Social Science

ARTICLE IN PRESS

Social

Social Science Research xxx (2007) xxx–xxx

Science

RESEARCH

www.elsevier.com/locate/ssresearch

Overlapping disadvantages and the racial/ethnic graduation gap among students attending selective institutions Sigal Alon

q

*

Department of Sociology and Anthropology, Tel Aviv University, Ramat Aviv, Tel Aviv 69978, Israel

Abstract Using a configurational approach, I examine the extent to which the intersection between background attributes can account for racial and ethnic gaps in graduation likelihood among students attending elite institutions in the United States. The results, which are based on the College & Beyond database, demonstrate the compounding effect of multiple disadvantages on students’ graduation likelihood, above and beyond the unique hardship associated with each background characteristic. Under-represented minority students are more likely to suffer from overlapping disadvantages than whites and Asians, but given similar constellations of disadvantages most minority students perform as well as whites. However, black students with overlapping disadvantages are slightly less likely to graduate than their white configuration-counterparts. About third of the overall race gap is attributed to the compounding effect of overlapping disadvantages on blacks’ achievement. That black male students with overlapping disadvantages are the most vulnerable group of all reveals an intersection between gender, race and class. Ó 2007 Elsevier Inc. All rights reserved. Keywords: Graduation gap; Overlapping disadvantages; Double disadvantage; Intersectionality; Configurations; Minorities; Elite institutions; Gender gap; Feminist theory

q

I thank the Mellon Foundation for permission to use the College & Beyond database and two anonymous reviewers for their insightful comments. * Fax: +972 3 6409215. E-mail address: [email protected] 0049-089X/$ - see front matter Ó 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.ssresearch.2007.01.006

Please cite this article in press as: Alon, S., Overlapping disadvantages and the ..., Social Sci. Res. (2007), doi:10.1016/j.ssresearch.2007.01.006

ARTICLE IN PRESS 2

S. Alon / Social Science Research xxx (2007) xxx–xxx

1. Introduction Given the increasing market returns to a college diploma it is disconcerting that the racial/ ethnic gap in the attainment of a bachelor’s degree grew over time (Kao and Thompson, 2003; Mare, 1995). While the racial/ethnic gap in college graduation rates in the United States is smaller at academically selective schools than in the overall student population, minority students attending these selective institutions are still less likely than whites or Asians to graduate (Bowen and Bok, 1998; Alon and Tienda, 2005). The societal need for a better representation of all groups in business, government, and the professions provides a compelling motivation for uncovering the roots of the racial/ethnic gap in attainment of a bachelor’s degree among this very select group of students. Such understanding is also important because the racial/ethnic gap in graduation is taken by some as an indication of the failure of affirmative action policies practiced by the most selective colleges and universities (see, for example, claims expressed by Sowell, 2003; Thernstrom and Thernstrom, 1997). The most straightforward explanation for the racial/ethnic gap in educational attainment is that minority students lack the resources needed for academic success. First, despite improvements in test scores in recent decades, black and Hispanic high school graduates still lag well behind Asians and whites in this regard (Camburn, 1990; Jencks and Phillips, 1998; Miller, 1995). Even among students enrolled at selective institutions, both blacks and Hispanics averaged test scores well below the respective institutional tier averages (Alon and Tienda, 2005). Since academic preparedness is specifically critical for coping with the rigors of selective schools’ learning environment, minority students are at a distinct disadvantage (Bowen and Bok, 1998; Bowen et al., 2005). Second, racial/ethnic variation in graduation rates can be traced to differences in parental socioeconomic status (Fischer et al., 1996; Kao et al., 1996; Mare and Winship, 1988; Karen, 2002). Financial capital is strongly related to college completion because it eliminates students’ need to work excessively to support themselves, thereby minimizing the danger of their dropping out for lack of funds (Alon, 2005, 2007). Moreover, the unique finances of selective private institutions pose a challenge to poor students, most of whom are minorities (Alon, 2007; Bowen et al., 2005). Despite the generous deployment of institutional resources, private institutions typically do not provide sufficient grants to fully cover students’ tuition costs (Trends in Student Aid, 2005). Consequently, even the lowest-income students typically borrow to help pay their tuition bills at private institutions (Kane, 1999; Bowen and Bok, 1998). Not surprisingly, Alon (2007) demonstrates that, among students attending elite private institutions, minority students’ graduation likelihood is more sensitive than that of their white counterparts to the amount of financial resources they are able to secure. By sharing their own college experiences and giving indispensable advice on what to anticipate and how to cope with academic pressures, college-educated parents are another important resource for persistence in college (Lareau, 2000; Lareau and Hovert, 1999; Massey et al., 2003; Farkas, 1996). Youth whose parents lack college exposure are at a decisive disadvantage in terms of graduation likelihood, even after controlling for academic ability and educational history (Choy, 2001). This is another graduation-enhancing resource that white students are more likely to have compared to both blacks and Hispanics. Yet, racial/ethnic differences in graduation likelihood still persist even after these factors—academic preparedness and parental socioeconomic status—are taken into account (Alon and Tienda, 2005; Bowen and Bok, 1998; Kao and Thompson, 2003). However, Please cite this article in press as: Alon, S., Overlapping disadvantages and the ..., Social Sci. Res. (2007), doi:10.1016/j.ssresearch.2007.01.006

ARTICLE IN PRESS S. Alon / Social Science Research xxx (2007) xxx–xxx

3

ascertaining ethno-racial disparities in academic outcomes is not straightforward since minority students, attending either selective or nonselective schools, often possess multiple disadvantages; black and Hispanic students typically not only have lower scholastic achievements but are also more likely to be socioeconomically disadvantaged than whites. As Mare and Winship (1988) note, any comparison of group-specific educational attainment among students as if they have equivalent social backgrounds can only be hypothetical, as only few members of disadvantaged minority groups have the benefit of family conditions equivalent to those of average whites. Within selective institutions, the overlap between race/ethnicity and multiple disadvantages is extremely pronounced. Massey et al. (2003) argue that the admissions processes to selective institutions operate to produce freshman classes composed of two very distinct subpopulations, which therefore have very different ‘‘starting lines’’ (see also Bowen and Bok, 1998). On the one hand are whites and Asians, who are typified by tremendous homogeneity: the typical white or Asian student attending an elite school is likely to be affluent, a second-generation college-goer, and academically prepared for college. On the other hand are black and Hispanic students, who are more likely to face an assortment of socioeconomic and academic disadvantages. As a result, a large share of black and Hispanic students attending elite institutions lack the resources that are routinely available to whites and Asians as they make their way through college. Since the overlap of performanceenhancing factors is unequally distributed among racial and ethnic groups, especially in the setting of elite schools, it is important to take the intersection between major status groupings into account when estimating the racial/ethnic divide in academic outcome. This idea is explicit in the premise of ‘‘double disadvantage’’ or ‘‘multiple disadvantages’’—a common theme in the ‘‘Intersectionality’’ perspective (Browne and Misra, 2003; Weber, 2001; O’Connor, 2001; King, 1988; Hill Collins, 1990; Cotter et al., 1999; Cookson and Persell, 1991; Acker, 1999). It is argued that disadvantages of social hierarchies are compounded so that individuals who suffer from multiple disadvantages because they occupy the lowest position in two (or more) social hierarchies—such as being black and socioeconomically disadvantaged—experience the most disadvantage of any group (Ransford, 1980; King, 1988; Pettigrew, 1981; Cookson and Persell, 1991). Although most of the intersectionality research focuses on women of color, there is some support for this notion with regard to minority students’ educational attainment. Analyzing the experiences of black students in elite prep schools, Cookson and Persell (1991) show that poor black students were doubly marginalized, being subjected to economic and social disempowerment that is difficult to overcome. Indicative of this problem is Miller’s (1995) finding that the test score gaps between minorities and whites were smaller at the high-income level than at the low-income level. In this study, I use the concept of intersectionality as a theoretical lens through which to study educational inequality. I flesh out the notion of intersectionality—which is not fully developed theoretically and is scantily embedded in empirical analyses of educational outcomes—by building on Ragin’s configurational approach (Ragin, 2000). This allows for a conceptualization of three types of intersections. First, I tackle the compounding effect of the overlap between social, economic, and academic resources on college graduation likelihood. Adding another layer, I consider the intersection of race with multiple disadvantages. In essence, the empirical analysis, which is the core of the paper, matches minority students with their white counterparts—those who have an equivalent level of social, economic, and academic resources—and compares their college outcomes. This makes it posPlease cite this article in press as: Alon, S., Overlapping disadvantages and the ..., Social Sci. Res. (2007), doi:10.1016/j.ssresearch.2007.01.006

