The issue in this case is whether the University of Maryland at College Park may maintain a separate merit scholarship program that it voluntarily established for which only African-American students are eligible. Because we find that the district court erred in finding that the University had sufficient evidence of present effects of past discrimination to justify the program and in finding that the program is narrowly tailored to serve its stated objectives, we reverse the district court's grant of summary judgment to the University. We further reverse the district court's denial of Podberesky's motion for summary judgment, and we remand for entry of judgment in favor of Podberesky.I

The facts and prior proceedings in this case are set forth at length in our earlier opinion, Podberesky v. Kirwan, 956 F.2d 52 (4th Cir.1992) (Podberesky I). In sum, Daniel Podberesky challenges the University of Maryland's Banneker scholarship program, which is a merit-based program for which only African-American students are eligible. The University maintains a separate merit-based scholarship program, the Francis Scott Key program, which is not restricted to African-American students. Podberesky is Hispanic; he was therefore ineligible for consideration under the Banneker Program, although he met the academic and all other requirements for consideration. Podberesky was ineligible for consideration under the Key program because his academic credentials fell just shy of its more rigorous standards.

In our earlier decision, we remanded the case because the district court had not made a specific finding on whether there was sufficient present effect of the University's past discrimination against African-Americans so as to justify the maintenance of the race-based restriction in the Banneker scholarship program. Podberesky I, 956 F.2d at 57. The district court allowed additional discovery to take place, after which cross-motions for summary judgment were filed. Podberesky v. Kirwan, 838 F.Supp. 1075, 1076-77 (D. Md.1993). The University claimed that four present effects of past discrimination exist at the University: (1) The University has a poor reputation within the African-American community; (2) African-Americans are underrepresented in the student population; (3) African-American students who enroll at the University have low retention and graduation rates; and (4) the atmosphere on campus is perceived as being hostile to African-American students. 838 F.Supp. at 1082. The district court reasoned that if a strong evidentiary basis existed to support any of the four present effects articulated by the University, the Banneker Program would be justified. The district court then found that there was a strong evidentiary basis to support the existence of each of those four present effects. 838 F.Supp. at 1083.

The district court also found that the Banneker Program was narrowly tailored to remedy those four present effects of past discrimination which it found at the University. 838 F.Supp. at 1094. The district court then granted the University's summary judgment motion and denied Podberesky's summary judgment motion. This appeal followed.

Because it chose the Banneker Program, which excludes all races from consideration but one, as a remedial measure for its past discrimination against African-Americans, the University stands before us burdened with a presumption that its choice cannot be sustained. As we have said before,

"Racial and ethnic distinctions of any sort are inherently suspect and thus call for the most exacting judicial examination." Wygant v. Jackson Board of Education, 476 U.S. 267, 273, 106 S.Ct. 1842, 1846, 90 L.Ed.2d 260 (1986) (plurality opinion) (quoting Regents of the University of California v. Bakke, 438 U.S. 265, 291, 98 S.Ct. 2733, 2748, 57 L.Ed.2d 750 (1978) (Powell, J.)). The rationale for this stringent standard of review is plain. Of all the criteria by which men and women can be judged, the most pernicious is that of race. The injustice of judging human beings by the color of their skin is so apparent that racial classifications cannot be rationalized by the casual invocation of benign remedial aims. City of Richmond v. J.A. Croson Co., 488 U.S. 469, 500, 109 S.Ct. 706, 724, 102 L.Ed.2d 854 (1989). While the inequities and indignities visited by past discrimination are undeniable, the use of race as a reparational device risks perpetuating the very race-consciousness such a remedy purports to overcome.... It thus remains our constitutional premise that race is an impermissible arbiter of human fortunes.

Maryland Troopers Ass'n v. Evans, 993 F.2d 1072, 1076 (4th Cir.1993) (parallel citations omitted).

