teachers, schools, racial differences

28
 Teachers College Record V olume 114, 070303, July 2012, 27 pages Copyright © by Teachers College, Columbia University 0161-4681 1 How Teachers and Schools Contribute to Racial Differences in the Realization of  Academic Potential TINA WILDHAGEN Smith College Background/Context: The fulfillment of academic potential is an underdeveloped area of inquiry as it relates to explaining racial differences in academic outcomes. Examining this issue is important for addressing not only differences in the typical outcomes for African American and White students but also the severe underrepresentation of African American students among the highest achieving students. Whereas other studies have operationalized lost academic potential as unfulfilled expectations for educational attainment, this study takes a different approach, measuring whether students earn higher or lower grades than the grades predicted by earlier tests of academic skills. Students whose grades are equal to or exceed those predicted by their earlier test scores are said to have fulfilled their academic poten- tial, whereas those whose grades are lower than predicted have not realized their potential. Purpose/Objective/Research Question/Focus of Study: This study finds that African American high school students are less likely than their White peers to realize their academic  potential. The analyses test several explanations for the racial gap in the realization of aca- demic potential, focusing on the students themselves, their teachers, and their schools. Research Design: This study uses hierarchical linear modeling to analyze data from the  Education Longitudinal Study of 2002. Conclusions/Recommendations: The results suggest that teachers perceive African American students as exerting less classroom effor t than White students, which accounts for a substantial proportion of the racial gap in unrealized academic potential, even with sev- eral student characteristics held constant. At the school level, there are larger racial gaps in unrealized academic potential in segregated schools and schools with strict disciplinary cli- mates. Strikingly, the negative effect of strict disciplinary climate exists net of students’ own receipt of disciplinary actions. That is, the negative association between strict disciplinary climate and the realization of academic potential for African American students applies to 

Upload: citizen-stewart

Post on 06-Oct-2015

15 views

Category:

Documents


0 download

DESCRIPTION

How teacher and schools contribute to differences in achievement.

TRANSCRIPT

  • Teachers College Record Volume 114, 070303, July 2012, 27 pagesCopyright by Teachers College, Columbia University0161-4681

    1

    How Teachers and Schools Contribute toRacial Differences in the Realization ofAcademic Potential

    TINA WILDHAGEN

    Smith College

    Background/Context: The fulfillment of academic potential is an underdeveloped area ofinquiry as it relates to explaining racial differences in academic outcomes. Examining thisissue is important for addressing not only differences in the typical outcomes for AfricanAmerican and White students but also the severe underrepresentation of African Americanstudents among the highest achieving students. Whereas other studies have operationalizedlost academic potential as unfulfilled expectations for educational attainment, this studytakes a different approach, measuring whether students earn higher or lower grades than thegrades predicted by earlier tests of academic skills. Students whose grades are equal to orexceed those predicted by their earlier test scores are said to have fulfilled their academic poten-tial, whereas those whose grades are lower than predicted have not realized their potential.Purpose/Objective/Research Question/Focus of Study: This study finds that AfricanAmerican high school students are less likely than their White peers to realize their academicpotential. The analyses test several explanations for the racial gap in the realization of aca-demic potential, focusing on the students themselves, their teachers, and their schools.Research Design: This study uses hierarchical linear modeling to analyze data from theEducation Longitudinal Study of 2002.Conclusions/Recommendations: The results suggest that teachers perceive AfricanAmerican students as exerting less classroom effort than White students, which accounts fora substantial proportion of the racial gap in unrealized academic potential, even with sev-eral student characteristics held constant. At the school level, there are larger racial gaps inunrealized academic potential in segregated schools and schools with strict disciplinary cli-mates. Strikingly, the negative effect of strict disciplinary climate exists net of students ownreceipt of disciplinary actions. That is, the negative association between strict disciplinaryclimate and the realization of academic potential for African American students applies to

  • Teachers College Record, 114, 070303 (2012)

    African American students regardless of whether they themselves have been in trouble atschool. This study reveals that characteristics of schools that lack immediately obvious racialimplications, such as a schools approach to student discipline, may be just as harmful asovertly racialized inequality within and between schools.

    An important element of closing mean differences between AfricanAmerican and White students in college matriculation and in the acade-mic criteria used in college admissions decisions (e.g., grades, advancedcoursework) is to identify and explain unrealized academic potentialamong African American high school students. This study focuses ongrades, an important criterion for college admissions, as indicators ofunrealized academic potential among high school seniors.I use nationally representative data from the Education Longitudinal

    Study of 2002 (ELS) to assess the extent to which behavioral and attitudi-nal differences between African American and White high school stu-dents, and differences in how teachers perceive the two groups, accountfor the racial gap in the realization of academic potential. I also examinethe extent to which this racial difference varies across high schools, focus-ing on the roles of school disciplinary climate, segregation, and racialinequality in access to advanced coursework.The fulfillment of academic potential is an underdeveloped area of

    inquiry as it relates to explaining racial differences in academic out-comes. Examining this issue is important for addressing not only differ-ences in the typical outcomes for African American and White studentsbut also the severe underrepresentation of African American studentsamong the highest achieving students. If it is the case that AfricanAmerican students are more likely than White students to possess stocksof unfulfilled academic potential, African Americans will remain under-represented in the upper tail of academic achievement distributions,which will perpetuate African American underrepresentation inAmericas most selective colleges and universities.

    PREVIOUS RESEARCH ON UNREALIZED ACADEMIC POTENTIALAMONG AFRICAN AMERICAN STUDENTS

    Most previous research that has tackled this issue of unrealized academicpotential has operationalized lost potential as it relates to students edu-cational plans. The idea is that students whose educational expectationscool out over a number of years have not realized their potential, result-ing in losses for both those individual students and society as a whole(Hanson, 1994). Using the nationally representative High School and

    2

  • TCR, 114, 070303 Racial Differences

    Beyond survey, Hanson (1994) measured changes in students educa-tional plans and whether students fulfilled their expectations over the 6years following high school, defining those who lowered their expecta-tions over time or did not fulfill plans of earning bachelors degrees ashaving lost talent. Interestingly, Hanson found that White students aremore likely to experience lost talent in the form of scaled-back educa-tional expectations than are non-White students. Studies by Trusty andHarris (Trusty, 2000; Trusty & Harris, 1999) yielded similar findingsusing the nationally representative National Education LongitudinalStudy.Although it is important to know the extent of and explanations for

    unfulfilled plans for educational attainment, we also need to focus onunfulfilled academic achievement. Indeed, relying only on unfulfilled edu-cational plans to measure racial differences in unrealized potential orlost talent may lead to underestimating the stock of unfulfilled potentialamong African American students. Among African American and Whitestudents who realize their plans for educational attainment, AfricanAmerican students still could possess a larger stock of lost potential foracademic achievement. In this study, then, I shift the question from DoWhite and African American students differ on the extent to which theytranslate plans for postsecondary education into educational attain-ment? to Do White and African American students differ on the extentto which they translate academic aptitude into academic achievement?This study widens our view of the extent of unrealized academic poten-tial among African American students compared with White students byexamining unfulfilled academic achievement rather than unfulfilledplans for educational attainment. Specifically, I examine the extent towhich students senior-year grades fall short of the grades predicted bytheir earlier achievement test scores.

