Rodriguez Article

Embed Size (px)

Citation preview

  • ! 1

    The Coexistence of Colorist and Categorical Inequality in Elite Institutions Natassia Rodriguez, Stanford University

    Abstract: Selective colleges and universities often tout their commitment to racial equality in policy and practice. An institutional push for racial egalitarianism coupled with heightened popular discourse around equality of opportunity fosters an environment that condemns racism in its more overt forms. Yet, the official discourse of egalitarianism does not necessarily render race irrelevant. This papers objective is to assess the form in which race exerts its effects in overtly egalitarian, elite settings and whether that type varies across racial groups. In particular, a distinction is made between (a) a colorist system in which the lightness or darkness of ones skin determines outcomes, and (b) a categorical system in which broad racial categories are the main determinants. By applying multilevel models to data from the National Longitudinal Study of Freshman, I examine the relationship between color, race, and graduation rates for Whites, Blacks, Asians, and Hispanics in selective institutions of higher education. Findings reveal that Blacks in elite institutions experience categorical inequality, whereas Asian, Hispanic, and even White males experience colorist inequality. By contrast, Asian, Hispanic, and White females are protected from colorist inequality, a result that may be a perverse outcome of gender-specific beauty norms. Keywords: Higher education, color, race, stratification Acknowledgements The author would like to thank David Grusky and Aliya Saperstein; workshop participants in the Stanford Sociology Department and School of Education; and Sam Lucas, Karolyn Tyson, and Michael Gaddis for their thoughtful suggestions and comments. This work was supported by the Institute for Education Sciences through grant award R305B090016. Correspondence Natassia Rodriguez, Department of Sociology, Stanford University 450 Serra Mall Sociology Dept, Building 120 Stanford, CA 94204 Email: [email protected] Phone:!215.828.9385!

  • ! 2

    Colleges and universities often tout their egalitarian ideals surrounding race. Indeed, most

    universities take highly visible steps toward reducing racism through diversity-based admissions

    policies, multicultural on-campus programming, and related support services (Pewewardy and

    Frey 2002; Anderson, Daugherty, and Corrigan 2005; Chace 2007). This commitment to visible

    egalitarianism is particularly true of selective institutions colleges and universities with large,

    highly qualified applicant pools as well as budgets to support diversity recruitment efforts and

    focused programming (Aries 2008). An institutional push for racial egalitarianism coupled with

    heightened popular discourse around equality of opportunity fosters an environment where fair

    treatment across racial lines is expected and the most overt forms of discrimination are

    delegitimized. There is indeed some evidence that the most educated people are the least likely to

    act on racial prejudice or hold racist views (Schuman, Steeh, Bobo 1997). But of course it would

    be incorrect to suggest that racial inequality has simply disappeared in elite institutions response to

    the practice of overt egalitarianism. As is well known, racial differences in graduation rates persist,

    and minority students continue to regard college campuses as less supportive settings than non-

    minority students (Hurtado and Carter 1997; Cabrera et al 1999; Kao and Thompson 2003). The

    puzzle, then, is how racial differences continue to be meaningful in an environment where unequal

    treatment is, at least to some extent, deemed problematic and even condemned.

    The answer to this puzzle is that racial inequality sometimes takes on a more covert

    colorist form in elite settings. Instead of overtly discriminating based on membership to broad

    racial categories, discrimination is instead based on skin color, a subtler type of discrimination

    than categorical discrimination. Because skin color is a prominent characteristic with deep

    cognitive connotations, what Omi and Winant term color-consciousness (1994, 70), can readily

    arise. Colorism is differential treatment is based on negative racial stereotypes and status

    expectations triggered by this skin tone (Jones 2009; Banks 2009; Banton 2012; Hall 2013). The

  • ! 3

    purpose of this paper is to show the reach of colorism: Does a colorist regime persist even in

    highly selective institutions that commit to visible egalitarianism? Does it obtain equally for all

    racial and ethnic groups? Do women and men experience it equally?

    Some evidence of skin color stratification in education does exist: Light-skinned Hispanics,

    Blacks, and Whites generally complete more years of education than their same-race counterparts

    with dark skin (Keith and Herring 1991; Hunter et al. 2001; Herring et al 2003; Frank et al. 2010,

    Branigan et al 2013). However, studies of color-based inequality pay little attention to the types of

    contexts in which color becomes salient. The result is that we simply do not know about the reach

    of colorism into highly selective institutions - settings that are arguably the leading edge of

    changing cultural commitments in higher education and where the commitment to overt

    egalitarianism is especially prominent (Chace 2007).

    I proceed by examining the extent to which apparent categorical stratification can be

    explained by an underlying system of colorist stratification in selective colleges universities. I first

    review research on the relevance of colorism for different racial groups, and I then develop a four-

    category typology of how color, race, and educational attainment might plausibly interact in elite

    colleges and universities. I next use data from the National Longitudinal Study of Freshman to

    assess whether this typology well characterizes educational patterns for males and females of four

    racial groups- Blacks, Asians, Whites, and Hispanics. The findings present a challenge to scholars

    of race and education to take a closer look at how race continues to be relevant in egalitarian

    contexts.

    THE RELEVANCE OF SKIN COLOR IN 21ST CENTURY AMERICA

    It is hardly novel to question whether broad racial categories suffice for understanding

    contemporary racial inequalities (Hall 2008; Bonilla-Silva 2004; Banton 2012; Hall 2013). It is

  • ! 4

    well appreciated that our fixation on broad categories in the sociology of race comes at the cost of

    a richer understanding of present-day inequality. Most importantly, a focus on categories

    constructed around the color line ignores individual-level divides within racial groups, divides that

    blur social outcomes within and between groups. This is obviously not to claim that traditional

    racial categories are no longer relevant (Hochschild 2012). Attention to splitting along the color

    spectrum is simply a way of understanding how race operates in a contemporary society where the

    lines between categorical groups are becoming less clear (Hunter 2007; Alba 2009).

    When taboos on categorical racial prejudice are dominant, as is the case in elite educational

    institutions, it is plausible that discrimination will instead take a colorist form. Indeed, many

    scholars argue that color is a salient racial marker and, at times, a stand in for categorical

    distinctions. In the United States, its been shown that skin color is more salient than Eurocentric

    features in inter-person interactions, an important demonstration of the significance of

    pigmentation (Phinney 1996; Stepanova and Strube 2011). The globalization of light skin as a

    symbol of status and beauty across cultural and political lines has been well documented (Glenn

    2009; Hall 2013). Likewise, Bailey and Saperstein (2013) find support for the simultaneous

    existence of a racial and a color hierarchy in the United States using novel General Social Survey

    data. Others suggest that the rising importance of color may result in a dual-status hierarchy,

    wherein individuals with both the lightest skin and most European lineage are situated in the top

    status positions, while those with both the darkest skin and African American lineage are in the

    bottom positions (Bonilla-Silva 2004; Roth 2012).

    In elite colleges and universities, the highly visible and overt commitment to egalitarian

    ideals pressure individuals to treat all racial groups fairly. However, while individuals are on guard

    against categorical inequality, covert racial markers may still operate as unconscious sources of

    bias, thus creating colorist inequality. This brings us to our key research question: Is color a

  • ! 5

    significant stratifying factor in elite colleges and universities? Has it fully replaced categorical

    inequality as a form of racial stratification?

    There are three ways in which color and race could be expected to come together (see

    Table 1). The first possibility, one that may be understood as the conventional understanding of

    race, implies that simple categorical inequality remains dominant even in elite colleges and

    universities. Under this ideal type, racial group differences in graduation rates take a wholly

    categorical form and the residual within-group inequalities are minor. I have argued here that, at

    least within some elite and selective institutions, categoric inequality is exposed to quite visible

    attack and, as a result, some headway in reducing it has been made.

