xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam a22 u 4500
controlfield tag 008 c20029999azu 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E11-00249
Educational policy analysis archives.
n Vol. 10, no. 1 (January 07, 2002).
Tempe, Ariz. :
b Arizona State University ;
Tampa, Fla. :
University of South Florida.
c January 07, 2002
Testing and diversity in postsecondary education : the case of California / Daniel Koretz, Michael Russell, Chingwei David Shin, Cathy Horn, [and] Kelly Shasby.
Arizona State University.
University of South Florida.
t Education Policy Analysis Archives (EPAA)
xml version 1.0 encoding UTF-8 standalone no
mods:mods xmlns:mods http:www.loc.govmodsv3 xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govmodsv3mods-3-1.xsd
mods:relatedItem type host
mods:identifier issn 1068-2341mods:part
mods:detail volume mods:number 10issue 1series Year mods:caption 20022002Month January1Day 77mods:originInfo mods:dateIssued iso8601 2002-01-07
1 of 39 Education Policy Analysis Archives Volume 10 Number 1January 7, 2002ISSN 1068-2341 A peer-reviewed scholarly journal Editor: Gene V Glass College of Education Arizona State University Copyright 2002, the EDUCATION POLICY ANALYSIS ARCHIVES Permission is hereby granted to copy any article if EPAA is credited and copies are not sold. Articles appearing in EPAA are abstracted in the Current Index to Journals in Education by the ERIC Clearinghouse on Assessment and Evaluation and are permanently archived in Resources in Education .Testing and Diversity in Postsecondary Education: The Case of California Daniel Koretz Harvard University Michael Russell Boston College Chingwei David Shin University of Iowa Cathy Horn Harvard University Kelly Shasby Boston CollegeCitation: Koretz, D., Russell, M., Shin, C.D., Horn C. & Shasby, K. (2002, January 7). Testing and diversity in postsecondary education: The case of C alifornia, Education Policy Analysis Archives 10 (1). Retrieved [date] from http://epaa.asu.edu/epaa /v10n1/.Abstract
2 of 39The past several years have seen numerous efforts t o scale back or eliminate affirmative action in postsecondary admis sions. In response, policymakers and postsecondary institutions in many states are searching for ways to maintain the diversity of student popul ations without resorting to a prohibited focus on race. In respons e to these changes, this study used data from California and a simplified mo del of the University of California admissions process to explore how var ious approaches to admissions affect the diversity of the admitted stu dent population. "Race-neutral" admissions based solely on test scor es and grades were compared with the results of actual admissions befo re and after the elimination of affirmative action. A final set of a nalyses explored the effects on diversity of alternative approaches that take into account factors other than grades and scores, but not race or ethnicity. Replacing the former admissions process that included prefere nces with a race-neutral model based solely on GPA and SAT-I sc ores substantially reduced minority representation at the two most sel ective UC campuses but had much smaller effects at the other six, less selective campuses. SAT-I scores contributed to but were not the sole c ause of the underrepresentation of African American and Hispani c students. A race-neutral model based solely on GPA also produce d an underrepresentation of minorities, albeit a less se vere one. None of the alternative admissions models analyzed could replic ate the composition of the student population that was in place before the termination of affirmative action in California. The only approach that substantially increased the representation of minority students w as accepting most students on the basis of within-school rather than statewide rankings, and this approach caused a sizable drop in both the ave rage SAT scores and the average GPA of admitted applicants, particularl y among African American and Hispanic students. Although admissions systems differ, the basic findings of this study are likely to appl y at a general level to many universities and underscore the difficulty of providing proportional representation for underserved minority students at highly selective institutions without explicit preferences. Over the past several years, the use of affirmative action to increase the representation of underserved minorities in postsecondary education h as faced increasingly widespread threats. Efforts to scale back or eliminate affirma tive action in admissions have taken numerous forms, including popular referenda, court decisions, and executive actions, and have led to its elimination in California, Flor ida, and Washington. Affirmative action was also terminated in Texas, although a rec ent court decision may permit its reinstatement. A university affirmative action prog ram recently survived legal challenge in Michigan, but additional litigation is pending i n other states, including cases on appeal in Michigan and Georgia.Faced with these events, policymakers and postsecon dary institutions in many states are searching for ways to maintain the diversity of stu dent populations without resorting to a prohibited focus on race. For example, several stat es are experimenting with what we call "X% rules," in which the best students in each high schoolÂ—that is, those who
3 of 39exceed a specified percentile in class rankÂ—are gua ranteed admission to a college campus or system.In response to these changes, we undertook to explo re how various approaches to admissions affect the diversity of the admitted stu dent population. We examined the effects of different stages in the admissions proce ssÂ—for example, the student's decision to take a college-admissions test and to apply, and the college's decision to accept an applicantÂ—on the composition of the student populat ion. In doing this, we focused not only on race and ethnicity, but also on other aspec ts of diversity, such as the educational background of accepted students. We modeled "race-n eutral" admissions based solely on test scores and grades and compared the results wit h actual admissions before and after the elimination of affirmative action. Finally, we explored the effects on diversity of alternative approaches that take into account facto rs other than grades and scores, but not race or ethnicity.Because California has been one of the primary focu ses of debate about the rollback of affirmative action, these analyses use data from th at state and are loosely modeled after the admissions procedures and student population of the University of California. We expect that many of the findings could be generaliz ed in broad brush to other states, but some patterns may differ depending on differences a mong states in demographics, the selectivity of state universities, and so on.Recent Trends in Postsecondary AdmissionsAlthough hard data on affirmative action are scanty most observers believe that selective institutions have widely employed it for several decades. The use of race as a factor in higher education admissions was legitimiz ed but limited by a 1978 Supreme Court decision, University of California Board of Regents v. Bakke (1978). The justices held that racial diversity was a legitimate goal fo r institutions of higher education but that creating a separate admissions process or a qu ota system was not the "least objectionable alternative" for achieving that goal. Consideration of race in the admissions process was deemed acceptable only if it was one of many factors considered.The policies endorsing affirmative action have some times been explicit. An example is the state system of higher education in Texas. Unti l 1996, the state of Texas maintained a concerted effort to recruit minority students int o higher education and to prepare its institutions to meet the demands of its growing min ority college-age population. This effort was due to an investigation of segregation i n Texas higher education by the U.S. Office of Civil Rights (OCR) between 1978 and 1981. As a result of that investigation, Texas was required to develop a plan to desegregate and to "increase the representation of blacks and Hispanics in institutions of higher e ducation" in order to avoid federal enforcement proceedings (Texas Higher Education Coo rdinating Board, 1997). Since Governor William Clements issued the first plan, ea ch subsequent governor has submitted a follow-up designed to continue increasi ng minority representation in Texas higher education. The most recent plan, Access and Equity 2000, sought increases reflecting the proportion of college-age minorities in the Texas population as well as increased minority representation on faculties and advisory boards. Similarly, until affirmative action was terminated in California, th e state of California engaged in numerous activities to increase minority representa tion on campus, ranging from
4 of 39academic preparation programs at community colleges to actively recruiting minority applicants to the campuses of the University of the California (UC). Nonetheless, many of the admissions policies implem enting affirmative action, particularly at the institutional level, have not b een made explicit, and their effect in practice has often been unclear. As Kane noted, "Ne arly two decades after the U.S. Supreme Court's 1978 Bakke decision, we know little about the true extent of affirmative action admissions by race or ethnicityÂ… Hard evidence has been difficult to obtain, primarily because many colleges guard their admissions practices closely" (Kane, 1998a, p. 17). One recent study used survey data to estimate the e xtent in practice of race-based preferences in higher education admissions. Kane (1 998b) estimated how race/ethnicity, high school grade-point average (GPA), scores on th e Scholastic Assessment Test (SAT), participation in student government and athl etics, and college selectivity affected the probability of acceptance to college for a nati onally representative sample of the high school class of 1982. Kane found that, holding cons tant the factors other than race, black and Hispanic applicants had an appreciable advantag e over white applicants, but only in selective colleges. In the most selective colleges (those in the top quintile of selectivity), Kane estimated that the average advantage of black applicants was "equivalent to nearly a full point increase in high school grade-point av erage (on a four-point scale), or to several hundred points on the SAT" (Kane, 1998b, p. 438). The data upon which Kane's estimates were based, however, are now quite dated, and the sample did not allow estimates specific to states or individual postseco ndary systems. Thus research leaves unclear how substantial preferences were in the sta tes that have been at the center of the debate about the elimination of affirmative action, such as California and Texas.Recent Policies Curtailing Group-Based PreferencesAlthough this report focuses on California, recent initiatives curtailing affirmative action have been proposed or enacted in several states.CaliforniaThe first recent major rollback of affirmative acti on in higher education was the enactment of SP1 in 1995 by the University of California Board of Regents. SP-1 stated that "the University of California shall not use race, religion, sex, color, ethnicity, or national origin as criteria for admission to the University or to any program of study." This resolution was a response to executive orders issued by Governor Pete Wilson that severely curtailed affirmative action in a broad ra nge of state procurement and administrative decisions.In November 1996, California voters approved Propos ition 209, which eliminated the consideration of race, ethnicity, and gender in pub lic employment, public contracting, and education. In effect, Proposition 209 provided constitutional backing for SP-1. The US Supreme Court refused to hear a challenge to Pro position 209 in November 1997, thus allowing the measure to stand. Admission decis ions based on these new policies went into effect with students seeking admission fo r the Spring quarter 1997-1998. These actions, taken together, represent a full rep eal of affirmative action policies in
