Don't tell, don't ask: Narrow tailoring after Grutter and Gratz

Texas Law Review

Volumn 85, Issue 3, 2007, Pages 517-583

  Ayres, Ian ^a   Foster, Sydney ^b  

^a YALE UNIVERSITY   (United States)

^b HARVARD LAW SCHOOL   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 33947718820     PISSN: 00404411     EISSN: None     Source Type: Journal    
DOI: None     Document Type: Review

References (274)

1. 539 U.S. 306 (2003).
2. 539 U.S. 244 (2003).
3. See infra notes 20-25 and accompanying text.
4. See infra notes 26-38 and accompanying text.
5. See infra notes 28-38 and accompanying text.
6. This is another way of stating the requirement that race-neutral alternatives be used when possible. See infra notes 25-26 and accompanying text.
7. See infra subpart IV(A).
8. See infra subpart IV(B).
9. See infra subpart IV(C).
10. See infra section V(A)(1).
11. See infra section V(B)(2).
12. See infra section V(B)(1).
13. 539 U.S. 244, 255-57 (2003).
14. See infra section V(B)(2).
15. See infra section V(B)(1).
16. The University of Michigan Law School and College of Literature, Science, and the Arts granted racial preferences to some, but not all, minority groups. See infra note 49. It is, therefore, not accurate to say that either school gave all minorities preferences. Because using the more accurate term "preferred minorities" is cumbersome, however, we frequently - but not always - use the term "minorities" throughout this Article when what we really mean is "preferred minorities."
17. See infra section IV(B)(2).
18. See infra section IV(B)(2).
19. See infra section V(B)(1).
20. Compare this list to the one Robert Post created to summarize what the Grutter Court stated should be considered as part of the narrow tailoring inquiry. Robert C. Post, The Supreme Court, 2002 Term - Foreword: Fashioning the Legal Constitution: Culture, Courts, and Law, 117 HARV. L. REV. 4, 66-67 (2003) ("The [Grutter] Court . . . holds [that the 'narrowly tailored' prong of the strict scrutiny test] has four components. A race-based affirmative action program (1) must 'not unduly harm members of any racial group'; (2) can be implemented only if there has been a 'serious, good faith consideration of workable race-neutral alternatives that will achieve the diversity the university seeks'; (3) 'must be limited in time'; and (4) must afford each applicant 'truly individualized consideration.'" (footnotes omitted) (quoting Grutter v. Bollinger, 539 U.S. 306, 341, 339, 342, 334 (2003))). Also compare this list to the one Richard Fallon created that describes the elements of the narrow tailoring inquiry in all contexts, not just the equal protection context. Richard Fallon, Strict Judicial Scrutiny, 54 UCLA L. REV. (forthcoming June 2007) (manuscript at 126-37, on file with authors) (listing four inquiries: (1) "[n]ecessity of infringement on a triggering right," (2) "underinclusiveness," (3) "overinclusiveness," and (4) "proportionality").
21. See, e.g., City of Richmond v. J.A. Croson Co., 488 U.S. 469, 506 (1989) (holding a set-aside benefiting individuals from several racial groups from all over the country not narrowly tailored to remedying past discrimination against African Americans in Richmond because, inter alia, the beneficiary group was overinclusive); Wygant v. Jackson Bd. of Educ., 476 U.S. 267, 284 n.13 (1986) (plurality opinion) (criticizing a school board's layoff provision that targeted certain groups of minorities as "undifferentiated" and noting that the board did not justify its selection of the minority groups that the plan favored). One of us has pointed out that "[a]s a theoretical matter, one might also conclude that an underinclusive program is not narrowly tailored [to the remedial interest] if victims of discrimination are arbitrarily excluded from the affirmative action preferences." Ian Ayres, Narrow Tailoring, 43 UCLA L. REV. 1781, 1786 n.13 (1996).
22. See, e.g., Croson, 488 U.S. at 508 (distinguishing the quota at issue in Fullilove v. Klutznick, 448 U.S. 448 (1980), from the one at issue in Croson by noting its flexible waiver and exemption elements and concluding that "programs [such as the one in Fullilove] are less problematic from an equal protection standpoint because they treat all candidates individually, rather than making the color of an applicant's skin the sole relevant consideration"); United States v. Paradise, 480 U.S. 149, 177 (1987) (plurality opinion) (noting approvingly that waivers built into the affirmative action plan at issue give it flexibility); cf. Grutter, 539 U.S. at 334 ("[T]ruly individualized consideration demands that race be used in a flexible, nonmechanical way.").
23. Regents of the Univ. of Cal. v. Bakke, 438 U.S. 265, 318 n.52 (1978) (Powell, J.); cf. Grutter, 539 U.S. at 334 ("To be narrowly tailored, a race-conscious admissions program cannot use a quota system . . . .").
24. See, e.g., Adarand Constructors, Inc. v. Pena, 515 U.S. 200, 238 (1995) (noting that affirmative action programs should not last longer than necessary); Croson, 488 U.S. at 510 ("[D]eviation from the norm of equal treatment of all racial and ethnic groups is a temporary matter . . . ."); cf. Grutter, 539 U.S. at 342 ("[R]ace-conscious admissions policies must be limited in time."). Note that conceptually, narrow tailoring does not entail a limited duration requirement when the compelling government interest is diversity rather than remedying past discrimination. The Sixth Circuit so recognized when it reversed the district court on this point in Grutter. Compare Grutter v. Bollinger, 288 F.3d 732, 751-52 (6th Cir. 2002) (en banc), aff'd, 539 U.S. 306 (2003), with Grutter v. Bollinger, 137 F. Supp. 2d 821, 851 (E.D. Mich. 2001), rev'd, 288 F.3d 732 (6th Cir. 2002) (en banc), aff'd, 539 U.S. 306 (2003).
25. See, e.g., Adarand, 515 U.S. at 237-38 (noting that strict scrutiny requires consideration of race-neutral means); Croson, 488 U.S. at 507 (noting disapprovingly that "there does not appear to have been any consideration of the use of race-neutral means"); see also Wygant, 476 U.S. at 280 n.6 (plurality opinion) ("[T]he term [narrowly tailored] . . . . may be used to require consideration of whether lawful alternative and less restrictive means could have been used."); cf. Grutter, 539 U.S. at 339 ("Narrow tailoring does . . . require serious, good faith consideration of workable race-neutral alternatives that will achieve the diversity the university seeks."). But see Ayres, supra note 21 (noting the tension between requiring that narrowly tailored means be used and encouraging the solution of race-specific problems with race-neutral means).
26. Note that the race-neutral alternatives requirement is an application of this minimum necessary preference principle to the case where racial preferences are not necessary to achieve the objectives of the government. When race-neutral means exist for accomplishing the objectives of the government, then racial preferences are not the minimum necessary and are therefore not allowed.
27. At least as a theoretical matter, narrow tailoring requires not only that preferences not be too large, but also that they not be too small so as to fail to achieve the goals of the relevant compelling government interest. See Ayres, supra note 21, at 1786 n.13.
28. See, e.g., Bakke, 438 U.S. at 357 (Brennan, White, Marshall, and Blackmun, JJ., concurring in the judgment in part and dissenting in part) ("Unquestionably we have held that a government practice or statute which restricts 'fundamental rights' or which contains 'suspect classifications' is to be subjected to 'strict scrutiny' and can be justified only if it furthers a compelling government purpose and, even then, only if no less restrictive alternative is available."); see also Wygant, 476 U.S. at 280 n.6 (plurality opinion) (noting that the narrow tailoring inquiry requires "consideration of whether lawful alternative and less restrictive means could have been used"). Some older cases include language suggesting that strict scrutiny does not demand use of the least restrictive means. See, e.g., United States v. Paradise, 480 U.S. 149, 184 (1987) (plurality opinion) ("Nor have we in all situations 'required remedial plans to be limited to the least restrictive means of implementation.'" (quoting Fullilove v. Klutznick, 448 U.S. 448, 508 (1980) (Powell, J., concurring))); Fullilove, 448 U.S. at 508 (Powell, J., concurring) ("[T]his Court has not required remedial plans to be limited to the least restrictive means of implementation."). In light of more recent cases demanding consideration of race-neutral alternatives and applying a stricter version of strict scrutiny, however, these cases are no longer good law with respect to this point. See supra note 25.
29. Bakke, 438 U.S. at 308 (Powell, J.) ("[T]he remedial action usually remains subject to continuing oversight to assure that it will work the least harm possible to other innocent persons competing for the benefit.").
30. See, e.g., Metro Broad., Inc. v. FCC, 497 U.S. 547, 630 (1990) (O'Connor, J., dissenting) (noting that to pass constitutional muster, programs should not "unduly burden individuals who are not members of the favored racial and ethnic groups"); Paradise, 480 U.S. at 171 (plurality opinion) (stating that the impact on third parties should be considered); Bakke, 438 U.S. at 308 (Powell, J.) (noting that continued oversight typically ensures "the least harm possible" to others); cf. Grutter, 539 U.S. at 341 ("To be narrowly tailored, a race-conscious admissions program must not unduly burden individuals who are not members of the favored racial and ethnic groups." (internal quotation marks omitted)). But see Jed Rubenfeld, Affirmative Action, 107 YALE L.J. 427, 432-67 (1997) (arguing that equal protection analysis should move from an examination of costs and benefits to an examination of impermissible purposes).