ARTICLE IN PRESS 4


sible to examine the extent to which racial/ethnic inequality in graduation rates stems from the unequal distribution of dis/advantages or whether there is a net ‘‘race’’ effect. Lastly, I examine the intersection between gender, race and overlapping disadvantages. Using the College & Beyond (C&B) database (Bowen and Bok, 1998), I assess the educational outcomes of white, black, Hispanic and Asian students attending selective institutions. The C&B data provide a unique window into the experience of the upper tier of the student population—those whose whereabouts nationally representative datasets cannot meaningfully depict. The substantial size of the database, combined with large minority shares, creates an exceptional opportunity to address the research objectives. In the following section, I describe the conceptual framework and outline several testable hypotheses. After presenting the empirical results, I discuss the theoretical, methodological and policy implications of the findings. 2. Conceptualizing the effect of ‘‘overlapping dis/advantages’’ One key concept in classifying stratification systems is the degree of crystallization, i.e., the correlation between the distribution of assets, goods and resources in a society (Grusky, 2001). A high level of crystallization is evident if the same individuals consistently appear at the top of all status hierarchies, while other individuals consistently appear at the bottom of the stratification system. A high degree of crystallization, i.e., a situation of overlapping dis/advantages, arises in the context of education because in reality there is no simple random assignment mechanism that is operating to assign some students of a given family income to college-educated parents and assign other students of the same income to less-educated parents. As a result, affluent students are more likely than non-affluent students to be academically prepared, secondgeneration college-goers. Focusing on such nonrandom patterns can facilitate our understanding of the structuring of class and the underlying stratification processes (Giddens, 1973). Translating this theoretical benefit into empirical wisdom, Lieberson (1985) contends that ‘‘Much can be learned about the basic causal force by simply examining the nature of the nonrandom combinations of variables for which we normally control’’ (Lieberson, 1985, p. 211). Charles Ragin’s diversity-oriented approach (Ragin, 1987, 2000) provides the theoretical and empirical leverage to deal with the crystallization of students’ background characteristics. The method is especially suited to the question at hand because it emphasizes the holistic nature of cases and the interaction of attributes. As originally developed in The Comparative Method (1987), qualitative comparative analysis was used primarily by comparative and historical sociologists. This is a very diverse method and in this application I incorporate the general configurational framework in conventional analysis. To my knowledge, this is the first application in the context of education.1

1

Recent articles using this method, for instance, have focused on homeless social movement organizations (Cress and Snow, 2000); workers’ resistance (Roscigno and Hodson, 2004); political mobilization and revolutions (Osa and Corduneanu-Huci, 2003; Wickham-Crowley, 1992; Goodwin, 2001); social spending and the social security system (Amenta and Poulsen, 1996; Amenta and Halfmann, 2000; Hicks, 1994); labor markets (Brown and Boswell, 1995; Brueggemann and Boswell, 1998); unions and labor policies (Ebbinghaus and Visser, 1999; Griffin et al., 1991; Coverdill et al., 1994); and political parties (Redding and Viterna, 1999).



5

This is also the first attempt to link Ragin’s approach with the notion of intersetionality. As shall be shortly established, the theoretical contribution of the configurational approach for the premise of intersectionality is that it helps specify the conditions under which intersections should appear and become meaningful. The number of adversities in one’s life is an important contextual factor—for example, Dumais (2005) finds that as more adversities crop up in one’s life, the less likely one is to attain a bachelor’s degree—but the configurational approach allows a more nuanced and in-depth appraisal of specific combinations of characteristics. Ragin’s approach emphasizes two related themes: (a) seeing cases as configurations of aspects and (b) disaggregating populations into types. Ragin’s main argument is that aspects of cases should not be viewed in isolation from each other. He suggests viewing each case as a configuration; the relevant aspects of a case have to be considered all at once, as an interpretable combination of aspects. This information should be brought into play to highlight similarities or differences between cases. Cases that have different configurations of aspects are taken to differ by type or kind. In essence, each type and configuration constitutes a different ‘‘population’’, which then makes possible the assessment of within-population group differences. These types and configurations serve as an important means for understanding and explaining differences between cases. The first type of intersection addressed in this study is that which obtains between social, economic, and academic resources required for academic success. Experiencing only one disadvantage may have few repercussions for future college outcomes, whereas a situation of clustered disadvantages has a totally different bearing on the educational attainment process. For example, financial problems may result in diverting valuable time to paid work, which can be detrimental to a student who also has too little academic and social preparation, but not to a student who is academically and socially prepared. The burden of overlapping disadvantages creates a practical problem that makes it necessary to fight simultaneously on several fronts (for example, by combining work and study) but can also build up psychological distress or disengagement, especially at bastions of privilege like elite institutions. Such stress can act to further undermine students’ performance. Given an overlap between several performance-enhancing attributes, there are three scenarios for the compounding effect of race and overlapping disadvantages. First, implicit in the intersectionality perspective is the notion that the racial/ethnic gap is contingent on the level of overlapping disadvantages, so that disadvantaged circumstances accentuate the effect of race/ethnicity (Browne and Misra, 2003). However, the theoretical justification for this is not straightforward because race and ethnicity are not causal agents in and of themselves (Zuberi, 2001; Hallinan, 2000). However, since the social construction of one’s status is not created in a vacuum, the fact that the C&B minority students attend elitist, selective and predominantly white institutions is imperative for the conceptualization of the intersection between race and overlapping disadvantages. Evidence documenting racial differences in adjustment to college life in predominantly white environments coincides with this point (Tinto, 1993; Jackson and Swan, 1991). Scholars of a feminist perspective differ in their emphasis on the mechanisms that underlie intersections. Some highlight the power relations that generate social closure. It is argued that the dominant group—heterosexual elite white males—control institutions to their benefit by consciously restricting access to resources and opportunities and excluding others (Hill Collins, 1999; Weber, 2001). Closure strategies include not only those of an exclusionary nature, but also strategies adopted by the excluded themselves as a direct Please cite this article in press as: Alon, S., Overlapping disadvantages and the ..., Social Sci. Res. (2007), doi:10.1016/j.ssresearch.2007.01.006

ARTICLE IN PRESS 6


response to their status as outsiders (Parkin, 1979). At predominantly white institutions, the evolution of race-based identity and oppositional culture among minority students can be seen as a collective response to preceding exclusions. According to the theory of oppositional culture, minority students may feel alienated from school and learning because, painfully aware of their disadvantaged status, they develop an oppositional culture that undermines academic achievement (Ogbu, 1978; Fordham and Ogbu, 1986). Specifically, black and Hispanic students’ fear of ‘‘acting white’’ leads them to reject mainstream behaviors and to disengage from the learning process. Massey et al. (2003) show that an oppositional culture is present even among blacks and Hispanics attending selective institutions.2 An alternative perspective posits that social hierarchies are created through perceptual, cognitive and behavioral processes without the awareness or intention of the participants (Browne and Misra, 2003). The evolution of prejudices against minority students and the corresponding stereotype threat are instances of such processes. Steele (1988) argues that minority students may develop stereotype threat as a psychological defense mechanism. Because of the unconscious fear of living up to negative stereotypes about their group’s intellectual capacity, they are prone to under-perform academically. Massey et al. (2003) and Fischer and Massey, (2007) find strong support for this threat among minority students attending selective institutions, plausibly because race-sensitive admissions practices nourish the perception that minorities are only admitted as a result of lowering academic standards. Taken together, both accounts predict a race effect that is above and beyond that of overlapping disadvantages. Even if the concrete problem of coping with overlapping disadvantages is the same for white and minority students attending these institutions, their attitudes and/or psychological vulnerability are clearly different. If minority students’ vulnerability is accentuated by the presence of other disadvantaged background characteristics then one can conclude that race intersect with overlapping disadvantages. Supporting evidence for this intersectionality explanation would need to demonstrate that minority students with overlapping disadvantages are falling behind their white classmates with similar constellations of characteristics, while minority students with overlapping advantages are performing as well as their white counterparts. In addition, since the institutional setting is perceived as important for generating this type of intersection, the empirical investigation will consider several institutional mechanisms that accentuate or alleviate the impact of race and overlapping disadvantages on students’ graduation likelihood. However, minority vulnerability must not be contingent on the level of disadvantages. Under the second scenario, a racial/ethnic gap in graduation could rise if minority students lag behind their white counterparts across all configurations. Corroborating evidence of this race explanation entails revealing a racial/ethnic gap within all configurations. Yet a third alternative is that minority students perform as well as whites, given similar constellations of disadvantages. That is, the overall racial/ethnic graduation gap is compositional, stemming from the groups’ unequal distribution of overlapping disadvantages. Results showing a gross race/ethnicity gap but no within-configuration gaps would reinforce the compositional explanation.