5 of 39California's state system of higher education.TexasAround the same time, the Fifth Circuit Court of Ap peals ended the use of any race-based consideration in admission decisions in the area under its jurisdiction. In 1992, four white students who had been denied entra nce to the University of Texas law school filed suit against the university, claiming that the partially race-based admission process violated their Fourteenth Amendment rights. Four years later, the Fifth Circuit Court of Appeals upheld their claims. "The case aga inst race-based preferences does not rest on the sterile assumption that American societ y is untouched or unaffected by the tragic oppression of its past. Rather, it is the ve ry enormity of that tragedy that lends resolve to the desire never to repeat it, and find a legal order in which distinctions based on race shall have no place" ( Hopwood v. Texas, (1996), as quoted in Feinberg, 1998, p.12).The principle enunciated in Hopwood appears to be inconsistent with the standards enforced by the OCR investigations of the state's e fforts to remedy the remaining vestiges of de jure segregation in public higher education. First in 1 980 and again in 1987, the OCR found that Texas had not made adequat e progress in eliminating such problems and required that additional plans be adop ted in order to avoid federal action. Texas was informed by the OCR in 1997 (after the Hopwood ruling) that its higher education system would once again be reviewed to en sure that an OCR-approved plan had been effectively implemented and that all trace s of segregation had been eliminated in compliance with Supreme Court precedent (Siegel, 1998). The standard for the OCR review was set by United States v. Fordice a 1992 case in which the U.S. Supreme Court held "that any state with a history of segreg ation in higher education must implement affirmative measures, including racial pr eferences to eliminate those vestiges." This standard differs from the one set o ut in Hopwood which allows the use of racial preferences, but only when a state entity is acting to remedy present effects of past discrimination at a specific institution (THEC B, 1997). The OCR review is still in progress and it is uncertain how OCR standard in Fordice may affect the interpretation or implementation of Hopwood The Hopwood case has been returned to the Court of Appeals thr ee times, and the most recent ruling suggests that race may play some role in admissions. In its most recent ruling, the Court of Appeals did not overturn the o riginal Hopwood ruling, but it did rule that the District Court injunction prohibiting the University of Texas Law School from any use of race in making admissions decisions was overly broad and excessive. The case has been remanded again to Circuit Court for a dditional action. The degree to which racial preferences will be allowable in Texas and in other states under the jurisdiction of the Fifth Circuit therefore remains uncertain. WashingtonInitiative 200 (I-200) was passed by the voters of Washington state in November 1998. Like Proposition 209 in California, it restricts th e use of race in employment, education, and contracting. In response to the initiative, Uni versity of Washington (UW) President Richard L. McCormick announced that UW would suspen d the use of race, ethnicity, and gender in admissions beginning in Spring 1999. It is important to note, however,
6 of 39that I-200 was passed as a law, not as an amendment to the state constitution as was Proposition 209 in California. It is therefore stil l uncertain whether I-200 will supersede existing laws that allow the use of race in employm ent and contracting decisions. It does not, for example, apply to federally funded state p rograms that must comply with federal nondiscrimination laws.FloridaThe Board of Regents of the State University System of Florida voted in favor of Governor Jeb Bush's "One Florida" Initiative in Feb ruary 2000. This plan eliminates the use of race as a factor in admission decisions in t he Florida University system and outlines an alternative, race-neutral admission pro cess. Florida planned to have its new admission criteria in effect for students graduatin g from high school in 2000. GeorgiaIn Georgia, a case was filed in federal court by th ree white women denied admission to the University of Georgia. The plaintiffs sued the University and State Board of Regents under Title VI of the Civil Rights Act of 1964, all eging that they were discriminated against because of their race. In a decision handed down in July of 2000, federal judge B. Avant Edenfield of Georgia ruled that the 1978 Bakke decision has been misinterpreted and that diversity is "an amorphous, unquantifiable goal" that cannot be constitutionally justified. He nullified the Univer sity of Georgia's now-discarded policy of maintaining lower admission standards for blacks In a non-binding opinion, he further criticized the university's use of race/eth nicity as a "plus" factor in the selection of 10 to 15% of the students in each year's enterin g class. The case is likely to be appealed to the 11th U.S. Circuit Court of Appeals (Denniston, 2000). MichiganMichigan has recently seen two challenges to affirm ative action in higher education in federal court. The first, Gratz v. Bollinger et. Al (2000), challenged the use of race in admissions at the undergraduate level. The plaintif fs were unsuccessful applicants to the College of Literature, Science, and the Arts in Fal l 1995 and Fall 1997, respectively. This case was recently decided in favor of the defe ndant. Grutter v. Bollinger (2001) was filed in 1997 against the University of Michiga n's law school, challenging the use of race in its admission policy. In March, 2001, Distr ict Court Judge Bernard Friedman ruled that the law school's admissions policies con sidering race were unconstitutional and a violation of Title VI of the 1964 Civil Right s Act. Both cases have been certified as class actions and are expected to be appealed.Policy Responses to Challenges to Affirmative Actio nPolicymakers in the university systems of Texas, Ca lifornia, and Florida have tried in various ways to maintain diversity in the face of l egal restrictions on affirmative action. In all three states, individual campuses have tried to recruit at high schools whose students are traditionally underrepresented in the college population. At the system level, outreach in California and Texas has focused on inf orming the public about the race-neutral admissions policies and on assuring mi norities that the higher education
7 of 39system is still hospitable.All three of these states have also instituted "X% plans"Â—that is, policies that admit a certain percentage of graduating public high school seniors automatically to their university systems, primarily on the basis of stude nts' academic ranks within their high schools. The Texas legislature passed House Bill No 588, also known as the 10% rule, in May of 1997. The measure mandates that public or private high school students whose GPA places them in the top 10% of their gradu ating class be admitted automatically to "each general academic teaching in stitution" if they graduated within the previous two years and filed the appropriate ap plications on time. The act also stipulates that the governing board of each such in stitution will decide on an institutional basis whether to automatically admit any student in the top 25% of his or her graduating class, but not in the top 10%. The legislature also outlined factors other than academic achievement that institutions were to take into con sideration when admitting the rest of their freshman classesÂ—factors related primarily to socioeconomic status, geographic region, and uncommon hardship. The admission criter ia for students not in the top 10% or 25% of their class were to be published in the a cademic catalogs and made available to the public not later than one year before the da te when they were to take effect. Similarly, the factors used in awarding competitive fellowships and scholarships were to be made public. The act has applied to all admissio ns and scholarship awards since the Fall semester of 1998.
8 of 39In California, the top 4% of graduating seniors fro m each public high school are now eligible for admission to a school in the UC system although not necessarily to the campus of their choice. Each school sets its own ad mission standards based on system policies, but a student deemed eligible is guarante ed admission to at least one of the UC campuses. The 4% plan in California has been termed ELC, eligibility in the local context. In addition to graduating in the top 4% of their class, students must fulfill a minimum course requirement that specifies the numbe r and level of courses to be taken in high school subject areas, and they must submit ACT or SAT scores if the institution of their choice requires test scores. The ELC 4% pl an is expected to be in effect for freshmen applicants in Fall 2001.Unlike Texas and California, Florida established an alternative policy concurrently with terminating of race-based admissions. Beginning wit h the class of 2000, the top 20% of graduating seniors from each public high school wil l automatically be admitted to the state university system under the Talented 20 Progr am. Because of enrollment caps, however, students are guaranteed admission only to one of the 10 Florida universities, not necessarily to their top choice. The "One Flori da" plan also calls for an additional $20 million in need-based financial aid. Under this initiative, universities are asked to address the financial-aid needs of students admitte d under the Talented 20 Program before those of other students.Washington state is in the process of reviewing its policies in response to Initiative 200. The UW Board of Regents is considering a proposal t o allow the use of race and gender as factors in choosing the recipients of privately funded scholarships. Applicants would undergo screening based on neutral factors such as merit and need. From among those who pass the screen, students would be matched with scholarships. The aim is to attract minority students to the UW system.Initial Effects of Policy ResponsesOnly Texas has fully implemented its "X% rule" admi ssion policy. Holley and Spencer (1999) found that in its first year, the 10% rule h ad no significant impact on the number of minority students enrolled as first-time freshme n at the University of Texas at Austin and Texas A&M University, the state's two flagship schools. Only eight more black students and one fewer Hispanic student enrolled at UT-Austin in 1998 than in 1997. At A&M, 19 more black students and 62 more Hispanic st udents enrolled in 1998. However, the results of the 10% rule were available only for the first year (academic year 1998) and may not be indicative of the long-te rm effects of the new program. Although no data on the effects of the 4% rule on f reshman enrollment patterns at the University of California schools are available, a s imulation study assessing the potential effects was conducted by Saul Geiser of the UC Offi ce of the President. 1 For all California public high-school graduates for whom SA T scores were available, Geiser (1998) calculated an Academic Index scoreÂ—an 8,000point scale that gives approximately equal weight to a student's high scho ol GPA and SAT scores. Students were then ranked by Academic Index score within eac h high school, and those in several top percentiles were combined into a simulated UC e ligibility pool. Geiser found that limiting admissions to only the top 4% of students within each high school would have a modest impact on the racial/ethnic composition of t he admitted population. Of the total eligible pool, 31% would be white, 47% Asian, 14% L atino, 3% black, and 5% "other."