    GRADES AS A MEASURE OF UNREALIZED ACADEMIC POTENTIAL

    Scholars have dedicated an incredible amount of energy to explainingracial differences in standardized test scores (Jencks & Phillips, 1998;Magnuson & Waldfogel, 2008), but as grades become an increasinglyimportant criterion for college admissions, we need to pay equal atten-tion to racial differences in grades. For example, with its Top Ten law,the state of Texas has made making good grades in high school a suffi-cient condition for acceptance at any of the states public universities,guaranteeing a seat for any student who graduates in the top 10% of hisor her high school class. In addition, according to the National Center forFair and Open Testing (2010), more than 800 colleges and universities

    3

  • Teachers College Record, 114, 070303 (2012)

    have made SAT I or ACT scores an optional component of students appli-cations for admission.This study finds that African American students earn grades that are

    lower than earlier test scores predicted, whereas Whites earn higher-than-predicted grades. In the following sections, I discuss potential explana-tions for this phenomenon, focusing first on the characteristics ofstudents and teachers perceptions of students, followed by a discussionof the potential role of schools.

    SCHOOL BEHAVIORS AND ATTITUDES AND THE REALIZATION OFACADEMIC POTENTIAL

    Research suggests that whereas African American students score lower onmeasures of proschool behaviors, such as time spent on homework,White students tend to hold less positive attitudes about school(Ainsworth-Darnell & Downey, 1998). Thus, to the extent that proschoolbehaviors and attitudes are positively related to the realization of acade-mic potential, racial differences in school behaviors and attitudes shouldcancel each other out as explanations for racial differences in realizingacademic potential:

    Hypothesis 1: Taken together, mean differences in school attitudes andschool behaviors between White and African American students do notaccount for the racial difference in realizing academic potential.

    Once students self-reported behaviors and attitudes are held constant,teachers perceptions of students engagement may still contribute toracial differences in unrealized academic potential. This could occureither because teachers are less likely to recognize the school engage-ment of African American students than White students, or becauseteachers misperceive African American students engagement as misbe-havior.Research demonstrates that some teachers misrecognize African

    American students academic alacrity as misbehavior (Ferguson, 2000;Lewis, 2003; Ogbu, 2003). For example, Lewis observed an AfricanAmerican student sent from a classroom for pumping his arms in a raisethe roof motion in celebration of giving a correct answer. This exampleillustrates the argument that schools and school officials hold all studentsto the same White, middle-class behavioral norms, even though not allstudents operate according to these norms (Bourdieu, 1977; Bourdieu &Passeron, 1990; Lewis). Ogbu argued that these kinds of cross-cultural

    4

  • TCR, 114, 070303 Racial Differences

    misunderstandings contributed to the disproportionate discipline prob-lems of Black students (p. 139).As key allocators of educational resources and rewards, teachers play a

    crucial role in the translation of student ability into academic achieve-ment. Once students self-reported school behaviors and attitudes areheld constant, teachers perceptions of students classroom effort shouldindicate the extent to which teachers recognize students proschool orien-tations. If teachers perceptions of classroom effort are related to racialdifferences in the realization of academic potential independent of stu-dents reports of their own behaviors and attitudes, this is consistent withthe argument that teachers misrecognition of African American studentsas less engaged with school contributes to racial differences in realizedacademic potential. The analysis tests the following hypothesis related toteachers perceptions:

    Hypothesis 2: Controlling for students self-reported school attitudesand behaviors, teachers perceptions of students classroom effort accountfor a portion of the racial difference in the realization of academic poten-tial.

    SCHOOLS AND THE REALIZATION OF ACADEMIC POTENTIAL

    Although student characteristics likely explain some portion of the differ-ence in unfulfilled academic potential between Whites and AfricanAmericans, schools probably play some role in the process as well.Indeed, Downey, von Hippel, and Broh (2004) found that test score gapsbetween young White and African American students widened duringthe school year but not during the summer months, suggesting thatschools play a role in racial differences in academic outcomes indepen-dent of family and neighborhood contexts. This article focuses on twoschool characteristics in particular: the disciplinary climate of the schooland the racial gap in Advanced Placement (AP) enrollment. On onehand, the racial gap in AP course enrollment provides an explicit mea-sure of school-level racial inequality by measuring the extent to which aschools academic opportunity structure is racialized. On the other hand,disciplinary climate is a characteristic that does not have immediatelyapparent racial implications. Typically, school discipline is thought of asa racialized process only when disciplinary actions disproportionately tar-get students of color, which research suggests is often the case (Arum,2003; Noguera, 1995). However, this study investigates whether strict dis-ciplinary climates are detrimental to the realization of academic poten-tial for African American students, regardless of any racially

    5

  • Teachers College Record, 114, 070303 (2012)

    disproportionate delivery of punishments.1 As I will discuss, this couldoccur because school disciplinary problems may be read through a racial-ized lens by students, teachers, and administrators. One purpose of theschool-level analysis, then, is to test the possibility that a school character-istic without immediately obvious racial implications can harm AfricanAmerican students academic progress relative to White studentsprogress to the same extent as (or more than) a racialized academicopportunity structure within the school.Research indicates that African American students are more likely than

    Whites to attend schools that focus on behavioral compliance anddeportment (Carter, 2003; Kelly, 2010; Tyson, 2003). Thus, disciplinaryclimate may affect the extent to which African American students fulfilltheir academic potential relative to Whites because African American stu-dents are more likely to attend schools that focus more on disciplinethan, say, critical thinking. However, it could also be the case that AfricanAmerican students are affected to a greater extent by strict disciplinaryschool climates than are White students. In other words, the same schoolcharacteristic, disciplinary climate, may stymie African American stu-dents progress to a greater extent than for White students.African American students could be differentially affected by strict dis-

    ciplinary climates at school because their own behaviors are morestrongly affected by attending such schools or because school officials aremore likely to perceive African American students as problem studentsin such schools. In an ethnographic study of an elementary school,Ferguson (2000) found that African American students, particularlyAfrican American boys, tend to be perceived by teachers as troublemak-ers who may even harbor nefarious intentions. Ferguson argued thatAfrican American students transgressions are made to take on a sinister,intentional, fully conscious tone that is stripped of any element of child-ish navet (p. 83).Because school violence and disciplinary problems are often thought

    of as urban (read Black) problems, harsh disciplinary environmentslikely carry a more consequential symbolic meaning for AfricanAmerican than White students (Noguera, 1995). These racialized sym-bolic meanings mean that all African American students, not just stu-dents who find themselves targets of disciplinary actions, may benegatively affected by strict disciplinary climates. Thus, the analysis willhold constant the number of times that each student has been the targetof disciplinary actions by the school.

    Hypothesis 3: Strict disciplinary climates are especially harmful to therealization of academic potential for African American students.

    6

  • TCR, 114, 070303 Racial Differences

    There is also reason to think that a strict disciplinary climate has alarger negative effect on the realization of academic potential for AfricanAmerican students in integrated schools than segregated schools. Thefrog pond effect, which occurs when equally capable students rate theirown academic abilities as higher in schools with few high achievers andlower in schools with many high achievers (Davis 1966), indicates thatstudents compare themselves with an immediate reference group with-out accounting for the relative standing of that reference group in thelarger population (Zell & Alicke, 2009). In integrated schools, AfricanAmerican students will assess their own capabilities relative not only tosame-race peers, but also to White peers. If school administrators andteachers treat White students as better school citizens than AfricanAmerican students in integrated schools with strict disciplinary climates,then African American students may downgrade their subjective assess-ments of their own academic potential in a way that they would not insegregated schools. In segregated schools with strict disciplinary climates,there is less opportunity to make direct racial comparisons between stu-dents. Thus, African American students in segregated schools may ratetheir own potential as higher than do their peers in integrated schoolswith similar disciplinary climates, even when comparing students ofequal academic aptitude.

    Hypothesis 4: Strict disciplinary climates will negatively affect the real-ization of academic potential for African American students to a greaterextent in integrated than segregated schools.