    [Table 1 About Here]

    But this of course does not mean that racial inequality disappears. Instead, a more probable

    outcome, even within highly egalitarian settings, is that colorist inequality remains strong. That is,

    racial group membership may not be significantly associated with educational success, net of skin

    color, but color does hold a significant association with success. As a third possibility, colorist and

    categorical inequality may coexist, producing whats labeled dual inequality in Table 1. Under a

    dual system, educational outcomes are affected by race and color simultaneously, meaning that

    within- and between-group inequalities are both in play. The analysis undertaken in this paper

    adjudicates between these possible scenarios in elite institutions of higher education.

    SKIN COLOR ACROSS RACIAL GROUPS

    The representation of Table 1 is overly simple because the extent to which skin color is

    relevant is likely to vary across racial groups in the United States. The lines between Blacks and

    Whites, for example, are drawn arguably much deeper than those between these groups and

    Latinos or Asians because of their extensive history (Alba 2009). In this section I briefly describe

  • ! 6

    how racial markers act as bases for discrimination separately from racial group membership for

    Latinos, Asians, Blacks, and Whites in the United States.

    Because the Latino category is a complicated construct, one based on an amalgam of

    ancestries in Spanish-speaking countries,1 Latinos must negotiate the U.S. color line and define

    their racial identity on a regular basis, regardless of the adequacy of such categories as self-

    descriptors (OBrien 2008; Frank et al. 2010). For example, Latinos are split in identifying racially

    as black, white, and other in their answers to survey questions (Rodriguez 1992). This raises

    the possibility that Latinos with different skin tones will end up with disparate outcomes. Indeed, it

    is known that lighter skinned Latinos are less likely to be unemployed or impoverished and

    generally earn higher incomes than darker skinned Latinos (Hunter 2007). In elite institutions, it is

    plausible the this type of color divide is quite important for Latinos, given that the elite egalitarian

    efforts have not typically been directed against within-group bias.

    On the other hand, this type of colorism may be less important for Asians. In practice,

    individuals from other racial backgrounds often categorize Asians as outside the color continuum,

    as foreigners with skin tones and features that do not fall along the black-white color spectrum

    (Takaki 1998; Rondilla and Spickard 2007; Alba 2009). As a result, we might expect color to be a

    larger diving factor among Latinos than Asians, if all Asians are treated similarly regardless of

    color. But skin tone may be more relevant in interactions among Asians themselves. That is,

    Asians prefer lighter skin tones for their ideal marriage partner as well as themselves, at times

    using artificial means to obtain this tint (Hunter 2003; Herring et al. 2003; Rondilla and Spickard

    2007; Rondilla 2009). These color preferences may extend to educational settings, suggesting a

    colorist system similar to that of Latinos, though it is unclear whether such colorism is as

    important as group status itself to other racial groups.

  • ! 7

    The role of colorism for blacks is yet more complicated. Russell and other scholars (1992)

    describe a color complex that creates divides within Black communities based on skin tone.

    Opinions about who is an authentic Black person lead darker-skinned Blacks to sometimes reject

    their lighter-skinned counterparts as group members (also see Herring et al 2003). These skin tone

    distinctions, however, may only arise in particular contexts. Hochschild and Weaver (2007) find

    that the strength of black identity and the desire to have a joint racial identity unites the Black

    community regardless of skin color. Other scholars have found that within-group divides based on

    color are less salient in settings where Blacks are in the minority (Harvey et al. 2005). Moreover,

    Hill (2002) finds that other racial groups, particularly Whites, do not observe skin tone variation

    among Blacks. Taken together, the foregoing results suggest that categorical inequality is likely

    more important than colorist inequality for blacks, at least in general. Within elite institutions, the

    expectation is more ambiguous, as categorical inequality has been exposed to quite aggressive

    attack that has to some extent delegitimized it. At the same time, because there are so few Blacks

    on the campuses of selective colleges and universities, the category itself may be hyper-salient and

    efforts to render it irrelevant may be undone.

    The role of skin color as a divider among Whites is less explored than that among other

    racial groups. Though Whites benefit from the valuation of light skin, this does not imply that

    Whites and non-Whites do not recognize within-group variation. To the contrary, it has been

    established that Whites perceive more variation in skin tone among themselves than among Blacks

    (Hill 2002; Ronquillo et al. 2007). This suggests that, while the continued significance of the color

    line maintains the group status position of Whites, individuals may be subject to color-based

    discrimination if they are perceived as too dark to be White. But the limited evidence of skin color

    bias among Whites leaves us mainly to speculate (see, however, Branigan and Freese 2013 for an

    important exception), and of course especially so in the context of elite colleges and universities.

  • ! 8

    THE VEILED ROLE OF GENDER

    Narratives delivered thus far ignore the possibly complicating effects of gender, a

    deficiency to which I turn now. In doing so, it is useful to briefly consider the mechanisms through

    which colorism may exert its effects. The most important one is the possibility that people with

    darker skin tones are more likely to experience discrimination than their lighter-skinned

    counterparts. Bruch and Loveman (2011) identify differences in perceptions of discrimination in

    school as well as in levels of self-esteem among students of dissimilar skin tones. Students with

    lighter skin possessed higher levels of self-esteem and reported fewer episodes of discrimination.

    Similarly, Massey and Owens (2013) find that students with darker skin are more susceptible to

    stereotype threat in college and receive lower GPAs.

    With this understanding of the key mechanism, there is good reason to believe that it plays

    out somewhat differently for women and men. This is because beauty norms and the

    commodification of women construct skin tone as a salient characteristic for females more so than

    males (Glenn 2009). For women, light skin is an asset, a unique form of capital that may be

    exchanged for high status. This lighter skin benefits women through boosts in perceived

    attractiveness and in personal self-esteem (Hunter 2002, Keith 2009). As this line of reasoning

    suggests, Hill (2002) finds this boost is weaker for males than women, in that skin tone is a less

    significant predictor of how attractive men are rated by interviewers than it is for women. The

    heightened salience of color as a beauty trait for women may lead skin tone to have a stronger

    relationship with womens attainment levels than with those of men. Indeed, some previous

    research suggests this is the case. While Gullickson (2005) identified no significance of color for

    Black men, Keith (2009) used the same data and found a significant relationship between years of

    education and color for Black women. Branigan and others (2013) also found that the relationship

  • ! 9

    between educational attainment and color was present among White women but not among White

    men.

    Yet, it is also possible that the gender interaction will take precisely the opposite form.

    Light skin may become a burden for women in contexts where romantic partnerships are at the

    forefront of social relations. In college campuses, where dating and hooking up are a central part

    of campus life and a way to fit in (Arum and Ropska 2011; Armstrong and Hamilton 2013),

    women with light skin may reap a social status benefit but not an academic benefit. Specifically,

    lighter-skinned women are seen as more desirable in heterosexual romantic partnerships, making it

    likely that these individuals spend more time dating or being courted than their darker-skinned

    counterparts (Hunter 2003; Rondilla and Spickard 2007; Rondilla 2009). We would expect this

    effect to be stronger for females than males because of the deeper associations between beauty and

    light skin for women. To the extent that lighter-skinned women are pulled into the social sphere

    more often than their darker-skinned counter parts, color may no longer be an asset for light-

    skinned females. By contrast, without as strong a countervailing social effect among males, lighter

    skin may still lead to reduced discrimination and the result being a stronger net effect for males in

    elite institutions.

    SKIN COLOR STRATIFICATION IN EDUCATION

    There is a growing body of research on skin-color stratification in educational institutions.

    For example, we know that Latinos with lighter skin tones complete more years of schooling, on

    average, than those with darker skin tones (Montalvo and Codina 2001; Hunter et al. 2001; Wilson

    and Senices 2008; Frank et al. 2010). This disparity persists among individuals of the same

    nationality and net of differences in acculturation and family background, suggesting that divides

  • ! 10

    in educational attainment are not simply a function of differences in ancestry or individual

    resources (Murguia and Telles 1996).