31. For a description of the race-neutral alternatives requirement, see supra note 25.
32. Paradise, 480 U.S. at 198-99 (O'Connor, J., dissenting) (citations omitted). Other Justices have recognized that a size inquiry is important but have not framed the inquiry in terms of whether a given size is the smallest necessary. In City of Richmond v. J.A. Croson, Co., the Court criticized the thirty-percent set-aside at issue because the thirty-percent figure rested on "the 'completely unrealistic' assumption that minorities will choose a particular trade in lockstep proportion to their representation in the local population." 488 U.S. 469, 507 (1989) (quoting Sheet Metal Workers v. EEOC, 478 U.S. 421, 494 (1986) (O'Connor, J., concurring in part and dissenting in part)); see also Paradise, 480 U.S. at 179-82 (plurality opinion) (upholding a fifty-percent promotion rate for African Americans, noting that the fifty-percent figure was not arbitrary relative to the twenty-five percent labor pool and the twenty-five percent representation of African Americans in the upper ranks); Fullilove, 448 U.S. at 513-14 (Powell, J., concurring) (noting that the ten-percent figure was "reasonable" because it fell approximately halfway between the percentage of the population who were minorities (seventeen percent) and the percentage of contractors who were minorities (four percent)). Note, however, that these discussions of weight in Paradise and Fullilove are from a plurality decision and a concurrence, respectively. Moreover, the scrutiny applied in these opinions was not full strict scrutiny.
33. See, e.g., Herman Schwartz, The 1986 and 1987 Affirmative Action Cases: It's All Over but the Shouting, 86 MICH. L. REV. 524, 547-48 (1987) (noting how Justice Powell adhered weakly to the strict scrutiny test in his Fullilove concurrence). So, to the extent that these opinions countenance upholding racial preferences without consideration of whether preferences are the minimum necessary, their analysis on this point may not still be controlling.
34. Paradise, 480 U.S. at 199 (O'Connor, J., dissenting). Other Justices have examined the weight given to race and its effect on nonbeneficiaries but have not articulated a standard for determining whether the weight is too much. For example, Justice Powell's concurrence in Fullilove noted the importance of examining "the effect of the set-aside upon innocent third parties." 448 U.S. at 514 (Powell, J., concurring). Justice Powell observed that the set-aside at issue would "reserve about 0.25% of all the funds expended yearly on construction work in the United States for approximately 4% of the Nation's contractors who are members of a minority group" and concluded that it was constitutionally permissible because "the effect of the set-aside [was] limited and so widely dispersed." Id. at 514-15; see also id. at 484 (plurality opinion) ("When effectuating a limited and properly tailored remedy to cure the effects of prior discrimination, such 'a sharing of the burden' by innocent parties is not impermissible."). To the extent that these opinions countenance upholding racial preferences without consideration of whether racial preferences are the minimum necessary, it should be remembered that Fullilove was decided in an era when strict scrutiny was far less strict than it is today, and therefore its analysis on this point may not still be good law.
35. See, e.g., R.A. Lenhardt, Understanding the Mark: Race, Stigma, and Equality in Context, 79 N.Y.U. L. REV. 803, 870 (2004) (describing the Fullilove standard as less "exacting" than the Croson standard); Schwartz, supra note 32, at 547-48 ("Justice Powell's Fullilove opinion was an especially revealing indicator of how weakly he adhered to the strict scrutiny test.").
36. Bakke, 438 U.S. at 317 (Powell, J.).
37. Note that to the extent that the diversity interest in Grutter and Gratz is different from the one articulated in Justice Powell's Bakke opinion, see infra notes 215-21 and accompanying text, the analysis on this point must be adjusted to take account of the differences, see infra note 221.
38. Richard Fallon argues that embedded in the doctrine is a third account of the function of strict scrutiny: "[s]trict [s]crutiny as a [n]early [c]ategorical [p]rohibition." Fallon, supra note 20 (manuscript at 80-85).
39. Croson, 488 U.S. at 493 (plurality opinion); see also Fallon, supra note 20 (manuscript at 89-97) (discussing "[s]trict [s]crutiny as an [i]llicit [m]otive [t]est"); Rubenfeld, supra note 30, at 428 ("One powerful function of strict scrutiny has always been that of 'smoking out' invidious purposes masquerading behind putatively legitimate public policy.").
40. See Adarand Constructors, Inc. v. Pena, 515 U.S. 200, 229-30 (1995) (noting that "whenever the government treats any person unequally because of his or her race, that person has suffered an injury that falls squarely within the language and spirit of the Constitution's guarantee of equal protection" and stating that "[t]he application of strict scrutiny . . . determines whether a compelling governmental interest justifies the infliction of that injury"); Rubenfeld, supra note 30, at 428 (stating that strict scrutiny has "become a cost-benefit test measuring whether a law that falls (according to the Court itself) squarely within the prohibition of the equal protection guarantee is justified by the specially important social gains that it will achieve"); see also Fallon, supra note 20 (manuscript at 85-89) (discussing "[s]trict [s]crutiny as a [w]eighted [balancing [t]est").
41. See infra section V(B)(1).
42. Indeed, as Part IV will argue, not only did the Grutter and Gratz Courts not develop methodologies for conducting the cost-benefit calculus, but also they effectively replaced the cost-benefit calculus with an "individualization" requirement. And, while the Grutter and Gratz Courts developed some methodologies for measuring the costs of affirmative action, those methodologies do not capture costs that are constitutionally relevant. See infra section V(A)(2).
43. We are not the first to suggest that the Law School may have placed more weight on race than the College, but we are the first to defend the proposition with rigorous empirical analysis. Richard Sander is the only scholar who has attempted to prove empirically that the Law School placed more weight on race than the College, see infra notes 84-87 and accompanying text, but he ignores the outcome dimension, and there are several problems with his analysis of the means dimension, see infra note 87. Other scholars have suggested that the Law School may have placed more weight on race but have provided no empirical support for this assertion. Abigail Thernstrom and Stephan Thernstrom state that the Law School's preferences were greater than the College's, but the source they cite for support of this proposition, Judge Boggs's dissenting opinion in the Sixth Circuit in Grutter, made no claims about whether the College or the Law School provided greater preferences.
44. See Abigail Thernstrom & Stephan Thernstrom, Secrecy and Dishonesty: The Supreme Court, Racial Preferences, and Higher Education, 21 CONST. COMMENT. 251, 259-60 & n.38 (2004) (citing Grutter v. Bollinger, 288 F.3d 732, 796 (6th Cir. 2002) (Boggs, J., dissenting), aff'd, 539 U.S. 306 (2003)). Thernstrom and Thernstrom's statement is therefore without empirical support. Similarly, Colin Diver states that "[b]ased on the statistics unearthed by the plaintiffs, and summarized in the dissenting opinion of Chief Justice Rehnquist [in Grutter], it appears that the advantage conferred on minority applicants by the law school was, in relative terms, as weighty as that conferred by the undergraduate point system."
45. Colin S. Diver, From Equality to Diversity: The Detour from Brown to Grutter, 2004 U. ILL. L. REV. 691, 718 (footnote omitted). Diver offers no direct support for this proposition, however, and Chief Justice Rehnquist's Grutter dissent never addressed the relative weight given to race at the Law School and the College. See Grutter v. Bollinger, 539 U.S. 306, 378-87 (2003) (Rehnquist, C.J., dissenting). Diver's statement is therefore also without empirical support. As far as we know, the University of Michigan's expert statistician did not directly compare the weight given to race at the Law School and the College. His analysis can, however, be used to compare the weight at the two schools along the outcome dimension. See infra notes 61-72 and accompanying text. The plaintiffs' expert statistician, Kinley Larntz, attempted to compare the weight given to race at the Law School and the College by comparing the overall "odds ratios" at the two schools. See Third Supplemental Expert Report of Kinley Larntz at 2-3, Grutter v. Bollinger, 137 F. Supp. 2d 821 (E.D. Mich. 2001) (No. 97-CV-75928-DT). Larntz calculated odds ratios for each school for each year from 1995-1999, controlling for GPA and test score. Id. at 2. For the College, Larntz also controlled for residency status. Id. For the Law School, Larntz compared white applicants with each of four subgroups of minority applicants: Native Americans, African Americans, Mexican Americans, and Puerto Ricans. Id. For the College, Larntz compared non-underrepresented minorities with underrepresented minorities. Id. Larntz found that for all five years, the odds ratios at the College were substantially higher than the odds ratios at the Law School for Native Americans, Mexican Americans, and Puerto Ricans, suggesting that the College placed more weight on race than the Law School with respect to these subgroups. Id. at 7-11. When Larntz compared the odds ratios at the College with the odds ratios at the Law School for African Americans, however, the results were different: for two years, including 1999, the odds ratios were substantially larger at the Law School, id. at 8, 11 (1996 and 1999); for one year, the odds ratio was substantially larger at the College, id. at 7 (1995); and for two years, the odds ratios were comparable, id. at 9-10 (1997 and 1998). Larntz's odds ratios did not compare the same groups at both schools, nor did he control for the same factors at each school. Moreover, Larntz's odds ratio measure does not satisfactorily measure weight assigned to race on either the outcome or the means dimension: it does not capture the number of students displaced on account of affirmative action, nor does it give a sense for the amount of boost but-for admits actually received. We therefore do not discuss Larntz's measure further. We note, however, that Larntz's results are not inconsistent with our ultimate finding that the program at issue in Gratz - the College's program implemented in 1999, see infra notes 58-59 and accompanying text-placed less weight on race than the program at issue in Grutter.
46. See Ayres, supra note 21, at 1803-04 (describing two dimensions to racial preferences in the contracting context); see also WILLIAM G. BOWEN & DEREK BOK, THE SHAPE OF THE RIVER: LONG-TERM CONSEQUENCES OF CONSIDERING RACE IN COLLEGE AND UNIVERSITY ADMISSIONS 31-39 (1998) (observing that the degree of racial preferences can be measured by examining both the number of but-for admits and also the qualification differentials between but-for admits and nonpreferred applicants who would have been admitted in the absence of affirmative action).