9 of 39In the second stage, the three models chosen in the first stage were applied to the College Board data to estimate the racial/ethnic co mposition of the admitted pool under race-neutral admissions rules. These analyses used 1998 and 1995 data and were limited to students who attended high school in California at the time they took the SAT. The models did not predict acceptance or rejection for individual students; rather, they predicted the probability of admission for students in a given range of SAT scores and GPA. These probabilities were applied to counts of tested students in each range to obtain estimated counts of admitted students. The r esulting estimates of racial/ethnic composition were compared with actual admission dat a from the class admitted in 1999 (after SP-1 and Proposition 209 had been implemente d) to confirm their reasonableness. The application and acceptance process in the UC sy stem can be seen as a sequence of filters, which are described in the following secti on: (1) taking the SAT; (2) meeting the UC system eligibility criteria, based on SAT scores and GPA; (3) applying to a campus at a given level of selectivity; and (4) being admi tted to that campus. Our models represented a simplified, race-neutral version of t he fourth of these filters. We examined the effects of the four filters, individually and i n various combinations, on the diversity of the surviving pool of students. For example, by removing the application filter, we estimated the racial/ethnic composition that would result if all students were successfully encouraged to apply to campuses at all three levels of selectivity. We also examined the effects of these filters on other characteristics o f the admitted groups: whether high school students attended an urban, suburban, or rur al high school, the type of high school attended (e.g., public, private, religiously affiliated), parents' level of education, and first language spoken at home. The results of t hese analyses were compared with actual admission data from three years: 1995 (to re present policy before the enactment of SP-1 and Proposition 209), 1999 (to represent full implementation of these policies), and 1997 (to represent the transitional period).In the third stage, a number of alternative admissi on models were applied to gauge their effects on the diversity of the admitted student po pulation. These models used both individual variables (such as parents' education) a nd school characteristics (such as the percentage of students receiving free or reduced-pr ice lunch). As part of this stage, we replicated the analyses performed by the UC office of the President to model the effects of 4%, 6% and 12.5% admission policies to ensure th at our data and methods were consistent with those used in that study.Steps in the Selection ProcessThe selection process entails a series of filters t hat progressively winnow the applicant pool. The filters in our models are the following.The decision to take the SAT (or ACT) Because admission to the University of California system requires that students take the S AT or ACT, those who fail to do so remove themselves from the pool of potential studen ts. This filter is nearly universal among selective colleges and universities nationwid e, although a few institutions (e.g., Bates and Bowdoin) either do not use it or make the submission of scores optional. In California, 98% of students who apply to the Uni versity of California take the SAT (Geiser, 1998), and our models accordingly simplifi ed this filter slightly by considering only whether students take the SAT, not either the SAT or the ACT.
10 of 39University system eligibility. The University of California screens students for eligibility to the entire UC system. Ineligible students are fo r the most part ineligible for admission to any of the eight campuses. However, at the outse t of the study period, each campus was allowed to allocate up to 6% of its slots to UC -ineligible students, and up to two-thirds of these slots could be used for admitti ng disadvantaged students ( 1996 Guidelines for Implementation of University Policy on Undergraduate Admissions http://www.ucop.edu/sas/exguides.html).UC system eligibility was based on three criteria. First, GPA and SAT-I scores were combined on a sliding scale to set minimum requirem ents. (SAT-I refers to the basic verbal and mathematics tests, while SAT-II refers t o a number of optional, subject-matter tests.) For example, students with G PAs of at least 3.29 were UC-eligible as long as their combined SAT-I scores were at leas t 570, while students with GPAs of 3.0 were required to have a combined SAT-I score of at least 1270. Second, students were required to take a set of required courses. Th ird, students had to take three SAT-II tests, "including writing, mathematics Level 1 or L evel 2, and one test in one of the following areas: English literature, foreign langua ge, science, or social studies," although they were not required to attain a specific score o n these tests ( Admission as a Freshman http://www.ucop.edu/pathways/impinfo/freshx.html) Our models simplified system eligibility by applyin g the UC GPA and SAT-I criteria but not the UC requirements for specific courses or for taking SAT-II tests. Because the system-eligibility filter is specific to the Univer sity of California system, we conducted parallel analyses that excluded it.Application to a campus at a given level of selecti vity Students who elect not to apply to an institution remove themselves from the pool of p otential students. We lacked data on actual applications, but we did have a record of al l institutions to which each student had his or her SAT-I scores sent. We treated sending a score as a proxy for application, thus overestimating by a presumably small amount the act ual number of applications. We established a flag indicating whether a student had sent scores to any of the campuses within three levels of selectivity (see Appendix A) : High selectivity: Berkeley and UCLA Moderate selectivity: Irvine, Davis, Santa Barbara, and San Diego Low selectivity: Santa Cruz and Riverside. Predicted admission based on GPA and SAT The probability that each student will be accepted to a campus at a given level of selectivit y was predicted using logistic regression models derived from published campus-lev el admission statistics (see Appendix A). 2 We refer to this as a race-neutral admissions mode l because the probabilities assigned to students were unaffected by race or ethnicity (or any characteristics other than SAT scores and GPA). The se models could not predict admission or rejection for individual students; rat her, they predicted the probability of admission for students within a given range of GPA and SAT-I scores. These probabilities were multiplied by the number of stud ents in each range to yield a count of "admitted" students.Limitations
11 of 39This study is limited to students who attended high school in California and who took the SAT, and it examines only the impact of race-ne utral admission decisions to University of California campuses. Analyses in othe r states might yield substantially different results. Because this study sorts campuse s into three categories and uses one model from one campus within each category to repre sent all campuses in that category, the findings do not necessarily apply to individual campuses. Moreover, this study is based almost entirely on data collected during the late 1990s, and patterns of application, test-taking, and acceptance may change with time. N onetheless, we expect that the findings generalize in broad stroke to numerous oth er state university systems. This study was also limited by the type of data to which we had access. For example, we had no access to individual-level data about accept ance or rejection, and the aggregate data on admissions probabilities were not available separately by race/ethnicity. That lack precluded more refined and powerful analysis.The Effects of Current and Race-Neutral Selection o n Racial/Ethnic CompositionEstimates of the effects of selection policies on d iversity are presented by the selectivity of the institutions, starting with the most highly selective.Effects in Highly Selective InstitutionsAdmissions to highly selective institutions were mo deled loosely on UCLA and Berkeley. As noted, while the application filter sh owed whether students applied to either of these two campuses, the regression model used for both was derived from Berkeley data. Use of the UCLA data would not have greatly changed the results (see Appendix A).The first screen applied, students' decision whethe r to take the SAT, substantially decreased the percentage of Hispanic students and i ncreased the percentage of Asian students. In 1998, 31% of California high school gr aduates, but only 19% of those taking the SAT, were Hispanic (Table 1). Conversely, 15% o f graduates but 23% of SAT-takers were Asian. This screen, however, only slightly red uced the representation of black students, who constituted roughly 7% of both gradua tes and SAT-takers. The decision to take the SAT also slightly reduced the representati on of white students, who constituted 45% of graduates and 42% of SAT-takers.It is important to note, however, that eliminating this screenÂ—that is, having all students take the SATÂ—would not fully eliminate its effects. The students who decide against taking the SAT are presumably lower-achieving on av erage than those who do take it. Thus if all students took the SAT, many of those wh o currently do not take it would fail to gain admission because of low scores; and if the SAT were no longer used in admissions, some would fail to gain admission becau se of weaker academic records. The UC system eligibility screen had a very differe nt effect: it reduced the percentage of black students substantially and the percentage of Hispanic students more modestly. In 1998, 7% of California students taking the SAT were black, in contrast to 4% of those system-eligible in terms of SAT scores and GPA (see italicized panel of Table 1.) Hispanics constituted 19% of all SAT-takers but 15% of those eligible. Applying the
12 of 39 eligibility screen slightly increased the represent ation of whites and Asians.Table 1 Racial/Ethnic Composition, Highly Selective Campuse s: Actual, and Estimated Using All Screens and SAT+GPA Admissions ModelAsian, Asian-American, Pacific Islander Black or African-American HispanicWhiteOtherDecline to State Graduates, 199815731451NASAT-takers,1998 227194263 UC eligible,1998 254154663 Eligible and applied tohigh-selectivity school, 1998 364153572 Admitted by neutral model,1998 38294273 1995 Admittedclass 367193125 1997 Admittedclass 386153326 1999 Admittedclass 413103529Note: Race/ethnicity is based on student self-repor ts for all rows except the "Graduates" row, which is based on reports by school administrators. Estimates are italicized; other numbers are actual counts. Percentages may not sum to totals be cause of the exclusion of American Indian students and rounding. Sources: Estimates reflect NBETPP analysis; admissi on figures are published figures from UC (http://www.ucop.edu/pathways/infoctr/introuc/prof_ engin.html); counts of SAT-takers are based on NBETPP tabulations of data provided by the Colle ge Board; counts of graduates are from California Department of Education, Educational Dem ographics Unit (http://data1.cde.gov/dataquest). Using the application screenÂ—that is, dropping all students who did not apply to Berkeley or UCLAÂ—did not affect the representation of Hispanics or blacks ("Eligible and applied to high selectivity school" row of Tabl e 1). The number of minority students dropped by nearly half when this filter was applied but that decrease was similar to the decrease in the total number of students in the poo l. The application filter did, however, increase the percentage of Asian students and decre ase the percentage of whites. Assuming that most of the students who requested th at scores be sent to a particular campus actually applied for admission to that campu s, it appears that Asians are particularly likely and whites less likely to apply to Berkeley and UCLA.
13 of 39The final screen, the race-neutral admissions funct ion based only on SAT and GPA, markedly reduced the representation of Hispanics an d more still that of blacks. Black students dropped from 4% to 2% of the pool at this stage (see "Admitted by neutral model" row of Table 1), while Hispanics dropped fro m 15% to 9%. The offsetting increase was among white students, not Asians.These screens have a cumulative effect, progressive ly reducing the representation of Hispanic and black students in the pool. That effec t for Hispanics can be seen in Figure 1, which graphically represents the percentages in Tab le 1. The second bar shows a dramatic reduction from all graduates to those who took the SAT. The next screen, UC eligibility (simplified, as noted earlier, to reflect only SAT scores and GPA), produced a more modest but still appreciable drop. Application to h igh-selectivity schools had no effect on the representation of Hispanics, but the race-neutr al admissions model reduced it substantially. Figure 1. Hispanics as Percentage of Group Admitted Highly Selective CampusesSOURCES: See Note, Table 1. In 1995, before the implementation of SP-1 or Propo sition 209, Hispanics constituted 19% of students admitted to Berkeley and UCLA, whic h was almost exactly equal to their representation in the population of SAT-taker s (Figure 1). Hispanics were thus overrepresented slightly relative to their numbers among UC-eligible students and substantially relative to a race-neutral policy. By 1999, after SP-1 and Proposition 209 were implemented, the representation of Hispanics a dmitted at these two campuses fell to roughly the percentage predicted by our race-neutra l model. As Karabel (1998) noted, the admission of minorities fell after the enactment bu t before the implementation of SP-1 and Proposition 209 (note the drop between 1995 and 1997 in Figure 1).