    The same type of result holds among Blacks: Blacks with lighter skin tones complete more

    years of schooling than those with darker skin tones (Hughes and Hertel 1990; Keith and Herring

    1991; Seltzer and Smith 1991; Keith 2009). But, scholars disagree over whether the significance of

    color for the educational attainment of Blacks has waned in the post-Civil Rights era. Whereas

    Gullickson (2005) reports that little to no skin tone stratification within Black cohorts born in later

    decades of the twentieth century, others revisiting this finding have shown that such skin tone

    differences do persist (Loury 2009), particularly for females (Keith 2009).

    Evidence of color-based differences in educational outcomes among Whites and Asians is

    scarce. Only in the past decade did surveys begin measuring skin tone for these groups. Branigan

    and Freese (2013) identify a gender-specific relationship for Whites: A significant relationship

    between years of education and skin color is evidenced among White women, but not White men.

    An educational transition analysis suggests that the association between skin tone and college

    graduation, given college attendance, is weak for White men but strong for White women. Among

    Asians, educational disparities remain largely unexplored, whereas income inequalities have been

    identified among Asians of varying skin tone (Kiang and Takeuchi 2009).

    Existing literature is silent on the form of racial inequality specifically within highly

    selective and elite educational institutions, where commentators argue egalitarian ideologies render

    the most overt forms of categorical discrimination unacceptable. The ensuing analysis explores this

    setting and, in particular, provides answers to the following questions: Has categorical inequality

    indeed been undermined in these institutions? Has it been supplanted by colorist inequality or do

    the two forms instead coexist in a dual system? And does the form of racial inequality vary across

    racial groups or gender?

  • ! 11

    METHODOLOGY

    Data

    To answer these questions, I analyze data from the National Longitudinal Study of

    Freshman (NLSF).2 The NLSF was designed to explore the experiences of minority students at

    selective colleges and universities in the United States.3 The NLSF selected 28 institutions,

    stratified by size of the Black student body, and sampled equal numbers of Latino, Asian, Black, or

    White students at each institution using information provided on their college applications

    (N=3924).4 Respondents were surveyed twice in their first year (1999-2000) and annually for the

    next three years about their social and academic experiences. The NLSF includes detailed data on

    pre-college experiences and family background in the first wave. Data from the National Student

    Clearinghouse is matched with the NLSF sample to determine the final graduation status of every

    respondent in the initial sample. Only first-time college students and U.S. citizens were eligible for

    the original sample.

    The NLSF is well suited for an analysis of skin color stratification because of the reliability

    of the skin tone measure. Interviewers measured each respondents skin color on a 10-point scale

    at the beginning of the first interview using a color palette. There is minimal inter-interviewer bias

    in skin tone measurement since a common scale of comparison is used, an important advantage

    relative to data sets where respondents skin color is rated without a static guide (e.g. Add Health,

    NSBA, NSC). Further, the survey structure reduces the possibility that skin color ratings are biased

    by the students socioeconomic status. Several studies find a whitening phenomenon, where

    interviewers rate respondents with higher reported incomes or wealthier residences as possessing

    lighter skin than those with lower reported incomes or less wealthy residences (Penner and

    Saperstein 2008; Freeman et al. 2011, Villarreal 2012). Because the question was asked prior to

  • ! 12

    the first round of NLSF interviews, we can assume that the respondents answers to questions

    about parental income and education or their high school experiences did not bias the interviewers

    choice of rating.

    An additional advantage of this data set is that it compares individuals exposed to the very

    same educational institutions. Because most analyses of skin color effects use years of education as

    the outcome, it is unclear where the disparities arise. That is, we cannot judge whether inequalities

    arise because darker skinned students do not pursue higher education, because they are more likely

    to drop out than their lighter skinned peers, or because they attend different types of institutions.

    The sampling of only college-goers at similarly high-status institutions allows for a comparison of

    students experiencing the same college context, eliminating the concern that students with dark

    and light skin are simply attending universities with disparate graduation rates.5

    The main limitation of this dataset is that interviewer characteristics are unreported. It is

    well known that perceived skin color varies by interviewer characteristics: In particular,

    interviewers do not detect as much color variation among individuals of a different racial group as

    they do within their own. For example, White interviewers tend to rate White individuals as having

    a larger variety of skin tones, relative to Black interviewers; and, similarly, Black interviewers rate

    Black individuals as having a larger variety of skin tones than White interviewers (Hill 2002). This

    limitation will likely make the estimates imprecise, rather than bias the coefficients themselves, as

    interviewers race is not directly related to students educational success.

    Analytic Strategy

    I use a multi-level logistic regression with random effects at the school level to compensate

    for the clustering of students within colleges and universities as well as unobserved differences

    across institutions.6 The models allow me to examine the relationship between educational success

    and skin tone in three ways. First I estimate the association using the full sample of all four racial

  • ! 13

    groups, to determine whether, overall, skin tone is significantly associated with educational

    success net of race in elite institutions. Interaction terms are also used to examine whether the size

    or presence of a relationship varies across the four racial groups. Second, I use a split sample

    approach to examine whether the significance of skin color for educational outcomes is a gender-

    specific phenomenon. Third, I estimate the association between skin tone and educational success

    independently for each racial group, allowing me to determine the degree to which skin tone

    effects vary by group. Throughout these analyses, the education outcome is the probability of

    attaining a bachelors degree conditional on enrolling in a selective college or university. This

    measure departs from the outcome most commonly used in skin tone studies, years of education,

    an approach that allows me to better eliminate cross-group heterogeneity.7

    The relationship between color and educational outcomes is sometimes claimed to be the

    result of a legacy effect. This refers to differences in the intergenerational transmission of

    advantages or disadvantages that result from racism in previous decades. For example, some posit

    that children with darker skinned ancestors may lack certain benefits because of structural

    disadvantages and racism incurred by past generations during the pre-Civil rights era (Loury 2009;

    Oliver and Shapiro 1995; Hill 2002). To reduce such possibly confounding effects, I examine

    whether color stratification is significant net of differences in family background, measured as (a)

    whether at least one parent holds a college degree, and (b) whether the students family earned

    more than $75,000 in annual income (the modal income category in the sample).

    The other individual level covariates are whether the student took AP courses in high

    school, immigrant generation, and gender. These characteristics are included because they may be

    associated with skin color and educational outcomes (see Table 1 for covariate means). The final

    sample without missing data on any covariates consists of 992 White, 913 Hispanic, 986 Black,

    and 949 Asian students in elite universities.

  • ! 14

    RESULTS

    I begin by establishing the extent to which skin color stratification exists within each racial

    group using simple descriptive measures (see Table 2). As shown in Table, 2, graduation rates are

    extremely high for the overall sample and all four racial groups 86 percent of students in each

    group graduate, compared to the national average of about 59 percent. This is of course to be

    expected in a sample of students attending highly selective institutions (National Center for

    Education Statistics 2012). Consistent with national data, Whites and Asians graduate at higher

    rates than Hispanics and Blacks. These students also come from relatively higher economic

    standing, as revealed by the proportion reporting family incomes of over 75,000 and with college

    educated parents. Unsurprisingly, Whites received the lowest average skin tone rating, denoting

    the lightest skin, and Blacks received the highest, denoting the darkest skin. Though the group

    means are very different, we also find substantial variation in skin tone ratings within each group.

    The range for Whites on this scale is 1 to 6, while that for Blacks is 1 to 10. Asians and Hispanics

    have means and variations in between those of Blacks and Whites.