    Another purpose of the school-level analysis is to explore the racial gapin academic achievement in integrated high schools, in particular. AsMuller, Riegle-Crumb, Schiller, Wilkinson, and Frank (2010) argued,school integration is not a panacea for racial inequality in academic out-comes. Academic opportunity structures that resegregate students withinintegrated schools pose a direct threat to African American students aca-demic progress. For instance, Kelly (2009) found that African Americanstudents are less likely to take advanced math courses in integrated andmajority-White schools. Analyzing a sample of integrated high schools,Muller and her colleagues (2010) found that in schools with an unequalacademic opportunity structure (as measured by the overrepresentationof White students in advanced math courses), African American studentsearn lower grade point averages (GPAs) on average, net of studentsprior achievement and course placement. This analysis tests for a similareffect of an unequal opportunity structure on the racial gap in the real-ization of academic potential. The analysis will hold constant each

    7

  • Teachers College Record, 114, 070303 (2012)

    students own course-taking to try to isolate the effect of a school-levelracialized academic opportunity structure from potential effects of stu-dents own course-taking patterns.Once an array of individual characteristics is held constant, why would

    African American students academic development be stymied in schoolswith large racial AP course gaps? Large racial AP course gaps may serveas a contextual trigger of stereotype threat. Stereotype threat occurswhen a person feels at risk of confirming a negative stereotype aboutones social group. Once threatened by the stereotype, the persons per-formance in the relevant domain suffers (Steele, 1997; Steele & Aronson,1995). As Steele wrote, it is important to keep in mind that students sus-ceptibility to [stereotype] threat derives not from internal doubts abouttheir ability (e.g., their internalization of the stereotype) but from theiridentification with the domain and the resulting concern they have aboutbeing stereotyped in it (p. 614). We can begin to see why, regardless ofstudents own course-taking patterns, an overall racialized AP course-tak-ing regime within the school could lead African American students to fallshort of realizing their academic potential.

    Hypothesis 5: Racial inequality in access to AP courses negativelyaffects the realization of academic potential for African American students.

    DATA AND SAMPLE

    The analyses used data from the Education Longitudinal Study of 2002,conducted by the National Center for Education Statistics (NCES). ELSis a nationally representative sample of high school students, who werefirst surveyed in 2002, when all the students were sophomores. The firstfollow-up wave of data collection occurred in 2004, when most studentswere seniors in high school. ELS used a two-stage sampling process. Inthe first stage, public and private high schools were sampled using a strat-ified probability proportional to size sampling methodology, yielding 752high schools. In stage 2, approximately 26 tenth-grade students fromeach school were sampled, yielding a sample size of 15,362 sophomores.Students were resurveyed in 2004, yielding a sample size of 12,427 stu-dents (Ingels, Pratt, Rogers, Siegel, & Stutts, 2005). The student-leveldata consist of student surveys, students transcript records, studentsscores on reading and math assessments conducted by ELS, and surveysof the students teachers and parents. In addition to student-level data,ELS includes school-level data, consisting of a school administrator sur-vey, a facilities checklist, and a librarian survey.

    8

  • TCR, 114, 070303 Racial Differences

    The analyses used the 10th- and 12th-grade waves of ELS. I limited thesample to self-identified African American and White students. The stu-dent sample size is 9,680, and the school sample size is 670.2 I used mul-tiple imputation to replace missing data on independent variablesbecause listwise deletion would have reduced the sample to 5,380.3 Allparameter estimates are averages of each of the parameters from the fivemultiply imputed data sets. Standard errors were computed by using theaverage of the squared standard errors over the set of analyses and thebetween-analysis parameter estimate variation, which accounts for theuncertainty introduced by missing data (Allison, 2002; Rubin, 1987).

    MEASURES

    DEPENDENT VARIABLE: UNREALIZED ACADEMIC POTENTIAL

    The dependent variableunrealized academic potentialmeasures theextent to which students senior-year grades were consistent with thegrades predicted by their academic skills, as measured in 10th grade. Themeasure was constructed in two steps, as presented in Table 1. First, anordinary least squares (OLS) equation was used to predict students 12th-grade GPAs across academic courses (taken from students official tran-scripts). Composite item response theory (IRT) scores on the math andreading tests administered by ELS in 10th grade were used to predictGPA. IRT scores are a particularly good measure for investigating unreal-ized academic potential because the score represents the students mas-tery of the material rather than how well the student performed relativeto other students, as with norm-referenced scores (Ingels et al., 2005).The tests conducted for ELS were designed to test curriculum-relatedmath and reading skills rather than general aptitude (Ingels, Pratt,Rogers, Siegel, & Stutts, 2004). The reliabilities were .92 for the math IRTand .86 for the reading IRT (Ingels et al., 2005). The test score coeffi-cient from the regression equation was used to predict the 12th-gradeGPA for each student in the sample.In the second step, I subtracted each students predicted GPA from his

    or her actual GPA, as reported on the students high school transcript.This difference yielded the residual GPA. Negative values indicate that astudent earned a GPA that was lower than his or her composite test scorepredicted, positive values indicate that a student earned a higher GPAthan predicted, and a value of 0 indicates that a student earned the exactGPA predicted by his or her test score.Some have argued that as measures of academic ability, test scores and

    grades are riddled with error, which can distort regression results in a

    9

  • Teachers College Record, 114, 070303 (2012)

    way that produces overprediction [lower actual than predicted grades]among lower-scoring groups and underprediction [higher actual thanpredicted grades] among higher-scoring groups (Zwick & Sklar, 2005,pp. 442443). However, as Zwick and Sklar (2005) noted, women tend toearn higher grades in college than predicted by their SAT scores eventhough as a group, they score lower on the SAT than do men. In addi-tion, as mentioned, the ELS IRT scores are highly reliable (.92 for math;.86 for reading).4 Thus, the overprediction and underprediction ofAfrican American students and White students grades, respectively,found here are likely not the result of the mean differences in IRT scoresacross the two groups.

    STUDENT-LEVEL VARIABLES

    Table 2 presents detailed descriptions and descriptive statistics for all stu-dent-level independent variables used in the analyses. Student-level vari-ables include measures of family background, school attitudes, schoolbehaviors, and teachers perceptions of the students effort. Several con-trol variables are also measured.

    SCHOOL-LEVEL VARIABLES

    The focal variables in the school-level analyses are disciplinary climateand the racial AP course gap. All school-level analyses controlled forschool sector, urbanicity, average studentteacher ratio, and the racialand social class composition of the student body. Table 3 describes theschool-level variables used in the analyses.

    10

    Table 1. Construction of the Dependent Variable: Realization of Academic Potential

    Note: African American residual GPA is not exactly equal to actual GPA predicted GPA because of round-ing.

  • TCR, 114, 070303 Racial Differences

    11

    Table 2. Means, Standard Deviations, and Descriptions for Student-Level Independent Variables Used inAnalyses

    Variable name Description Metric Mean SD Alpha

    Dependent Variable:Realization ofAcademic Potential

    Actual GPA minus predicted GPA (See textand Table 1 for detailed description ofconstruction of this variable.)