Although the district court correctly recited in its opinion that the standard of review of such an overtly open racial yardstick was strict scrutiny, and despite the fact that that standard has been adopted time and again both by the Supreme Court and by this circuit, e.g., City of Richmond v. J.A. Croson Co., 488 U.S. 469, 109 S.Ct. 706, 102 L.Ed.2d 854 (1989); Hayes v. North State Law Enforcement Officers Ass'n, 10 F.3d 207, 212 (4th Cir.1993), its restlessness in compliance with that standard showed through as it rejected explicitly cases from four other Courts of Appeals and a reference pool of those high school graduates who were in fact eligible for admission to the University: "There is a danger (created in part by the images of microscope and magnifying glass which the term 'strict scrutiny' brings to mind) that a judge will become myopic when confronted with statistics such as these and assume that a single reference pool must be selected. In fact, such a narrowing of perspective is neither necessary nor proper." 838 F.Supp. at 1089. When we add (as we demonstrate below) that both the Supreme Court and this court have used a pool of qualified applicants as a reference pool in other than employment context, it is not out of order to note that this restlessness carried through its opinion. Indeed, in the penultimate section thereof, the court justified this analysis by its statement that "I have reached the conclusion that in our earlier opinions both I and the Fourth Circuit may have construed too rigid a framework of analysis," and this "[b]ecause I have come to believe that (1) precedents involving employment disputes provide imperfect analogies for determining the constitutionality of an affirmative action program in an education context, and (2) focusing solely upon past discrimination in education cases blurs vision and obstructs understanding...." 838 F.Supp. at 1097. Altogether, these insights into the way the district court arrived at its conclusion amount to little more than tacit acknowledgement by that court that the Banneker Program would not withstand strict scrutiny analysis, which it does not.

We have established a two-step analysis for determining whether a particular race-conscious remedial measure can be sustained under the Constitution: (1) the proponent of the measure must demonstrate a " 'strong basis in evidence for its conclusion that remedial action [is] necessary;' " and (2) the remedial measure must be narrowly tailored to meet the remedial goal.1 Maryland Troopers, 993 F.2d at 1076 (citing and quoting City of Richmond v. J.A. Croson Co., 488 U.S. 469, 500, 109 S.Ct. 706, 724, 102 L.Ed.2d 854 (1989); Wygant v. Jackson Bd. of Educ., 476 U.S. 267, 277, 106 S.Ct. 1842, 1849, 90 L.Ed.2d 260 (1986) (plurality opinion)). The purpose of our earlier remand in this case was to allow the district court to determine whether the University could prove that there were present effects of past discrimination which warranted such race conscious remedial action. Podberesky I, 956 F.2d at 56.

At the outset, we note that the district court held that any present effect of past discrimination found by the University would be sufficient under our Maryland Troopers decision and the Supreme Court's opinion in City of Richmond v. J.A. Croson Co., 488 U.S. 469, 109 S.Ct. 706, 102 L.Ed.2d 854 (1989), to justify the program: "UMCP's finding that the consequences of its segregative past continues to be felt ipso facto establishes the necessity for relief." 838 F.Supp. at 1094. However, Croson itself makes clear that the district court's assumption is incorrect. To have a present effect of past discrimination sufficient to justify the program, the party seeking to implement the program must, at a minimum, prove that the effect it proffers is caused by the past discrimination and that the effect is of sufficient magnitude to justify the program. As to the effect justifying the remedial measure, " '[a]bsent searching judicial inquiry into the justification for such race-based measures, there is simply no way of determining what classifications are "benign" or "remedial" and what classifications are in fact motivated by illegitimate notions of racial inferiority or simple racial politics.' " Maryland Troopers, 993 F.2d at 1076 (4th Cir.1993) (quoting Croson, 88 U.S. at 493, 109 S.Ct. at 721 (plurality opinion)). Therefore, the district court was incorrect in stating that if the University found strong evidence to support any of its proffered effects, the program would be justified. The effects must themselves be examined to see whether they were caused by the past discrimination and whether they are of a type that justifies the program. Only then could we consider affirming the district court's grant of summary judgment to the University on this issue.