2

These students internalized stereotypes of their group as less achievement-oriented, in-group acceptance was important to them, and excelling in school was associated with ‘‘acting white’’.



7

In the following empirical analysis I examine these three alternative accounts for the racial/ethnic gap in graduation. I start by creating configurations of dis/advantages based on students’ academic preparation, economic status and parental education. Within these configurations I assess the racial/ethnic gap in graduation. The pattern of the within-configuration racial/ethnic gaps is the evidence required to adjudicate between the three alternative accounts of the racial/ethnic gap in graduation. After describing the data used for the empirical implementation of this approach, I examine the extent to which disadvantages overlap, whether this situation is more pervasive among minority students than among whites, and whether and how this may impinge on students’ likelihood of timely graduation. 3. Data and methods 3.1. Data and sample This study uses the unique C&B database, a restricted-access database built by the Andrew W. Mellon Foundation that consists of individual records of undergraduate students who enrolled at one of 28 academically selective 4-year colleges and universities (Bowen and Bok, 1998).3 According to the Barron’s classification (Barron’s Profiles of American Colleges, 1992) the C&B schools are classified as ‘‘very selective’’ (median SAT 1050–1150); ‘‘highly selective’’ (median SAT 1150–1250); and ‘‘most selective’’ (median SAT above 1250).4 Only three C&B institutions are public, and all three are classified as very selective. Institutional records were collected for all students who enrolled in the fall of 1989 at all of the C&B institutions.5 The file consists of information drawn from students’ applications and transcripts, including students’ race, sex, test scores, college grade point average, major field of study, and whether and when the student graduated. The information drawn from students’ transcripts provides an exceptional source for accurate persistence data, and unlike the commonly used survey-based data, it circumvents the need to rely on students’ self-reports of their college experience.6 The dependent variable— 6-year graduation status—is obtained from these institutional records.

3 Barnard College, Bryn Mawr College, Columbia University, Denison University, Duke University, Emory University, Georgetown University, Hamilton College, Kenyon College, University of Michigan (Ann Arbor), University of Notre Dame, Northwestern University, Oberlin College, University of Pennsylvania, Princeton University, Pennsylvania State University, Rice University, Smith College, Stanford University, Swarthmore College, Tuft College, Tulane University, University of North Carolina (Chapel Hill), Washington University, Wellesley College, Wesleyan University, Williams College, and Yale University. 4 The Barron’s classification ranks institutions by their academic ‘‘competitiveness’’ (from ‘‘Noncompetitive’’ to ‘‘Most Competitive’’) – I use the term ‘‘selective’’ instead because it is more intuitive and common. Each school receives a Barron’s exclusive academic rating, which is based on the median SAT or median composite ACT entrance exam score; students’ high school class rank; average grade point average (GPA) of enrolled students, and the percentage of applicants accepted. For more details see Barron’s, 1992. 5 However, for two public institutions, the data were derived from samples. In these cases, sample weights are equal to the inverse of the probability of being sampled. All descriptive statistics presented use appropriate sample weights so that the results accurately represent the entire entering cohort at each institution. 6 These individual student records are linked to several other sources, including a survey conducted among those students and files provided by the College Entrance Examination Board (CEEB) and the Higher Education Research Institute (HERI) at the University of California, Los Angeles.


ARTICLE IN PRESS 8


The final sample consists of 24,521 students, of whom 19,364 are white, 1950 black, 1078 Hispanic and 2129 of Asian origin.7 All analyses are weighted and the multivariate analyses are adjusted to account for the clustered survey design of the data set.8 The large C&B sample size is critical to ensure a sufficient sample for each configuration. Moreover, such a sizable sample is also essential for analyzing group-based heterogeneity in order to understand the experience of minority students—a group sometimes too small to analyze in national datasets, especially when only those attending elite institutions are being considered. Another aspect of this database that makes it especially attractive for assessing the research question is the similarity among the C&B institutions in terms of quality of scholarship, academic rigor, culture, values and norms, and admissions practices.

3.2. Analytical strategy Implementing an intersectional approach is challenging because of the multidimensionality that is involved and the multiple comparisons required (Browne and Misra, 2003). The configurational approach has several built-in advantages over regression analysis as a method of group comparison by avoiding the homogenizing assumptions common in regression, such as the unrealistic assumption that all students are drawn from the same population or that the meaning of a category or value of a variable is the same across all cases. First, while the regression model’s default assumptions are of linear and additive relationships, the configurational approach automatically allows all covariates to interact with each other as well as with the group indicator.9 A second advantage of the configurational approach is that by creating discrete configuration classes, it requires continuous variables to first be recoded into categorical variables, which automatically allows the researcher to check for non-linearity or non-monotonic effects. Finally, the conventional regression model is based on an often unchecked assumption that the groups have sufficient overlap in covariates distribution. If this assumption is violated, as may be the case for students attending elite schools, estimates of group differences will be largely based on extrapolation and have little observed data support. By design, the configurational approach facilitates the examination of the required condition of group comparability, and prevents the researcher from making group comparisons within configurations where one group is extremely overrepresented relative to the other. One drawback of this method is the limited number of independent variables that can be used because of the number of configurations generated. This methodological weakness can also be seen as a theoretical advantage because it forces the researcher to carefully choose variables 7

I limit the analysis to native-born U.S. citizens for whom racial/ethnic information and graduation status were available. 8 The C&B sample design (with its large number of students in a small number of institutions) requires that all multivariate analyses be adjusted to account for the clustering of observations in primary survey units. This correction affects the estimated standard errors and the variance-covariance matrix of the estimators, but not the estimated coefficients. 9 To relax the regression’s additive assumption a multiplicative approach can be utilized. However, unless discrete classes are created, the multiplicative model also assumes linearity. What is more, in situations of complex interactions (i.e., more than two variables and/or with several levels in each variable) interpretation may be difficult.



9

(Roscigno and Hodson, 2004). This methodological limitation becomes evident in this study when I include intersections with gender and with institutional characteristics. I use three variables to capture the three resources mentioned. Economic resources are measured by parental income. Family academic resources are measured by parental education to ascertain whether the offspring are among the first-generation of collegegoers. Students’ academic resources are gauged by their SAT scores. The combination of parents’ education and income is common (i.e., SES), but an explanation about the inclusion of test scores in the configurations is warranted. First, given the focus of the paper on racial and ethnic gaps a net ‘‘race’’ effect can emerge only after all important hierarchies that breed crystallization in higher education—test scores among them—are taken into account. Second, it is important to include test scores in the configurational analysis because their distribution is racially and socioeconomically skewed. Poor students, among whom blacks and Hispanics are overrepresented, average lower test scores than their wealthy and non-minority counterparts (Sacks, 2000; Fischer et al., 1996).10 Consequently, this may result in insufficient overlap in group-specific test scores distributions. Among the C&B students, 75% of blacks have test scores below 1170, while more than 75% of whites and Asians have test score above this mark.11 This would put conventional regression practice (namely, controlling for SAT scores) in violation of the assumption about sufficient overlap in covariates distribution. Complicating this situation is the fact that SAT scores gaps between minorities and whites are smaller at the high-income level than at the low-income level (Miller, 1995). Thus, incorporating SAT scores in the construction of the configurations ensures group comparability and prevents the researcher from making what Mare and Winship (1988) call a hypothetical comparison. To ease the construction of configurations and the assessment of non-linearity, I divide the variables into categories. Categorizing a continuous variable—one—of course compromises the variation within categories. Since this could have implications for the assessment of the racial/ethnic gap in graduation, some specifications also take into account this variation within configurations by also including in the specification continuous measures of SAT scores and family income and an ordinal measure of parental educational attainment (high school education, some college, B.A., advanced degree). In categorizing these variables I follow both theoretical and analytical guidelines. I want to ensure that the categories differentiate between students in a meaningful way; that they take into account the selective nature of the C&B student population; and that each category is provided with enough cases. For example, I transform parental education into a dichotomy indicating whether either parent obtained a bachelor’s degree. Theoretically, this is the principal difference in considering whether parents could offer advice in navigating college. Collapsing categories like high school dropout, high school diploma, or some college— categories that discriminate among the general population—also makes sense

10 For example, Sacks (2000) reports that among students with SAT scores above 1100 (a minimum for highly selective colleges), one-third came from the upper-income brackets while less than one in ten came from the lower economic rungs. A recent study by Buchmann et al. (2006) finds that students from high SES families are the ones most likely to use test preparation courses and private tutors, which boost their score by about 60 points. 11 The interquartile range (Q1–Q3, the central 50% of a distribution) for blacks is 930–1170; 1040–1260 for Hispanics; 1200–1390 for Asians; and 1140–1350 for whites. Interestingly, the top quartile for blacks (1170) is lower than Q1 for Asians and similar to Q1 for whites.