14 of 39The cumulative effects of these admissions screens on blacks present a somewhat different picture. While the sharpest drops in the representation of Hispanics arose from self-selection to take the SAT and the use of a rac e-neutral model, the declines for blacks arose primarily from the UC system eligibility scre en and the race-neutral model (Figure 2). In 1995, blacks constituted 7% of students admi tted to Berkeley or UCLA, almost exactly matching their representation among SAT-tak ers and high school graduates (Figure 2). Their representation among students act ually admitted to Berkeley or UCLA, like that of Hispanics, dropped in both 1997 and 19 99. Their representation among actual admissions in 1999, however, while very low, was su bstantially higher than was predicted by our simple GPAand SAT-based race-neu tral admissions model. Figure 2. Blacks as Percentage of Group Admitted, H ighly Selective CampusesSOURCES: See Note, Table 1. Our model did not match as well the representation of Asian and white students in the admitted pool. In 1999, 41% of the students admitte d to Berkeley or UCLA were Asian, and 35% were white. Our race-neutral model predicte d slightly fewer Asians (38%) and appreciably more whites (42%) than were actually ad mitted. We suspect but cannot verify that this is due to differences in the proportions of white and Asian students applying to selective private institutions in California and to colleges outside the state. Because most states lack the system-eligibility scr een used in California, we tested the generality of these findings. We applied the race-n eutral admissions model based on SAT scores and GPA to all students who had sought admis sion to either UCLA or Berkeley and eliminated the UC system eligibility screen. Dr opping that screen had almost no effect on the racial/ethnic composition of the grou p "admitted by neutral model" presented in Table 1. Recall that our simplified sy stem eligibility rule is based solely on SAT scores and GPA, and the admissions model for th e highly selective campuses
15 of 39 applies such stringent requirements for those score s that the system eligibility screen is simply irrelevant.In principle, one simple way to address the underre presentation of minority students would be to encourage all students to apply to the highly selective campuses. Therefore, in a second simplification, we applied the race-neu tral admissions model to all students who took the SAT, regardless of system eligibility and of the schools to which they had their SAT scores sent. This too affected the racial /ethnic composition of the "admitted" pool only slightly. The total number of "admitted" students went up by more than half; the increases in the numbers of "admitted" blacks a nd Hispanics, however, were roughly proportional to that overall increase.Some observers have argued that admissions tests su ch as the SAT should be abandoned in order to produce a student body more nearly repr esentative of the racial/ethnic composition of the entire population. For example, in 1997, a university task force recommended that the University of California drop the SAT as an admission requirement to avoid a precipitous decline in the e nrollment of minorities at the university's flagship campuses (Fletcher, 1997). Th us we estimated a second set of race-neutral models, based solely on GPA with no co nsideration of SAT, to explore how that criterion would affect the diversity of the ac cepted student population. (See Appendix A.)A race-neutral model based solely on GPA creates a substantial underrepresentation of both black and Hispanic students, though a somewhat less severe one than that generated by the race-neutral model based on SAT and GPA toge ther. This is shown in Table 2, which presents the racial/ethnic composition of the groups admitted by the SAT+GPA and GPA-only models. The second panel in Table 2, SAT+GPA," presents estimates (discussed earlier) based on the SAT+GPA race-neutr al model. 3 For example, using the SAT+GPA model with students who applied to either B erkeley or UCLA, 9% of those "admitted" would be Hispanic. In contrast, 13% of s tudents selected by the GPA-only model would be Hispanic (third panel of Table 2). T hus the GPA-only model increased by roughly half the percentage of admitted students who are Hispanic; but even with this model, Hispanics were substantially underrepresente d relative to the 31% of graduates and 19% of SAT-takers who were Hispanic. Similarly, using a GPA-only model increased the percentage of the admitted group who are black by about half, but would still admit less than half as many blacks as there were either among graduates or among SAT-takers.Table 2 Racial/Ethnic Composition, Highly Selective Campuse s, Estimated Using SAT+GPA and GPA-Only Admissions ModelsAsian, Asian-American, Pacific Islander Black or African-American HispanicWhiteOtherDecline to State Graduates,1998 15731451NA
16 of 39 SAT-takers,1998 227194263 SAT+GPAApplied andadmitted 38294273 GPA onlyApplied andadmitted 363133772Note: Race/ethnicity is based on student self-repor ts for all rows except the "Graduates" row, which is based on reports by school administrators. Estim ates are italicized; other numbers are actual counts. Percentages may not sum to totals because o f the exclusion of American Indian students and rounding. Sources: Estimates reflect NBETPP analysis; admissi on figures are published figures from UC (http://www.ucop.edu/pathways/infoctr/introuc/prof_ engin.html); counts of SAT-takers are based on NBETPP tabulations of data provided by the Colle ge Board; counts of graduates are from California Department of Education, Educational Dem ographics Unit (http://data1.cde.gov/dataquest). The improved representation of minority students ac hieved by using a GPA-only model rather than a GPA+SAT model, however, would come at a price. Grading standards are inconsistent from high school to high school, and t here is evidence that they vary across types of school. For example, grading tends to be m ore lenient in schools with high poverty rates (U.S. Department of Education, 1994). Absent a measure standardized across schools, these inconsistencies would introdu ce additional arbitrariness into the admission process and could lower the overall level of academic preparedness of the admitted group.Effects in Moderately Selective InstitutionsWe classified as moderately selective the campuses at Irvine, Davis, Santa Barbara, and San Diego. Our race-neutral admissions model for th ese campuses was based on data from Irvine.The first two screens applied in examining admissio ns to moderately selective institutions were the same as for highly selective schoolsÂ—that is, deciding to take the SAT and meeting UC system eligibility requirements. Thus, t he representation of non-Asian minority students fell substantially before use of the filter of application to a moderately selective campus: Hispanic representation was reduc ed by the SAT-taking screen, and black representation by the UC system eligibility s creen. Although the self-selection of eligible students to apply to moderately selective campuses reduced the pool by about 40%, it affected the raci al/ethnic composition of the pool only modestly (Table 3). Hispanic students constituted 1 5% of the eligible pool but 13 % of the eligible students who applied to such an instit ution. Black students, who constituted a meager 4% of eligible students, made up only 3% of those who were eligible and applied. The representation of Asian-American students incre ased appreciably with this filter, and the representation of whites dropped.Table 3
17 of 39 Racial/Ethnic Composition, Moderately Selective Cam puses: Actual, and Estimated Using All Screens and SAT+GPA Admissi ons ModelAsian, Asian-American, Pacific Islander Black or African-American HispanicWhiteOtherDecline to State Graduates,1998 15731451NA SAT-takers,1998 227194263 UC eligible,1998 254154663 Eligible and applied tomoderately selective school, 1998 323134262 Admitted by neutral model,1998 332124372 1995 Admittedclass 353144242 1997 Admittedclass 363134435 1999 Admittedclass 362114128Note 1: Race/ethnicity is based on student self-rep orts for all rows except the "Graduates" row, which is based on reports by school administrators. Estimates are italicized; other numbers are actual counts. Percentages may not sum to totals be cause of the exclusion of American Indian students and rounding. Note 2: An earlier version of this table was correc ted on January 17, 2002. Sources: Estimates reflect NBETPP analysis; admissi on figures are published figures from UC (http://www.ucop.edu/pathways/infoctr/introuc/prof_ engin.html); counts of SAT-takers are based on NBETPP tabulations of data provided by the Colle ge Board; counts of graduates are from California Department of Education, Educational Dem ographics Unit (http://data1.cde.gov/dataquest). When used with UC-eligible students who applied to one of the four moderately selective institutions, the race-neutral admissions model had little effect on the racial/ethnic composition of the student pool. The number of stud ents fell by roughly one-fourth with use of the admissions model, but the reduction was nearly proportional to the racial/ethnic groups. The percentage of Hispanics d ecreased only from 13% to 12%; that of blacks dropped from 2.6% to 2.2 %.For moderately selective campuses as well, our mode ls suggest that by 1999 admissions in all four campuses taken together were largely ra ce-neutral. The composition of the group admitted in 1999 was very similar to that pre dicted by our race-neutral model (Table 3; compare the "Admitted by neutral model" r ow with the actual figures for 1999
18 of 39 admissions). However, at the moderate-selectivity campuses taken togetherÂ—in contrast to the highly selective campusesÂ—the composition of the classes a dmitted changed only modestly from 1995 to 1999. Between 1997 and 1999, the perce ntage of admitted students who were black declined from 3% to 2%, and the percenta ge who were Hispanic from 14% to 11% (Table 3). These small changes after affirmativ e action was terminated suggest that racial/ethnic preferences had been much less substa ntial at the moderately selective campuses, taken together, than at the highly select ive campuses. We again examined the effects of removing the UC el igibility requirements and the application screen. Removing the former only slight ly increased the size of the pool of students, and had only trivial effects on the ethni c composition of the accepted student group. In other words, for the most part students w ho were not UC system eligible either did not apply to any of these colleges or were pred icted to be rejected by our admissions model. This was due mainly to students who took the SAT but did not apply to any of the four institutions. Removing the application scr eenÂ—in effect, having all students who took the SAT applyÂ—increased the number of stud ents "accepted" by more than half. This increase, however, was roughly proportio nal to racial/ethnic groups, and so would raise the percentage of students who were bla ck or Hispanic only slightly.Effects in the Least Selective InstitutionsIn many respects, admission to the least and the mo derately selective UC campuses was similar. In both cases, the main reduction in the r epresentation of Hispanics occurred though the self-selection of students to take the S AT (Table 4). The UC system eligibility screen brought a modest further reducti on, but the application screen and the race-neutral model had little effect. In contrast, blacks were proportionately represented among SAT-takers, and the primary reduction in the representation resulted from the application of the UC eligibility screen.Table 4 Racial/Ethnic Composition, Least Selective Campuses : Actual, and Estimated Using all Screens and SAT+GPA Admissions ModelAsian, Asian-American, Pacific Islander Black or African-American HispanicWhiteOtherDecline to State Graduates,1998 15731451NA SAT-takers,1998 227194263 UC eligible,1998 254154663 Eligible and applied tolow-selectivity school, 1998 333163872