* Turning to the present effects articulated by the University, we disagree with the district court that the first effect, a poor reputation in the African-American community, and the fourth effect, a climate on campus that is perceived as being racially hostile, are sufficient, standing alone, to justify the single-race Banneker Program. As the district court's opinion makes clear, any poor reputation the University may have in the African-American community is tied solely to knowledge of the University's discrimination before it admitted African-American students. There is no doubt that many Maryland residents, as well as some citizens in other States, know of the University's past segregation, and that fact cannot be denied. However, mere knowledge of historical fact is not the kind of present effect that can justify a race-exclusive remedy. If it were otherwise, as long as there are people who have access to history books, there will be programs such as this one. Our decisions do not permit such a result. See, e.g., Maryland Troopers, 993 F.2d at 1079.

The hostile-climate effect proffered by the University suffers from another flaw, however. The main support for the University's assertion that the campus climate is hostile to African-American students is contained in a survey of student attitudes and reported results of student focus groups.2 For an articulated effect to justify the program, however, there must be a connection between the past discrimination and the effect. United States v. Fordice, --- U.S. ----, ---- n. 4, 112 S.Ct. 2727, 2736 n. 4, 120 L.Ed.2d 575 (1992). The district court recognized this and reasoned that there was a nexus because racial incidents have occurred with some frequency and regularity, which the district court called a "stream," 838 F.Supp. at 1092, since 1970, which is when the district court found that de facto segregation ended at the University. The district court found that "[t]he very nature of the college experience is that younger students learn from older ones.... Since 1970, both black and white students have been handing down racial attitudes that perpetuate a hostile racial climate." 838 F.Supp. at 1093. The district court appears to have found the connection between the University's previous discriminatory acts and the present attitudes obvious, but we have not so found it. The frequency and regularity of the incidents, as well as claimed instances of backlash to remedial measures, do not necessarily implicate past discrimination on the part of the University, as opposed to present societal discrimination, which the district court implicitly held.

Podberesky argues that the claimed hostility did not have its genesis in the University's discriminatory acts of the past. He points to several northern universities that suffer from comparable racial problems.3 The district court rejected this argument for the reason that it found that most northern universities had experienced de facto segregation, and it held that racial hostility on the northern universities' campuses was the present effect of those universities' past de facto, not de jure, discrimination. 838 F.Supp. at 1090-91.

The district court's analysis cannot be sustained on this point. When we begin by assuming that every predominately white college or university discriminated in the past, whether or not true, we are no longer talking about the kind of discrimination for which a race-conscious remedy may be prescribed. Instead, we are confronting societal discrimination, which cannot be used as a basis for supporting a race-conscious remedy. Podberesky I, 956 F.2d at 55 (citing Wygant v. Jackson Bd. of Educ., 476 U.S. 267, 276, 106 S.Ct. 1842, 1848, 90 L.Ed.2d 260 (1986) (plurality opinion)). There is no doubt that racial tensions still exist in American society, including the campuses of our institutions of higher learning. However, these tensions and attitudes are not a sufficient ground for employing a race-conscious remedy at the University of Maryland. See City of Richmond v. J.A. Croson Co., 488 U.S. 469, 498, 109 S.Ct. 706, 724, 102 L.Ed.2d 854 (1989) (majority opinion).

We next turn to the two effects that rely on statistical data: underrepresentation of African-American students at the University and low retention and graduation rates for African-American students. The district court found that there was strong evidence of African-American underrepresentation in the University's entering-student classes. With respect to the low retention and graduation rates, the district court found that the statistics showed that African-American students had higher attrition rates than any other identifiable group on campus.

The district court erred in its analysis of the underrepresentation evidence and the attrition evidence for a fundamental reason: the posture of the case before the district court was that cross-motions for summary judgment had been filed. We review grants of summary judgment de novo. Higgins v. E.I. DuPont de Nemours & Co., 863 F.2d 1162, 1167 (4th Cir.1988).