ARTICLE IN PRESS 10


analytically, because in the C&B sample most students (about 60%) had at least one parent with an advanced degree and an additional 20% had at least one parent with a bachelor’s degree. Only 20% of the students in the sample were first-generation college-goers. Parental income was originally arranged into 15 categories, which I divided further into three roughly equal thirds: low income, up to $37,500 (comprising 28% of the sample); middle income, between $37,500 and $67,500 (comprising 32%); and high income, above $67,500 in yearly earnings (comprising 40%). Once again, an external benchmark for parental income cannot make a real distinction among the mostly affluent C&B students. Similarly, SAT scores, a continuous variable, were divided into three equal thirds: lowSAT, up to a score of 1150 (31% of sample); middle-SAT, between 1150 and 1300 (35%); and high-SAT, scores above 1300 (34%). Appendix A provides detailed definitions and descriptive statistics for all variables. The resulting 18 combinations are presented in Table 1. These configurations can be aggregated into several types of ‘‘overlapping dis/advantages’’. Configurations with at least two disadvantages are referred to as Overlapping Disadvantages (OD); those with at least two advantages are designated as Overlapping Advantages (OA); and the residual category, Non-Overlapping Dis/advantages, aggregates configurations that include neither more than one disadvantage nor more than one advantage. Using Ragin’s terminology, there are several ‘‘hypothetical’’ combinations, within which there are only a few cases. The six hypothetical configurations are 5, 9, 11, 13, 15, and 17 (in gray). These configurations capture unrealistic combinations of socioeconomic attributes, like being a first-generation college-goer, affluent and a high-achieving student (configuration 17). Moreover, very few minority students were classified into these configurations.12 Accordingly, the following analyses are restricted to non-hypothetical configurations—those with more than 500 students. Thus, the racial/ethnic gap is estimated only among groups that have more or less equivalent levels of parental education, income and test scores.13 A sensitivity analysis assured that no selection bias is caused by limiting the analysis to the sample of nonhypothetical configurations.14 4. Results 4.1. The unequal distribution of overlapping dis/advantages I start by examining whether black and Hispanic students are more likely than white and Asian students to suffer from overlapping disadvantages in terms of economic, social, 12

For example, in configuration 17 there are zero blacks, two Hispanics and four Asians. In configuration 5, the largest group that is being dropped, there are 11 blacks, 25 Hispanics, and 47 Asians. 13 This resembles the notion of ‘‘region of common support’’ of the widely used matching estimator (Heckman et al., 1998) and the related Propensity Score method advocated by Rosenbaum and Rubin (1983). In these methods the treatment effect is estimated only within strata that have sufficient treatment and control cases. Despite this and other conceptual similarities, these methods are not applicable to the current research question because I am not assessing the causal effect of a treatment on a certain outcome (and thus there are no treatment and control groups). 14 I estimated students’ graduation likelihood controlling for group dummies, family income, parental education, and SAT scores. I fitted this model (Model 2 of Table 3) twice using the full and the restricted samples. The point estimates and the standard errors for all covariates are practically identical across the two samples. Results are available upon request.



11

Table 1 Typology of students’ overlapping dis/advantages

Table 2 Distribution of overlapping dis/advantage types and configurations, C&B students, 1989 Configuration #

Family income

SAT scores

Parental B.A.

White (%)

Black (%)

Hispanic (%)

Asian (%)

Overlapping disadvantages (total 20%) 1 Low Low 2 Low Low 3 Low Middle 7 Middle Low

No Yes No No

15.7 4.8 5.2 3.2 2.5

56.3 28.3 18.0 4.8 5.3

43.8 19.8 9.9 10.2 3.9

12.4 3.8 3.8 4.1 0.8

Non-overlapping dis/advantages (total 25%) 4 Low Middle 8 Middle Low 10 Middle Middle

Yes Yes Yes

25.6 6.2 8.7 10.7

26.7 6.5 13.8 6.4

23.7 7.3 10.1 6.4

21.4 6.6 4.9 9.8

Overlapping advantages (total 55%) 6 Low High 12 Middle High 14 High Low 16 High Middle 18 High High

Yes Yes Yes Yes Yes

58.8 5.1 10.0 10.9 16.6 16.1

16.9 1.4 1.8 7.2 4.6 2.0

32.5 3.7 5.2 9.2 8.9 5.6

66.2 8.0 16.1 5.4 14.1 22.5

N

18,145

1909

1003

2026

and academic resources. Table 2, which presents the distribution of overlapping types and configurations, shows that more than half of students attending the C&B institutions are described as enjoying multiple advantages, about one in five students is classified as sufferPlease cite this article in press as: Alon, S., Overlapping disadvantages and the ..., Social Sci. Res. (2007), doi:10.1016/j.ssresearch.2007.01.006

ARTICLE IN PRESS 12


ing from overlapping disadvantages, and the rest (25%) experienced a mix of disadvantages and advantages. The data in this table corroborate Massey et al.’s (2003) claim that the perception of the C&B population as being elitist pertains only to white and Asian students but does not correctly describe their black and Hispanic classmates. Fifty-six percent of blacks attending these schools suffered from overlapping disadvantages, compared with 44% of Hispanics, 16% of whites, and only 12% of Asians. Students with overlapping advantages exhibit a mirror image of the same finding: 66% and 59% of Asians and whites, respectively, are classified as enjoying multiple advantages, compared with 33% and 17% of Hispanics and blacks, respectively. Even more noteworthy is the fact that 28% of blacks and 20% of Hispanics suffered from all three disadvantages (configuration 1), i.e., low family income, low SAT scores, and non-college-educated parents, as opposed to only 5% of whites and 4% of Asians. At the other end of the spectrum, while only a tiny share of underrepresented minorities—2% of blacks and 6% of Hispanics—had high levels of all three resources (configuration 18), this desirable situation was the status of 16% of whites and almost a quarter of Asian students. In sum, underrepresented minorities (blacks more than Hispanics) who attend selective institutions of higher education have a completely different profile from that of their white and Asian classmates. Not only do they tend to be socioeconomically or academically disadvantaged, they are also more likely to experience intersections of these disadvantages. The following analysis is aimed at assessing whether and how this unequal distribution of overlapping disadvantages can account for the racial/ethnic gap in the attainment of a bachelor’s degree. 4.2. The graduation gap Consistent with prior findings, the graduation rates of Hispanic and black C&B students were significantly lower than those of their white and Asian classmates. Only 77% of blacks and 84% of Hispanics graduated within six years of matriculating, compared with 88% of whites and 91% of Asians. I start examining these gross gaps by modeling 6-year graduation status. The first logistic regression model is a baseline estimation of the racial/ethnic gap. The second model adds three control variables: family income (continuous), parental education (ordinal), and SAT scores (continuous). To gauge the configuration-specific graduation likelihood, the third specification includes dummies for all non-hypothetical configurations (configuration 18, all three advantages, is the reference category) as well as group dummies. To account for within-configuration differences, Model 4 also controls for family income, parental education, and SAT scores. Model 5 contains all product terms between race/ethnicity and configurational positions, and Model 6 adds controls for family income, parental education, and SAT scores. The corresponding odds ratios are reported in Table 3. Model 1 suggests that the gross racial/ethnic gaps are statistically significant: the odds ratio for blacks is 0.428, while for Hispanics it is 0.694. The racial and ethnic graduation gap is substantially reduced by the inclusion of family income, parental education and SAT scores in Model 2, indicating that a large part of the race gap in graduation is attributable to heterogeneity in background and preparation. Nonetheless, a significant racial gap in graduation likelihood still persists, even after the introduction of these controls (the odds ratio for blacks is 0.696, significant at 1%). It is theoretically and analytical challenging that part of the race gap cannot be accounted for by the additive effect of the backPlease cite this article in press as: Alon, S., Overlapping disadvantages and the ..., Social Sci. Res. (2007), doi:10.1016/j.ssresearch.2007.01.006


13

Table 3 Odds ratios of student 6-yr graduation status, C&B students, 1989, N = 23,083 (1)

(2)

(3)

(4)

(5)

(6)

0.428 ** 0.694** 1.342 *

0.696** 0.937 1.223

0.634** 0.916 1.233

0.707** 0.953 1.224

0.918 0.787 1.333

0.942 0.805 1.299

0.273** 0.350** 0.362** 0.456** 0.739* 0.383** 0.475** 0.648** 0.865 0.533** 0.681**