47. There are several ways to aggregate these two dimensions, but one is to multiply the two dimensions together. If race is decisive for x% of all admitted applicants, and the effective credit for race is y units, we can say that the overall weight placed on race is xy units. Using this logic, we might find, for example, that an affirmative action program that gives minority applicants an effective credit of 15 LSAT points and is decisive for 10% of the admitted applicants places the same amount of weight on race as a program that gives minority applicants an effective credit of 10 LSAT points and is decisive for 15% of the admitted applicants. See Ayres, supra note 21, at 1804.
48. It may be the case that the percentage of admits who accept offers of admission - a statistic known as the "yield" - is different for applicants who are given a racial preference and applicants who receive no racial preference. See BOWEN & BOK, supra note 42, at 33-34 (describing evidence that the yield for African American students is much lower than for white students, at least at highly selective colleges); Goodwin Liu, The Causation Fallacy: Bakke and the Basic Arithmetic of Selective Admissions, 100 MICH. L. REV. 1045, 1076 (2002) (citing BOWEN & BOK, supra note 42). If the yield is in fact different for preferred applicants and nonpreferred applicants, then the outcome dimension should probably be measured not by the percentage of admits who are but-for admits, but rather by the percentage of enrollees who are but-for admits. But, this becomes a more difficult normative calculus, in part because there might be a constitutional cost to racial preferences even if none of the but-for admits accept their offers of admission. Yield data were not available to us, and we therefore did not calculate the percentage of enrollees who were but-for admits. We note, however, that data were available to us from which we could compute yields for the College in 1995 and the Law School in 2000. One of the reports of the defendants' expert statistician included yield data for several subgroups of applicants to the College in 1995.
49. See Supplemental Expert Report of Stephen W. Raudenbush at 11 tbl.3, Gratz v. Bollinger, 122 F. Supp. 2d 811 (E.D. Mich. 2000) (No. 97-CV-75231-DT)
50. [hereinafter Raudenbush, Supplemental Expert Report (July 13, 2000)], available at http://www.vpcomm.umich.edu/admissions/research/swrjul13. pdf. Using these data, we calculated that for the College in 1995, the yield for minorities was 41.3% and the yield for nonminorities was 33.9%. See infra note 68. The Grutter district court opinion included data that permitted us to compute yields for the Law School in 2000. See Grutter v. Bollinger, 137 F. Supp. 2d 821, 839 (E.D. Mich. 2001), rev 'd, 288 F.3d 732 (6th Cir. 2002) (en banc), aff'd, 539 U.S. 306 (2003). The opinion stated that in 2000, there were 170 minority admits and 58 minority enrollees. Id. The minority yield was therefore 34.1%. As described in note 65, we calculated that the number of nonminority admits in 2000 was 1,240. Because the 58 minority enrollees constituted 14.5% of the entering class, see id., it follows that the number of nonminority enrollees was 342. Thus, the yield for nonminorities was 27.6%. So, the yield was approximately 7 percentage points larger for minorities than it was for nonminorities at both the College in 1995 and the Law School in 2000. To the extent that the yield differs for applicants receiving a racial preference and those not receiving a racial preference, it is likely that the differences in the yields are sufficiently similar at the Law School and the College that if the percentage of admits who were but-for admits is larger at the Law School than at the College, the percentage of enrollees who were but-for admits is also larger at the Law School than at the College. This principle certainly holds true when applied to the data from the Law School in 2000 and the College in 1995. Compare infra text accompanying note 66 (establishing that using Raudenbush's calculations, 8.8% of 2000 Law School admits were but-for admits), and infra text accompanying note 68 (establishing that using Raudenbush's calculations, 7.8% of 1995 College admits were but-for admits), with infra note 66 (establishing that using Raudenbush's calculations, 10.5% of 2000 Law School enrollees were but-for admits), and infra note 67 (establishing that using Raudenbush's calculations, 9.3% of 1995 College enrollees were but-for admits). We therefore assumed that comparing the percentages of admits who were but-for admits at the Law School and the College was a good proxy for comparing the percentages of enrollees who were but-for admits at the two schools.
51. In Gratz, the Court found it relevant that the affirmative action program had "the effect of making 'the factor of race . . . decisive' for virtually every minimally qualified underrepresented minority applicant." 539 U.S. 244, 272 (2003) (omission in original) (quoting Regents of the Univ. of Cal. v. Bakke, 438 U.S. 265, 317 (1978) (Powell, J.)); see also infra note 142 and accompanying text. For a critique of this measure of weight, see infra subsection V(A)(2)(c).
52. Data were available in many different forms, but we took our data from the expert reports of Kinley Larntz, the plaintiffs' expert. Supplemental Expert Report of Kinley Larntz, Gratz, 122 F. Supp. 2d 811 (No. 97-CV-75231-DT)
53. [hereinafter Larntz, Supplemental Export Report, Gratz];
54. Revised Second Supplemental Expert Report of Kinley Larntz, Gratz, 122 F. Supp. 2d 811 (No. 97-CV-75231-DT)
55. [hereinafter Larntz, Revised Second Supplemental Expert Report, Gratz];
56. Supplemental Expert Report of Kinley Larntz, Grutter, 137 F. Supp. 2d 821 (No. 97-CV-75928-DT)
57. [hereinafter Larntz, Supplemental Expert Report, Grutter] ;
58. Second Supplemental Expert Report of Kinley Larntz, Grutter, 137 F. Supp. 2d 821 (No. 97-CV-75928-DT)
59. [hereinafter Larntz, Second Supplemental Expert Report, Grutter];
60. Fourth Supplemental Expert Report of Kinley Larntz, Grutter, 137 F. Supp. 2d 821 (No. 97-CV-75928-DT)
61. [hereinafter Larntz, Fourth Supplemental Expert Report, Grutter]. Larntz took the data from the University of Michigan and created tables that broke down the data as described above.
62. See, e.g., Larntz, Supplemental Expert Report, Gratz, supra, at 1-2.
63. College data for 1995-1997 included both a "GPA 1" and "GPA 2," and College data for 1998 and 1999 included only one GPA. Larntz, Supplemental Expert Report, Gratz, supra note 46, at 1-2;
64. Larntz, Revised Second Supplemental Expert Report, Gratz, supra note 46, at 1-2. The GPA 1 is the actual GPA of applicants, while the GPA 2 is a number calculated by starting with the actual GPA and then adding or subtracting points for other factors such as quality of high school and strength of the curriculum. See Gratz, 122 F. Supp. 2d at 827 n.15; Brief for the Petitioners at 5-6, Gratz, 539 U.S. 244 (No. 02-516), 2003 WL 164186. In 1997, the GPA 2 also included points for minority status. See Joint Appendix at 111-12, Gratz, 539 U.S. 244 (No. 02-516); see also Gratz, 539 U.S. at 255. Our calculations used the actual GPA of the applicants: GPA 1 for 1995-1997 and the only GPA provided for 1998-1999.
65. The test for which College test score ranges were available was the SAT. See Larntz, Supplemental Expert Report, Gratz, supra note 46, at 1;
66. Larntz, Revised Second Supplemental Expert Report, Gratz, supra note 46, at 1-2. The test for which Law School test score ranges were available was the LSAT.
67. See Larntz, Supplemental Expert Report, Grutter, supra note 46, at 1;
68. Larntz, Second Supplemental Expert Report, Grutter, supra note 46, at 1;
69. Larntz, Fourth Supplemental Expert Report, Grutter, supra note 46, at 1.
70. For the Law School, data were available on admissions statistics for particular races such as African Americans and Native Americans, but such data were not further broken down by residency status. Data were available, however, for "Selected Minorities" and "Majority Applicants" that were further broken down by residency status (and the other factors described above), and these were the data we used. See Larntz, Supplemental Expert Report, Grutter, supra note 46;
71. Larntz, Second Supplemental Expert Report, Grutter, supra note 46;
72. Larntz, Fourth Supplemental Expert Report, Grutter, supra note 46. Inspection of the data revealed that four subgroups made up the larger group of "Selected Minorities": African Americans, Mexican Americans, Puerto Ricans, and Native Americans.
73. Cf. Larntz, Supplemental Expert Report, Grutter, supra note 46, at 1 (stating that the report analyzes the relationship between Law School acceptance and membership in the group of "selected minorities (Native American, African American, Mexican American, or Puerto Rican)");
74. Larntz, Second Supplemental Expert Report, Grutter, supra note 46, at 1 (same);
75. Larntz, Fourth Supplemental Expert Report, Grutter, supra note 46, at 1 (same). Inspection of the data also revealed that five subgroups made up the larger group of "Majority Applicants": Caucasians, "Other Hispanic Americans," Asian Americans, Foreign Applicants, and "Unknown Ethnicity" applicants. So, when we refer to "preferred minority" Law School applicants, we mean African Americans, Mexican Americans, Puerto Ricans, and Native Americans. For the College, data were not available for particular minority subgroups but were available only for "Underrepresented Minorities" and "Not Underrepresented Minorities."
76. Larntz, Supplemental Expert Report, Gratz, supra note 46, at 1-2;
77. Larntz, Revised Second Supplemental Expert Report, Gratz, supra note 46, at 1-2. While the data in the reports do not indicate which subgroups were included in the larger group of "Underrepresented Minorities," the College's brief stated that the College "consider[ed] African-Americans, Hispanics, and Native Americans to be underrepresented minorities for purposes of considering race or ethnicity in admissions." Brief for Respondents at 9 n.13, Gratz, 539 U.S. 244 (No. 02-516), 2003 WL 402237 [hereinafter Brief for Respondents, Gratz]. So, when we refer to "preferred minority" College applicants, we mean African Americans, Hispanics, and Native Americans.
78. Data for the Law School were available for the years 1995 through 2000. Larntz, Supplemental Expert Report, Grutter, supra note 46;
79. Larntz, Second Supplemental Expert Report, Grutter, supra note 46;
80. Larntz, Fourth Supplemental Expert Report, Grutter, supra note 46. Data for the College were available for the years 1995 through 1999.