19 of 39 Admitted by neutral model,1998 333153973 1995 Admittedclass 324193842 1997 Admittedclass 353153935 1999 Admittedclass 333153927Note: Race/ethnicity is based on student self-repor ts for all rows except the "Graduates" row, which is based on reports by school administrators. Estimates are italicized; other numbers are actual counts. Percentages may not sum to totals be cause of the exclusion of American Indian students and rounding. Sources: Estimates reflect NBETPP analysis; admissi on figures are published figures from UC (http://www.ucop.edu/pathways/infoctr/introuc/prof_ engin.html); counts of SAT-takers are based on NBETPP tabulations of data provided by the Colle ge Board; counts of graduates are from California Department of Education, Educational Dem ographics Unit (http://data1.cde.gov/dataquest). Admission of black and Hispanic students to the lea st selective campuses changed little from 1995 to 1999 and matched our race-neutral mode l reasonably closely in all years. This suggests that students' preferences played lit tle role in admission to these institutions.The Effects of Admission Filters and Race-Neutral S election on Other Aspects of DiversityThe diversity of the student body has numerous aspe cts in addition to race and ethnicity. In this section, we examine the effect of each filt er in the admission process on other aspects of diversity: the geographic location and t ype of the secondary schools students attended, the education level of students' parents, and the languages students speak at home.Information on these variables was obtained from th e Student Descriptive Questionnaire (SDQ) that students complete when registering for t he SAT and thus is subject to the errors common to survey data of this sort. For exam ple, students may not consistently characterize the language used in their homes. Beca use the effects on racial/ethnic diversity are largest at the highly selective campu ses, these analyses are limited to them.Geographic LocationThe SDQ offers six options for classifying the loca tion of students' high schools: large city, medium city, small city, suburban, rural, and other. In 1998, 30% of SAT takers in California reported attending a secondary school lo cated in a large city, and 60% in cities of all sizes (Table 5). 4 Thirty percent attended school in a suburban area, and only 5% in a rural area.Table 5 Geographic Composition, Highly Selective Campuses, Using System
20 of 39 Eligibility, Application, and SAT+GPA Admissions Mo delLarge City Medium City Small City SuburbanRuralOther SAT-takers, 19983016143055UC Eligible, 19982817153263Eligible and applied tohigh-selectivity schools, 1998 3216123532 Admitted by neutral model, 19982816124031 Although the effects of the application filters on geographic representation are modest compared with those on racial/ethnic composition, t hey do somewhat increase the percentage of students who are from suburban school s. All three filtersÂ—UC system eligibility, application to a highly selective camp us, and predicted admissionÂ—contribute to this effect; taken together they increased the representation of suburban students from 30% of all SAT-takers to 40% of those admitted by a race-neutral model (Table 5). This increase was off set by smaller decreases in the percentages from schools in other locations. Surpri singly, the admission filter had only very small and inconsistent effects on the represen tation of students from large cities.School TypeThe SDQ allowed students to specify four types of h igh school: public school, religiously affiliated school, independent school w ithout religious affiliation, and other. Of California students who took the SAT in 1998, 81 % attended a public school, 13% attended religiously affiliated schools, 2% attende d a non-religious independent school, and 4% attended alternative types of school (Table 6).Table 6 School Type, Highly Selective Campuses, Using Syste m Eligibility, Application, and SAT+GPA Admission ModelPublicIndependentReligiousOther SAT-takers, 1998812134UC Eligible, 1998823133Eligible and applied to high-selectivity schools,1998 833131 Admitted by neutral model, 1998834121 The effects of the admission filters on the mix of school types were minor. At all stages of selection, between 81% and 83% of students were from public schools, and 12% or 13% were from religious schools (Table 6). The filt ers reduced the representation of students from "other" schools and increased that of students from non-religious independent schools, but both of these groups const ituted only a small percentage of the total group at each stage.
21 of 39 Parents' EducationStudents were asked to report the highest education level attained by their fathers and mothers. The SDQ offered the following response opt ions: grade school, some high school, high school diploma, business school, some college, associate's degree, BA degree, some graduate school, and graduate degree. We collapsed these nine categories into five: No high school diploma; High school diploma; Some higher education; College degree (BA degree); Beyond BA. Each of the filters increased the representation of students whose parents had at least a bachelor's degree. The three filters taken together increased the representation of children of college-educated mothers by 50% or more (Table 7). For example, 18% of SAT-takers but 30% of "admitted" students had mothe rs with more than a BA degree. The two categories of mothers with a BA or beyond w ere roughly of equal size and showed approximately the same effects. While all th ree screens contributed to this pattern, use of the race-neutral admissions model h ad the largest effect. These increases were offset by decreases in the representation of c hildren of the three categories of less-educated mothers, with proportionately the gre atest reduction occurring for children of mothers with no high school diploma.Table 7 Mother's Education, Highly Selective Campuses, Usin g System Eligibility, Application, and SAT+GPA Admissions Mo delNo HS Diploma HS Diploma Some Higher Ed BA Degree Beyond BA SAT-takers, 19981416341918UC Eligible, 1998 1114322121 Eligible and applied tohigh-selectivity schools, 1998 1213282323 Admitted by neutral model,1998 611252830 The same general pattern appeared with fathers' edu cation, for students who reported that as well, although some specific effects were d ifferent. The admission filtersÂ—particularly the race-neutral admission mod elÂ—had more impact on the representation of children whose fathers had post-g raduate education. Students who reported that their fathers had more than a BA cons tituted 24% of all SAT-takers but 43% of "admitted" students (Table 8).Table 8 Father's Education, Highly Selective Campuses, Usin g System
22 of 39 Eligibility, Application, and SAT+GPA Admissions Mo delNo HS Diploma HS Diploma Some Higher Ed BA Degree Beyond BA SAT-takers, 19981314282024UC Eligible, 1998 1112262229 Eligible and applied tohigh, 1998 1110222333 Admitted by neutralmodel, 1998 67192543Home LanguageStudents responding to the SDQ were given three opt ions for describing the first language they speak at home: only English, English and another language, and another language. As Table 9 indicates, most students who t ook the SAT in 1998 spoke only English at home; 21% of the SAT takers primarily sp oke a language other than English; and 16% spoke a combination of English and another language. The effects of the admission filters on the represe ntation of these three groups were inconsistent. For example, use of the UC system eli gibility filter slightly decreased the representation of students who speak other language s at home, from 21% to 19%; use of the application filter increased their representati on to 25%; and use of the race-neutral admission model reduced it again to precisely the l evelÂ—21%Â—shown among all SAT-takers (Table 9). We suspect that these effects stem from the fact that the categories "English and other language" and "anothe r language" include some of the Asian students who are overrepresented at the most selective UC campuses.Table 9 Representation of Home Languages, Highly Selective Campuses, Using System Eligibility, Application, and SAT+GPA Admiss ion ModelEnglish OnlyEnglish & Other Language Another Language SAT-takers, 1998631621UC Eligible, 1998 651619 Eligible and applied to high, 1998 542125 Admitted by neutral model, 1998 592021The Effects of Alternative Selection Models on Dive rsityThe "X%" policies adopted by California, Texas, and Florida are intended to capitalize on the unequal distribution of low-scoring minority students among high schools in order to increase the representation of racial/ethn ic minorities in the pool of admitted students. An alternative approach to that end is to give weight to other demographic variables that may act as a proxy for race/ethnicit y. In this section, the effects of both
23 of 39 approaches are examined.Top X % Policies and Other Aspects of DiversityAs summarized above, Geiser (1998) simulated the im pact four different X% policies would have on both the racial/ethnic composition an d on the average academic preparedness of groups admitted to the University o f California. Here, we examine the effect of these policies on other measures of ethni c diversity as well. Unlike Geiser (1998), however, we ranked students solely on their GPAs, an approach that is more consistent with the X% policies actually implemente d to date. 5 We modeled four X% rules. The first ranked all stud ents in public high schools statewide by their GPAs and admitted the top 12.5%. The second ranked all students within each school and admitted the top 12.5% from each school. 6 The third rule admitted the top 6% within each school, and the fou rth admitted the top 4% within each school. To yield an admitted group of students that represents 12.5% of graduating public school students, the third and fourth models also accept the top 6.5% and 8.5% of students statewide after removing the top 6% and 4% from within each school, respectively.The baseline rule is to attract the top 12.5% acros s the state. Automatically accepting the top 4% from each school before accepting the remain ing top 8.5% would not appreciably affect the academic qualifications of a dmitted students overall or the proportion of black or Hispanic students. Accepting the top 6% within each school would likewise have little effect on diversity and academic qualifications, but it would reduce the mean SAT of accepted black and Hispanic students appreciably relative to the baseline, by 45 and 33 points respectively.Table 10 Modeled Results of Top 4%, 6% and 12.5% Admission P oliciesWhiteAsianBlackHispanicOtherTotal Top 12.5% Across State (Baseline)% of Admitted Group492921010Mean SAT121112221136112612101204Mean GPA3.873.893.903.933.883.89Top 4% Within HS% of Admitted Group492921010Mean SAT121412231117111512101204Mean GPA3.883.903.873.933.883.89Top 6% Within HS% of Admitted Group482921110Mean SAT121612211091109312081200Mean GPA3.893.903.843.893.883.89
24 of 39 Top 12.5% Within HS% of Admitted Group42274189Mean SAT11981173100199911681145Mean GPA3.873.883.603.683.853.83 In contrast, accepting the top 12.5% within each sc hool would have dramatic effects compared with the baseline condition of accepting t he top 12.5% statewide. The percentage of admitted students who are black would double, from 2% to 4%, and the percentage of Hispanic students would increase from 10% to 18%. This increase in diversity, however, would occur at the cost of a la rge drop in the academic qualifications of admitted minority students. The mean SAT scores of black and Hispanic students would drop 135 and 127 points, respectively, and th eir mean GPAs would drop by .3 and .25, respectively. The academic qualifications of the total admitted pool would drop as well, although less markedly. The mean SAT would drop 59 points relative to the baseline rule.Accepting the top 12.5% within each school rather t han the top 12.5% across the state to would have similar effects on other aspects of dive rsity. Table 11 shows that moving from 12.5% statewide to accepting the top 4% or 6% within schools would have little impact on the distribution of geographic location, first language, or education level of a student's mother. However, accepting the top 12.5% within schools increases the representation of urban students, students who spea k a language other than English at home, and students whose mothers have limited educa tion.Table 11 Impact of Top X% Policies on Other Aspects of Diver sity12.5% AcrossState 4% WithinSchool 6% WithinSchool 12.5% WithinSchool LocationUrban.22.214.171.124Suburban.126.96.36.199Rural.06.06.06.07Home LanguageEnglish.188.8.131.52Eng. and OtherLanguage .184.108.40.206 Other than English.220.127.116.11Mother's EducationNo HS diploma.07.08.09.15HS Diploma.18.104.22.168Some Higher Ed..22.214.171.124College Degree.126.96.36.199
25 of 39 Beyond BA.188.8.131.52Missing.03.03.03.03Giving Preference to Other Aspects of DiversityThe race-neutral admission models presented above r esult in an over-representation of white and Asian students. The model for highly sele ctive schools also results in an overrepresentation of suburban students, students w ho speak only English at home, and students whose parents are highly educated. In this section, we explore the impact on racial/ethnic composition of giving preference to s tudents who are from low-income families, whose mothers have little education, who attended high school in urban or rural areas, who attended high schools with low gra duation rates, or who attended high schools in which a high percentage of students rece ived free or reduced-price lunch. 7 In each of these analyses, we awarded the equivalen t of a 200-point SAT bonus to students who came from the most disadvantaged backg round in terms of one of these variables. The preference awarded for each step on a variable depended on the number of categories the variable had. For example, in thi s analysis, mother's education had only two categories (BA or beyond versus no BA), while i ncome was broken into 14 categories (Table 12). Accordingly, while all stude nts whose mothers lacked a BA were given the full 200-point preference, each decrease of one step on the income variable warranted an additional 1/14 of the total 200 point s, or roughly 17 points per step.Table 12 Variables, Number of Levels, and Preference per Ste p Applied in Alternative ModelsVariableLevelsEffective SAT Point Boost Per LevelIncome1416.67 per stepMother's EducationBA or Beyond No BA 0 200 HS Graduation Rate>75% 50 to 75% <50% 0 100 200 Free/Reduced LunchContinuous, from 0% to 95% 2.1 for each 1 percent increasein percent free lunch School LocationSuburban/Other Urban/Rural 0 200 Table 13 displays the effect of giving preference t o each variable, first individually and then in combination. When combinations of variables were used, the maximum impact of the combination was set to 200 points. Since sch ool-level data were available only for public schools, these analyses were run on a reduce d data set. The first two rows of Table 13 compare the results of the SAT and GPA-onl y models and show that the reduced and full data set yielded about the same et hnic mix of admitted students. Giving preference to students based on any of the v ariables decreases the representation