    [Table 2 About Here]

    Table 3 presents students average skin tone by their graduation status for the whole sample

    and separately by gender. Because skin tone was measured several years prior to these outcomes,

    we may assume that the rating is uninfluenced by biases about the higher ability of lighter skin

    students. Across all racial groups, students who obtain a bachelors degree possess lighter skin on

    average in comparison to their same-race counterparts who do not obtain a degree. The differences

    are statistically significant for Hispanic and White students as well as generally in the sample,

    results for which there is a precedent in past literature. At the same time, there is not a significant

  • ! 15

    difference among Asians and Blacks, although the effect is still clearly in favor of lighter skinned

    students.

    [Table 3 About Here]

    When we disaggregate by gender, we see that male students are the main drivers of the

    differences in skin color. Male Hispanics, Whites, and Asians who graduate have lighter skin on

    average than those who do not graduate. There is no significant difference in skin color ratings for

    females within each race, suggesting that skin tone is a more significant dividing factor for males

    in college settings. Additionally, we see that there is no difference among Black graduates and

    non-graduates in average skin tone for either males or females.

    Next we turn to the multivariate analysis to establish whether these relationships persist

    after accounting for a possible legacy effect and for differences among students or their

    institutions. Table 4 displays coefficients for multi-level logistic regression models predicting the

    log odds of completing a Bachelors degree among students who enrolled in 4-year programs,

    relative to not graduating after enrolling. Model 1 for each group displays the likelihood of

    graduating using only skin tone and race as predictors. I add all other individual covariates in

    Model 2 and the interaction between skin tone and race in Model 3. This final model gauges

    whether the relationship differs by racial group and will be the main focus of this discussion. The

    analysis is conducted for the whole sample and separately by gender.

    [Table 4 About Here]

    In Table 4 we see that skin tone is indeed associated with the probability of graduation:

    students with darker skin tones are less likely to graduate from college than their lighter skinned

    counterparts. This difference persists net of racial group differences and covariates associated with

    a legacy effect or other individual-level variation (Model 3). There is a negative association for

    Whites and Hispanics, but an almost null association for Blacks and Asians. There are significant

  • ! 16

    negative coefficients for Whites and Hispanics, but the interaction terms for Asians and Blacks

    bring the effect size to virtually zero for these groups.

    This result is seen more clearly in Figure 1, which displays the predicted probabilities for

    each racial group over skin tone. We see downward trending lines for Whites and Hispanics,

    indicating that the probability of graduation is lower for students with darker skin tone ratings in

    each group. For Asians and Blacks, however, we see fairly flat lines that suggest no difference in

    the probability of graduation across individuals with different skin tones in these groups.

    [Figure 1 About Here]

    The next two panels present the models for females and males separately. Here it is

    revealed that the multivariate results in the full sample only capture part of the story. There is a

    negative association between skin tone and probability of graduation for males, net of race and

    other covariates, but no apparent association for females. Figure 2 displays the predicted values for

    each racial group using the gender-specific results with the race-skin tone interaction terms. The

    interactions between race and skin tone reveal a negative association for Hispanic, White, and

    Asian males. Though the coefficient for Asian males is slightly closer to zero, yielding a flatter

    slope across the skin tone spectrum, T-tests for differences in the magnitude of coefficients reveal

    that the estimate for Asians is not statistically different from those of Hispanics and Whites.

    Among Black males, on the other hand, the relationship is significantly less negative than it is for

    the other racial groups and not statistically different from zero. The interaction terms in the female-

    only models suggest null or weak relationships for all racial groups.

    We have thus far learned that skin tone has significant effects for students at selective

    colleges and universities. There is a negative relationship between dark skin and likelihood of

    graduation for Whites, Hispanics, and Asians and no relationship between these racial categories

    independent of color, implying the undermining of categorical inequality by colorist inequality

  • ! 17

    within these groups. We have also found that male students are the main drivers of this effect. But

    for blacks categorical inequality remains strong. That is, we have not identified a skin color effect

    for Blacks, while the categorical effect of being Black is very strong.

    [Figure 2 About Here]

    However, since each racial group has a unique range of skin tone ratings, it is possible that

    the estimated effects of being darker are misleading in the above analysis. That is, effects may be

    over- or underestimated in the models since the specified effects are relative to the whole

    distribution of possible skin color ratings. We can address this possibility by examining differences

    in outcomes by skin tone for each group independently. Table 5 displays coefficients for models

    predicting the probability of graduation with and without covariates (Models 1 and 2) individually

    for each racial group. I focus on the final model (Model 3) in which I add the interaction term

    between gender and the skin color measure.

    The expectation is that the estimates will be fairly similar to those in Table 4, with

    coefficients and standard errors perhaps slightly smaller due to the change in range of skin color.

    Indeed, we do observe the same patterns. While most coefficients do not reach statistical

    significance, likely due to the small sample size, we see coefficients of similar magnitude and

    similar trend lines. Both Hispanic males and females with darker skin are less likely to graduate

    than their lighter skinned counterparts. White and Asian males with dark skin are also less likely to

    graduate than their counterparts with light skin, though the same is not true of White and Asian

    females. There is a slight but statistically insignificant uptick in probability of graduation across

    skin color for Asian females, suggesting that darker skinned Asian females are more likely to

    graduate than their lighter skinned counterparts. Finally, among Blacks, there is no difference in

    the likelihood of graduation across skin tone levels. Also notable is that the effect of being Black is

    negative and large in magnitude.

  • ! 18

    [Table 5 About Here]

    It is important to examine the probability of graduation at values where group members are

    likely to be rated. Figure 3 displays the predicted probability of graduation for each racial group by

    gender (based on the final model), both on separate graphs and all on the same graph. Also shown

    in Figure 3 is the proportion of students within a racial group who received a particular rating. We

    see that, although in principle a White male student with a rating of 9 or 10 has less than a 0.6

    probability of graduating, no White students ever actually received this rating. There is a decrease

    of approximately 0.1 in the likelihood of graduation for Hispanics males between those rated a 2

    versus a 5, the span over which the majority of Hispanics lie. Comparable drops are evident for

    Asian and White males when we examine the difference in probability of graduation across the

    range of skin tones that each respective group characteristically occupies.

    [Figure 3 About Here]

    It is equally important to consider the predicted probability of graduation for students of

    one race relative to another. Figure 4 presents the same results, but with all racial groups on the

    same graph. The first point of interest here is that even the darkest Asian males and females still

    have higher probabilities of graduating than the lightest Blacks of the same gender. This suggests

    that the negative relationship between skin color and graduation rates for Asians creates a small

    disadvantage relative to the disadvantage of being Black, regardless of skin color. The second

    point of interest is that the darkest Hispanic men have probabilities of graduation near to or lower

    than those of Blacks, while the lightest ones have probabilities near those of Asians and Whites.

    This is indicative of sizeable stratification among Hispanics with different skin tones that results

    from differences in treatment or experiences during college.

    [Figure 4 About Here]

  • ! 19

    A final result of interest is that female students across almost all skin tones are more likely

    to graduate than their male counterparts of the same race. The result is consistent with other

    research showing females graduating at higher rates than males. Female students from all racial

    groups have a probability of graduation of above 0.8, regardless of skin tone; male students with

    relatively dark skin as well as Black males of all skin tones are the least likely to graduate.

    DISCUSSION

    Several key points arise from these analyses. First, color-based stratification in educational

    success does exist in elite colleges and universities, and it is not limited to racial groups known to

    possess relatively dark skin tones. For Asians and Latinos, there is no overall association between

    racial group membership and likelihood of graduation once color is considered, implying that

    differences in color are the main source of inequality. It follows that the colorist form of inequality

    is indeed a relevant one in elite educational institutions. Yet, at the same time, the colorist form is

    only important for men. For women the dominant racial regime appears to be egalitarian, with

    evidence of neither categorical nor colorist differences in graduation rates. Finally, the racial

    regime for Blacks is anything but egalitarian, with the dominant regime of inequality in this case

    taking on a categorical form. These findings are summarized in Table 6 and the specific

    implications for each racial group are reviewed in detail in the remainder of this section.