81. Larntz, Supplemental Expert Report, Gratz, supra note 46;
82. Larntz, Revised Second Supplemental Expert Report, Gratz, supra note 46. Throughout this Article, when we refer to data for year x, we are referring to data for the class entering in the fall of year x.
83. Our method for computing the number of but-for admits is identical to one of the methods used by Linda Wightman in her 1997 study of admissions to the 173 ABA-approved law schools in the 1990-1991 application year. In Wightman's method that differs from ours, Wightman performs a logistic regression on the data for white applicants. She then uses this regression to estimate the probabilities of admission for all nonwhite applicants, had they been evaluated by the standards used to evaluate white applicants. She sums these probabilities to estimate the number of nonwhite applicants who would have been admitted without affirmative action and compares that number to the actual number of nonwhite applicants accepted in order to compute the number of but-for admits. Linda F. Wightman, The Threat to Diversity in Legal Education: An Empirical Analysis of the Consequences of Abandoning Race as a Factor in Law School Admission Decisions, 72 N.Y.U. L. REV. 1, 6-8 (1997). The defendants' expert statistician, Stephen Raudenbush, performed calculations that could be used to compute the number of but-for admits at the College and the Law School. When analyzing the 1998 College data, for example, Raudenbush first performed a regression to predict the probability of admission for nonpreferred applicants and then used this regression to predict the probability of admission for applicants under a race-blind system.
84. Supplemental Expert Report of Stephen W. Raudenbush at 4, Gratz, 122 F. Supp. 2d 811 (No. 97-CV-75231-DT) [hereinafter Raudenbush, Supplemental Expert Report (Feb. 24, 2000)], available at http://www.vpcomm.umich.edu/ admissions/research/swrfeb24.pdf. Raudenbush then departed from Wightman's method and rank-ordered the applicants by their probability of admission and assumed that the applicants who would have been accepted would have been those at the top of the list; he used this assumption to estimate the number of preferred minorities who would have been admitted in the absence of racial preferences. Id. For the results of this analysis, see infra notes 62-72 and accompanying text.
85. If there were no preferred minority applicants or no nonpreferred applicants, we defined the probability enhancement to be zero.
86. Note that our method will approximate the number of but-for admits only if other admissions factors - in the aggregate - are distributed evenly across preferred and nonpreferred applicants for each combination of GPA range, test score range, residency status, and year. Other individual factors may not be distributed evenly. For example, legacy status may be distributed unevenly so as to favor white applicants. See, e.g., John D. Lamb, The Real Affirmative Action Babies: Legacy Preferences at Harvard and Yale, 26 COLUM. J.L. & SOC. PROBS. 491, 508 (1993) (noting that an investigation by the U.S. Department of Education's Office for Civil Rights found that legacy preferences at Harvard and Yale disproportionately advantaged white applicants over Asian American applicants). Athletic ability and experience may be distributed unevenly so as to favor preferred minority applicants.
87. See JAMES L. SHULMAN & WILLIAM G. BOWEN, THE GAME OF LIFE: COLLEGE SPORTS AND EDUCATIONAL VALUES 55 (2001) (concluding that athletic preferences increase racial diversity modestly for males - in the study's 1989 cohort, the percentage of African American males would have "declined . . . from 6 percent to 5 percent, if the athletic contribution to racial diversity had been eliminated"); see also Liu, supra note 44, at 1069 & n.101 (describing Shulman and Bowen's findings that "although athletics helps promote racial diversity, the impact is modest" (internal quotation marks omitted) (quoting SHULMAN & BOWEN, supra, at 55)). Low socioeconomic status may also be distributed unevenly so as to favor preferred minority applicants.
88. See Thomas J. Kane, Racial and Ethnic Preferences in College Admissions, in THE BLACK-WHITE TEST SCORE GAP 431, 449-50 (Christopher Jencks & Meredith Phillips eds., 1998) (noting that socioeconomic status is distributed unevenly by race among students with SAT scores in the top ten percent); see also Wightman, supra note 51, at 24-25 & n.51 (finding that socioeconomic status was distributed unevenly by race among applicants to law school for a sample of applicants in 1991). If these other factors have a net effect of advantaging either preferred or nonpreferred applicants, then the differences in admissions rates we observe cannot be attributed solely to racial preferences. Most admissions factors (such as extracurricular activities and special talents) are, however, likely distributed evenly among preferred and nonpreferred applicants with the same GPAs and test scores. And, of the factors that are unevenly distributed, some factors favor preferred applicants while others favor nonpreferred applicants. Thus, as a first cut, it seems reasonable to assume that the other factors in combination do not substantially favor either preferred minority applicants or nonpreferred applicants. Moreover, the Law School represented that "soft" variables at least do not systematically favor minority applicants. See Grutter v. Bollinger, 288 F.3d 732, 800 (6th Cir. 2002) (Boggs, J., dissenting) (noting that during oral argument in the Sixth Circuit, counsel for the Law School "responded with a firm 'no'" when asked, "Do you assert that under-represented minorities systematically have stronger [soft variables] than non-minority students?"), aff'd, 539 U.S. 306 (2003); see also BOWEN & BOK, supra note 42, at 31 (noting that comparing admissions rates of African American and white applicants with the same SAT range could provide a measure of the amount of affirmative action); Liu, supra note 44, at 1070 (concluding that "differences in admission rates based on SAT scores provide a reasonably valid measure of the admissions advantage black applicants receive through affirmative action"); Wightman, supra note 51, at 6-9 (using a model that compared admissions rates for white and nonwhite applicants with similar LSAT scores and GPAs and noting that it "was built under the assumption that if race were not a factor in the decisions, patterns of admission decisions observed for white applicants would also hold for applicants of color").
89. See supra note 50.
90. We found that there were 755 but-for admits and 7,284 total admits. The total number of minority admits was 921, so the percentage of minority admits who were but-for admits was 82.0%.
91. See supra note 50.
92. We found that there were 3,373 but-for admits and 50,055 total admits. The total number of minority admits was 6,341, so the percentage of minority admits who were but-for admits was 53.2%.
93. For a summary of the changes, see Gratz v. Bollinger, 539 U.S. 244, 253-57 (2003) and Gratz v. Bollinger, 122 F. Supp. 2d 811, 827-33 (E.D. Mich. 2000), rev'd in part, 539 U.S. 244 (2003).
94. The district court struck down the policies in place between 1995 and 1998, and upheld the policy that was implemented in 1999. Gratz, 122 F. Supp. 2d at 814. The College's brief explained why only the policy that was implemented beginning in 1999 was at issue before the Supreme Court: The admissions programs governed by the 1995-98 guidelines included three race-conscious practices that the University undisputedly has discontinued and disavowed: (1) the use of grids that take race into account by setting forth admissions options for applicants with various combinations of qualifications; (2) the exemption of minority students from the practice of rejecting candidates with very low grades and test scores without counselor review; and (3) a procedure known as "protected seats" that used projections of expected applications from groups known to apply late in the process (including minorities) to pace the rolling admissions process to permit consideration of such applications. The district court concluded that, while the use of grids, standing alone, was not necessarily unlawful, the combination of the three practices was impermissible. Petitioners devote much of their brief to attacking these abandoned admissions practices. However, because the University did not cross-petition to seek review of the district court's determination that these practices, taken together, were impermissible, those practices are not properly before this Court. Brief for Respondents, Gratz, supra note 49, at 5 n.7 (citations omitted). Indeed, the Supreme Court seemed to limit its holding to the policy implemented beginning in 1999. See Gratz, 539 U.S. at 271 (noting that the "current . . . policy does not provide such individualized consideration" (emphasis added)).
95. We found that there were 445 but-for admits and 11,228 total admits. The total number of minority admits was 1,228, so the percentage of minority admits who were but-for admits was 36.2%.
96. In his first method, Raudenbush used a mixed model for logistic regression that allowed for random effects and calculated the percentage of admittees who would have been admitted if the only factors the Law School considered were GPA, LSAT scores, residency status, and sex. Richard O. Lempert et al., Michigan's Minority Graduates in Practice: The River Runs Through Law School, 25 LAW & SOC. INQUIRY 395, 492 (2000). In his second method, Raudenbush "placed [applicants] in cells on a grid based on the conjunction of LSAT scores and [GPAs], and assumfed] that the proportion of admitted minorities in a cell would be the proportion of all applicants in the cell who [were] admitted." Id. at 493.
97. Id. at 492-93.
98. Id.
99. Grutter v. Bollinger, 137 F. Supp. 2d 821, 839 (E.D. Mich. 2001), rev'd, 288 F.3d 732 (6th Cir. 2002) (en banc), aff'd, 539 U.S. 306 (2003).
100. The opinion stated that Raudenbush found that without affirmative action, the percentage of nonminority applicants who were admitted would have been 44% rather than the actual percentage of 40%. Id. It follows that 4% of the nonminority applicants is equal to 124, the number of but-for admits. Because 40% of all nonminority applicants were actually admitted, the actual number of nonminority admits was 1,240. The total number of admits was therefore 1,410, the sum of 1,240 and 170.
101. The district court opinion and the Supreme Court's Grutter opinion reported that when examining the Law School data for 2000, Raudenbush found that without affirmative action, minorities would have constituted 4% of the entering class rather than the actual figure of 14.5%. Grutter v. Bollinger, 539 U.S. 306, 320 (2003); Grutter, 137 F. Supp. 2d at 839. Using these figures, we would say that the percentage of enrollees who were but-for admits was 10.5%. Cf. supra note 44 (distinguishing between the percentage of enrollees who are but-for admits and the percentage of admits who are but-for admits).
102. Raudenbush, Supplemental Expert Report (July 13, 2000), supra note 44, at 3. Using these figures, we would say that the percentage of enrollees who were but-for admits was 9.3%. Cf. supra note 44 (distinguishing between the percentage of enrollees who are but-for admits and the percentage of admits who are but-for admits).