Rule 56 provides, "The judgment sought shall be rendered forthwith if the pleadings, depositions, answers to interrogatories, and admissions on file, together with the affidavits, if any, show that there is no genuine issue as to any material fact and that the moving party is entitled to a judgment as a matter of law." Fed.R.Civ.P. 56(c). We have held that the rule's language is clear that it is not enough for the district court to determine that the moving party has the winning legal argument; in accepting that argument, the district court must also ensure that there is no genuine issue as to any material fact before a grant of summary judgment is proper. See, e.g., Charbonnages de France v. Smith, 597 F.2d 406 (4th Cir.1979) (reversing grant of summary judgment because of genuine issues of material fact); Stevens v. Howard D. Johnson Co., 181 F.2d 390, 394 (4th Cir.1950) ("The motion for summary judgment ... should be granted only where it is perfectly clear that no issue of fact is involved and inquiry into the facts is not desirable to clarify the application of the law."); see also Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 248, 106 S.Ct. 2505, 2510, 91 L.Ed.2d 202 (1986). "[T]he fact that both parties simultaneously are arguing that there is no genuine issue of fact does not establish that a trial is unnecessary thereby empowering the court to enter judgment as it sees fit." 10A Charles A. Wright, Arthur R. Miller & Mary Kay Kane, Federal Practice and Procedure Sec. 2720 (2d ed.1983).

Taking the facts in the light most favorable to Podberesky, the non-moving party, we find that the district court erred in granting the University's motion for summary judgment. As to the low retention and graduation rates, there is a dispute in the evidence about why African-American students leave the University of Maryland in greater numbers than other students. Podberesky offered evidence tending to show that the attrition rate revealed by the statistics was the result of economic and other factors and not because of past discrimination. The district court rejected Podberesky's study by reasoning that economic concerns are often more pressing for African-American students because many of those students come from less wealthy backgrounds. The district court then reasoned that the disproportionate number of less wealthy African-American families is the result of past discrimination in society. The district court also found some evidence in some of the University's exhibits that showed that the University's poor reputation and hostile climate have an effect on attrition rates. 838 F.Supp. at 1091-92.

As to the underrepresentation, our decisions and those of the Supreme Court have made clear that the selection of the correct reference pool is critical. The district court must first determine as a matter of law whether it is appropriate to apply a pool consisting of the local population or whether another pool made up of people with special qualifications is appropriate. In the employment context, this determination is made by looking at the job requirements. If the job is an unskilled one, the general population is more likely the relevant pool. If, however, the job requires some special skills or training, the relevant pool is made up of only those people who meet the criteria. E.g., Johnson v. Transportation Agency, 480 U.S. 616, 631-32, 107 S.Ct. 1442, 1451-52, 94 L.Ed.2d 615 (1987); Maryland Troopers Ass'n v. Evans, 993 F.2d 1072, 1076-77 (4th Cir.1993). The method of determining the relevant pool by looking at the qualifications needed to take advantage of the opportunity from which minorities historically have been excluded and the prevalence of those qualifications in the population is not limited to the employment context. See City of Richmond v. J.A. Croson Co., 488 U.S. 469, 501-02, 109 S.Ct. 706, 725-26, 102 L.Ed.2d 854 (1989) (majority opinion) (applying analysis to Richmond's minority set-aside program).

The district court rejected a pool which consisted of all graduating high school seniors because that pool "does not take into account even flexible minimum admission requirements." 838 F.Supp. at 1089. Thus, the district court correctly determined the legal issue of whether the appropriate pool was the general population or a smaller qualification specific pool. The district court erred, however, in its attempt to resolve the factual dispute4 about what are the effective minimum admission criteria.5 The district court declined to decide the requisite qualification for membership in the reference pool, but mentioned the percentage of students taking the SAT in Maryland, the minimum course curriculum required, and minimum math and verbal SAT scores. It later found that the percentage of African-American incoming freshmen at UMCP (13%) was less than any of them (17.9% for required course curriculum to 22% of students taking the SAT in Maryland who were African-American).6 838 F.Supp. at 1088-89. It rejected Podberesky's proposed effective minimum criterion for admission, which was based on a combination of SAT scores, high school curriculum requirements, and grade-point averages, because the use of those numbers "ignores the variables in the admissions process and the intergenerational effects of segregated education on the applicant pool." 838 F.Supp. at 1089.7 We are of opinion that the goal of the program, remedying any present effects of past discrimination, cannot be used to lower the effective minimum criteria needed to determine the applicant pool.8 Additionally, the district court erred in resolving, on a summary judgment motion, the factual dispute about the effective minimum criteria for admission.