    -3.054 = highest degree of unrealized potential;2.313 = highest degree of exceeded potential

    0 .700

    School Attitudes

    Importance ofschool

    Standardized mean of students evaluationof importance of education for getting a joblater on (14), a good education (13), andgood grades (14) (10th-grade survey)

    -5.013 = loweststandardized score;1.843 = higheststandardized score

    0 1.000 .725

    Educational expectations

    Number of years of education studentexpects to complete (10th-grade survey)

    11 = not finish high school;20 = PhD or MD

    16.623 2.190

    Student-ReportedBehaviors

    Hours homework

    Number of hours spent on homeworkoutside of school per week (10th-gradesurvey)

    026 5.579 5.870

    Disciplinaryactions

    Standardized mean of students report ofhow many times student got in trouble fornot following school rules, was suspendedor put on probation, and put on in-schoolsuspension (10th-grade survey)

    -2.944 = loweststandardized score;10.271 = higheststandardized score

    0 1.000 .585

    Teachersevaluations of classroomeffort

    Standardized mean of teachers responsesto how often the student works hard in class(15), how often the student completeshomework (15), and whether the studentworks hard for good grades (0, 1) (Eachitem is the average response for the 10th-grade English and math teachers.)

    -3.783 = loweststandardized score;3.076 = higheststandardized score

    0 1.000 .874

    Family Background

    Parent education Highest number of years of educationcompleted by either of the students parents(10th-grade survey)

    11 = did not graduatehigh school;20 = PhD or MD

    15.035 2.370

    nl family income

    Students total family income (10th-gradeparent survey)

    012.206 10.759 1.020

    Two-parent family

    Student resides with two parents (10th-grade survey)

    0 = no; 1 = yes .760 .420

    Additional VariablesAcademic courses Number of Carnegie units in academic

    subjects (.5 is equivalent to a coursemeeting one class period per day week forone semester.) (transcript)

    030.973 17.929 4.520

    Female Student is female (10th-grade survey) 0 = no; 1 = yes .495 .500

    African American Student identifies as African American(10th-grade survey)

    0 = no; 1 = yes .181 .380

  • Teachers College Record, 114, 070303 (2012)

    ANALYTIC METHOD

    To properly model potential effects of school-level characteristics onracial mean differences in unrealized academic potential, I used hierar-chical linear modeling (HLM). All analyses used HLM6 software. HLMallowed me to examine, first, whether racial differences in unrealizedacademic potential vary significantly across high schools, and second,whether the school characteristics discussed earlier contribute to thesemean differences. Each analysis was performed on each of five multiply-imputed data sets, and parameters were averaged across the five data sets.The standard errors of fixed effects were not averaged, but estimatedwith regard to both sampling and measurement error (Raudenbush,Bryk, Cheong, & Congdon, 2004), which takes into account the uncer-tainty introduced by imputing missing data (Allison, 2002; Rubin, 1987).

    12

    Table 3. Means and Descriptions for School-Level Variables Used in Analyses, Standard Deviations inParentheses

    Description Metric Mean

    Urban School in urban area 0 = no; 1 = yes .310(.460)

    Private religious Private religious school 0 = no; 1 = yes .190(.390)

    Private secular Private secular school 0 = no; 1 = yes .040(.190)

    % free lunch Percent of students eligible for freelunch

    0%100% 26.840(20.980)

    Studentteacher ratio Ratio of number of students to full-time teachers

    6.50078.800 18.030(4.970)

    0%24% minority Student body is 0%24% minority 0 = no; 1 = yes .550(.500)

    25%49% minority Student body is 25%49% minority 0 = no; 1 = yes .210(.410)

    75%100% minority Student body is 75%100% percentminority

    0 = no; 1 = yes .140(.340)

    Disciplinary climate School mean of students reports offrequency of disciplinary actions(Disciplinary actions measure isdescribed in Table 2.)

    -.933 = loweststandardized score;3.167 = higheststandardized score

    -.170(.390)

    Racial AP course gap Proportion of White students enrolledin AP courses minus proportion ofBlack students enrolled in AP courses

    -1.0001.000 .493(1.133)

  • TCR, 114, 070303 Racial Differences

    There were two sets of analyses. The first set of analyses used the entiresample and proceeded in a series of models, in which sets of predictorvariables were progressively entered to isolate the contribution of eachset of variables to racial differences in unrealized academic potential. Ifany mean differences between African American and White students inunrealized academic potential (as indicated by the African Americancoefficient) diminished once sets of predictors were added to the equa-tion, this suggests that racial differences in these variables contributed toracial differences in unrealized academic potential. I examined changesin the African American coefficient to decide whether particular sets ofvariables helped explain racial differences in the predictive validity ofearlier high school test scores on later high school GPAs.5 The final level1 model (Model 6 in Table 4) estimated the coefficients for the student-level variables for each school j (Equation 1). The literature suggests cen-tering variables around the group mean when cross-level interactions areof interest (Bryk & Raudenbush, 2002; Enders & Tofighi, 2007). Thus,the African American coefficient was centered around its group mean inmodels testing cross-level interactions, which means that AfricanAmerican and White students were compared within the same schools.All other variables are centered around their grand means.

    Yij = 0j1j (African American) + 2j (Female) + 3j (Academic Courses) (1)+ 4j (Parent Education) + 5j (Natural Log of Family Income) + 6j (Two Parents)+ 7j (Importance of Education) + 8j (Educational Expectations)+ 9j (Time Spent on Homework) + 10j (Disciplinary Actions)+ 11j (Teacher-Reported Effort) + r1j

    The level 2 model in the first set of analyses predicted the relationshipsbetween school characteristics and the level of unrealized academicpotential for all students who had the school-level average for each of thelevel 1 variables (the intercept, 0j) and the level of unrealized academicpotential for African Americans relative to Whites (the African Americancoefficient, 1j). The measurement strategy for school racial compositionallowed for the possibility of nonlinear effects. Whereas a continuous per-cent minority variable would indicate whether increases in the percent-age of minority students in a school were associated with increases ordecreases in the realization of academic potential, the dummy variableapproach employed here allowed for the detection of negative effects ateither very high or low levels of minority student representation.6 Allschool-level variables were centered around their grand means.

    13

  • Teachers College Record, 114, 070303 (2012)

    0j = 00 + 01(Urban) + 02(Private) + 03(Percent Free Lunch) (2)+ 04(Student-Teacher Ratio) + 05(0-24% Minority) + 06(25-49% Minority)+ 07(75-100% Minority) + 07(Disciplinary Climate) + u0j

    1j = 10 + 11(Urban) + 12(Private) + 13(Percent Free Lunch) (3)+ 14(Student-Teacher Ratio) + 15(0-24% Minority) + 16(25-49% Minority)+ 17(75-100% Minority) + 18(Disciplinary Climate) + u1j

    The second set of analyses splits the sample into integrated and segre-gated schools. The subsample analyses focus on potential variations inschool-level effects across segregated and integrated schools. The level 1model in these analyses is the same as Equation 1, except that the num-ber of AP courses that students have taken was controlled in the inte-grated subsample analysis. The level 2 models are essentially the same asEquations 2 and 3. The racial AP course gap variable is included only inthe integrated schools analysis because inclusion of this variable wouldhave resulted in the loss of 276 schools from the full sample or segregatedschools sample.

    RESULTS

    Studies measuring unrealized potential as lowered educational expecta-tions or unrealized plans for educational attainment have found thatunrealized potential is more likely to exist among White students thanstudents of color (Hanson, 1994; Trusty, 2000; Trusty & Harris, 1999).Here, however, the mean residual GPA is -.163 for African American stu-dents and .076 for White students. Thus, African American students fallshort of their academic potential by .163 grade points, whereas Whitesexceed their potential by .076 grade points. Although .163 grade pointsmay seem inconsequential in absolute terms, in the world of selective col-lege admissions, admissions officers are charged with using very small dif-ferences between students to make decisions about who will be admitted.Thus, small quantitative differences can have important consequencesfor students.