26 of 39 of Asian students and increases that of white, blac k, and Hispanic students. Giving preference to students from schools that have low g raduation rates and that are located in either urban or rural settings has the largest impa ct on the representation of white students and the smallest on that of Hispanic and b lack students. However, giving preference to students whose mothers are less welleducated or whose families are poor has the largest impact on the representation of His panic and black students. Perhaps most important, even the largest effects of giving preference based on demographic variables do not come close to making t he representation of black and Hispanic students in the admitted groups proportion ate to their numbers in the pool of potential students. As Table 13 shows, even giving preference to urban and rural students with low family incomes, whose mothers hav e not completed college, and who attend high schools with low graduation rates and h igh free and reduced lunch rates still results in a dramatic underrepresentation of black and Hispanic students and an overrepresentation of white students.Table 13 Students Admitted (%)Modeled VariableAsianBlackHispanicWhiteOtherDecline to state SAT + GPA Full Sample184.108.40.206220.127.116.11SAT + GPA Alt. Sample18.104.22.1680.07.52.4SAT + GPA + Income32.62.311.246.07.10.9SAT + GPA + Mother's Education30.02.211.747.97.01.3SAT + GPA + Location22.214.171.124126.96.36.199SAT + GPA + Graduation Rate188.8.131.52.87.12.6SAT + GPA + Free Lunch184.108.40.206.47.02.5SAT + GPA + Income + Free Lunch+ Mother's Ed + Location 220.127.116.1118.104.22.168DiscussionOur analyses addressed two broad questions: what st ages of the admissions process produce the under representation of minorities, and what effects might different admissions processesÂ—including both a strict race-n eutral policy and alternative preferences based on variables other than race and ethnicityÂ—have on the diversity of the student population? Investigating these general questions shed light on several others as well: the extent of racial/ethnic preferences in place before the end of affirmative action, the relationships between preferences and t he selectivity of campuses, and the effects of alternative admissions policies on the c haracteristics of admitted students, both minority and non-minority.Replacing the former admissions process that includ ed preferences with a race-neutral model based solely on GPA and SAT scores had major effects at the two most selective campuses in the UC system but much smaller effects at both moderateand
27 of 39low-selectivity campuses. Kane (1998b) found a simi lar pattern in national data, but the present analyses show that this pattern is maintain ed even within a single university system in which some admissions criteria are common across campuses. Both black and Hispanic populations were also noticeably underrepr esented in our moderately and least selective environments, but this under representati on stemmed primarily from factors other than the actual admissions processÂ—in particu lar, whether a student decided to take the SAT and whether the student met the minimu m eligibility criteria for the UC system. Once students passed these two hurdles, the actual admission decision had a substantial impact on the representation of black a nd Hispanic students only for highly selective campuses.The adverse effects of a race-neutral admissions pr ocess were complex. An under-representation of Hispanics (but not of black s) arose because of the large percentage of the former group who decided not to t ake the SAT. (Because we lacked scores for those students, we could not estimate ho w many would have been admitted had they taken the test.) Scores had an adverse imp act at two stages, the UC eligibility stage and the campus-level race-neutral admissions process. However, these effects were in some ways duplicative, and eliminating the UC el igibility screen had little impact on the composition of the groups admitted to highly se lective campuses if the campus-level admissions process remained unaltered. The decision to apply to selective campuses had little impact on diversity; the race-neutral admiss ions model would have produced a similar mix of students even if all students who ha d taken the SAT had applied. The adverse impact of a race-neutral admissions pol icy was not solely the result of group differences in scores on admissions tests. A race-n eutral model based solely on GPA also produced an under-representation of minorities albeit a less severe one. The effects of using GPA alone are smaller because the gap betw een groups in grades is smaller than the gap in average scores. The reasons for thi s difference and the potential effects of relying more on GPA, however, remain uncertain.None of the alternative admissions models we analyz ed could replicate the composition of the student population that was in place before the termination of affirmative action in California. Giving preference to students on the ba sis of other socioeconomic or demographic variables had only modest effects on th e representation of black and Hispanic students; none that we examined brought mi nority students to proportional representation. Some of these preferences, however, increased the representation of whites at the cost of Asians. Guaranteeing admissio n to top students within each schoolÂ—the "X-percent rules"Â—would substantially in crease the representation of minority students only if the percentage within eac h high school guaranteed admission is large. Of the models we examined, only admitting th e top 12.5 percent of students from each high schoolÂ—in effect, basing admission to the UC system solely on rank within high schoolsÂ—led to a large increase in the represe ntation of black and Hispanic students. Applying the 12.5% rule, however, had a l arge cost: it caused a sizable drop in both average SAT scores and average GPA, and that d ecline was particularly large for black and Hispanic students. As Geiser (1998, p. 4) noted, "Redefining the UC eligibility pool to include the top 12.5% of each school would, in short, produce a bifurcated eligibility pool with severe academic disparities a long racial/ethnic lines." Admissions systems differ greatly, and the UC syste m studied has elements not shared by many othersÂ—in particular, the dual screening, f irst for UC system eligibility and subsequently for admission to a specific campus. Mo reover, the effects of preferences
28 of 39and other admissions policies depend on the charact eristics of the student populations from which universities draw. For example, the Hisp anic population in Florida is unlike that in California in several important respects, a nd the effects of admissions policies on access for Hispanic students therefore could be sub stantially different in Florida. Nonetheless, we expect that many of the basic concl usions we reached in examining the California system will apply at a general level to many university systems nationwide because group differences in prior academic perform ance and test scores are typically large. The task of providing access to postsecondar y education for underrepresented minorities without frank preferences is likely to b e difficult and complex throughout the nation. Many of the alternatives will have unintend ed effects, such as lowering the average level of qualification among admitted minor ity students. However, some important details of the impact of alternative admi ssions policies will vary from one system and population to another. Therefore, it wou ld be prudent to examine proposed alternatives carefully before implementation and to monitor their effects once implemented in order to maximize their positive eff ects and minimize unintended outcomes.NotesThis research was conducted under the auspices of t he National Board on Educational Testing and Public Policy (NBETPP), and is publishe d here with the Board's permission. The NBETPP is located in the Lynch School of Educat ion at Boston College and is an independent body created to monitor assessment in A merican education. The NBETPP provides research-based information for policy deci sion making, with special attention to groups historically underserved by the education al system. In particular, the Board a) monitors testing programs, policies, and products; b) evaluates the benefits and costs of specific testing policies; and c) evaluates to what extent professional standards for test development and use are met in specific contexts. Note that Geiser's simulation modeled a policy tha t differs from the actual 4% policy implemented in California. Rather than selecting th e top 4% of students from within each high school based on their GPAs, Geiser based selection on students' combined high school GPA and SAT scores. More precisely, the models were weighted least squ ares regressions of logits of admissions probabilities on GPA and SAT. This is eq uivalent except in estimation method to a logistic regression of the probability of admission on GPA and SAT. See Appendix A. The estimates in Table 2, unlike those in Table 1, do not use the UC system-eligibility screen. That screen is based in part on SAT scores; applying it here would therefore not provide a clear contrast between admissions models that do and do not use SAT scores. The estimates in Table 2 reflect only students who have taken the SAT, however, as we have data for only those students. 4 Because could not locate data that describe the dis tribution of all high school graduates in terms of school location, school type, or househ old language use, we were compelled to use student self-reports from the SDQ for this i nformation. Therefore, the tables in
29 of 39this section consider no groups larger than the poo l of SAT-takers and lack the "Graduates" rows that appear in the tables in the p revious section. 5 Geiser calculated an "Academic Index score" using t he following formula: AI = 1,000GPA + 2.5SAT. Geiser then ranked students with in schools based on AI. 6 To determine the top 12.5 % within a graduating cl ass, we used data from the CDE to establish the number of students who graduated and multiplied this by .125. Students were then ranked by their GPA within schools, and t he number of students representing the top 12.5% within the graduating class were admi tted. The same procedure was repeated for the 6% and 4% models. 7 These analyses roughly follow the approach Wightman (1997) took in examining the result of giving preference to students with disadv antaged backgrounds for law school decisions. 8 Because they are simpler to interpret, we also esti mated linear probability models. As expected, however, they were problematic. In some c ases, they yielded considerably weaker fits, gave impossible estimates for some cel ls, and showed inappropriate residuals. 9 Values of 0 and 1 were set to .001 and .999, respec tively, to calculate logits. 10 In the case of Berkeley, the interactive model pred icted somewhat better than the non-interactive model, but nonetheless yielded unre asonable estimates for some cells. In the case of Irvine and Santa Cruz, the interaction term added little to prediction.ReferencesBakke v. Regents of the University of California, 5 53 P. 2d 1152 (Cal., 1976). Denniston, L. (2000, July 28). Judge dismisses inte rpretation of high court's Bakke decision Antiaffirmative action ruling could be n ext test for justices' current view. The Baltimore Sun p. 3A. Feinberg, W. (1998). On higher ground: Education and the case for affirm ative action Columbia University, New York: Teachers College Pre ss. Fletcher, M. A. (1997). UC may drop SAT entry requi rement; California system fears sharp decline in Black, Latino enrollment. The Washington Post September 20, p. A3. Holley, D. and Spencer, D. (Winter, 1999). The Texa s 10 percent plan. Harvard Civil Rights-Civil Liberties Law Review.Geiser, S. (1998). Redefining UC's Eligibility Pool to Include a Perce ntage of Students from Each High School: Summary of Simulation Result s. Oakland, CA: University of California Office of the President.Gratz and Hamacher v. Bollinger and the University of Michigan, 97-CV-75231-DT (2000).