    [Table 6 About Here]

    The clearest evidence for color-based stratification is among Hispanic males. Color seems

    to split the Hispanic population into two groups throughout college: The lighter skinned students

    receive honorary White status, while the darker skinned students are grouped with Blacks. This

    divide amongst Latinos has been afore hypothesized (see Bonilla-Silva 2004). Is this result simply

    the expression of differences in outcomes by country of origin? This seems unlikely because, if it

  • ! 20

    were, we would expect to find the same relationship between skin color and educational outcomes

    for both males and females.

    It is important, perhaps even surprising, that skin color is significantly related to

    educational attainment for the two groups that have received the least attention in colorism

    research: Whites and Asians. We can conclude that skin tone is related to life chances even within

    groups with historically high attainment levels. Although dark Asian males continue to maintain

    high graduation rates, the same is not true of dark White males, for whom there is a much steeper

    drop in the likelihood of graduation.

    This evidence suggests that colorism has risen in importance as some categorical forms

    have been delegitimized. When racial group differences are driven into the background, the

    formerly protected groups become vulnerable as markers loosely associated with race become

    more covert bases for discrimination. The story however, is very different for Blacks. Blacks are

    the only racial group for which we find no evidence of colorist inequality. This result may reflect a

    rejection, at least among Blacks, of color-based discrimination among recent birth cohorts (see

    Gullickson 2005). Another possibility is that the black category is seen as so overwhelmingly

    negative that distinctions within that category do not surface as easily (Smith 2004; Smith-

    McLellan et al 2006). Moreover, ones status as a Black may be especially overwhelming when

    Blacks are so few in numbers, as is the case in elite institutions. Blacks make up only 3.5 percent

    of the student body at elite colleges and universities (Reardon, Baker, Klasik 2012).

    The final noteworthy finding is that the colorist form of inequality does not appear for

    women. This finding, which does not align with previously-reported results pertaining to years of

    education, may well reflect the way in which light skin can produce at once social benefits and

    academic costs specifically in elite institutions. For women, the benefits of light skin may be

  • ! 21

    neutralized by the disadvantages associated with (a) engaging in more romantic relations, and (b)

    the presumption that beautiful women are not academically strong.

    If the former hypothesis were on the mark, we would expect differences across skin tone in

    romantic or sexual relations for women but not for men. This is precisely what we find in

    supplementary analyses not reported here. That is, t-tests for group differences reveal that women

    who engage in intercourse and date are significantly lighter skinned than those who do not engage

    in these activities (p

  • ! 22

    colorist form for other groups is harder to track. More over, differential graduation rates among

    students in elite institutions are likely only the tip of the iceberg. The findings suggest that, in the

    presence of a push for racial equality, differential treatment may still arise based on covert factors

    aligned with deeply ingrained racial biases. This shift toward colorist inequality is likely endemic

    to a wide range of settings, including workplaces, clubs and voluntary organizations, and other

    everyday settings. We are thus presented with a challenge of identifying where, for whom, and

    why skin color becomes more significant than race as a source of stratification and basis for

    differential treatment. The broader implication is clear: color is an assiduous, complex source of

    racial stratification and its role in perpetuating inequality is in need of further research.

  • ! 23!

    REFERENCES Alba, Richard. 2009. Blurring the Color Line: The New Chance for a More Integrated America.

    Cambridge, MA: Harvard University Press.

    Aries, Elizabeth. 2008. Becoming a More Diverse College: Challenges and Benefits. Chapter 1,

    1-12. Race and Class Matters at an Elite College. Philadelphia, PA: Temple University Press.

    Armstrong, Elizabeth and Laura Hamilton. 2013. Paying for the Party. Cambridge, MA: Harvard

    University Press.

    Anderson, Gregory R., Eleanor JB Daugherty, and Darlene M. Corrigan. 2005. The Search for a

    Critical Mass of Minority Students: Affirmative Action and Diversity at Highly Selective

    Universities and Colleges. The Good Society, 14(3): 51-57.

    Arum, Richard and Josipa Roksa. 2011. Academically Adrift: Limited Learning on College

    Campuses. Chicago, IL: University of Chicago Press.

    Bailey, Stephen R., and Saperstein, Aliya. 2013. Skin Color and Social Inequality: A

    Comparative Exploration of Races Multidimensionality. Presented at the Population

    Association of America Annual Meeting, April 2013: New Orleans, LA.

    Banks, Taunya L. 2009. Multi-layered Racism: Courts Continued Resistance to Colorism

    Claims. (Glenn, E. N., ed.) Shades of Difference: Why Skin Color Matters. Stanford

    University Press.

    Banton, Michael. 2012. The Colour Line and the Colour Scale in the Twentieth Century. Ethnic

    and Racial Studies, 35(7): 110931.

    Bonilla-Silva, Eduardo. 2004. From Bi-racial to Tri-racial: Towards a New System of Racial

    Stratification in the USA. Ethnic and Racial Studies, 27(6): 93150.

  • ! 24!

    Bowman, Phillip J., Ray Muhammad, and Mosi Ifatuni. 2003. Skin Tone, Class, and Racial

    Attitudes among African Americans. (Herring, C., Keith, V. M., and Hayward, D., Eds) Skin

    Deep: How Race and Complexion Matter in the Color Blind Era. Chicago, IL: University

    of Illinois Press.

    Branigan, Amelia J., Jeremy Freese, Assaf Patir, Thomas W. McDade, Kiang Liu and Catarina

    Kiefe. 2013. Skin Color, Sex, and Educational Attainment in the Post-Civil Rights Era. Social

    Science Research, 42(6): 1659-74.

    Bruch, Sarah K. and Mara Loveman. 2011. Measuring and Modeling Race as a

    Multidimensional Construct: Evidence from Research on Racial Disparities in Education.

    Presented at the Population Association of America Annual Meeting, April 2011: Washington

    DC.

    Cabrera Alberto F., Amaury Nora, Patrick T. Terenzini, Ernest Pascarella and Linda Serra

    Hagedorn. 1999. Campus Racial Climate and the Adjustment of Students to College: A

    Comparison between White Students and African American Students. Journal of Higher

    Education, 70(2): 134-160.

    Chace, William. 2007. The Pursuit of Meritocratic and Egalitarian Ideals on College Campuses.

    Diverseeducation.com, July 15. Retrieved Dec 2 2013.

    (http://diverseeducation.com/article/8621/)

    Charles, Camille Z., Mary J. Fischer, Margarita A. Mooney, and Douglas Massey. 2009. Taming

    the River: Negotiating Academic, Financial and Social Currents in Selective Colleges and

    Universities. Princeton, NJ: Princeton University Press.

    England, Paula. 2010. The Gender Revolution: Uneven and Stalled. Gender and Society, 24:149.

  • ! 25!

    Frank, Reanne, Ilana R. Kresh, and Bo Lu. 2010. Latino Immigrants and the U.S. Racial Order:

    How and Where Do They Fit In? American Sociological Review, 75: 378-401.

    Fischer, Mary J. 2010. A Longitudinal Examination of the Role of Stereotype Threat and Racial

    Climate on College Outcomes for Minorities at Elite Institutions. Social Psychology of

    Education, 13(1): 1940.

    Freeman, Jonathan B., Andrew M. Penner, Aliya Saperstein, Matthias Scheutz, and Nalini

    Ambady. 2011. Looking the Part: Social Status Cues Shape Race Perception. Journal of

    Experimental Psychology, 46: 179-85.

    Glenn, Evelyn N. 2009. Shades of Difference: Why Skin Color Matters. Stanford, CA: Stanford

    University Press.

    Gullickson, Aaron. 2005. The Significance of Color Declines: A Re-Analysis of Skin Tone

    Differentials in Post-Civil Rights America. Social Forces, 84(1): 157180.