The factual disputes in this case are not inconsequential and could have been resolved only at trial. A district court may not resolve conflicts in the evidence on summary judgment motions, and the district court erred in so doing here.

We next turn to the denial of Podberesky's motion for summary judgment. An order denying summary judgment is ordinarily not appealable. See, e.g., President & Directors of Georgetown College v. Madden, 660 F.2d 91, 96-97 (4th Cir.1981) (per curiam); Valdosta Livestock Co. v. Williams, 316 F.2d 188 (4th Cir.1963) (per curiam). In the circumstances of this case, however, we may review the district court's denial of Podberesky's motion for summary judgment as well as its grant of summary judgment to the University. The district court's orders on the summary judgment motions disposed of all claims between the parties, and the orders are therefore within our appellate jurisdiction under 28 U.S.C. Sec. 1291. See 10 Wright, Miller & Kane, supra Sec. 2715, at 626.

Even if we assumed that the University had demonstrated that African-Americans were underrepresented at the University and that the higher attrition rate was related to past discrimination,9 we could not uphold the Banneker Program. It is not narrowly tailored to remedy the underrepresentation and attrition problems, and the district court erred in its analysis of this issue as well.

It is difficult to determine whether the Banneker scholarship program is narrowly tailored to remedy the present effects of past discrimination when the proof of present effects is so weak. See Croson, 488 U.S. at 507, 109 S.Ct. at 729 (majority opinion). In determining whether the Banneker Program is narrowly tailored to accomplish its stated objective, we may consider possible race-neutral alternatives and whether the program actually furthers a different objective from the one it is claimed to remedy.10 See Croson, 488 U.S. at 507, 109 S.Ct. at 729 (majority opinion).

A. Attraction to Only High-Achieving Black Students

The district court found that the Banneker Program attracted "high-achieving black students" to the University, which "directly increases the number of African-Americans who are admitted and likely to stay through graduation. Even more importantly, the Program helps to build a base of strong, supportive alumni, combat racial stereotypes and provide mentors and role models for other African-American students. Continuation of the Program thus serves to enhance [the University's] reputation in the African-American community, increase the number of African-American students who might apply to the University, improve the retention rate of those African-American students who are admitted and help ease racial tensions that exist on campus." 838 F.Supp. at 1094-95. In sum, the district court found that the Banneker Program is employed by the University as an effective recruiting tool that draws high-achieving African-Americans to the University. The district court further noted that the University's "success in curing the vestiges of its past discrimination depends upon it attracting high-achieving African-Americans to the College Park campus." 838 F.Supp. at 1095. As we demonstrate below, in conducting its analysis, the district court did not sufficiently connect the problems the University purports to remedy to the Banneker Program: low retention and graduation rates and underrepresentation. If the purpose of the program was to draw only high-achieving African-American students to the University, it could not be sustained. High achievers, whether African-American or not, are not the group against which the University discriminated in the past.

The district court also erred in giving no weight to Podberesky's argument that the Banneker Program is not narrowly tailored because the scholarships are open to non-Maryland residents.11 The district court stated that the goals of the program would be served "whether Banneker Scholars are Maryland natives or not." 838 F.Supp. at 1095 n. 74. It is at once apparent that the Banneker Program considers all African-American students for merit scholarships at the expense of non-African-American Maryland students.

The University, throughout this case, has taken the position that the pool from which the students eligible to enter UMCP is drawn are from "qualified African-American high school students in Maryland," A. 3476, and "the University expects that the racial composition of its student body will reflect the racial composition of qualified college-eligible high school graduates." A. 3476. While all of the prerequisites for membership in the pool were a matter of dispute between the parties, that the University measured its desired number of black students against Maryland high school graduates who are qualified to attend the University is not a matter of dispute. That being true, it is obvious that awarding Banneker Scholarships to non-residents of Maryland is not narrowly tailored to correcting the condition that the University argues, that not enough qualified African-American Maryland residents attend at College Park. Cf. Croson, 488 U.S. at 508, 109 S.Ct. at 729.

