciousness among Black Americans. Social Forces, 68, 1105-1120.

12.

Jordan

Hernandez-Reif

(2009). Reexamination of young children’s racial attitudes and skin tone preferences. Journal of Black Psychology, 35, 388-403.

13.

Keith

V. M.

Herring

(1991). Skin tone and stratification in the Black community. American Journal of Sociology, 97, 760-778.

14.

Loevinger

(1957). Objective tests as instruments of psychological theory. Psychological Reports, 3, 635–694.

15.

Maddox

K. B.

Chase

S. G.

(2004). Manipulating subcategory salience: Exploring the link between skin tone and social perception of Black Americans. European Journal of Social Psychology, 34, 533-546.

16.

Maddox

K. B.

Gray

S. A.

(2002). Cognitive representations of Black Americans: Reexploring the role of skin tone. Personality and Social Psychology Bulletin, 27, 250-259.

17.

Meade

A. W.

Watson

A. M.

Kroustalis

C. M.

(2007, April). Assessing common methods bias in organizational research. Paper presented at the 22nd Annual Meeting of the Society for Industrial and Organizational Psychology, New York.

18.

Norwood

K. J.

(2014). Color matters: Skin tone bias and the myth of a postracial society. New York: Routledge.

19.

Vivek

P. H.

Singh

S. N.

Mishra

Donavan

D. T.

(2007). Parallel analysis engine to aid determining number of factors to retain [Computer software]. Retrieved from http://smishra.faculty.ku.edu/parallelengine.htm

20.

Vivek

P. H.

Singh

S. N.

Mishra

Donovan

D. T.

(2008). Efficient theory development and factor retention criteria: A case for abandoning the “eigenvalue greater than one” criterion. Journal of Business Research, 61, 162-170.

21.

Podsakoff

P. M.

MacKenzie

S. B.

Lee

J. Y.

Podsakoff

N. P.

(2003). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88, 879-903.

22.

Robinson

T. L.

Ward

J. V.

(1995). African American adolescents and skin color. Journal of Black Psychology, 21, 256-274.

23.

Rosenberg

(1965). Society and the adolescent self-image. Princeton, NJ: Princeton University Press.

24.

Sellers

R. M.

Rowley

S. A.

Chavous

T. M.

Shelton

J. N.

Smith

M. A.

(1997). Multidimensional inventory of Black identity: A preliminary investigation of reliability and construct validity. Journal of Personality and Social Psychology, 73, 805-815.

25.

Sellers

R. M.

Smith

M. A.

Shelton

J. N.

Rowley

S. A.

Chavous

T. M.

(1998). Multidimensional model of racial identity: A reconceptualization of African American racial identity. Personality and Social Psychology Review, 2, 18-39.

26.

Simms

L. J.

(2007). Classical and modern methods of psychological scale construction. Social and Personality Psychology Compass, 1, 1-20.

27.

Sonnentag

(2003). Recovery, work engagement, and proactive behavior: A new look at the interface between nonwork and work. Journal of Applied Psychology, 88, 518-528.

28.

Tatum

B.D.

(2003). Why are all the Black kids sitting together in the cafeteria? New York: Basic Books.

29.

Thompson

M. S.

Keith

V. M.

(2001). The blacker the berry: Gender, skin-tone, self-esteem and self-efficacy. Gender & Society, 15, 336-357.

30.

Wilder

(2010). Revisiting “color names and color notions”: A contemporary examination of the language and attitudes of skin color among young Black women. Journal of Black Studies, 41, 184-206.

31.

Wilder

Cain

(2011). Teaching and learning color consciousness in Black families: Exploring family processes and women’s experiences with colorism. Journal of Family Issues, 32, 577-604.