30 of 39Grutter v. Bollinger and the University of Michigan Civil Action # 97-75928. Hopwood v. Texas, 78 F. 3d 932, 946 (5th Cir. 1996) Kane, T. J. (1998a). Misconceptions in the debate o ver affirmative action in college admissions. In G. Orfield and E. Miller (Eds.). Chilling Admissions: The Affirmative Action Crisis and the Search for Alternatives. Cambridge, MA: Harvard Education Publishing Group, pp. 17-32.Kane, T. J. (1998b). Racial and ethnic preferences in college admissions. In C. Jencks and M Phillips, The Black-White Test Score Gap Brookings, pp. 431-456. Karabel, J. (1998). No alternative: The effects of color-blind admissions in California. In G. Orfield and E. Miller (Eds.). Chilling Admissions: The Affirmative Action Crisis and the Search for Alternatives. Cambridge, MA: Harvard Education Publishing Group, pp. 33-50.Siegel, R. (1998). The racial rhetorics of colorbli nd constitutionalism: The case of Hopwood v. Texas In Post, R. and Rogin, M. (Eds.), Race and representation: Affirmative action New York: Zone Books. Texas Higher Education Coordinating Board. (1997). Alternative diversity criteria: Analyses and recommendations. Austin, Texas: Texas Higher Education Coordinating Board.Tracy v. Board of Regents. CV 497-45 (2000).United States v. Fordice, No. 90-1205 (1992).U.S. Department of Education (1994). What do student grades mean? Differences across schools Research Report (January). University of California Board of Regents v. Bakke, 438 U.S. 265 (1978). Wightman, L. F. (1997). The threat to diversity in legal education: An empirical analysis of he consequences of abandoning race as a factor i n law school admission decisions. New York University Law Review 72(1), 1-53.About the AuthorsDaniel Koretz Harvard Graduate School of Education 415 Gutman Library 6 Appian Way Cambridge, MA 02138Phone: (617) 384-8090 Fax: (617) 496-3095 E-mail: firstname.lastname@example.orgDaniel Koretz is a professor at the Harvard Graduate School of E ducation and
31 of 39Associate Director of the Center for Research on Ev aluation, Standards, and Student Testing (CRESST). His work focuses primarily on edu cational assessment and recently has included studies of the validity of gains in hi gh-stakes testing programs, the effects of testing programs on schooling, the assessment of students with disabilities, and international differences in the influences on math ematics achievement. Michael Russell is a senior research associate with the National B oard on Educational Testing and Public Policy and a Research Professor at the Center for the Study of Testing, Evaluation and Educational Policy at Bosto n College. His research interests include educational testing and technology. E-mail: Mike Russell email@example.comChingwei David Shin is currently a doctoral student enrolled in the Ed ucational Statistics and Measurement program at University of Iowa. His research interests include validity issues in alternative assessments, item re sponse theory, equating, scaling, and statistical methodology. E-mail: firstname.lastname@example.orgCathy Horn is a research associate with the Civil Rights Proj ect at Harvard University. She is currently working on a project assessing the effects of race-neutral admissions policies on diversity at public universities. E-mail: email@example.comKelly Shasby Burling is working on her PhD in Educational Research, Mea surement, and Evaluation at Boston College. She is currently working at Harvard University's Center for International Development within the Joh n F. Kennedy School of Government on projects using technology in educatio nal settings to promote sustainable development in developing countries.Appendix A Constructing Race-Neutral Admissions ModelsThe baseline admission models presented in this rep ort are intended to reflect some of the most important characteristics of the Universit y of California system, but it was not possible to match that system precisely. Unlike Bow en and Bok (1998), we lacked data from campuses on the characteristics of individual students, and we lacked important aggregate variables separately by race, such as the probability of acceptance given SAT scores and HSGPA. We lacked campus-level data on th e characteristics of the relatively few students admitted despite failing to meet the U C eligibility requirements ("admitted by exception"). More important, we lacked campus-le vel information on the more numerous UC-eligible students who were admitted for reasons other than only their academic performance, variously measured. Diverse f actors, including personal disadvantage and school characteristics, can be use d in deciding whether to admit these students, who can constitute 25% to 50% of an admit ted class (http://www.ucop.edu/pathways/infoctr/introuc/selec t.html). Even without this information, however, it was poss ible to create an approximation to admission in the UC system as it would operate with out racial preferences. The steps we followed are presented here.
32 of 39 The starting point for our baseline models was data showing the numbers of total and accepted applicants by SAT score and HSGPA, separat ely by campus, for all programs except Engineering. In the data we obtained, SAT sc ores were broken into five ranges, and HSGPA was broken into six. Table A.1 shows the data we obtained for one of the campuses; these were not further broken into racial /ethnic categories. We analyzed the probability of admission in each cell of this matri xÂ—that is, the ratio of the number admitted to the total number of applicantsÂ—separate ly for each of the eight UC campuses.Table A.1 Admissions Probabilities, Berkeley All Programs Except Engineering, 1999GPASAT Composite Score 490 790 800 9901000 1190 1200 13901400 1600 Overall 2.82 2.99 149 / 10115 / 416 / 0280 / 14 6.70%3.50%6.70%5.00% 3.00 3.29 61 / 6288 / 23831 / 65730 / 48138 / 92048 / 151 9.80%8.00%7.80%6.60%6.50%7.40% 3.30 3.59 65 / 11408 / 331336 / 941620 / 116423 / 453852 / 29 9 16.90%8.10%7.00%7.20%10.60%7.80% 3.60 3.89 52 / 4421 / 571726 / 1753025 / 414830 / 2596054 / 9 09 7.70%13.50%10.10%13.70%31.20%15.00% 3.90 3.99 01-May59 / 14353 / 48798 / 181198 / 1071413 / 351 20.00%23.70%13.60%22.70%54.00%24.80% 421 / 5210 / 591673 / 5064775 / 21003179 / 2405 9858 / 5075 23.80%28.10%30.20%44.00%75.70%51.50% Overall204 / 271386 / 186 6068 / 89811063 / 2863 4784 / 2825 25796 / 7072 13.20%13.40%14.80%25.90%59.10%27.40%SOURCE: University of California, Office of the Pre sident (www.ucop.edu/pathways/infoctr/introuc/prof_except. html).As expected, these data show tremendous differences in the selectivity of the UC campuses. The probability of admission to Berkeley along with UCLA, the most
33 of 39selective, is low for all applicants other than tho se with both high SAT scores and high HSGPA (Figure A.1). For those with low SAT scores, only very high grades increase the probability of admission at all. Similarly, only ve ry high SAT scores help students with grades as low as B (3.0), and even high scores do n ot increase the probability of admission greatly. Figure A.1. Probability of Admission to Berkeley, b y SAT and GPA Riverside presents a dramatically different picture (Figure A.2). Most students with either high GPA or high SAT scores are admitted, and accep tance rates are above 50% for most groups in the graph. Note that as a result, th e relationship between admissions probabilities and both SAT scores and grades is rel atively weak; lines drawn between most points in planes parallel to either the GPA or the SAT axis in Figure A.2 have shallow slopes.