0.494** 0.657** 0.440** 0.601** 0.748 0.670 0.870 0.854 0.868 0.957 0.894

0.266** 0.346** 0.356** 0.447** 0.706* 0.424** 0.510** 0.671** 0.888 0.528** 0.695** 0.692 0.741 0.698 1.116 0.951 0.390* 0.497! 0.851 0.548 0.988 1.061

0.481* 0.637** 0.448** 0.598** 0.722! 0.748 0.935 0.891 0.895 0.942 0.913 0.800 0.842 0.691 1.119 0.974 0.432* 0.539! 0.843 0.549 1.051 1.049

(disad == 1)*Hispanic (disad == 2)*Hispanic (disad == 3)*Hispanic (disad == 4)*Hispanic (disad == 6)*Hispanic (disad == 7)*Hispanic (disad == 8)*Hispanic (disad == 10)*Hispanic (disad == 12)*Hispanic (disad == 14)*Hispanic (disad == 16)*Hispanic

1.613 1.275 0.947 0.924 1.574 2.022 1.195 0.633 0.847 1.367 0.825

1.628 1.457 0.922 0.916 1.561 2.137 1.211 0.625 0.856 1.452 0.815

(disad == 1)*Asian (disad == 2)*Asian (disad == 3)*Asian (disad == 4)*Asian (disad == 6)*Asian (disad == 7)*Asian (disad == 8)*Asian (disad == 10)*Asian (disad == 12)*Asian (disad == 14)*Asian (disad == 16)*Asian

0.988 0.890 2.243* 0.911 1.358 1.091 0.780 0.737 0.906 0.759 0.782

0.969 0.987 2.266* 0.929 1.361 1.089 0.799 0.751 0.914 0.797 0.795

Black Hispanic Asian disad == 1 disad == 2 disad == 3 disad == 4 disad == 6 disad == 7 disad == 8 disad == 10 disad == 12 disad == 14 disad == 16 (disad == 1)*black (disad == 2)*black (disad == 3)*black (disad == 4)*black (disad == 6)*black (disad == 7)*black (disad == 8)*black (disad == 10)*black (disad == 12)*black (disad == 14)*black (disad == 16)*black

Parental income (in $1000) SAT scores (in 100)

1.005** 1.204**

1.000 1.200**

1.000 1.199** (continued on next page)


ARTICLE IN PRESS 14


Table 3 (continued) (1) Parental educational attainmenta a * ** !

(2)

(3)

1.085**

(4) 0.955

(5)

(6) 0.962

Ordinal variable: HS education, some college, B.A., advanced degree. Significant at 5%. Significant at 1%. Significant at 10%.

ground characteristics included. The configurational approach has the potential to elucidate some of the non-additive mechanisms producing the race gap. Since no statistically discernible difference was found between the graduation patterns of whites and Hispanics or Asians after introducing these controls, the configurational approach is less informative with regard to these (negligible) gaps. A between-configuration comparison (Model 3), assessing the influence of the first type of intersectionality on graduation likelihood, reveals a clear hierarchy: students with multiple advantages are more likely to graduate than those with overlapping disadvantages. All four categories with the relatively lowest graduation likelihood (1, 2, 3, and 7) were composed of students with overlapping disadvantages, whereas students with overlapping advantages constituted all four categories with the highest relative graduation likelihood (18, 12, 6, and 16). The decisive hardship of students with overlapping disadvantages remained intact (and statistically significant) even after controlling for within-configuration variation in the exact levels of family income, SAT scores and parental education (Model 4). Specifically, students classified in categories 1–4 were less likely to graduate in comparison to their classmates with all three advantages, even after taking into account the additive impact of family income, SAT scores and parental education. This points to the burden of overlapping disadvantages above and beyond the unique hardship associated with each background characteristic. It is noteworthy that the racial gap still exists even after controlling for background characteristics and configurational positions.15 Models 5 and 6, specifications in which race/ethnicity interacts with the configuration dummies, provide the main test for the merit of the intersectionality, race and compositional explanations. As mentioned above, evidence that the race gap is largest among students with overlapping disadvantages would support the intersectionality explanation; similar race gaps across all configurations would concur with the race explanation; and no within-configuration race gaps would demonstrate that the gross race gap is, in the main, compositional. Overall, the results do not provide any definite support for the notion that race intersects with overlapping disadvantages (intersectionality explanation). Although black students with overlapping disadvantages are slightly less likely to graduate than their white configuration-counterparts, in only two configurations (7 and 8) the race gap is statistically significant. The common theme characterizing these two configurations is mid-level

15

To emphasize the added value in the configurational approach over the common multiplicative model, I fit a specification that contains all 2- and 3-way interaction terms between family income, parental education, and SAT scores. None of these product terms are found to be statistically significant. So is the product term between these terms and the race/ethnicity dummies.



15

family income (between $37,500 and $67,500) combined with low test scores (lower than 1150). The racial gap in graduation found in these configurations could capture withinconfiguration variation in the exact level of resources. For example, if black students classified as ‘‘middle-income’’ are poorer, on average, in comparison to ‘‘middle-income’’ whites (or, alternatively, black students classified as ‘‘low-SAT’’ have lower scores, on average, than ‘‘low-SAT’’ whites), then the race gap in graduation may simply reflect these variances. To assess this possibility, I estimate the within-configuration black/white graduation gap, while controlling for variation in SAT scores, parental income and education. The results (Model 6) show that the race gap found in configurations 7 and 8 remained intact (although the within-configuration black/white graduation gaps narrow slightly in the face of these controls). This suggests that within-configuration variation is not the sole cause for these differences. The results also do not completely coincide with the race explanation. There is a race disadvantage within all configurations, but except for configurations 7 and 8, these gaps are statistically insignificant and decline in magnitude once within-configuration heterogeneity is accounted for. Taken together, the weak evidence in support of both the intersectionality and race explanations suggests that there is merit to the compositional explanation. Given similar constellations of disadvantages, most black students are as likely as whites to graduate from these elite schools. As for the Hispanic-white gap, the results in Table 3 show that there are almost no within-configuration gaps. Similar results are obtained for the Asian-white gap, except for Asian students’ over-achievement in configuration 3 (which was not altered by the within-configuration variation in SAT scores and income). A visualization of the results may further help adjudicate between these explanations. Fig. 1 depicts the configuration-specific graduation probability, by group. The configurations are sorted by whites’ configuration-specific graduation probability, starting from the configuration with the lowest probability (configuration 1). The top panel compares black and white students. This visual depiction indicates that the intersectionality explanation cannot be entirely discarded. Blacks with overlapping disadvantages lag behind whites, and the gap is especially apparent in configurations 7 and 8, as noted by the multivariate results. Yet in most cases within-configuration gaps are negligible. This suggests that, to a large extent, the black-white gap in graduation likelihood stems from blacks’ compositional disadvantage. The middle and bottom panels provide a clear visual depiction of the within-configuration equality between whites and Hispanics/Asians. To illustrate the extent to which the composition can account for the graduation gap between the groups, I simulated the counterfactual graduation rate for minority students had they had the same configurational composition as whites, while retaining their observed configuration-specific graduation likelihoods. The results of this exercise are presented in Table 4. Blacks’ simulated overall graduation rate, had they had the same configurational composition as whites, is 85%, an increase of 8% points from the observed 77%. Thus, the composition of overlapping disadvantages accounts for about 70% of the observed 12 percentage-point graduation gap between black and white students. The remaining share of the gap can be attributed to the compounding effect of overlapping disadvantages on blacks’ achievement. Hispanic students’ graduation rate rises to 86% (from 84%) once I adjust their configurational composition, leaving a trivial unexplained gap of about 2% points. The graduation rate of Asian students declines slightly to 90% (from 91%), approaching whites’ graduation rate. These results provide additional support Please cite this article in press as: Alon, S., Overlapping disadvantages and the ..., Social Sci. Res. (2007), doi:10.1016/j.ssresearch.2007.01.006

ARTICLE IN PRESS 16


6-yr grad probability

Black-White Gap 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1

2

3

7

4

8 14 10 Configuration white

16

6

12

18

16

6

12

18

16

6

12

18

black


Hispanic-White Gap 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1

2

3

7

4

8

14

10

Configuration white

Hispanic


Asian-White Gap 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1

2

3

7

4

8

14

10

Configuration white

Asian

Fig. 1. Gaps in 6-yr graduation probability, by configurations of overlapping dis/advantages-sorted by whites’ graduation rate, C&B students, 1989.