103. The expert report provided yield data for several subgroups. See Raudenbush, Supplemental Expert Report (July 13, 2000), supra note 44, at 11 tbl.3. Using these yield data, we calculated that the yield for minority applicants was 41.3% and the yield for nonminority applicants was 33.9%. Raudenbush's calculations suggested that the ratios of the number of but-for minority enrollees to the number of non-but-for minority enrollees to the number of nonminority enrollees were 9.3:5:85.7. We used these ratios along with the yield statistics to calculate that the ratios of the number of but-for minority admits to the number of non-but-for minority admits to the number of nonminority admits were 22.5:12.1:252.9. Thus, the percentage of admits who were but-for admits was 7.8%.
104. Brief for Respondents, Gratz, supra note 49, at 4 n.5.
105. Id.
106. Raudenbush, Supplemental Expert Report (Feb. 24, 2000), supra note 51, at 3-5.
107. All percentages for the Law School are greater than all percentages for the College with one exception: the percentage of 2000 Law School admits who were but-for admits is 8.8% while the percentage of 1996 College admits who were but-for admits is 10.3%. It is significant, however, that the Law School percentages are larger than the College percentages for all other pairs of data. Moreover, the 1996 College program was not before the Court in Gratz; only the program implemented beginning in 1999 was before the Court. See supra note 59 and accompanying text.
108. See Ayres, supra note 21, at 1803-04 (arguing that qualified nonminorities are more burdened when they lose out to substantially less qualified minorities than to only slightly less qualified minorities).
109. Some literature has suggested other measures for capturing the weight given to race on what we have called the means dimension. A common approach is to compare the average test scores of preferred minority and nonpreferred matriculants. See, e.g., Peter H. Schuck, Affirmative Action: Past, Present, and Future, 20 YALE L. & POL'Y REV. 1, 18-19 (2002) (comparing the differences in test scores and GPAs between African American and white applicants); see also Liu, supra note 44, at 1063 & n.79 (criticizing studies that take this approach). As Goodwin Liu points out, however, differences in average test scores cannot be attributed solely to affirmative action and are therefore not a good measure of the means dimension. See Liu, supra note 44, at 1064. William Bowen and Derek Bok propose to capture part of what we call the means dimension by examining the differential in qualifications between but-for African American admits and white applicants who would have been admitted in the absence of affirmative action. BOWEN & BOK, supra note 42, at 37. Because of problems identifying the white applicants who would have been admitted in the absence of affirmative action, they propose approximating their credentials with the lowest decile of white admits. Id. at 37-38. Linda Wightman follows this approach in a study of selective law schools.
110. See Linda F. Wightman, Are Other Things Essentially Equal? An Empirical Investigation of the Consequences of Including Race as a Factor in Law School Admission, 28 SW. U. L. REV. 1, 16-25 (1998) (comparing LSAT scores and undergraduate GPAs of minorities who would have been denied admission based on a numbers-only admission model with LSAT scores and undergraduate GPAs of the lowest ten percent of white students that had been admitted to their respective schools).
111. For each set of data falling in a particular GPA range and test score range, we assigned the data a GPA corresponding to the mean of the GPA range and a test score corresponding to the mean of the test score range. In addition, we dropped data corresponding to combinations that included at least one of the following: no test score or GPA, or the lowest possible test score range or GPA range.
112. To compute the GPA enhancement, we divided the minority coefficient by the GPA coefficient from the group probit.
113. Judge Boggs attempted to estimate the GPA enhancement for the Law School's admissions program in his dissent from the Sixth Circuit opinion in Grutter. Grutter v. Bollinger, 288 F.3d 732, 796 (6th Cir. 2002) (Boggs, J., dissenting), aff'd, 539 U.S. 306 (2003). Examining the 1997 data for applicants with LSAT scores between 167 and 169, he noted that "under-represented minorities with a high C to low B undergraduate average are admitted at the same rate as majority applicants with an A average" and concluded that the Law School's GPA enhancement was over 1 full GPA point. Id. Using a similar technique, Judge Boggs concluded that the LSAT enhancement was approximately 11 LSAT points. Id.
114. These results are consistent with the results of the plaintiffs' expert, Kinley Larntz, who used a different method to compute GPA enhancement for the College in 1995 and 1996. For combinations of test score ranges, race, and residency status, Larntz computed the lowest GPA with a probability of admission of greater than fifty percent. Expert Report of Kinley Larntz at 2, Gratz v. Bollinger, 122 F. Supp. 2d 811 (E.D. Mich. 2000) (No. 97-CV-75231-DT). Then, Larntz compared these GPAs for preferred and nonpreferred applicants and computed a GPA enhancement for each test score range. Id. at 3-4. Larntz concluded that the GPA enhancements in 1995 were approximately 1 GPA point, while the GPA enhancements in 1996 were approximately 1.2 GPA points. Id. Larntz did not compute GPA enhancements for other years for the College, nor did he compute GPA enhancements for the Law School.
115. These results are consistent with the results we get when we calculate the GPA enhancement using a different method in which we examine the point values in the College's point system. Beginning in 1998, the College assigned each applicant a "selection index." See infra notes 90-94 and accompanying text. Applicants could earn up to 80 points for their GPA and 20 points for membership in an underrepresented minority group. See infra note 92. While minorities could theoretically earn 20 points for their race, not all minorities earned these points for their race. See infra note 92. This is because the 20 points available for race could be earned on the basis of many factors such as socioeconomic status and athletic ability, but these points could be earned on the basis of one - and only one - of these factors. See infra note 92. So, if a racial minority would have earned 20 points for one of these other factors, then in some sense she did not earn any points on account of her race. In addition, some minority applicants effectively received fewer than 20 points for their race because applicants could earn no more than 40 points for nonacademic factors, so if an applicant earned more than 20 points in the other nonacademic factor categories, then she effectively did not earn the full 20 points for her race. See infra note 92. Minority applicants therefore received, on average, fewer than 20 points for their race. Because a 4.0 GPA was worth 80 points, we would expect the GPA enhancement to be less than 1 GPA point. Our method described in the text accompanying this footnote produces a GPA enhancement of 0.67 GPA points, so the two results are consistent.
116. For a robustness check, we performed the same computations using a logit model rather than a probit model, and our results were similar: (Table Presented)
117. The GPA enhancement measure probably underestimates the amount of weight placed on race at the Law School relative to the College because the Law School likely weighted academic factors more heavily than the College. See BOWEN & BOK, supra note 42, at 25 (noting that compared to undergraduate institutions, "[professional schools place little emphasis on assembling a diversity of talents for the sake of enriching extracurricular life" and observing that "law schools . . . place the greatest weight on the traditional measures of academic achievement").
118. We normalized test score data on a 1,000-point scale and found the following results: (Table Presented) These data suggest the College gave minority applicants a bigger boost than the Law School, and the 1999 College policy provided an even bigger boost to minorities than earlier College policies. Note that a large SAT point boost is consistent with the point system that was in effect starting in 1998 at the College, which awarded applicants up to 20 points for being a racial minority and up to 12 points for SAT scores. See infra notes 90-92 and accompanying text. The 20 points minorities could earn for race correspond to 2,000 SAT points. Given that not all minorities could earn all 20 points for race, we would expect the SAT enhancement to be less than 2,000 SAT points. See supra note 79. We also note that the Law School minority coefficient was higher than the College minority coefficient for 1999, but lower than the College minority coefficient for 1995-1999. The raw minority coefficients were as follows: (Table Presented) These data suggest that during 1995-1999, the College gave minority applicants a bigger boost than the Law School, but that the 1999 College policy gave minority applicants a smaller boost than the Law School.
119. In the point system the College began using in 1998, applicants could earn up to 80 points for their GPA but only 12 points for their test scores. See infra notes 90, 92.
120. Richard H. Sander, A Systemic Analysis of Affirmative Action in American Law Schools, 57 STAN. L. REV. 367, 402 tbl.2.1, 405 tbl.2.2 (2004).
121. Id. at 402-05.
122. Id.
123. Sander's method suffered from three principal flaws. First, Sander was not comparing the same groups at the Law School and the College. Instead, he compared nonpreferred and preferred minority nonresidents at the College, id. at 401-02, and white and African American applicants (residents and nonresidents) at the Law School, id. at 404-05. Second, while Sander ran a logistic regression on the College data to create an index formula for the College, id. at 402 tbl.2.1, he did not run his own regression on the Law School data to come up with an index formula for the Law School. Instead, he used an index formula that he says is typical of law schools, and thus it is unclear how well the index formula he used for the Law School is a good match for the data. Id. at 405 tbl.2.2. Third, Sander estimated the index point enhancement figure by eyeballing the data - he created tables that showed that, at several different ranges of index scores, there was a 140-point difference at the Law School and a 120-point difference at the College. Id. at 402-05. He does not use more sophisticated analysis such as a group probit to determine if these observed differences are characteristic of the data as a whole. When we ran a group probit on the data Sander examined in his tables, we found that the Law School had a 141-point enhancement and the College had a 160-point enhancement.
124. See Grutter v. Bollinger, 539 U.S. 306, 313-16 (2003) (noting the Law School policy's focus on grades, test scores, and "other criteria that are important to the Law School's educational objectives"); Gratz v. Bollinger, 539 U.S. 244, 253 (2003) (summarizing the admissions factors considered by the College); Brief for Respondents at 3-4, Grutter, 539 U.S. 306 (No. 02-241), 2003 WL 402236 [hereinafter Brief for Respondents, Grutter] (noting that while the Law School's admissions program considered race as a factor, the program's "hallmark [was] a focus on academic capabilities coupled with a flexible assessment of every individual applicant's talents, experiences and potential to contribute to the learning of those around them" (internal quotation marks omitted)); Brief for Respondents, Gratz, supra note 49, at 5-10 (describing the admissions factors considered by the College).