    Hall, Ronald E. (Ed.) 2008. Racism in the 21st Century: An Empirical Analysis of Skin Color.

    New York: Springer.

    - (Ed.) 2013. The Melanin Millennium: Skin Color as 21st Century International

    Discourse. New York: Springer.

    Harvey, Richard D., Nicole LaBeach, Ellie Pridgen, and Tammy M. Gocial. 2005. The

    Intragroup Stigmatization of Skin Tone Among Black Americans. Journal of Black

    Psychology, 31(3): 237253.

    Herring, Cedric, Verna M. Keith, and Hayward Derrick Horton (Eds) 2003. Skin Deep: How

    Race and Complexion Matter in the Color Blind Era. Chicago, IL: University of Illinois

    Press.

  • ! 26!

    Hersch, Joni. 2006. Skin Tone Effects among African Americans: Perceptions and Reality. The

    American Economic Review, 96: 251255.

    Hill, Mark E. 2002. Race of the Interviewer and Perception of Skin Color: Evidence from the

    Multi-City Study of Urban Inequality. American Sociological Review, 67(1), 99108.

    - 2002. Skin Color and the Perception of Attractiveness Among African Americans: Does

    Gender Make A Difference? Social Psychology Quarterly, 65(1): 77-91.

    Hochschild, Jennifer. 2006. When Do People Not Protest Unfairness? The Case of Skin Color

    Discrimination. Social Research, 73(2).

    - 2012. Lumpers or Splitters!: Analytic and Political Choices in Studying Colour

    Lines and Colour Scales. Ethnic and Racial Studies, 35(7), 11321136.

    Hunter, Margaret. 2003. Light, Bright and Almost White: The Advantages and Disadvantages of

    Light Skin. Skin Deep: How Race and Complexion Matter in the Color Blind Era.

    (Herring, Cedric, Verna M. Keith, and Hayward Derrick Horton, Eds). Chicago, IL:

    University of Illinois Press.

    - Walter R. Allen, and Edward Telles. 2001. The Significance of Skin Color among

    African Americans and Mexican Americans. African American Research Perspectives,

    (Winter): 173184.

    Hurtado, Sylvia and Deborah F. Carter. 1997. Effects of College Transition and Perceptions of

    the Racial Climate on Latino College Students Sense of Belonging. Sociology of Education,

    70(4), 324-345.

    Jones, Trina. 2009. The Case for Legal Recognition of Colorism Claims. (Glenn, E. N., ed.)

    Shades of Difference: Why Skin Color Matters. Stanford, CA: Stanford University Press.

  • ! 27!

    Kao, Grace and Jennifer S. Thompson. 2003. Racial and Ethnic Stratification in Education

    Achievement and Attainment. Annual Review of Sociology, 29: 417-42.

    Keith, Verna M. and Cedric Herring. 1991. Skin Tone and Stratification in the Black

    Community. American Journal of Sociology, 97(3): 760778.

    - 2009. A Colorstruck World: Skin Tone, Achievement, and Self-Esteem Among African

    American Women. (Glenn, E. N., ed.) Shades of Difference: Why Skin Color Matters.

    Stanford, CA: Stanford University Press.

    Kiang, Lisa and David T. Takeuchi. 2009. Phenotypic Bias and Ethnic Identity in Filipino

    Americans. Social Science Quarterly, 90(2), 428.

    Loury, Linda D. 2009. Am I still too Black for You?: Schooling and Secular Change in Skin

    Tone Effects. Economics of Education Review, 28(4): 428433.

    Massey, Douglas S., Camille Z. Charles, Garvey F. Lundy, and Mary J. Fischer. 2006. The

    Source of the River: The social Origins of Freshman at Americas Selective Colleges and

    Universities. Princeton, NJ: Princeton University Press.

    Massey, Douglas and Jayanti Owens 2013. Mediators of Stereotype Threat among Black

    Students. Ethnic and Racial Studies, 36:1-19.

    Montalvo, Frank F. and G. Edward Codina. 2001. Skin Color and Latinos in the United States.

    Ethnicities, 1(3): 321341.

    Murguia, Edward and Edward Telles. 1996. Phenotype and Schooling among Mexican

    Americans. Sociology of Education, 69: 276-89.

    National Center for Education Statistics. 2012. Digest of Education Statistics 2012, Chapter 4,

    Table 376. U.S. Department of Education, (Spring).

  • ! 28!

    OBrien, Eileen. 2008. The Racial Middle: Latinos and Asian Americans Living Beyond the

    Racial Divide. New York: New York University Press.

    Office of Management and Budget. 1997. Revisions to the Standards for the Classification of

    Federal Data on Race and Ethnicity. Federal Register 62: 5878158790. Accessed online

    (http://www.whitehouse.gov/omb/fedreg_1997standards/)

    Oliver, Melvin L. and Thomas M. Shapiro (Eds). 2006. Black Wealth / White Wealth: A New

    Perspective on Racial Inequality. New York: Routledge.

    Omi, Michael and Howard Winant. 1994. Racial Formation in the United States. New York:

    Routledge.

    Penner, Andrew. M. and Aliya Saperstein. 2008. How social status shapes race. Proceedings of

    the National Academy of Sciences of the United States of America, 105(50): 1962830.

    Pewewardly, Cornel and Bruce Frey. 2002. Surveying the Landscape: Perceptions of

    Multicultural Support Services and Racial Climate at a Predominantly White University. The

    Journal of Negro Education, 71(1/2): 77-95.

    Phinney, Jean S. 1996. When we talk about American ethnic groups, what do we mean?

    American Psychologist, 51: 918927.

    Reardon, Sean F., Rachel Baker, Daniel Klasik. 2013. Race, Income, and Enrollment Patterns in

    Highly Selective Colleges, 1982-2004. Working Paper. Stanford, CA: Center for Education

    Policy Analysis.

    Rodriguez, Clara E. 1992. Race, Culture, and Latino "otherness" in the 1980 Census. Social

    Science Quarterly, 73(4): 930-37.

    Rondilla, Joan L. 2009. Filipinos and the Color Complex: Ideal Asian Beauty. (Glenn, E. N., ed.)

    Shades of Difference: Why Skin Color Matters. Stanford, CA: Stanford University Press.

  • ! 29!

    - and Paul Spickard. 2007. Is Lighter Better?: Skin Color Discrimination among

    Asian Americans. Lanham, MD: Rowman and Littlefield.

    Ronquillo, Jaclyn, Thomas F. Denson, Brian Lickel, Zhong-Lin Lu, Anirvan Nandy, and Keith

    B. Maddox. 2007. The Effects of Skin Tone on Race-Related Amygdala Activity: An fMRI

    Investigation. Social Cognitive and Affective Neuroscience, 2(1): 3944.

    Roth, Wendy. 2012. Race Migrations: Latinos and the Cultural Transformation of Race.

    Stanford, CA: Stanford University Press.

    Russell, Kathy, Midge Wilson, and Ronald Hall (Eds). 1992. The Color Complex: The Politics of

    Skin Color among African Americans. New York: Harcourt.

    Seltzer, Richard and Robert C. Smith. 1991. Color Differences in the Afro-American

    Community and the Difference They Make. Journal of Black Studies, 21(3): 279286.

    Schuman, Howard, Charlotte Steeh, and Lawrence Bobo. 1985. Racial Attitudes in America.

    Cambridge, MA: Harvard University Press.

    Smith, William. 2004. Black Faculty Coping with Racial Battle Fatigue: The Campus Racial

    Climate in a Post-Civil Rights Era. ed Darrell Cleveland. A Long Way to Go: Conversations

    about Race by African American Faculty and Graduate Students. New York, NY: Peter-Lang

    Publishing, Inc.