    EXPLANATORY ANALYSES

    The analyses for the full sample focus on two questions: (1) Which student-level characteristics help explain racial differences in the fulfillment of academic potential? (2) Which school-level characteristics help explainvariation across high schools in this racial difference? Table 4 presents theresults of the HLM models for the full sample. Models 2 through 5 add one

    14

  • TCR, 114, 070303 Racial Differences

    or more student-level variables. The intercept and the African Americancoefficient are allowed to vary across schools. Model 6 adds school-levelvariables to the equations for the intercept (level of realized academicpotential for all students who have the school-level mean for each level 1variable) and African American coefficient (level of realized academicpotential for African American students relative to White students).

    Note: For Models 1 through 5, all level 1 predictors are centered around their grand means, except for theAfrican American variable. For Model 6, the African American variable is centered around its group meanto properly model the cross-level interaction; all other variables are centered around their grand means.ICC for the dependent variable = .1077. nl = natural log.*p < .05. **p < .01. ***p < .001.

    15

    Table 4. HLMs Predicting Realization of Academic Potential, Standard Errors in Parentheses

    Model 1 Model 2 Model 3 Model 4 Model 5 Model 6Intercept .035* (.014) -.075*** (.017) -.137*** (.026)-.118*** (.025) -.097*** (.025) .017 (.012)Variance .046*** .057*** .056*** .054*** .047*** .044***

    Urban -.092** (.028)Private religious -.202** (.048)Private secular .032 (.070)% receiving free lunch .005*** (.001)

    Studentteacher ratio .007* (.003)0%24 % minority .110** (.040)

    25%49 % minority .046 (.044)75%100 % minority -.186** (.058)Disciplinary climate -.035 (.047)

    African American x _____Intercept -.225*** (.031) -.134*** (.031) -.112*** (.032)-.123*** (.032) -.071* (.031) -.053 (.035)Variance .101*** .097*** .096*** .086*** .082*** .074***Urban .009 (.076)Private religious .234* (.115)Private secular -.005 (.119)% receiving free lunch -.004 (.002)

    Studentteacher ratio -.009 (.009)0%24 % minority -.235* (.098)

    25%49 % minority -.102 (.092)75%100 % minority -.402** (.141)Disciplinary climate -.324*** (.082)

    Female .232*** (.016) .234*** (.016) .206*** (.016) .169*** (.016) .166*** (.016)Academic courses .048*** (.002) .047*** (.002) .038*** (.002) .027*** (.003) .027*** (.002)Family backgroundParent education .000 (.004) .002 (.004) -.001 (.004) .001 (.004)nl family income .006 (.010) .007 (.010) .004 (.010) .006 (.010)Two parents .075** (.024) .067** (.025) .058* (.023) .052* (.023)

    Student-reported behaviors and attitudesImportance of education .066*** (.010) .032*** (.010) .031** (.009)Educational expectations -.020*** (.005) -.024*** (.004) -.023*** (.004)Time spent on homework -.001 (.001) -.002 (.001) -.001 (.001)Disciplinary actions -.090*** (.013) -.038* (.013) -.038** (.013)

    Teacher-reported effort .196*** (.011) .195*** (.011)BIC 20559.951 19146.472 19147.178 18902.616 18301.956 18173.242

  • Teachers College Record, 114, 070303 (2012)

    Model 1 serves as a baseline for comparing the size of the racial differ-ence in realization of academic potential with successive models. On aver-age, African American students GPAs are .225 points lower than predictedrelative to Whites. This racial difference varies significantly across highschools, as indicated by the significant variance of the African Americancoefficient. Within one standard deviation (.317) the African Americancoefficient ranges from -.542 to .092. Once gender and the number of aca-demic courses completed by students are held constant in Model 2, theAfrican American coefficient falls to -.134. Socioeconomic and familystructure differences between White and African American studentsexplain about 16% of the racial difference in fulfillment of academicpotential, as the African American coefficient falls to -.112 in Model 3.Model 4 tests Hypothesis 1that, taken together, students self-

    reported school behaviors and attitudes account for none of the racialdifference in the realization of academic potential. However, the addi-tion of these variables leads to a reduction of the African American coef-ficient by about 14%. Thus, mean differences between White and AfricanAmerican students on school attitudes and behaviors do account for amodest portion of the racial difference in the realization of academicpotential. Interestingly, the coefficient for educational expectations isnegative, indicating that as educational expectations increase, studentsare less likely to reach their academic potential.Model 5 introduces teachers perceptions of students effort in class.

    The addition of this variable accounts for the largest drop (42%) in racialdisparity in realizing academic potential. This result is striking becausedifferences in students own reports of their conduct, time spent onhomework, and school attitudes are held constant. Hypothesis 2, thatteachers perceptions of students matter independent of students self-reported behaviors, thus receives support.Model 6 introduces school-level variables. The coefficients for the

    intercept equation indicate how these school characteristics are relatedto the realization of academic potential for students who have the school-level average for each of the level 1 variables. On average, students in75%100% minority schools fall short of their predicted GPAs by .186grade points. However, in schools that are 76%100% White, studentswith the school-level average for all level 1 variables are predicted toexceed their academic potential by .11 grade points. Thus, predomi-nantly minority schools appear to dampen the extent to which bothWhite and African American students reach their academic potential,whereas at predominantly White schools, both groups of students arepredicted to exceed their potential. The coefficient for disciplinary climate is close to zero and not significant, suggesting that it is unrelated,

    16

  • TCR, 114, 070303 Racial Differences

    on average, to the realization of academic potential.Although the African American coefficient (-.05) is not statistically sig-

    nificant, indicating that, on average, there is not a significant racial dif-ference in realization of academic potential across these schools, thereare several significant cross-level interactions. The cross-level interactionsfor the African American coefficient show how school-level characteris-tics affect the racial gap in the realization of academic potential. Negativevalues indicate that the school-level variable is associated with largerracial differences in the realization of academic potential, with AfricanAmerican students being less likely to fulfill their academic potential rel-ative to White students. Although disciplinary climate was unrelated tothe realization of academic potential for students with school-level aver-ages for the level 1 variables, the negative cross-level interaction (-.324)indicates that the racial disparity in the realization of academic potentialincreases as the school disciplinary climate becomes increasingly puni-tive. In other words, African American students are less likely to realizetheir academic potential relative to White students in schools with strictdisciplinary climates. Converting the coefficient into an effect size givesan idea of the substantive contribution of strict disciplinary climates tothe racial gap in the realization of academic potential. The effect size is.7, meaning that the racial gap in the realization of academic potential is.7 standard deviation units higher in schools that are one standard devi-ation above the sample mean for the disciplinary climate variable as com-pared with high schools that are at the sample mean. Thus, the datastrongly support Hypothesis 3.Figure 1 shows predicted values for the realization of academic poten-

    17

    Figure 1. Predicted fulfillment of academic potential for African American Students, by school disciplinaryclimate

    Note: All predictors are held at their sample grand means. Negative values for the fulfillment of academicpotential indicate lost potential; positive values indicate exceeded potential.

  • Teachers College Record, 114, 070303 (2012)

    tial for African American students as the disciplinary climate becomesincreasingly punitive. All else equal, African American students areexpected to fulfill their academic potential at schools that hover aroundthe average school disciplinary climate, but at schools that are morepunitive than the average school, African American students are pre-dicted to fall short of their academic potential.The cross-level interactions for the racial composition dummy variables

    show that the advantages of attending predominantly White schoolsaccrue to White students to a greater extent than to their AfricanAmerican peers, and the disadvantages of attending predominantlyminority schools accrue to a greater extent to African Americans thanWhites.7 The racial gap in the realization of academic potential at pre-dominantly minority schools is .402 grade points (.650 standard deviationunits), and at predominantly White schools, .235 grade points (.411 stan-dard deviation units). These results could stem from two phenomena:(1) internal stratification processes that privilege Whites relative toAfrican Americans in mostly White schools or (2) the burden of beingone of the few African American students in a predominantly Whiteschool, which could stymie African American students ability to achieveto their fullest potential.