125. Grutter, 539 U.S. at 315 (internal quotation marks omitted).
126. The College used the point system in 1998 and subsequent years. See Gratz, 539 U.S. at 255.
127. Grutter, 539 U.S. at 315.
128. Brief for Respondents, Gratz, supra note 49, at 7, 9. The 110 points for academic factors were available as follows: 80 points for tenth and eleventh grade GPA, 12 points for standardized test scores, 10 points for "academic strength of . . . high school," and 8 points for a rigorous course of study. Id. at 7-8. The 40 points for nonacademic factors were available as follows: 10 points for Michigan residents, 6 points for being from an underrepresented Michigan county, 2 points for being from an underrepresented state, 4 points for children of an alumnus or 1 point for close relative of an alumnus, 3 points for outstanding essay, 5 points for leadership or public service, 5 points for personal achievement, and 20 points for one and only one of five characteristics or factors - (1) socioeconomic disadvantage, (2) membership in an underrepresented minority group, (3) predominantly minority or socioeconomically disadvantaged high school, (4) athletic recruitment, and (5) provost's discretion. In calculating the selection index, a maximum of 40 points could come from these nonacademic factors, even if a student earned more than 40 points in the preceding categories. Id. at 8-9. The College considered "African-Americans, Hispanics, and Native Americans" to be underrepresented minorities for purposes of their admissions policy. Gratz, 539 U.S. at 253-54.
129. The majority opinion in Gratz stated that "[t]his index was divided linearly into ranges generally calling for admissions dispositions as follows: 100-150 (admit); 95-99 (admit or postpone); 90-94 (postpone or admit); 75-89 (delay or postpone); 74 and below (delay or reject)." Gratz, 539 U.S. at 255. The brief for the University of Michigan stated that "[t]o avoid overenrollment, [the College] sets and adjusts, when necessary, the selection index levels that trigger the three possible admissions outcomes - admittance, deferral, and denial." Brief for Respondents, Gratz, supra note 49, at 10.
130. The admissions counselor could "flag" an application as long as the applicant met three criteria. The applicant must: (1) have been academically prepared; (2) have had a selection index of at least 75 for nonresidents of Michigan and 80 for residents of Michigan; and (3) have "possessed] at least one of a variety of qualities or characteristics important to the University's composition of its freshman class." Brief for Respondents, Gratz, supra note 49, at 10.
131. Gratz, 539 U.S. at 271; see also Post, supra note 20, at 70 (stating that the Court used the individualized consideration requirement to strike down the admissions program in Gratz).
132. Post, supra note 20, at 67 (footnotes omitted) (quoting Grutter, 539 U.S. at 341, 339, 342, 334); see also Parents Involved in Cmty. Sch. v. Seattle Sch. Dist., No. 1, 426 F.3d 1162, 1180 (9th Cir. 2005) (en banc) (listing five "hallmarks" of a narrowly tailored affirmative action program: the four listed by Post and a "no-quotas" hallmark, which Post presumably folded into the individualized consideration hallmark), cert. granted, 126 S. Ct. 2351 (2006).
133. Grutter, 539 U.S. at 341 ("[I]n the context of its individualized inquiry into the possible diversity contributions of all applicants, the Law School's race-conscious admissions program does not unduly harm nonminority applicants.").
134. Id. at 340 ("[T]he Law School sufficiently considered workable race-neutral alternatives.").
135. Id. at 342 (noting that "race-conscious admissions policies must be limited in time" and that "the durational requirement can be met by sunset provisions in race-conscious admissions policies and periodic reviews to determine whether racial preferences are still necessary to achieve student body diversity").
136. Id. at 343 ("We expect that 25 years from now, the use of racial preferences will no longer be necessary to further the interest approved today.").
137. See, e.g., Post, supra note 20, at 70-71 ("[T]he Court never makes clear whether the Michigan undergraduate program fails the individualized consideration requirement because it quantifies the contribution of race to diversity by 'a specific and identifiable' measure, or instead because the program employs a measure that is 'decisive.'" (quoting Gratz v. Bollinger, 539 U.S. 244, 271 (2003)); Richard A. Primus, Equal Protection and Disparate Impact: Round Three, 117 HARV. L. REV. 493, 561 (2003) ("Exactly what it means to treat applicants as individuals is, and will surely continue to be, a contested question. After Grutter, individual treatment can include some consideration of race, though it is difficult to say how much and what kind.").
138. Justice Souter noted that the majority objected either to "the use of points to quantify and compare characteristics, or to the number of points awarded due to race." Gratz, 539 U.S. at 295 (Souter, J., dissenting).
139. Id. at 270 (majority opinion).
140. Id. (emphasis added).
141. Id.
142. Robert Post reads the opinions as susceptible to two, rather than three, interpretations of what satisfies the individualized consideration requirement. His first interpretation seems to be a combination of our first and second interpretations; his second interpretation maps onto our third interpretation: Gratz offers two distinct accounts of the individualized consideration requirement. It states, on the one hand, that the requirement is inconsistent with any program in which "any single characteristic automatically ensure[s] a specific and identifiable contribution to a university's diversity." But it also notes, on the other hand, that the Michigan undergraduate affirmative action program is unconstitutional because the "automatic distribution of 20 points has the effect of making 'the factor of race . . . decisive' for virtually every minimally qualified underrepresented minority applicant." Post, supra note 20, at 70 (alteration in original) (omission in original) (footnote omitted) (quoting Gratz, 539 U.S. at 271, 272).
143. Gratz, 539 U.S. at 295 (Souter, J., dissenting).
144. Even after Grutter and Gratz, it appears that a university could still employ a nonracial point system - that is, a point system that does not directly assign points on the basis of race - as part of its process for selecting students. See Smith v. Univ. of Wash., 392 F.3d 367, 376 (9th Cir. 2004) (upholding a law school admissions program that assigned applicants an index score on the basis of their grades and test scores). As Reva Siegel has pointed out, "[a]pplicants who are evaluated as individuals can be categorized and valued on the basis of any trait (for example, grades, standardized test scores, parental income, residence, high school, alumni affiliations, or musical or athletic ability) except race." Reva B. Siegel, Equality Talk: Antisubordination and Anticlassification Values in Constitutional Struggles over Brown, 117 HARV. L. REV. 1470, 1540 n.240 (2004). Indeed, if the Court meant to make even nonracial point systems unconstitutional, then use of GPAs and standardized test scores would presumably not be allowed, as they assign points to reflect particular attributes.
145. For example, Robert George explains the difference between the outcomes of Grutter and Gratz as turning on the point system qua point system. Robert P. George, Grutter and Gratz: Some Hard Questions, 103 COLUM. L. REV. 1634, 1634 (2003) ("The admissions policy of the undergraduate college, which formally awarded valuable 'points' to certain candidates based on race and ethnicity, was judged to fail the test of narrow tailoring; the Law School's policy of taking race and ethnicity into account without the formal awarding of points was judged to pass." (footnote omitted));
146. see also Marvin Krislov, Affirmative Action in Higher Education: The Value, the Method, and the Future, 72 U. CIN. L. REV. 899, 903 (2004) ("Although [the College's] system, like the law school, did not employ quotas or set-asides, the Court found the weighting of race 'mechanistic' and 'automatic,' and thus concluded that the program was not narrowly tailored."); Post, supra note 20, at 70-71 ("The upshot is that the Court never makes clear whether the Michigan undergraduate program fails the individualized consideration requirement because it quantifies the contribution of race to diversity by 'a specific and identifiable' measure, or instead because the program employs a measure that is 'decisive.'" (quoting Gratz, 539 U.S. at 271));
147. Cass R. Sunstein, Problems with Minimalism, 58 STAN. L. REV. 1899, 1901 (2006) (noting that in Gratz, "[t]he Court did not rule that the twenty points [awarded for race] were too high; it ruled instead that a point system, in the context of racial preference, is invalid as such").
148. Gratz, 539 U.S. at 271-72. "LSA" refers to the College, which is called the College of Literature, Science, and the Arts. Id. at 251.
149. Id. at 271.
150. Of course, one might respond to Chief Justice Rehnquist by noting that just because a point system assigns the same number of points to each preferred minority applicant, it does not embody the notion that race "automatically ensure[s] a specific and identifiable contribution to a university's diversity." Id. Rather, such a system stands for the notion that, on average, minority students will contribute to the diversity of the student body on account of their race.
151. Id. at 279 (O'Connor, J., concurring).
152. Id. at 277 (citation omitted).
153. Grutter v. Bollinger, 539 U.S. 306, 337 (2003) (citation omitted).
154. Id. at 334.
155. Justice Souter's Gratz dissent provides clues as to whether the majority meant to prohibit quantification. He noted that the Court's "objection goes to the use of points to quantify and compare characteristics, or to the number of points awarded due to race, but on either reading the objection is mistaken." Gratz, 539 U.S. at 295 (Souter, J., dissenting). In other words, he interpreted the Court to be taking a position against either quantification or excessive weight.
156. To be sure, narrow tailoring was originally meant to focus on the closeness of fit between the means and ends of affirmative action, while a possible statistical badness of fit requirement refers to a disjunct between the qualifications of an applicant and her probability of being admitted. Badness of fit in this statistical sense suggests the degree to which randomness or unknown variables are driving admissions behavior. But it would still be perverse if a generalized disjunct between qualifications and admission might add to the constitutionality of a program.
157. Cf. Sander, supra note 84, at 406 (finding that 88% of admissions outcomes can be correctly predicted for the Law School in 1999 on the basis of academic factors and race, while only 82% of admissions outcomes can be correctly predicted for the College in 1999 on the basis of academic factors, race, and residency status).