    Smith-McLellan, Aaron, Blair T. Johnson, John F. Dovidio, Adam R. Pearson. 2006. Black an

    White: The Role of Color Bias in Implicit Race Bias. Social Cognition, 24(1):46-73.

    Stepanova, Elena V. and Michael J. Strube. 2012. The Role of Skin Color and Facial

    Physiognomy in Racial Categorization: Moderation by Implicit Racial Attitudes. Journal of

    Experimental Social Psychology, 48(4): 867-78.

  • ! 30!

    Takaki, Ronald. 1998. Strangers from a Distant Shore: A History of Asian Americans (2nd ed).

    New York, NY: Penguin.

    Villarreal, Andres. 2012. Flawed Statistical Reasoning and Misconceptions about Race and

    Ethnicity. American Sociological Review, 77(3): 495502.

    Wilson, Keith B. and Julissa Senices. 2008. Skin Color and Latinos with Disabilities: Expanding

    What We Know About Colorism in the United States. (Hall, Ronald, ed.) Racism in the 21st

    Century: An Empirical Analysis of Skin Color. New York: Springer.

    ! !

  • ! 31!

    Table 1. Racial Inequality Typology Gradational Inequality Categorical Inequality High Low

    High Dual system Categoric Low Colorist Egalitarian

    Table 2. Variable summaries for covariates by racial/ethnic group in the NLSF: N=3911

    Mean (SD) Whites Hispanics Blacks Asians

    Skin tone (1=lightest; 10=darkest) 1.67 (1.27)

    2.73 (1.71)

    4.97 (2.18)

    3.18 (1.76)

    Graduated with BA? 0.92 0.86 0.80 0.90 Demographic Characteristics

    Male 0.47 0.42 0.36 0.43 First generation immigrant 0.05 0.19 0.08 0.31 Second generation immigrant 0.10 0.49 0.19 0.63 High income household (>$75,000) 0.68 0.44 0.39 0.85 Parent(s) college educated? 0.91 0.72 0.76 0.85 Took AP courses? 0.92 0.89 0.84 0.94 N 954 873 944 911

    Table 3. Average skin color by graduation status for each racial group and by gender within racial groups

    All White Hispanic Black Asian

    Graduated Not graduated

    3.70 3.05a

    1.66 1.90 a

    2.67 3.13a

    4.96 5.03

    3.14 3.22

    M F M F M F M F M F Graduated 3.08 3.03 1.84 1.49 2.70 2.67 5.10 4.89 3.44 2.91 Not graduated 3.97 a 3.42 a 2.23 a 1.55 3.33a 2.92 5.22 4.85 3.88 a 2.67

    Note: 1 denotes the lightest rating, 10 the darkest. Skin color measurement was conducted in Wave 1, while educational success reflects data collected in later waves of data collection. a Difference in average skin color across students who did and did not graduate is statistically significant. ! !

  • ! 32!

    Table 4. Multi-level Logit models predicting the likelihood of graduating with a Bachelors degree by gender: Log Odds [SEs] All Males Females (1) (2) (3) (1) (2) (3) (1)! (2)! (3)!

    Hispanic -0.59*** -0.49** -0.48 -0.67** -0.51* -0.53 -0.55* -0.47+ -0.32 (0.16) (0.18) (0.30) (0.22) (0.25) (0.44) (0.23) (0.25) (0.40)

    Black -0.90*** -0.85*** -1.33*** -1.04*** -0.83*** -1.69***

    -0.90*** -0.82*** -0.99*

    (0.18) (0.18) (0.30) (0.25) (0.26) (0.45) (0.25) (0.25) (0.41) Asian -0.14 -0.15 -0.54 0.01 0.01 -0.26 -0.24 -0.24 -0.57 (0.17) (0.20) (0.34) (0.25) (0.29) (0.54) (0.24) (0.27) (0.45) Skin tone -0.06* -0.04 -0.17+ -0.10** -0.10** -0.27* 0.00 0.01 -0.04 (0.03) (0.03) (0.09) (0.04) (0.04) (0.12) (0.04) (0.04) (0.16)

    xHispanic 0.04 0.06 -0.04 (0.11) (0.14) (0.17) xBlack 0.18+ 0.26* 0.07 (0.10) (0.13) (0.16) xAsian 0.18 0.15 0.13

    (0.11) (0.15) (0.18) Controls? X X X X X X Note: All models include school level random effects. Controls include indicators of immigrant generation, AP course completion, family income of over $75,000, college educated parent(s), and belief that self will graduate. Model with full sample also controls for gender. For full table, see Table 4b in Appendix. +p

  • ! 33!

    0.500.550.600.650.700.750.800.850.900.951.00

    Pred

    icted

    pro

    babil

    ity

    1 2 3 4 5 6 7 8 9 10Skin tone rating

    Blacks HispanicsWhites Asians

    Note: Predicted values based off of Model 3 estimates in Table 2 for all students.

    Figure 1. Probability of graduation over skin tone: All students

  • ! 34!

    ! ! !

    0.500.550.600.650.700.750.800.850.900.951.00

    Pred

    icted

    pro

    babil

    ity

    1 2 3 4 5 6 7 8 9 10Skin tone rating

    Females

    0.500.550.600.650.700.750.800.850.900.951.00

    Pred

    icted

    pro

    babil

    ity1 2 3 4 5 6 7 8 9 10

    Skin tone rating

    Males

    Note: Predicted values based off of Model 3 estimates in Table 2 for each subgroup. Horizontal graylines denote average graduation rate for each gender

    Figure 2. Probabilty of graduation over skin tone for males and females

    Blacks HispanicsWhites Asians

  • ! 35!

    Table 5. Multi-level Logit models predicting the likelihood of graduating with a Bachelors degree by race: Log Odds [SEs] White Hispanic Black Asian (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3)

    Skin tone -0.17+ -0.16+ -0.06 -0.17** -0.13* -0.09 -0.00 0.01 0.02 -0.03 -0.02 0.12 (0.10) (0.10) (0.15) (0.06) (0.06) (0.08) (0.04) (0.04) (0.06) (0.07) (0.07) (0.10) Skin tone -0.18 -0.10 -0.02 -0.27* X Male (0.20) (0.11) (0.08) (0.14) Controls Included X X X X X X X X

    Note: All models include school level random effects. Controls include indicators of immigrant generation, AP course completion, family income of over $75,000, college educated parent(s), and belief that self will graduate. Model with full sample also controls for gender. For full table, see Table 5b in Appendix. +p

  • ! 36!

    0.1.2.3.4.5

    Prop

    . rec

    eived

    ratin

    g

    0.500.600.700.800.901.00

    Pred

    icted

    pro

    babil

    ity

    1 2 3 4 5 6 7 8 9 10Skin tone rating

    Whites

    0.1.2.3.4.5

    Prop

    . rec

    eived

    ratin

    g

    0.500.600.700.800.901.00

    Pred

    icted

    pro

    babil

    ity

    1 2 3 4 5 6 7 8 9 10Skin tone rating

    Hispanics

    0.1.2.3.4.5

    Prop

    . rec

    eived

    ratin

    g0.500.600.700.800.901.00

    Pred

    icted

    pro

    babil

    ity

    1 2 3 4 5 6 7 8 9 10Skin tone rating

    Blacks

    0.1.2.3.4.5

    Prop

    . rec

    eived

    ratin

    g

    0.500.600.700.800.901.00

    Pred

    icted

    pro

    babil

    ity1 2 3 4 5 6 7 8 9 10

    Skin tone rating

    Asians

    Note: Predicted values based off of Model 3 estimates in Table 3 for each subgroup.

    Figure 3. Probability of graduation over skin tone by racial group and gender

    Males FemalesProportion given a rating

  • ! 37!