The district court found the program to be narrowly tailored to increasing representation because an increase in the number of high-achieving African-American students would remedy the underrepresentation problem. The district court so found because it reasoned that the Banneker Scholars would serve as mentors and role models for other African-American students, thereby attracting more African-American students. The Supreme Court has expressly rejected the role-model theory as a basis for implementing a race-conscious remedy, as do we. Wygant v. Jackson Bd. of Educ., 476 U.S. 267, 276, 106 S.Ct. 1842, 1848, 90 L.Ed.2d 260 (1986) (plurality opinion).

Furthermore, the district court's analysis of underrepresentation, although it relied on various academic criteria to determine eligibility, relied on each relevant criterion item by item instead of in combination. It is axiomatic that if all of the relevant criteria (270 verbal SAT score, 380 math SAT score, UMCP general course-curriculum requirements, and 2.0 GPA) were applied simultaneously, as indeed UMCP itself claims it most commonly does in determining admissions qualifications, the percentage of eligible Maryland residents who are African-American might well be significantly lower than the percentage satisfying the least burdensome of those criteria relied upon.12 In other words, even within the confines of its own analysis, the district court's conclusion is based on flawed reasoning, and results in a series of inconclusive and possibly inflated figures regarding the makeup of the reference pool.

Moreover, and more important, eligibility is not the only relevant criterion in determining the reference pool in this case. We note the critical fact that application for admission to college is voluntary rather than obligatory. In addition, the choice of which institution to attend is voluntary, and is dependent upon many variables other than race-based considerations. Further, economic concerns and other factors, offered by Podberesky below, may induce many otherwise-eligible African-American high school graduates not to enter college in numbers which are proportionately higher than those of their non-African-American peers.

In short, the district court failed to account for statistics regarding that percentage of otherwise eligible African-American high school graduates who either (1) chose not to go to any college; (2) chose to apply only to out-of-state colleges; (3) chose to postpone application to a four-year institution for reasons relating to economics or otherwise, such as spending a year or so in a community college to save money; or (4) voluntarily limited their applications to Maryland's predominantly African-American institutions.13 What if, for example, in some year only 2/3 of those academically eligible African-American Maryland high school graduates applied to any college, while 90% of eligible non-African-American Maryland high school graduates did? What then would be the relevance of measuring the percentage of those eligible against the percentage of African-Americans in the UMCP student body?

We will not speculate as to what extent these variables might reduce the size of the reference pool, since no definitive information regarding these types of statistics is in the record.14 We can say with certainty, however, that the failure to account for these, and possibly other, nontrivial variables cannot withstand strict scrutiny. In analyzing underrepresentation, disparity between the composition of the student body and the composition of a reference pool is significant in this case only to the extent that it can be shown to be based on present effects of past discrimination. In more practical terms, the reference pool must factor out, to the extent practicable, all nontrivial, non-race-based disparities in order to permit an inference that such, if any, racial considerations contributed to the remaining disparity. This the district court simply has not done. The result is no more than a collection of arbitrary figures upon which it held UMCP may rely in its efforts to recruit African-Americans using facially racial classifications.

The Supreme Court has stated in Croson, which involved a 30% racial set-aside quota, that "the 30% quota cannot be said to be narrowly tailored to any goal, except perhaps outright racial balancing." Croson, 488 U.S. at 507, 109 S.Ct. at 729. Although the percentage of African-Americans that was expected to be in each entering class was disputed, see supra Part II.B., it is clear that the district court has implicitly approved the use by UMCP of a similar quota, because it "rests on ... unsupported assumption[s]," Croson, 488 U.S. at 502, 109 S.Ct. at 726, as to the appropriate levels, if any, to remedy the present effects of past discrimination.