158. Some scholars have suggested that the Court's main concern was that admissions systems be sufficiently nuanced. See, e.g., Michael Rosman, Uncertain Direction: The Legacy of Gratz and Grutter, JURIST LEGAL INTELLIGENCE F., Sept. 5, 2003, http://jurist.law.pitt. edu/forum/symposium-aa/rosman.php (implying that the lack of nuance might have been a fatal flaw of the College program). Robert Post's discussion of the individualized consideration requirement notes two possible meanings of individualized consideration, and it is unclear whether Post sees his first interpretation as consistent with what we have called the quantification meaning or the differentiation meaning. See Post, supra note 20, at 70-71 ("The upshot is that the Court never makes clear whether the Michigan undergraduate program fails the individualized consideration requirement because it quantifies the contribution of race to diversity by 'a specific and identifiable' measure, or instead because the program employs a measure that is 'decisive.'" (quoting Gratz, 539 U.S. at 271)).
159. Gratz, 539 U.S. at 271 (emphasis added).
160. Id.
161. Id. at 272 (internal quotation marks omitted) (quoting Regents of the Univ. of Cal. v. Bakke, 438 U.S. 265, 324 (1978) (Powell, J.)).
162. Id. at 273. This example demonstrates that Chief Justice Rehnquist seemed to be bothered by the fact that the College made no distinctions on class grounds among minority applicants. Indeed, he was correct that the point system itself operated in this way. As described supra in note 92, applicants could earn 20 points for race or low socioeconomic status, but not both. Had the point system allowed applicants to earn points for both race and low socioeconomic status, then Chief Justice Rehnquist's concerns would seemingly be alleviated. Note that this illustrates that a point system can differentiate among racial minorities and award a constant number of points to race as long as it also awards points for the other relevant dimensions. Thus, an award of a constant amount of points for race does not necessarily indicate that a system is undifferentiated.
163. To the extent that these opinions suggest that one's race can be considered only in context with other characteristics, they raise several questions. First, what other characteristics must universities consider when deciding the amount of preference to give in conjunction with race? Chief Justice Rehnquist's example implied that class and parental education are factors that should be considered, but what other factors are relevant? Second, and related, how is it decided which combinations of race and other characteristics receive more preference than others? We will argue in section V(B)(1) that admissions programs should favor combinations that promote the goals of the diversity interest of (1) creating a lively campus atmosphere with robust discussions by people of diverse backgrounds; (2) breaking down racial stereotypes; and (3) ensuring that education, the pathway to leadership, is open to people of all races. See infra notes 215-19 and accompanying text for a discussion of the goals of the diversity interest.
164. Gratz, 539 U.S. at 277 (O'Connor, J., concurring) (emphasis added).
165. Id. at 276-77.
166. Id. at 279.
167. Id. at 281 (Thomas, J., concurring).
168. Id. ("An admissions policy, however, must allow for consideration of these nonracial distinctions among applicants on both sides of the single permitted racial classification."). Justice Thomas is the only Justice in Gratz to address expressly the issue of whether it is permissible to consider racial distinctions among applicants who are members of underrepresented racial minority groups. Justice Thomas stated emphatically that "[u]nder today's decisions, a university may not racially discriminate between the groups constituting the critical mass." Id. Some of the Grutter opinions also addressed this issue, at
169. One circuit court case decided since Grutter and Gratz also suggested that differentiation is a consideration when determining whether admissions programs are narrowly tailored. See Smith v. Univ. of Wash., 392 F.3d 367, 375-77 (9th Cir. 2004) (holding that a law school's consideration of "ethnicity substantiation letter[s]," in which applicants explain the role race or ethnicity has played in their lives, "supports rather than undermines the constitutionality of the Law School's program" because the letters allow the law school to "give more weight to those minority candidates who ha[ve] more to contribute to the diversity of the classroom").
170. For example, the Court stated that "the Law School engages in a highly individualized, holistic review of each applicant's file, giving serious consideration to all the ways an applicant might contribute to a diverse educational environment." Grutter, 539 U.S. at 337. Later, the Court stated that "the Law School's admissions policy 'is flexible enough to consider all pertinent elements of diversity in light of the particular qualifications of each applicant, and to place them on the same footing for consideration, although not necessarily according them the same weight.'" Id. (quoting Regents of the Univ. of Cal. v. Bakke, 438 U.S. 265, 317 (1978) (Powell, J.)). The Court continued, saying that the "Law School's race-conscious admissions program adequately ensures that all factors that may contribute to student body diversity are meaningfully considered alongside race in admissions decisions." Id.
171. Id. at 337-39; cf. infra note 179 and accompanying text.
172. See Grutter, 539 U.S. at 338. Not only does this statistic fail to support the proposition that the Law School's admissions program was sufficiently nuanced, but also it fails to measure the extent to which the Law School placed weight on race. See infra subsection V(A)(2)(d).
173. Robert Post states that the Court is "ambiguous" about the meaning of individualized consideration. He reads the Court to be offering two "accounts of the individualized consideration requirement," one of which is the excessive preferences meaning. Post, supra note 20, at 70-71 ("The upshot is that the Court never makes clear whether the Michigan undergraduate program fails the individualized consideration requirement because it quantifies the contribution of race to diversity by 'a specific and identifiable' measure, or instead because the program employs a measure that is 'decisive.'" (quoting Gratz, 539 U.S. at 271)). Lani Guinier and Richard Primus agree that the weight inquiry was important to the Court, but they are silent as to whether they read the Court's weight inquiry to be part of the individualized consideration inquiry. Guinier states that the cases mean that "as long as institutions do not weight race so heavily that it overdetermines admissions outcomes, their good faith is essentially presumed." Lani Guinier, The Supreme Court, 2002 Term - Comment: Admissions Rituals as Political Acts: Guardians at the Gates of Our Democratic Ideals, 117 HARV. L. REV. 113, 198 (2003). According to Primus, the cases stand for the proposition that "universities can prefer members of racially disadvantaged groups as long as the value of the racial preference is, inter alia, not too large as compared to the value of other admissions criteria." Primus, supra note 101, at 548. Primus emphasizes this point: In Grutter and Gratz . . . the validity of the University of Michigan's affirmative action plans turned substantially on the relative importance of the racial criterion for admission as compared to other admissions factors. Where the Court found race to be predominant, it disallowed affirmative action. Where it found the racial motive to be merely one factor among several, the Court permitted affirmative action. Id. at 547 (footnote omitted);
174. see also Mark W. Cordes, Affirmative Action After Grutter and Gratz, 24 N. ILL. U. L. REV. 691, 715-16 (2004) (observing that a weight inquiry was part of the individualized consideration inquiry).
175. Gratz, 539 U.S. at 272 (quoting Bakke, 438 U.S. at 324 (Powell, J.)). We discussed this hypothetical earlier. See supra notes 123-24 and accompanying text.
176. Gratz, 539 U.S. at 273 (citation omitted). Note that Chief Justice Rehnquist gave short shrift to how the flagging system mitigated what seems like a disproportionate point distribution. After all, if applicant C truly had artistic talent that rivaled Monet or Picasso, then as long as that applicant met the requirements for being flagged for further review by the admissions review committee, he would most certainly have been admitted by the review committee. To be flagged, the student would need to have been academically qualified and earned a selection index score of at least 75 if he was not a resident of Michigan and 80 if he was a resident of Michigan. See supra note 94. The applicant would have been awarded 5 points for his personal achievement due to his "extraordinary talent," and surely an applicant with this much talent would have earned the 20 points available at the provost's discretion. See supra note 92. So, this future Monet or Picasso would have needed to cobble together only 50 or 55 more points to be flagged. Given the multitude of other ways he could have earned points and the fact that many points were available for even a low GPA, it seems that such a student would have been flagged as long as his credentials were not unduly low, in which case it might be appropriate to reject him.
177. See infra subsection V(A)(2)(a) (discussing this measure).
178. Gratz, 539 U.S. at 270 (emphasis added). Chief Justice Rehnquist's figure of one-fifth comes from the fact that preferred minorities could earn 20 points for their race, see supra note 92, and the majority opinion stated that an index of 100 points or more guaranteed admission, see supra note 93.
179. See infra subsection V(A)(2)(b) (criticizing this measure).
180. Gratz, 539 U.S. at 272 (omission in original) (citation omitted) (quoting Bakke, 438 U.S. at 317 (Powell, J.)).
181. Id. Two circuit court opinions have applied the Gratz outcome test as part of their narrow tailoring analysis. See Smith v. Univ. of Wash., 392 F.3d 367, 376 (9th Cir. 2004) (stating, without providing empirical support, that "unlike the program found unconstitutional in Gratz, the racial and ethnic pluses here did not 'ha[ve] the effect of making the factor of race decisive for virtually every minimally qualified underrepresented minority applicant'" (alteration in original) (quoting Gratz, 539 U.S. at 272)); Petit v. City of Chicago, 352 F.3d 1111, 1117 (7th Cir. 2003) (applying a more lenient version of the Gratz outcome test, stating that "[i]t cannot be said that the process affected every 'minimally qualified' candidate as did the blanket award of 20 points per candidate, the procedure found to be unconstitutional in Gratz").
182. See infra subsection V(A)(2)(c) (criticizing this measure).
183. Justice Souter's dissent also addressed how he thinks weight should be measured: [I]t [is not] possible to say that the 20 points convert race into a decisive factor comparable to reserving minority places as in Bakke. Of course we can conceive of a point system in which the "plus" factor given to minority applicants would be so extreme as to guarantee every minority applicant a higher rank than every nonminority applicant in the university's admissions system. But petitioners do not have a convincing argument that the freshman admissions system operates this way. . . . It suffices for me, as it did for the District Court, that there are no Bakke-like set-asides and that consideration of an applicant's whole spectrum of ability is no more ruled out by giving 20 points for race than by giving the same points for athletic ability or socioeconomic disadvantage. Gratz, 539 U.S. at 295-96 (Souter, J., dissenting). So, for Justice Souter, an admissions system should not weigh race so heavily that either it effectively operates as a quota or all minority applicants are ranked higher than all nonminority applicants. Note, however, that this latter test is absurd - only the most extreme admissions programs would rank all preferred minority applicants above all other applicants. This test does not offer a helpful benchmark for separating constitutional systems from unconstitutional ones.