17

Table 4 Actual and simulated 6-year graduation rate, by group Graduation rate Actual

Simulateda

Black Hispanic Asian White

76.5 84.1 91.1 88.4

84.6 86.0 90.0

N

23,083

a

White’s configurational composition/group-specific graduation rate.

for the merit of the compositional explanation in accounting for the graduation gap between whites and blacks. A substantial share of the overall graduation gap can be explained by minority students’ concentration in the configurations that feature overlapping disadvantages. In what follows I extend the scope of the configurational framework in two directions. I start by examining the intersection between gender, race, and overlapping disadvantages, and subsequently I consider whether and how institution-level characteristics influence the performance of minority and disadvantaged students. 4.3. Intersection between gender, race and overlapping disadvantages Extending the configurational framework to assess gender differences in the impact of overlapping disadvantages on college completion is especially intriguing because the very notion of intersectionality was developed within feminist theories focusing on the interdependencies between gender, race and class in the labor market (Acker, 1999). However, the consideration of gender carries new nuances in the context of higher education. First, in the United States there is a growing female advantage in college completion (Peter and Horn, 2005; Goldin, 2006; Buchmann and DiPrete, 2006). Germane to the question at hand are findings that this ‘‘new gender gap’’ in higher education is related to social class. Buchmann and DiPrete (2006) find that, for cohorts born after the mid-1960s, the female advantage was largest mainly in households with parents lacking college education. Moreover, although the female advantage in college completion exists across all racial and ethnic groups, it is largest among black students. Pertinent to this investigation is Buchmann and DiPrete’s finding that blacks with less-than-college educated parents are the group within which the female advantage was largest. These general trends concerning the relationship between gender, race and class in higher education as a backdrop, motivate an appraisal of race-specific gender differences in the influence of overlapping disadvantages on the graduation likelihood among students attending elite institutions. Specifically, I examine several questions: (1) whether there are gender differences in the prevalence of overlapping disadvantages; (2) whether there is a female advantage in college completion; and (3) to what extent there is a complex interplay between overlapping disadvantages, gender and race at these elite schools. Table 5 shows the gender-specific distribution of types of overlapping dis/advantages. There is a gender imbalance in the student body of these elite schools for all groups, except Asian students, and it is is largest among black students (female/male ratio of 1.5). Nevertheless, females are more likely than males to suffer from overlapping disadvantages Please cite this article in press as: Alon, S., Overlapping disadvantages and the ..., Social Sci. Res. (2007), doi:10.1016/j.ssresearch.2007.01.006

ARTICLE IN PRESS 18


Table 5 Distribution of overlapping dis/advantage types, by gender, C&B students, 1989 White % Overlapping advantages (%) Non-overlapping dis/ advantages (%) Overlapping disadvantages (%) N

Black

Hispanic

Asian

Female 50.7

Male 49.3

Female 54.3

Male 45.7

Female 59.5

Male 40.5

Female 53.1

Male 46.9

Female 49.4

Male 50.6

48.9

55.9

52.2

58.7

15.6

18.0

26.2

35.2

62.2

64.7

27.7

26.2

28.9

26.0

24.7

31.1

25.7

26.7

21.3

23.6

23.3

17.9

18.9

15.3

59.7

50.9

48.1

38.1

16.4

11.7

12,170

12,351

9909

9455

1165

785

572

506

970

1159

(23% vs. 18%, respectively). This gender disadvantage is present among all racial and ethnic groups, and it is largest among Asian students. Yet, despite this higher propensity of females to be disadvantaged, results from a multivariate analysis of six-year graduation likelihood reveal a significant gender advantage in college graduation.16 The female advantage in college completion is smallest among whites and largest among Hispanics. Moreover, the female graduation advantage was smaller among students with overlapping disadvantages than among those with overlapping advantages. To further examine the complex intersection between gender, race and overlapping disadvantages, I fit a model of six-year graduation likelihood, including the 3-way interaction between gender, race and the types of overlapping disadvantages.17 Based on these results, I calculated the group-specific predicted probabilities of 6-year graduation. Predicted probabilities for students with overlapping advantages (OA) and those with overlapping disadvantages (OD) are graphed in the top panel of Fig. 2. This graph demonstrates several effects already discussed: within all race and gender groups, students with overlapping disadvantages are less likely to graduate than their counterparts with overlapping advantages; within all race and type groups, females are more likely to graduate than males; and within all gender and type groups, black are less likely to graduate than whites. Especially evident is the very low graduation likelihood of blacks with overlapping disadvantages. The bottom graph summarizes the group-specific female graduation advantage. The results demonstrate a complex interplay between gender, race and overlapping disadvantages. While for white students, the female advantage is greatest among those with overlapping advantages (OA), the reverse is true for minority students. Specifically, for black, Hispanic and Asian students, the female advantage in college completion is larger among the disadvantaged than among the privileged. These findings not only support Buchmann and DiPrete’s (2006) conclusion about the vulnerability of male students from low SES 16 Several models were estimated. The full model includes indicators for race/ethnicity, a female dummy, and all 2-way product terms between gender, race and the configurations of overlapping disadvantages. All models control for family income, parental education and SAT scores (results available upon request). 17 The C&B sample size is insufficient to support all 3-way product terms between race/ethnicity, gender and all configurations. The model estimated includes indicators for race/ethnicity, female, and two dummies for types: overlapping disadvantages and non-overlapping dis/advantages (overlapping advantages is the reference group), and all 2- and 3-way product terms between race, gender and types.



a

19

0.95 0.9 0.85 0.8 0.75 0.7 OD White

OA

OD Black

OA

Female

b

OD Hispanic

OA

OD Asian

OA

OA

OD Asian

OA

Male

0.06 0.05 0.04 0.03 0.02 0.01 0 OD White

OA

OD Black

OA

OD Hispanic

Fig. 2. The intersection between gender, race, and overlapping disadvantages, C&B students, 1989. (a) Predicted probabilities of 6-yr graduation likelihood, by types of overlapping dis/advantages, C&B 1989 cohort. (b) Groupspecific female advantage in graduation probabillities.

families, but also reveal that minority males hailing from such families are multiply disadvantaged. In sum, among students attending elite institutions, minority male students with overlapping disadvantages constitute the most vulnerable group. 4.4. Is intersection contingent on institutional characteristics? The theoretical underpinnings of the intersectionality perspective motivate a supplementary inquiry as to whether and which institutional characteristics accentuate the impact of race and overlapping disadvantages on students’ graduation likelihood. For example, overlapping disadvantages may be more salient in determining students’ performance in settings where disadvantaged students are a marginal fraction among a bulk of privileged classmates than in settings within which most students are underprivileged. In the former setting, either because of students’ attitudes (i.e., oppositional culture in response to prejudices and exclusion) or because of an aggravated psychological burden (i.e., stereotype threat), intersectionality may lead to underperformance. Alternatively, when disadvantaged students are fewer, institutional support mechanisms may be more conducive to the adjustment of these students. Please cite this article in press as: Alon, S., Overlapping disadvantages and the ..., Social Sci. Res. (2007), doi:10.1016/j.ssresearch.2007.01.006

ARTICLE IN PRESS 20


Table 6 Selected institutional characteristics of C&B institutions in 1989, by selectivity tier Selectivity tier Very

Highly

Most

% of total sample No. of institutions

37.6 5

42.1 15

20.3 8

Selectivity Tier median SATa Institutional mean (SD)

1050–1150 1131.5 (153.0)

1150–1250 1232.1 (128.2)

1250–1600 1333.1 (121.9)

85.5 7.2 2.4 4.9 25.4

80.3 6.5 4.1 9.1 13.8

75.4 7.3 5.6 11.8 8.8

Tuition (SD)

11007.6 (4219.2)

14949.9 (871.9)

13978.4 (2906.5)

Endowment, in $K

198,947 (180,929)

538,241 (485,333)

1,149,134 (974,587)

Demographics % White % Black % Hispanic % Asian % students with overlapping disadvantages

a

Based on the Barron’s data for entering freshmen.