The district court has approved the use of the Banneker Program to affirmatively admit African-American students solely on the basis of race until the composition of African-Americans on the University campus reflects the percentage of African-American Maryland high school graduates who potentially might participate in higher education at UMCP, without an accurate determination of either the extent to which the present disparity exists, see supra, or the extent to which that disparity flows from past discrimination, see supra part I. The program thus could remain in force indefinitely based on arbitrary statistics unrelated to constitutionally permissible purposes. Without specific determination of what measure should be used, if any, to remedy the effects of past discrimination that still exist, " 'relief' ... could extend until the percentage of [African-American students at UMCP] mirrored the percentage of [African-Americans] in the population as a whole." Croson, 488 U.S. at 498, 109 S.Ct. at 724. We are thus of opinion that, as analyzed by the district court, the program more resembles outright racial balancing than a tailored remedy program. As such, it is not narrowly tailored to remedy past discrimination. In fact, it is not tailored at all.

The district court also suggested that an increase in the number of high-achieving African-American students would remedy the low retention and graduation rates for African-American students at the University. Podberesky submitted a 1993 study by two University of Maryland professors which indicates that after the freshman year, in which grades are the principal problem, students leave the University for financial and other reasons.15 Specifically, students who left the University "tended to be more likely to provide their own expenses, live off campus with long commutes, have a job with long hours, spend few free hours on campus, and have few friends on campus." Roger W. McIntire & Sandra Smith, Work and Life Styles Among Dropouts and Ongoing College Students, 4 J.A. 1062, 1067 (survey of 455 drop-out and 455 returning University of Maryland students). "[M]ales, minority groups and transfer students show greater attrition because they are more likely to provide their own expenses and have little time for campus activities and friends due to off campus living and work." 4 J.A. at 1068. That study suggests that the best remedy is "campus job opportunities and convenient, attractive, and economically reasonable campus housing ... available to a greater proportion of students." 4 J.A. at 1070-71.

The district court rejected Podberesky's argument because it found that, in addition to economic hardship,

[i]n given cases an absence of commitment to the school because of its poor reputation in the community from which a student comes, the lack of shared experience with family members to help the student through the arduous process of higher education, the absence of African-American members of the faculty to serve as mentors and the existence of a hostile racial atmosphere on campus are other significant contributing factors.

838 F.Supp. at 1091-92.

The causes of the low retention rates submitted both by Podberesky and the University and found by the district court have little, if anything, to do with the Banneker Program. To the extent that the district court's opinion can be read as having found a connection between the University's poor reputation and hostile environment and the Banneker Program, it is on either a role model theory or a societal discrimination theory, neither of which can be sustained. In addition, there is no connection between the Banneker Program and shared experience with family members, African-American faculty members, or jobs and housing. Even if there is some connection between the two, the University has not made any attempt to show that it has tried, without success, any race-neutral solutions to the retention problem. Thus, the University's choice of a race-exclusive merit scholarship program as a remedy cannot be sustained.

Because we find that the University has not shown that its programs and quota goals are narrowly tailored, we reverse the district court's grant of summary judgment to the University. We also reverse the district court's denial of Podberesky's summary judgment motion.

In our first opinion in this case, we required that should no further evidence be available upon remand, summary judgment for Podberesky should be granted.

Simply put, since the summary judgment motion of the University has failed, it may be argued that we might well enter judgment in favor of Podberesky. Because such failure, however, has consisted at least in part of the district court's weighing evidence on a motion for summary judgment, and we expressed no opinion on the subject previously, we have not without more directed entry of judgment in favor of Podberesky. Instead, we have examined the Banneker Program and have concluded it is not narrowly tailored so as to justify its principal feature, which is its mandatory reliance on race as a qualification for participation therein.

The University has had two opportunities to justify its position and has failed.

Accordingly, on remand, the district court will enter its order denying the University's motion for summary judgment, granting Podberesky's motion for summary judgment, and requiring the University to re-examine Podberesky's admission to the Banneker Program as of the date it was made. On such re-examination, the University will be enjoined from enforcing that part of the qualifications for entry into the Banneker Program which require that the applicant be of the African-American race. Following such re-examination, the district court will award appropriate relief if required.

The judgment of the district court is vacated, and the case is remanded for action consistent with this opinion.

VACATED AND REMANDED WITH INSTRUCTIONS.*