    SPLIT SAMPLE ANALYSES: SEGREGATED AND INTEGRATED HIGHSCHOOLS

    The second set of analyses takes Muller and colleagues (2010) point thatit is crucial that we examine variations among racially diverse schools(p. 1058) by splitting the sample by the level of school segregation. Theintegrated subsample includes schools that are 25%74% minority,because the full sample analysis revealed no significant differencesbetween 25%50% minority schools and 51%74% minority schools inthe realization of academic potential for African Americans or Whites.The results of the full sample analysis also showed that AfricanAmericans realization of academic potential was harmed relative toWhites in schools that were mostly White or mostly non-White. Thus, thesegregated school sample includes schools that are either 0%24% or75%100% minority. The focus in these analyses is, first, to test whetherdisciplinary climate affects the racial gap in the realization of academicpotential to a greater extent in integrated schools than segregatedschools (Hypothesis 4), and second, to test whether racialized opportu-nity structures within integrated schools affect the racial gap in the real-ization of academic potential (Hypothesis 5). Results are presented inTable 5.

    18

  • TCR, 114, 070303 Racial Differences

    19

    Table 5. HLM Predicting Realization of Academic Potential by Percent of Minority Students in School,Standard Errors in Parentheses

    Note: The same level 1 coefficients were estimated as those shown in Equation 1. The number of AP coursestaken by students was added as a level 1 variable in the integrated sample analysis to control for studentsown AP course-taking once the school-level racial AP course gap variable was added. All variables are cen-tered around their grand means, except for the African American variable, which is centered around itsgroup mean. For the segregated sample, Nstudents =7370; Nschools = 500. For the integrated sample,Nstudents =2310; Nschools =170.a Racial AP course gap was excluded from the segregated sample analysis because it would have resulted inthe loss of 276 schools from the subsample.b The African American coefficient is constrained to be equal across schools in the segregated samplebecause the variance was not significant. The BIC for the model with the fixed African American coeffi-cients was significantly lower than the BIC for the model in which the coefficient was allowed to vary, indi-cating that the fixed-coefficient model was a better fit.*p < .05. **p < .01. ***p < .001.

  • Teachers College Record, 114, 070303 (2012)

    The results are consistent with Hypothesis 4. Disciplinary climate has anegative effect on the realization of academic potential for AfricanAmerican students only in integrated schools.8 These results imply that astrict disciplinary climate takes on a racialized symbolic meaning partic-ularly when there enough African American and White students presentto make racial comparisons. In such contexts, students, school officials,or both may come to understand the disciplinary climate in racializedterms.The cross-level interaction between the African American coefficient

    and the measure of inequality in opportunities for AP coursework is notsignificant. The racial AP course gap thus does not negatively affect therealization of academic potential for African American students, asHypothesis 5 predicted. The absence of this predicted relationship couldreflect that AP racial enrollment gaps align closely with racial differencesin mean test scores. Prior research suggests that African American stu-dents are more likely to take advanced courses once prior achievement iscontrolled (Attewell, 2001; Gamoran & Mare, 1989). If AfricanAmericans with high test scores are just as likely (or more likely) to takeAP courses as high-scoring Whites, all else equal, then we might notexpect large WhiteBlack gaps in AP enrollment to affect the realizationof academic potential. They may instead reflect large racial differences inmean test scores prior to 11th grade, when most students first have thechance to enroll in AP courses.

    DISCUSSION

    Researchers should make no mistake about the importance of examiningracial differences in high school grades alongside racial differences intest scores. As an illustration, I divided the full sample analyzed hereaccording to whether students had enrolled in a four-year college or uni-versity by 2006, two years after high school. For African American stu-dents who had enrolled in a four-year college, there is virtually nodifference in predicted and actual 12th-grade GPAs, but AfricanAmerican students who had not enrolled in college fell short of their pre-dicted GPAs by .34 grade points on average. White students who had notenrolled in college had earned grades that were just .08 points lower thanexpected. This underscores the role that supporting African Americanstudents in reaching their GPA potential could play in clearing the path-way from high school to college.The results have yielded several notable findings regarding the extent

    of and explanations for African American students unrealized academicpotential. First, African American 12th graders are more likely than

    20

  • TCR, 114, 070303 Racial Differences

    White 12th graders to earn grades that are lower than their 10th-gradescores on math and reading tests predicted. The high reliabilities of thetests bolster the conclusion that many African American students areearning grades, an increasingly important criterion for college admis-sions, that are not commensurate with their actual academic abilities.The biggest reason for this unrealized stock of academic potential at

    the student level is that, on average, teachers perceive White students asexerting more effort and conforming more to classroom expectations ascompared with African American students. The analyses controlled forseveral student-reported school behaviors and attitudes, which suggestssome level of disconnect between teachers perceptions of students andstudents reports of what they actually do and believe.These data do not allow for an empirical exploration of why teachers

    would rate African American students classroom effort as lower thanthat of White students who reported the same amount of time spent onhomework, frequency of disciplinary actions, and attitudes about school.Given that the number of times that students were formally disciplinedby the school was held constant, it is unlikely that the teachers percep-tions effect reflects large differences in students actual classroom behav-iors. However, smaller, nuanced differences in the ways that studentspositively engage with school could have amplified ramifications. Inother words, even when comparing African American and White studentswho spend the same amount of time on homework, behave equally wellat school, and value education equally, it could be that White studentsstill are better able to play the game of school. White students, forinstance, may be more comfortable interacting with their teachers, whichteachers may perceive as stronger engagement. In addition, as discussedearlier in the article, teachers may interpret classroom behaviors that donot comply with White cultural norms as misbehavior.Downey and Pribesh (2004) found that African American students

    received more favorable evaluations of classroom effort from AfricanAmerican teachers than from White teachers. Although the measure ofteachers perceptions of student effort used in this study is an average oftwo teachers evaluations, I conducted an auxiliary analysis focusing onwhether students were paired with at least one same-race teacher. If theteacher perceptions effect found in this study were due to a cultural mis-match between White teachers and African American students, then con-trolling for student-teacher racial congruence should have decreased theextent to which teachers perceptions accounted for the racial gap in therealization of academic potential. However, the results showed thataccounting for studentteacher racial congruence did not affect the rolethat teachers perceptions played in the racial gap in the realization of

    21

  • Teachers College Record, 114, 070303 (2012)

    academic potential. Although Downey and Pribesh found potentiallynegative effects of studentteacher racial incongruity for AfricanAmerican kindergarteners and eighth graders, these results suggest thatAfrican American high school students are not as positively perceived astheir White peers, regardless of racial congruity with teachers. It could bethat social class mismatches between students and teachers, rather thanracial incongruity alone, would help explain why teachers perceptions ofstudents contribute such a great deal to the racial gap in the realizationof academic potential. The second set of findings center on the role of school characteristics

    in African American students unrealized academic potential. The resultsshow that attending segregated high schools that are 75%100% minor-ity is particularly detrimental to the realization of full academic potentialfor African American students. These results are consistent with paststudies that have found negative effects of attending predominantlyminority schools on African American students academic achievement(e.g., Bankston & Caldas, 1996; Mickelson & Heath, 1999). This studyalso indicates, however, that although African American students aremore likely to reach their academic potential in predominantly Whiteschools than predominantly minority schools, there still exists aWhiteBlack racial gap in the realization of academic potential in major-ity-White schools. This finding is consistent with Goldsmith (2004),whose study found that African American students in predominantlyWhite schools are less optimistic and hold less positive attitudes towardeducation than their same-race peers in predominantly minority schools.Simply sending African American students to predominantly Whitesschools, then, does not guarantee that they will see themselves as fullyenfranchised students or benefit from the environment to the sameextent as their White peers.Integrated high schools with strict disciplinary climates have wider

    racial gaps in the realization of academic potential. One interpretation ofthis result is that African American students feel more alienated by strictdisciplinary climates than do Whites because racial meanings areascribed to the disciplinary climate. Even when African American stu-dents are not themselves the targets of disciplinary actions, attending aschool like this could affect African American students concentrationand effort, thus making it more difficult to fulfill academic potential. Itis important to note that African American students in schools with harshdisciplinary climates are not likely developing resistance stances towardeducation, because this effect of strict disciplinary climates exists control-ling for students own behaviors and attitudes. A second interpretation of the negative effect of strict disciplinary climates on the racial gap in