34 of 39 Figure A.2. Probability of Admission to Riverside, by SAT and GPA To estimate the probability of admission for indivi dual students, we fit models estimating admission probability as a function of S AT scores and GPA, separately for each campus. Given the dichotomous outcome and the distribution of probabilities (the number of cells with either very high or very low p robability), we used logistic models to estimate both non-interactive and interactive mo dels. All models were weighted by the number of applicants in each cell. 8 Because all of the variables were categorical, the logistic models could be estimated as ordinary leas t squares models by taking the logits of the probabilities for each cell 9 : These are equivalent to logistic probability models but are simpler to estimate. For example, model 1 is equivalent to:
35 of 39 These simple logistic models fit the data closely. The R2 values for the non-interactive models, adjusted for shrinkage, were all greater th an or equal to.79, and six of the eight were greater than or equal to .90. The interaction added appreciably to the fit in the case of Berkeley and UCLA but had little impact elsewher e. Examination of the data and the models suggested th at the UC campuses fell into the following three levels of selectivity. High selectivity. Berkeley and UCLA were clearly mo re selective than any of the other campuses. Although the models for these two s chools had substantially different parameter estimates, the probabilities th ey predicted were very similar. Moderate selectivity. This group includes four scho ols: Irvine, Davis, Santa Barbara, and San Diego. They appeared to place some what different weights on GPA and SAT scores. Irvine and San Diego showed gre ater effects of GPA than did Davis and Santa Barbara. Santa Barbara and San Diego showed stronger effects of SAT scores than did Irvine or, especiall y, Davis. As a group, however, they were distinct from the highand low-selectivi ty schools. Low selectivity. Santa Cruz and Riverside appeared to be the least selective of the eight campuses. They gave similarly little weight t o low SATs; Riverside gave less weight to low GPAs. These three groups are the basis for our high-, mod erate-, and low-selectivity scenarios. The high-selectivity scenario was based on the Berk eley campus. The mid-selectivity scenario was based on the Irvine campus, and the lo w-selectivity scenario was based on the Santa Cruz campus. The non-interactive logistic model was used in all cases. 10 Because SAT scores and GPA are correlated in the UC data, we used separate regression models to estimate the effects of selection models based on solely GPA or SAT rather than both. These models were simply: These were estimated using marginal percentages in the data tables such as Table 1. Using data from the College Board, these models wer e applied to records of all California high school seniors who took the SAT in 1995 and 1998. Students' SAT-Total and HSGPA were used to place them in cells correspo nding to the UC admissions probability matrix, and on that basis each student was assigned a probability of admission to a campus at each of the three levels. The models estimated logits, so the estimated probabilities were simply the anti-logits of the model estimates, that is:
36 of 39 These probabilities were multiplied by the counts i n each cell and rounded to get counts of "admitted" students. For certain purposes, the c ounts provided by each model were adjusted to approximate total admissions for all of the campuses (either two or four) at that level of selectivity, but in most cases, only the characteristics of the "admitted" group (i.e., the percentage of admitted students wh o were black) were used. A series of additional flags was created for each s tudent in the College Board database. The data contain no information about actual applic ations but do include the identities of all schools to which students had their SAT scores sent. Students who sent scores to any of the UC campuses were assumed to have applied to that campus. Four flags were created in this way: sent scores to any campus; sen t scores to one of the two high-selectivity campuses; sent scores to any moder ate-selectivity campus; and sent scores to any of the two low-selectivity schools. T hese were treated as application flags but may overestimate applications, presumably by a modest amount. An additional eligibility flag was created using the UC systemwid e eligibility criteria for SAT scores and GPA. The UC requirement that SAT-II scores be s ubmitted was not used in creating this flag. All of these flags were set to 0 when th e condition was not met and to 1 when the condition was met.Applying these screens and our baseline admission m odels in various combinations allowed us to examine the effects of various stages of the admission process and to simulate the effects of alternatives on the composi tion of the accepted group. For example, removing the application flag provides an upper-bound estimate of the effects of efforts to encourage all UC-eligible students to apply to all campuses; removing the eligibility screen estimates the impact of moving t o a system in which students apply directly to UC campuses without first being filtere d by a systemwide eligibility screen.Appendix B Comparison of Merged and Full DatabasesAs noted in the body of this report, data on high s chool characteristics were unavailable for many of the California students for whom we had data from the College Entrance Examination Board. When school characteristics were not needed, we used the full database, but analyses involving any school charact eristics were necessarily conducted with a reduced database. This Appendix briefly desc ribes the two databases. The full database was defined as all students in th e College Board database who had valid data on GPA (variable = RECUMGPA) and SAT sco res (variable = SATTOTAL). In 1998, that selection criterion left a total of 1 31,406 out of 152,680 students in the College Board California data. A key school variabl e obtained from the California Department of Education (Educational Demographics U nit, http://data1.cde.gov/dataquest) was counts of stude nts in grade 12. We were able to merge this variable into the records of 93,027 stud ents who also had valid data on SAT scores and GPA. These 93,027 students are represent ed in the merged database. Thus, merging school data caused a loss of 29% of the ful l database.
37 of 39 Although this sample loss was large, tabulations su ggest that it did not materially affect our analyses. Table B1 provides a comparison of the racial/ethnic composition of the full and merged databases at four stages of the process of admission to the highly selective campuses. In all cases, the percentages are similar More important for our purposes, the change in percentages caused by each of the filters is similar in the merged and full databases. The conclusions presented in the paper w ould not differ greatly if one of these databases were substituted for the other.Table B1 Racial/Ethnic Composition, Merged and Full Database s Model Based on SAT+GPA (Row Percents) Asian, Pacific Islander Black or African-American HispanicWhiteOtherDecline to State Merged Data SAT-takers, 199824.26.619.522.214.171.124UC eligible, 199827.43.515.3126.96.36.199Eligible and applied to highselectivity schools, 1998 38.43.514.7188.8.131.52 Admitted by neutral model,1998 184.108.40.2060.06.92.4 Full Database Without Merge SAT-takers, 199822.46.619.041.95.83.2UC eligible, 199825.33.6220.127.116.11.9Eligible and applied to highselectivity schools, 1998 36.03.714.818.104.22.168 Admitted by neutral model,1998 22.214.171.1241.77.02.8Copyright 2002 by the Education Policy Analysis ArchivesThe World Wide Web address for the Education Policy Analysis Archives is epaa.asu.edu General questions about appropriateness of topics o r particular articles may be addressed to the Editor, Gene V Glass, firstname.lastname@example.org or reach him at College of Education, Arizona State University, Tempe, AZ 8 5287-0211. The
38 of 39Commentary Editor is Casey D. Cobb: email@example.com .EPAA Editorial Board Michael W. Apple University of Wisconsin Greg Camilli Rutgers University John Covaleskie Northern Michigan University Alan Davis University of Colorado, Denver Sherman Dorn University of South Florida Mark E. Fetler California Commission on Teacher Credentialing Richard Garlikov firstname.lastname@example.org Thomas F. Green Syracuse University Alison I. Griffith York University Arlen Gullickson Western Michigan University Ernest R. House University of Colorado Aimee Howley Ohio University Craig B. Howley Appalachia Educational Laboratory William Hunter University of Calgary Daniel Kalls Ume University Benjamin Levin University of Manitoba Thomas Mauhs-Pugh Green Mountain College Dewayne Matthews Education Commission of the States William McInerney Purdue University Mary McKeown-Moak MGT of America (Austin, TX) Les McLean University of Toronto Susan Bobbitt Nolen University of Washington Anne L. Pemberton email@example.com Hugh G. Petrie SUNY Buffalo Richard C. Richardson New York University Anthony G. Rud Jr. Purdue University Dennis Sayers California State UniversityÂ—Stanislaus Jay D. Scribner University of Texas at Austin Michael Scriven firstname.lastname@example.org Robert E. Stake University of IllinoisÂ—UC Robert Stonehill U.S. Department of Education David D. Williams Brigham Young University EPAA Spanish Language Editorial BoardAssociate Editor for Spanish Language Roberto Rodrguez Gmez Universidad Nacional Autnoma de Mxico email@example.com
39 of 39 Adrin Acosta (Mxico) Universidad de Guadalajaraadrianacosta@compuserve.com J. Flix Angulo Rasco (Spain) Universidad de Cdizfelix.firstname.lastname@example.org Teresa Bracho (Mxico) Centro de Investigacin y DocenciaEconmica-CIDEbracho dis1.cide.mx Alejandro Canales (Mxico) Universidad Nacional Autnoma deMxicocanalesa@servidor.unam.mx Ursula Casanova (U.S.A.) Arizona State Universitycasanova@asu.edu Jos Contreras Domingo Universitat de Barcelona Jose.Contreras@doe.d5.ub.es Erwin Epstein (U.S.A.) Loyola University of ChicagoEepstein@luc.edu Josu Gonzlez (U.S.A.) Arizona State Universityjosue@asu.edu Rollin Kent (Mxico)Departamento de InvestigacinEducativa-DIE/CINVESTAVrkent@gemtel.com.mx email@example.com Mara Beatriz Luce (Brazil)Universidad Federal de Rio Grande do Sul-UFRGSlucemb@orion.ufrgs.brJavier Mendoza Rojas (Mxico)Universidad Nacional Autnoma deMxicojaviermr@servidor.unam.mxMarcela Mollis (Argentina)Universidad de Buenos Airesmmollis@filo.uba.ar Humberto Muoz Garca (Mxico) Universidad Nacional Autnoma deMxicohumberto@servidor.unam.mxAngel Ignacio Prez Gmez (Spain)Universidad de Mlagaaiperez@uma.es Daniel Schugurensky (Argentina-Canad)OISE/UT, Canadadschugurensky@oise.utoronto.ca Simon Schwartzman (Brazil)Fundao Instituto Brasileiro e Geografiae Estatstica firstname.lastname@example.org Jurjo Torres Santom (Spain)Universidad de A Coruajurjo@udc.es Carlos Alberto Torres (U.S.A.)University of California, Los Angelestorres@gseisucla.edu