Table 6 characterizes the C&B schools with regard to selected institution-level variables. Organized by tiers of institutional selectivity, the data in Table 6 shows that student bodies of the top-tier institutions are predominantly white, despite concerted efforts to diversify selective college campuses through the use of race-sensitive admission criteria. However, racial and ethnic diversity is positively related to institutional selectivity, as affirmative action practices are more prevalent at the most selective schools (Alon and Tienda, 2005). The opposite is true for the share of students with overlapping disadvantages. Only 8% of the students at the most selective schools had overlapping disadvantages, while they comprised about a quarter of the student body at very selective institutions. Plausibly the competitiveness of admission combined with the high tuition level prevents poor students with average academic preparation from gaining access to these schools. Because institutional selectivity is also related to the school’s endowment, the bulk of the C&B disadvantaged students attended institutions with relatively fewer resources. To test how these settings shape minority and disadvantaged students’ graduation likelihood, I estimated several multi-level models using Hierarchical Linear Models (HLM) (Raudenbush and Bryk, 2002).18 I do not report these results here but summarize their 18

Institution-level attributes are selectivity (freshmen’ mean SAT scores); tuition and endowment levels; the share of students with overlapping disadvantages, and the share of under-represented minority students (URM), i.e., blacks and Hispanics. At the student-level I include, in addition to dummies for race/ethnicity, two dummies for types of overlapping disadvantages (overlapping advantages is the reference group) because these complex multi-level models could not be fitted with all 12 configurations. For the same reason, the analysis is limited to assessing the effect of the first type of intersectionality, i.e., between multiple disadvantages. Because of high multicolinearity I estimated two multi-level multivariate models (one with percent disadvantaged, percent URM, and endowment level; the other with percent URM, selectivity, and tuition levels).



21

conclusions. The findings underscore the fact that at these elite schools the intersection between several overlapping disadvantages is ubiquitous, although some settings are more conducive to disadvantaged students’ graduation likelihood. Disadvantaged students are better off attending a more selective institution with fewer disadvantaged students. Plausibly, in such a setting, institutional support mechanisms are more effective in identifying the complex matrix of their disadvantages and tailoring programs to assist them. Further research is required to test whether these institutional characteristics differentially affect disadvantaged minority and white students. 5. Discussion This paper offers a new outlook on the racial and ethnic disparities in academic outcomes among students attending the nation’s most selective institutions. Using Ragin’s configurational approach, I examine whether the constellation of background attributes has an independent effect on students’ college graduation likelihood. The results clearly demonstrate the compounding effect of multiple disadvantages on students’ graduation likelihood, above and beyond the unique hardship associated with each background characteristic. Under-represented minority students are more likely to suffer from overlapping disadvantages than whites and Asians. The empirical analysis further supply evidence regarding whether and how this unequal distribution of overlapping disadvantages can account for the racial/ethnic gap in the attainment of a bachelor’s degree. The graduation disadvantage of Hispanics is mostly explained by these compositional differences. A similar conclusion pertains to Asian students’ overachievement. Regarding blacks the situation is more complex. About 70% of the race gap in graduation from elite institutions is related to the socioeconomic composition of their student body. The remaining share of the gap can be attributed to the compounding effect of overlapping disadvantages on blacks’ achievement. Furthermore, the results show that black male students with overlapping disadvantages are the most vulnerable group of all. These findings highlight the benefits of a multidimensional understanding of inequality that an intersectional lens brings into focus. The study of inequalities in educational processes can surely gain from such a unique approach. A comprehensive assessment of how institutional processes shape students’ experiences is a promising direction, from both theoretical and policy perspectives. This requires expanding the scope of the assessment to students attending less selective schools so that the full variation among postsecondary institutions may be exploited to disentangle the complex interplay between students and institutional characteristics. In addition, given the twist that the findings of this study add to the emerging literature on female advantage in higher education, one promising direction is an extensive consideration of the intersection between gender, race, and social class. This line of research has the potential to better identify disadvantaged groups in higher education and detect the common roots for the ‘‘old race gap’’ and the ‘‘new gender gap’’ in higher education. In addition to providing a more sophisticated and nuanced empirical understanding of the race gap in college graduation, the paper contributes more broadly to theories of intersectionality. First, by linking the configurational approach to the notion of intersectionality, it provides a quantitative approach to assessing intersections which can be easily implemented to address a broad range of research questions. Second, the results contribute Please cite this article in press as: Alon, S., Overlapping disadvantages and the ..., Social Sci. Res. (2007), doi:10.1016/j.ssresearch.2007.01.006

ARTICLE IN PRESS 22


to the feminist literature on intersectionality by supporting black and multiracial feminist claims that individuals do not experience race, social class and gender as separate or additive, but as simultaneous and linked. However, the finding that the outcome of these intersections is a ‘‘female advantage’’ in graduation rates, particularly among black students, should push feminist theories of intersectionality to consider the of the conditions under which being female might serve as an advantage (King, 1988). The results of the current study demonstrate that the proportion of disadvantaged students attending the C&B selective institutions was very low. Expectedly, more and more scholars and policy makers think that elite, private, well-endowed institutions could do a better job of enrolling disadvantaged students by implementing class-based affirmative action programs (Bowen et al., 2005). However, directing the efforts to diversification of student bodies is insufficient because the reality is that disadvantaged students are at a greater risk of dropping out of college compared to their privileged classmates. Policies that encourage enrollment but fail to provide guidance and resources that allow disadvantaged students to translate participation into college credentials not only do not alleviate class inequality, but are also in many ways a waste of both public and personal resources (Bowen et al., 2005). These students need more academic, social, emotional and financial support to facilitate their adjustment to campus life and to help them overcome multiple barriers. What is clear from this undertaking is that students’ chances of success should be increasingly understood as a synthesis of their characteristics and institutional support mechanisms should be designed to address this multidimensionality in students’ profiles.

Appendix A Variables descriptive statistics and definitions, College & Beyond, 1989 entering cohort xðSDÞ=% Variable Definition Graduation status 6-yr grad

6-year graduation of those students who 87.6% graduated from the same school at which they matriculated as first-time freshmen

Race White Black Hispanic Asian

Student’s racial or ethnic background White, not of Hispanic origin Black, not of Hispanic origin Hispanic, regardless of race Asian or Pacific Islander

80.6% 7.2% 3.8% 8.5%

Female

Female

50.7%

SAT scores

Standardized test scores

1215.32 (156.19)

Line missing Please cite this article in press as: Alon, S., Overlapping disadvantages and the ..., Social Sci. Res. (2007), doi:10.1016/j.ssresearch.2007.01.006


23

Appendix A (continued) Variable

Definition

xðSDÞ=%

Parentaleducational attainment B.A.

Mother or father with B.A.

81.7%

Parental Income Mean (in $1000)

In 15 categories

59.4 (28.54)

N

24,521

References Acker, J., 1999. Rewriting class, race, and gender: problems in feminist rethinking. In: Lorber, J., Hess, B., Ferree, M.M. (Eds.), Revisioning Gender. Sage, Thousand Oaks, CA. Alon, S., 2005. Model mis-specification in assessing the impact of financial aid on academic outcomes. Research in Higher Education 46 (1), 109–125. Alon, S., 2007. The influence of financial aid in leveling group differences in graduating from Elite institutions. Economics of Education Review 26 (3), in press. Alon, S., Tienda, M., 2005. Assessing the ‘‘Mismatch’’ hypothesis: differentials in college graduation rates by institutional selectivity. Sociology of Education 78 (4), 294–315. Amenta, E., Halfmann, D., 2000. Wage wars: institutional politics, WPA wages, and the struggle for U.S. social policy. American Sociological Review 65, 506–528. Amenta, E., Poulsen, J.D., 1996. Social politics in context: the institutional politics theory and social spending at the end of the new deal. Social Forces 75, 33–60. Barron’s, 1992. Profiles of American Colleges. Barron’s Educational Series, Inc., Hauppauge, NY. Bowen, W.G., Bok, D., 1998. The Shape of the River: Long-Term Consequences of Considering Race in College and University Admissions. Princeton University Press, Princeton, NJ. Bowen, W.G., Kurzweil, M.A., Tobin, E.M., 2005. Equity and Excellence in American Higher Education. University of Virginia Press, Charlottesville, VA. Brown, C., Boswell, T., 1995. Strikebreaking solidarity: the great steel strike of 1919. American Journal of Sociology 100, 1479–1519. Browne, I., Misra, J., 2003. The intersection of gender and race in the labor market. Annual Review of Sociology 29, 487–513. Brueggemann, J., Boswell, T., 1998. Realizing solidarity: sources of interracial unionism during the great depression. Work and Occupations 25, 436–482. Buchmann, C., DiPrete, T.A., 2006. The growing female advantage in college completion: the role of family background and academic achievement. American Sociological Review 71, 515–541. Buchmann, C., Roscigno, V.J., Condron, D.J., 2006. The myth of meritocracy? SAT preparation, college enrollment, class and race in the United States. In: Paper presented at the ASA meetings in Montreal, August 2006. Camburn, E.M., 1990. College completion among students from high schools located in large metropolitan areas. American Journal of Education 98, 551–569. Choy, S.P., 2001. Students whose parents did not go to college: postsecondary access, persistence, and attainment. Findings from the Condition of Higher Education, 2001, NCED 2001-126. U.S. Department of Education, NCES. U.S. Government Printing Office, Washington, DC. Cookson, P.W., Persell, C.H., 1991. Race and class in America’s Elite preparatory boarding schools: African Americans as the outsiders within. Journal of Negro Education 60 (2), 219–228. Cotter, D.A., Hermsen, J.M., Vannwman, R., 1999. System of gender, race, and class inequality: multilevel analyses. Social Forces 78 (2), 433–460. Coverdill, J.E., Finlay, W., Martin, J.K., 1994. Labor management in the southern textile industry. Sociological Methods and Research 23, 54–85.