    22

  • TCR, 114, 070303 Racial Differences

    fulfilling academic potential is that school administrators and teachersperceptions of African American students are negatively affected by thedisciplinary climate, regardless of African American students actualbehaviors and attitudes.

    POLICY RECOMMENDATIONS

    Programs that focus on relationships between teachers and AfricanAmerican students seem a promising avenue for increasing the fulfill-ment of African American students potential, given that once teachersperceptions of student effort were held constant, African Americanunderfulfillment of potential relative to Whites narrowed by 42%. This isespecially important considering the evidence that African American stu-dents are more strongly affected by their relationships with teachers thanare White students (Jussim, Eccles, & Madon, 1996). Although teachersperceptions are not totally inaccurate, neither are they perfect reflectionsof students actual effort (Jussim et al., 1996). Some of the noise inmeasures of teachers perceptions of African American students effortmay stem from the use of a seemingly neutral lens that is actually cali-brated by the expectations of White middle-class culture.At the school level, redoubling our efforts to fight the resegregation of

    Americas schools is an important piece of making sure that AfricanAmerican students realize their academic potential. Desegregating pre-dominantly minority schools, in particular, is an important goal. Theresults showed that both African American and White students fall shortof their academic potential, on average, at schools that are 75%100%minority. However, African American students fall short of their acade-mic potential to a greater extent than Whites in these schools. In fact, theracial disparity in realization of academic potential is quite large at theseschools, at .65 standard deviation units. But the results suggest that inte-gration efforts should target predominantly White schools as well.Although on average, students at predominantly White schools areexpected to exceed their academic potential by a small amount (.11grade points), there is still a sizeable racial gap in the extent to which stu-dents exceed their academic potential (.411 standard deviation units). High schools that are racially integrated must actively work to ensure

    that African American students are not disadvantaged within integratedschools, either through the internal resegregation of students or by fos-tering school climates that are particularly nonconducive to AfricanAmerican students reaching their full academic potential. Several recentstudies have drawn attention to the pernicious effects of segregatedcourse-taking regimes within integrated schools (Mickelson & Velasco,

    23

  • Teachers College Record, 114, 070303 (2012)

    2006; Muller et al., 2010; Tyson, 2006; Tyson, Darity, & Castellino, 2005).This study reveals that characteristics of schools that lack immediatelyobvious racial implications, such as a schools approach to student disci-pline, may be just as harmful as overtly racialized inequality withinschools. Administrators in integrated schools need to recognize that strictdisciplinary climates can be detrimental to the fulfillment of academicpotential for African American students, regardless of students individ-ual experiences of disciplinary actions. The results showed that forschools that are one standard deviation above the sample mean forschool disciplinary practices, the WhiteBlack gap in the realization ofacademic potential is .7 standard deviations, a large effect size by anystandard. Thus, the perceived benefits of strict disciplinary policies needto be weighed very carefully against the drawbacks of such policies forAfrican American students. A promising line for future research wouldbe to identify the specific aspects of punitive disciplinary approaches thatdisadvantage African American students.

    Acknowledgments

    The author would like to thank Jennifer Glanville, Mary Campbell, Mary Noonan, Kevin Leicht, andDavid Bills for their feedback on an earlier version of this article.

    Notes

    1. In this study, I operationalized school disciplinary climate as the average number ofdisciplinary actions administered by the school. See Table 3 for a more detailed description.

    2. The National Center for Education Statistics (NCES) requires all sample sizes to berounded to the nearest ten to protect anonymity.

    3. I used NORM (Schafer, 2000) to create five data sets with randomly drawn imputedvalues for the missing cases.

    4. Unfortunately, ELS does not report test reliabilities by racial groups.5. The psychometric literature has a long tradition of predictive validity analyses.

    Although most of these studies focus on describing predictive validity rather than explain-ing it, a few studies do investigate explanations for group differences in predictive validity.For example, Stricker, Rock, and Burton (1993) investigated explanations for the differen-tial prediction of college grades for men and women. To determine what accounted for thisunderprediction, they checked whether the squared semipartial correlation (sr2) of genderwith residual college GPA decreased significantly as other independent variables are addedto the equation. This method of comparing correlations, however, is limited when attempt-ing to detect mediations. As Kenny (2009) has noted, there are situations in which examin-ing changes in correlations may suggest that mediation has occurred, when in fact there isno mediation. For example, even if the proposed mediator has no effect on the outcome,the partial correlation will still decline, leading to the erroneous conclusion of mediation.Kenny therefore recommended checking for mediation not by examining changes in cor-relations, but by examining changes in coefficients.

    6. Although the percent of African American students would have been preferable to

    24

  • TCR, 114, 070303 Racial Differences

    the percent of minority students, ELS does not include school-level data on the percentageof specific minority groups by school. Aggregated student data would not have been a reli-able measure of the true representation of African American students in a given schoolgiven the within-school sample sizes. Alternative versions of all models were estimated usingthe continuous percent minority measure. Use of this alternative measure did not substan-tively affect other model parameters.

    7. Modeling percent minority as a continuous variable would have masked this rela-tionship between school racial composition and realization of academic potential forAfrican American students. In an auxiliary analysis, a cross-level interaction betweenAfrican American and a continuous percent minority variable did not reach statistical sig-nificance (b = .002; SE = .001), though it was very close. Additionally, Schwarzs Bayesianinformation criterion (BIC) for this auxiliary model was more than 10 points higher thanthe BIC for Model 6, indicating a better model fit when measuring racial composition asdummy variables.

    8. To check the sensitivity of these results, I ran an auxiliary subsample analysis withfour subsamples: 0%24% minority, 25%49% minority, 50%74% minority, and75%100% minority. The pattern of results was consistent with the split sample analysesshown in Table 5. School disciplinary climate had a significant negative effect on theAfrican American coefficient only in the 25%49% and 50%74% minority subsamples.

    References

    Ainsworth-Darnell, J., & Downey, D. (1998). Assessing the oppositional culture explanationfor racial/ethnic differences in school performance. American Sociological Review, 63,536553.

    Allison, P. D. (2002). Missing data, series in quantitative applications in the social sciences.Thousand Oaks, CA: Sage.

    Arum, R. (2003). Judging school discipline: The crisis of moral authority. Cambridge, MA:Harvard University Press.

    Attewell, P. (2001). The winner-take-all high school: Organizational adaptations to educa-tional stratification. Sociology of Education, 74, 267295.

    Bankston, C., & Caldas, C. (1996). Majority African American schools and social injustice:The influence of de facto segregation on academic achievement. Social Forces, 75,535552.

    Bourdieu, P. (1977). Cultural reproduction and social reproduction. In J. Karabel & H. A.Halsey (Eds.), Power and ideology in education (pp. 487511). New York: Oxford UniversityPress.