    0.50

    0.60

    0.70

    0.80

    0.90

    1.00Pr

    edict

    ed p

    roba

    bility

    1 2 3 4 5 6 7 8 9 10Skin tone rating

    White Males White FemalesHispanic Males Hispanic FemalesBlack Males Black FemalesAsian Males Asian Females

    Note: Predicted values based off of Model 3 estimates in Table 3 for each subgroup

    Figure 4. Probabilty of graduation over skin tone by subgroup

  • ! 38!

    Table 6. Summary of results

    Type of Inequality

    Racial group Male Female Black Categorical Categorical Asian Colorist Egalitarian Hispanic Colorist Egalitarian White Colorist Egalitarian

  • ! 39!

    APPENDIX. Full Model Tables

    Table 4b. Multi-level Logit models predicting the likelihood of graduating with a Bachelors degree by gender: Log Odds [SEs] All Males Females (1) (2) (3) (1) (2) (3) (1)! (2)! (3)!

    Hispanic -0.59*** -0.49** -0.46 -0.70** -0.52* -0.52 -0.57* -0.48+ -0.25 (0.16) (0.18) (0.30) (0.23) (0.26) (0.45) (0.23) (0.25) (0.40) Black -0.90*** -0.88*** -1.34*** -1.06*** -0.90*** -1.71*** -0.93*** -0.85*** -0.93* (0.18) (0.18) (0.30) (0.25) (0.26) (0.46) (0.25) (0.25) (0.41) Asian -0.14 -0.16 -0.52 -0.01 -0.01 -0.28 -0.26 -0.26 -0.49 (0.17) (0.2) (0.34) (0.25) (0.29) (0.54) (0.24) (0.28) (0.45) Skin tone -0.06* -0.04 -0.16+ -0.10** -0.11** -0.27* 0.01 0.01 0.01 (0.03) (0.03) (0.09) (0.04) (0.04) (0.12) (0.04) (0.04) (0.16)

    xHispanic 0.03 0.05 -0.09 (0.11) (0.15) (0.17) xBlack 0.17 0.25+ 0.02 (0.10) (0.13) (0.16) xAsian 0.16 0.14 0.08

    (0.12) (0.16) (0.18) Male -0.42*** -0.42*** - - - - (0.10) (0.10) - - - - First gen immigrant -0.12 -0.10 -0.15 -0.16 -0.08 -0.06 (0.17) (0.17) (0.24) (0.24) (0.24) (0.24) Second gen immigrant 0.09 0.09 0.11 0.10 0.09 0.09 (0.14) (0.14) (0.21) (0.21) (0.19) (0.19) Took any AP classes 0.26+ 0.26+ 0.53* 0.54* 0.09 0.09 (0.15) (0.15) (0.22) (0.22) (0.22) (0.22) Family income over 75K 0.12 0.12 0.09 0.08 0.14 0.13

  • ! 40!

    (0.11) (0.11) (0.17) (0.17) (0.15) (0.15) Parent(s) are college graduates 0.48*** 0.46*** 0.43* 0.40* 0.53** 0.53** (0.12) (0.12) (0.19) (0.19) (0.16) (0.16) Self-perceived likelihood of graduation 0.17** 0.17** 0.31*** 0.30*** 0.03 0.03 (0.05) (0.05) (0.07) (0.08) (0.09) (0.09) Level 1 Constant 2.70*** 0.45 0.71 2.72*** -1.24 -0.80 2.70*** 1.76+ 1.83+ (0.17) (0.56) (0.59) (0.22) (0.76) (0.82) -0.22 (0.94) (0.97) Level 2 sd(Constant) -0.71*** -0.79*** -0.78*** -0.63** -0.83** -0.78** -0.59** -0.63** -0.62**

    (0.19) (0.20) (0.20) (0.22) (0.26) (0.25) (0.23) (0.23) (0.23)

    Note: All models include school level random effects. +p

  • ! 41!

    immigrant -0.57 -0.58 -0.29 -0.29 -0.31 -0.31 -0.53 -0.53 Second gen immigrant -0.42 -0.43 0.12 0.11 0.29 0.28 -0.06 -0.07 -0.37 -0.37 -0.23 -0.23 -0.25 -0.25 -0.51 -0.51 Took any AP classes 1.19*** 1.19*** -0.09 -0.08 0.12 0.12 0.67 0.72+ -0.34 -0.35 -0.32 -0.32 -0.23 -0.23 -0.41 -0.41 Family income over 75K 0.36 0.33 0.12 0.13 0.15 0.16 -0.21 -0.24 -0.27 -0.27 -0.23 -0.23 -0.19 -0.19 -0.26 -0.26 Parent(s) are college graduates 0.26 0.27 0.48* 0.46* 0.56** 0.56** 0.44 0.46 -0.39 -0.40 -0.23 -0.23 -0.19 -0.19 -0.31 -0.31 Self-perceived likelihood of graduation 0.10 0.09 0.20* 0.20* 0.24* 0.245* 0.15 0.13 -0.13 -0.14 -0.08 -0.08 -0.11 -0.11 -0.17 -0.17 Constant 2.85*** 0.48 0.35 2.46*** 0.28 0.11 1.56*** -1.20 -1.32 2.42*** 0.26 0.07 -0.27 -1.40 -1.42 -0.24 -0.87 -0.88 -0.24 -1.08 -1.12 -0.27 -1.75 -1.77

    Constant -0.79 -1.07 -0.90 -0.59* -0.61* -0.61* -0.65** -0.75** -0.76** -0.82* -1.05+ -0.99+

    -0.48 -0.68 -0.56 -0.29 -0.30 -0.30 -0.24 -0.27 -0.27 -0.41 -0.57 -0.53

    Note: All models include school level random effects. Controls include indicators of immigrant generation, AP course completion, family income of over $75,000, college educated parent(s), and belief that self will graduate. Model with full sample also controls for gender. +p

  • ! 42!

    !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!ENDNOTES 1!Latinos are labeled as a racial group as part of this argument that we not only have race-based inequality but also skin-color based inequality in education. However, this is a contested label for Latinos; some consider Latino an ethnicity rather than a race and still others consider it neither type of classification (Rodriguez 2000). I recognize this point of ambiguity and make no claims that Latino is one type of group rather than another; this label choice is used for consistency when in the discussion of existing research.!2 For more detail on the NLSF and analyses of differences between racial groups, see work by Massey (2006) and Charles (2009). !3!The value of this focus is that it provides a glimpse at the role of color in one type of elite setting. However, it is not necessarily indicative of processes in other educational settings. Inequalities amongst elite college-goers could perhaps serve as a lower bound for inequalities in the broader college student population since these students are more likely to graduate on average. That is, if we identify stratification within institutions exhibiting very little variation in graduation rates, we might expect this stratification to be small in comparison to that within schools possessing more variation in student success.!4 Respondents from the sole Historically Black College or University (N=60) are excluded from this analysis because the unique racial distribution makes it likely that the experience of skin color differs from the rest of the sample. 5!Note that it is certainly possible that skin color stratification does occur through selection into disparate quality colleges. The focus in this paper is just on stratification that results from experiences during college. Overall stratification levels may be higher than those estimated in the current analysis if selection-based stratification were also taken into account. 6!One alternative model specification might allow intercepts at the institution level to vary. This is perhaps unnecessary in a sample where universities possess remarkably similar average graduation rates (minimum=0.78, standard deviation=0.06) and were selected because of similarities in selectivity. Indeed, analyses using this alternative yielded an insignificant intercept term and coefficient estimates not statistically different from those estimated in models without random intercepts. As an additional check for intra-institutional differences in the relationship between color and attainment, I conducted the analysis separately by institution and compared coefficients across universities. I identified no significant differences in the estimated associations across institutions.!!7!Students who graduated in six rather than four years and who transferred to other institutions are included in the analytic sample as having graduated. Sensitivity analyses showed no differences in skin tone between those who took longer and graduated on time or between who left and those who stayed in their original universities, suggesting that their inclusion does not bias the results. !