ARTICLE IN PRESS 24


Cress, D., Snow, D., 2000. The outcomes of homeless mobilization: the influence of organization, disruption, political mediation, and framing. American Journal of Sociology 105, 1063–1104. Dumais, S.A., 2005. Accumulating adversity and advantage on the path to postsecondary education: an application of a person-centered approach. Social Science Research 34 (2), 304–332. Ebbinghaus, B., Visser, J., 1999. When institutions matter: union growth and decline in Western Europe, 1950– 95. European Sociological Review 15, 135–158. Farkas, G., 1996. Human Capital or Cultural Capital? Ethnicity and Poverty Groups in an Urban School District. Aldine de Gruyter, New York. Fischer, C.S. et al., 1996. Inequality by Design: Cracking the Bell Curve Myth. Princeton University Press. Fischer, M.J., Massey, D.S., 2007. The effects of affirmative action in higher education. Social Science Research. Available from: . Fordham, S., Ogbu, J.U., 1986. Black students school success: coping with the burden of acting white. The Urban Review 18, 176–206. Giddens, A., 1973. The Class Structure of the Advanced Societies. HarperCollins Publisher, Inc.. Goldin, C., 2006. The quiet revolution that transformed women’s employment, education, and family. The American Economic Review 96 (2), 1–20, Richard T. Ely Lecture. Goodwin, J., 2001. No Other Way Out: States and Revolutionary Movements, 1945–1991. Cambridge University Press, New York. Griffin, L.J., Botsko, C., Wahl, A.-M., Isaac, L., 1991. Theoretical generality, case particularity: qualitative comparative analysis of trade union growth and decline. International Journal of Comparative Sociology 32, 110–136. Grusky, D.B., 2001. The past, present and future of social inequality. In: Grusky, D.B. (Ed.), Social Stratification, second ed. Westview Press, Boulder, CO. Hallinan, M.T., 2000. On the linkages between sociology of race and ethnicity and sociology of education. In: Hallinan, M.T. (Ed.), Handbook of the Sociology of Education. Kluwer Academic/Plenum Publishers, New York. Heckman, J.J., Ichimura, H., Todd, P., 1998. Matching as an econometric evaluation estimator. The Review of Economic Studies 65 (2), 261–294. Hicks, A., 1994. Qualitative comparative analysis and analytic induction: the case of the emergence of the social security state. Sociological Methods & Research 23, 86–113. Hill Collins, P.C., 1990. Black Feminist Thought: Knowledge, Consciousness and the Politics of Empowerment. Routledge, New York. Jackson, K.W., Swan, A.L., 1991. Institutional and individuals factors affecting black undergraduate students performance: campus race and student gender. In: Walter, R.A., Epps, E.G., Haniff, N.Z. (Eds.), College in Black and White: African American Students in Predominantly White and in Historically Black Public Universities. State University of New York Press, New York, pp. 127–141. Jencks, C., Phillips, M., 1998. The Black-White test scores gap: an Introduction. In: Jencks, C., Phillips, M. (Eds.), The Black-White Test Score Gap. Brookings Institution Press, Washington, DC. Kane, T.J., 1999. The Price of Admission. Russell Sage Foundation, New York. Kao, G., Tienda, M., Schneider, B., 1996. Racial and ethnic variation in educational achievement. Research in Education and Socialization 11, 263–297. Kao, G., Thompson, J.S., 2003. Racial and ethnic stratification in educational achievement and attainment. Annual Review of Sociology 29, 417–442. Karen, D., 2002. Changes in access to higher education in the United States: 1980–1992. Sociology of Education 75, 191–210. King, D.K., 1988. Multiple Jeopardy, multiple consciousness: the context of a black feminist ideology. Signs: Journal of Woman in Culture and Society 14, 42–72. Lareau, A., 2000. Home Advantage: Social Class and Parental Intervention in Elementary Education. Roman and Littlefield, Oxford. Lareau, A., Hovert, E.M., 1999. Race, class and cultural capital in family–school relationships. Sociology of Education 72, 37–53. Lieberson, S., 1985. Making It Count: The Improvement of Social Research and Theory. University of California Press, Berkeley CA. Mare, R.D., 1995. Changes in educational attainment and school enrollment. In: Farley, R. (Ed.), State of the Union: America in the 1990s. Russell Sage Foundation, New York, pp. 155–213.



25

Mare, R.D., Winship, C., 1988. Ethnic and racial patterns of educational attainment and school enrollment. In: Sandefur, G.D., Tienda, M. (Eds.), Divided Opportunities: Minorities, Poverty, and Social Policy. Plenum Press, New York. Massey, D.S., Charles, C.Z., Lundy, G., Fischer, M.J., 2003. Source of the River: The Social Origin of Freshmen at America’s Selective Colleges and Universities. Princeton University Press, Princeton, NJ. Miller, S., 1995. An American Imperative: Accelerating Minority Educational Advancement. Yale University Press, New Haven, CT. O’Connor, C., 2001. Making sense of the complexity of social identity in relation to achievement: a sociological challenge in the new millennium. Sociology of Education, 159–168, Extra Issues 2001. Ogbu, J.U., 1978. Minority Education and Caste: The American System in Cross-Cultural Perspective. Academic Press, New York. Osa, M.J., Corduneanu-Huci, C., 2003. Running uphill: political opportunity in non-democracies. Comparative Sociology 2, 605–619. Parkin, F., 1979. Marxism and Class Theory: A Bourgeois Critique. Taylor & Francis Books. Peter, K., Horn, L., 2005. Gender differences in participation and completion of undergraduate education and how they have changed over time. NCES 2005-169, U.S. Department of Education, Institute of Education Sciences. Pettigrew, T.F., 1981. Race and class: an interactive view. Daedalus 110, 133–155. Ragin, C., 1987. The Comparative Method: Moving Beyond Qualitative and Quantitative Strategies. University of California Press, Berkeley. Ragin, C., 2000. Fuzzy-Set Social Science. University of Chicago Press, Chicago. Ransford, H.E., 1980. The prediction of social behavior and attitudes: the correlates tradition. In: Jeffries, V., Ransford, H.E. (Eds.), Social Stratification: A Multiple Hierarchy Approach. Allyn & Bacon, Boston, pp. 265–303. Raudenbush, S.W., Bryk, A.S., 2002. Hierarchical Linear Models: Applications and Data Analysis Methods. Sage Publications, Thousand Oaks. Redding, K., Viterna, J.S., 1999. Political demands, political opportunities: explaining the differential success of left-libertarian parties. Social Forces 78, 491–510. Roscigno, V.J., Hodson, R., 2004. The organizational and social foundations of worker resistance. American Sociological Review 69, 14–40. Rosenbaum, P.R., Rubin, D.B., 1983. The central role of the propensity score in observational studies for casual effect. Biometrika 70 (1), 41–55. Sacks, P., 2000. Standardized Minds: The High Price of America’s Testing Culture and What We Can Do to Change It. Perseus Books, Cambridge, Mass. Sowell, T., 2003. Damaging admissions: increasing faculty diversity. Capitalism Magazine. Retrieved July 31, 2003 (http://www.capmag.com/article.asp?ID=2448). Steele, C.M., 1988. The psychology of self-affirmation: sustaining the integrity of the self. In: Berkowitz, L. (Ed.), Advances in Experimental Social Psychology. Academic Press, New York, NY, pp. 261–302. Thernstrom, S., Thernstrom, A., 1997. America in Black and White: One Nation Indivisible. Simon and Schuster, New York. Tinto, V., 1993. Leaving College: Rethinking the Causes of Student Attrition, second ed. University of Chicago Press, Chicago, IL. Trends in Student Aid, 2005. The College Board, New York. Weber, L., 2001. Understanding Race, Class, Gender and Sexuality: A Conceptual Framework. McGraw-Hill, New York. Wickham-Crowley, T.P., 1992. Guerrillas and Revolution in Latin America: A Comparative Study of Insurgents and Regimes Since 1956. Princeton University Press, Princeton, NJ. Zuberi, T., 2001. Thicker than Blood: How Racial Statistics Lie. University of Minnesota Press, Minneapolis, MN.