    Bourdieu, P., & Passeron, J.-C. (1990). Reproduction in education, society, and culture (2nd ed.).London: Sage.

    Bryk, A. S., & Raudenbush, S. W. (2002). Hierarchical linear models: Applications and dataanalysis methods (2nd ed.). Thousand Oaks, CA: Sage.

    Carter, P. L. (2003). Black cultural capital, status positioning and schooling conflicts for low-income African American youth. Social Problems, 50, 136155.

    Davis, J. A. (1966). The campus as a frog pond: An application of the theory of relativedeprivation to career decisions of college men. American Journal of Sociology, 72, 1731.

    Downey, D. B., von Hippel, P. T., & Broh, B. A. (2004). Are schools the great equalizer?Cognitive inequality during the summer months and the school year. American SociologicalReview, 69, 613635.

    25

  • Teachers College Record, 114, 070303 (2012)

    Downey, D. B., & Pribesh, S. (2004). When race matters: Teachers evaluations of studentsclassroom behavior. Sociology of Education, 77, 267282.

    Enders, C. K., & Tofighi, D. (2007). Centering predictor variables in cross-sectional multi-level models: A new look at an old issue. Psychological Methods, 12, 121138.

    Ferguson, A. A. (2000). Bad boys: Public schools in the making of Black masculinity. Ann Arbor:University of Michigan Press.

    Gamoran, A., & Mare, R. D. (1989). Secondary school tracking and educational inequality:Compensation, reinforcement, or neutrality? American Journal of Sociology, 94, 11461183.

    Goldsmith, P. A. (2004). Schools racial mix, students optimism, and the Black-White andLatino-White achievement gaps. Sociology of Education, 77, 121147.

    Hanson, S. L. (1994). Unrealized educational aspirations and expectations among U.S.youths. Sociology of Education, 67, 159183.

    Ingels, S. J., Pratt, D. J., Rogers, J. E., Siegel, P. H., & Stutts, E. S. (2004). EducationLongitudinal Study of 2002: Base-year data file users manual (NCES 2004-405). U.S.Department of Education. Washington, DC: National Center for Education Statistics.

    Ingels, S. J., Pratt, D. J., Rogers, J. E., Siegel, P. H., & Stutts, E. S. (2005). EducationLongitudinal Study of 2002: Base-year to first follow-up data documentation (NCES 2006-344).U.S. Department of Education. Washington, DC: National Center for EducationStatistics.

    Jencks, C., & Phillips, M. (1998). The Black-White test score gap: An introduction. In C.Jencks & M. Phillips (Eds.), The Black-White test score gap (pp. 151). Washington, DC:Brookings Institution Press.

    Jussim, S., Eccles, J., & Madon, L. (1996). Social perception, social stereotypes, and teacherexpectations: Accuracy and the quest for the powerful self-fulfilling prophecy. Advancesin Experimental Social Psychology, 28, 281388.

    Kelly, S. (2009). The Black-White gap in mathematics coursetaking. Sociology of Education,82, 4769.

    Kelly, S. (2010). A crisis of authority in predominantly Black schools? Teachers CollegeRecord, 112, 12471274.

    Kenny, D. (2009). Mediation. Retrieved from http://davidakenny.net/cm/mediate.htmLewis, A. E. (2003). Race in the school yard: Negotiating the color line in classrooms and communi-

    ties. New Brunswick, NJ: Rutgers University Press.Magnuson, K., & Waldfogel, J. (2008). Steady gains and stalled progress: Inequality and the Black-

    White test score gap. New York: Russell Sage Foundation.Mickelson, R. A., & Heath, D. (1999). The effects of segregation on African American high

    school seniors academic achievement. Journal of Negro Education, 68, 566586.Mickelson, R. A., & Velasco, A. E. (2006). Bring it on! Diverse responses to acting White

    among academically-able Black adolescents. In E. Horvat & C. OConnor (Eds.), Beyondacting White: Reframing the debate on Black student achievement (pp. 2756). Lanham, MD:Rowman and Littlefield.

    Muller, C., Riegle-Crumb, C., Schiller, K. S., Wilkinson, L., & Frank, K. A. (2010). Race andacademic achievement in racially diverse high schools: Opportunity and stratification.Teachers College Record, 112, 10381063.

    National Center for Fair and Open Testing. (2010). Schools that do not use SAT or ACT scoresfor admitting substantial numbers of students into bachelor degree programs. Retrieved fromhttp://www.fairtest.org/files/OptionalPDFHardCopy.pdf

    Noguera, P. A. (1995). Preventing and producing violence: A critical analysis of responsesto school violence.Harvard Educational Review, 65, 189212.

    Ogbu, J. (2003). Black students in an affluent suburb: A study of academic disengagement.Mahwah, NJ: Erlbaum.

    26

  • TCR, 114, 070303 Racial Differences

    Raudenbush, S. W., Bryk, A. S., Cheong, Y. F., & Congdon, R. T., Jr. (2004). HLM 6:Hierarchical linear and nonlinear modeling. Lincolnwood, IL: Scientific SoftwareInternational.

    Rubin, D. B. (1987). Multiple imputation for nonresponse in surveys. New York: Wiley.Schafer, J. L. (2000). NORM: Multiple imputation of incomplete multivariate data under a

    Normal Model (Version 2.03) [Computer software for Windows 95/98/NT]. Retrievedfrom http://www.stat.psu.edu/~jls/misoftwa.html

    Steele, C. M. (1997). A threat in the air: How stereotypes shape intellectual identity and per-formance. American Psychologist, 52, 613629.

    Steele, C. M., & Aronson, J. (1995). Stereotype threat and the intellectual test performanceof African Americans. Journal of Personality and Social Psychology, 69, 797811.

    Stricker, L. J., Rock, D. A., & Burton, N. W. (1993). Sex differences in prediction of collegegrades from Scholastic Aptitude Test scores. Journal of Educational Psychology, 85, 710718.

    Trusty, J. (2000). High educational expectations and low achievement: Stability of educa-tional goals across adolescence. Journal of Educational Research, 93, 356365.

    Trusty, J., & Harris, M. B. C. (1999). Lost talent: Predictors of the stability of educationalexpectations across adolescence. Journal of Adolescent Research, 14, 359382.

    Tyson, K. (2003). Notes from the back of the room: Problems and paradoxes in the school-ing of young Black students. Sociology of Education, 76, 326343.

    Tyson, K. (2006). The making of a burden: Tracing the development of a burden of act-ing White in schools. In E. Horvat & C. OConnor (Eds.), Beyond acting White: Reframingthe debate on Black student achievement (pp. 5788). Lanham, MD: Rowman and Littlefield.

    Tyson, K., Darity, W., Jr., & Castellino, D. (2005). Its not a Black thing: Understanding theburden of acting White and other dilemmas of high achievement. American SociologicalReview, 70, 582605.

    Zell, E., & Alicke, M. D. (2009). Contextual neglect, self-evaluation, and the frog-pondeffect. Journal of Personality and Social Psychology, 97, 467482.

    Zwick, R., & Sklar, J. C. (2005). Predicting college grades and degree completion using highschool grades and SAT Scores: The role of student ethnicity and first language. AmericanEducational Research Journal, 42, 439464.

    TINA WILDHAGEN is assistant professor of sociology at Smith College.She conducts research on racial, ethnic, and social class inequality in theAmerican education system. Wildhagens recent articles on racial andclass inequality in the education system can be found in recent or forth-coming volumes of Sociological Quarterly, Sociological Perspectives, and theJournal of Negro Education.

    27