184. Id. at 295.
185. The previous sentence is as follows: "The Court nonetheless finds fault with a scheme that 'automatically' distributes 20 points to minority applicants because 'the only consideration that accompanies this distribution of points is a factual review of an application to determine whether an individual is a member of one of these minority groups.'" Id.
186. We discuss both of these shortcomings infra in section V(A)(2).
187. Grutter v. Bollinger, 539 U.S. 306, 336-37 (2003) (emphasis added).
188. Id.
189. Justice O'Connor's views on this matter are consistent with her views in the voting redistricting cases - just as consideration of race is permissible (in that it passes strict scrutiny) in the admissions context as long as race is not the "defining feature" of an application, consideration of race is permissible (in that it does not trigger strict scrutiny) in the redistricting context as long as race is not the "predominant factor" motivating a redistricting decision. See, e.g., Bush v. Vera, 517 U.S. 952, 959 (1996) (O'Connor, J.) ("For strict scrutiny to apply, the plaintiffs must prove that other, legitimate districting principles were 'subordinated' to race. By that, we mean that race must be 'the predominant factor motivating the legislature's [redistricting] decision.'" (alteration in original) (citation omitted) (quoting Miller v. Johnson, 515 U.S. 900, 916 (1995))); cf. Easley v. Cromartie, 532 U.S. 234, 241 (2001) ("Race must not simply have been 'a motivation for the drawing of a majority-minority district,' but 'the "predominant factor" motivating the legislature's districting decision.'" (citation omitted) (quoting Vera, 517 U.S. at 959 (O'Connor, J.) and Hunt v. Cromartie, 526 U.S. 541, 547 (1999))).
190. Grutter, 539 U.S. at 338 (citations omitted).
191. For a description and discussion of what the Court means by the word "frequently," see infra notes 201-04 and accompanying text.
192. One circuit court has applied a more lenient version of the Grutter outcome test. See Smith v. Univ. of Wash., 392 F.3d 367, 376 (9th Cir. 2004) (stating, without empirical support, that "[t]he Law School also accepted nonminority applicants with grades and test scores lower than underrepresented minority applicants who were rejected, thus showing that the Law School 'seriously weigh[ed] many other diversity factors besides race that [could] make a real and dispositive difference'" (quoting Grutter, 539 U.S. at 338)). See infra note 204 for a discussion of Smith's use of this measure.
193. Grutter, 539 U.S. at 341.
194. See supra note 30 and accompanying text.
195. See infra subsection V(A)(2)(d) (criticizing this measure).
196. Cf. Girardeau A. Spann, The Dark Side of Grutter, 21 CONST. COMMENT. 221, 247 (2004) ("The real difference between the law school program upheld in Grutter and the undergraduate program invalidated in Gratz seems to be that the Supreme Court believes that the Gratz program gave too much weight to the factor of race, and it did so in a manner that was too transparent.");
197. cf. also PETER H. SCHUCK, MEDITATIONS OF A MILITANT MODERATE 12-18 (2006) (arguing that the strict scrutiny applied by the Grutter Court was anything but strict);
198. Nelson Lund, The Rehnquist Court's Pragmatic Approach to Civil Rights, 99 NW. U. L. REV. 249, 281 (2004) ("'Strict scrutiny' now means virtually no scrutiny, at least as to university admissions policies that discriminate against certain races, such as whites and Asians. To put the point another way, Grutter creates a safe harbor for such discrimination that extends over the whole ocean, except for one little cove that contains strictly unbending quotas and absolutely mechanical preferences like those at issue in Bakke and Gratz"). Another way of conceptualizing the "Don't Tell, Don't Ask" position of the Court is that not only should universities not "tell" by quantifying their admissions programs, but also universities should not "tell" by placing so much weight on race that the university effectively "tells" the Court something about the weight it places on race.
199. See supra notes 131-34 and accompanying text (criticizing the Court in Grutter for "tak[ing] the Law School at its word that its admissions program is nuanced" without a showing of actual differentiation while striking down the College's program in Gratz because of its "mechanical nature").
200. We do not conduct similar analysis on the Grutter test because the test itself is not well defined. See infra notes 201-05 and accompanying text. While we do not fully analyze how the College data fare under the Grutter test, we do report on some calculations that have some bearing on this question. See infra note 205.
201. Gratz v. Bollinger, 539 U.S. 244, 272 (2003) (omission in original) (quoting Regents of the Univ. of Cal. v. Bakke, 438 U.S. 265, 317 (1978) (Powell, J.)).
202. The Court cited the Appendix to the Petition for Certiorari. Id. at 254 (citing Petition for Writ of Certiorari app. at 111a, Gratz, 539 U.S. 244 (No. 02-516)).
203. When computing statistics for the College for 1999, we did not aggregate across years; we examined only the 1999 data. Also, when performing these computations for the College data, we were not able to aggregate all five years of data because the test score ranges used in the 1995 data were different from the ranges used in the 1996-1999 data. Because we were already separately performing computations for the 1999 data, we created separate grids for the 1995 data, the 1996-1998 data, and the 1999 data. After coming up with a different definition for "qualified" for each grid, we computed the relevant numbers for each grid and then aggregated those numbers to come up with the figures for the "College: all years" row of Table 5.
204. The Court stated that all applicants who were accepted were "qualified." See Grutter v. Bollinger, 539 U.S. 306, 338 (2003) (noting that "all underrepresented minority students admitted by the Law School have been deemed qualified"); Gratz, 539 U.S. at 303 (Ginsburg, J., dissenting) ("Every applicant admitted under the current plan, petitioners do not here dispute, is qualified to attend the College." (citing Petition for Writ of Certiorari app. at 111a, Gratz, 539 U.S. 244 (No. 02-516))); Petition for Writ of Certiorari app. at 111a, Gratz, 539 U.S. 244 (No. 02-516) ("According to the University, all of the students admitted to the [College] are qualified to attend the University, and for purpose of these motions, Plaintiffs assume this proposition to be true."). While all admits were qualified, it does not follow that all applicants with the same academic credentials as all admits were qualified. Indeed, it is likely that some of the admits with low qualifications were deemed qualified because of factors beyond their GPAs and test scores, such as strong letters of recommendation. We therefore did not want to count as qualified every applicant sharing a cell with an admit. We tried to strike a balance by deeming an applicant qualified if and only if greater than or equal to three applicants in her cell were admitted, and they made up greater than or equal to five percent of all applicants in that cell. This measure is, of course, to some degree arbitrary in that we could have chosen other figures besides the three admit and five-percent figures. Nonetheless, this measure seems to be a good attempt at approximating a definition of "qualified."
205. See supra notes 51-53 and accompanying text for a description of how we computed the number of but-for admits. Note that when computing the number of but-for admits, our data were broken down by year, residency status, test score range, and GPA range. We also computed the percentage of qualified preferred minority applicants who were accepted and, for comparative purposes, the percentage of nonpreferred qualified applicants who were accepted: (Table Presented) Thus, while the proportion of minimally qualified preferred minority applicants who were admitted was larger at the College, the proportion of minimally qualified preferred minority applicants who were admitted because of their race was larger at the Law School.
206. Our results were robust. To check the robustness, we used two closely related methods for approximating a definition for qualified. For the first alternative method, we defined an applicant to be "qualified" if the school admitted at least one person with the same characteristics as the applicant (ignoring race) - that is, the same GPA range, test score range, residency status, and year of application. This method differs from the method used in the text in several ways. First, we do not aggregate data (except race) before determining whether particular cells contain applicants who are qualified. Second, as long as any applicant with a given combination of characteristics (ignoring race) is admitted, all applicants with that same combination of characteristics are deemed "qualified," even if only a small number or percentage of such applicants are accepted. Third, this method does not deem qualified those applicants in cells where there are no admits, even if those cells have test score ranges and GPA ranges that are greater than or equal to other cells whose applicants have been deemed "qualified." Significantly, the results from this method are almost identical to the results we obtain when using the method described in the text:
207. Early in the opinion, Chief Justice Rehnquist noted that "it is undisputed that the University admits 'virtually every qualified . . . applicant' from these groups [underrepresented minorities]." Gratz, 539 U.S. at 254 (omission in original) (quoting Petition for Writ of Certiorari, supra note 163, app. at 111a). His statement that this fact was "undisputed" seems accurate - the brief for the University of Michigan stated that "LS&A ends up admitting virtually all minority applicants with competitive academic credentials." Brief for Respondents, Gratz, supra note 49, at 4-5. And, our calculations indicate that 90% of the qualified minority applicants were admitted for the years 1995-1999, and 87% were admitted in 1999; both figures are close to "virtually every" qualified applicant.
208. Perhaps Chief Justice Rehnquist meant to say that the College's admissions program was defective not because race was decisive for virtually every qualified minority applicant, but rather because virtually every qualified minority applicant was admitted. If this was the test that he meant to apply, then the College would have done worse under this measure than the Law School because the College admitted 90% of all qualified minority applicants in all years and 87% in 1999, while the Law School admitted only 63% of all qualified minority applicants. But, this cannot be a sensible metric because it penalizes less selective institutions like the College, which admit a higher percentage of all applicants. If an institution admits around 70% of all qualified applicants and gives a nontrivial boost for race, then it follows that "virtually every" qualified minority applicant will be admitted. In contrast, if an institution admits fewer applicants, such as, for example, 40% of applicants, and grants a boost of a similar size, the boost will not result in admitting "virtually every" qualified minority applicant. In addition, if the applicant pool of minority applicants who fall into the "qualified" category happens to be especially strong, then "virtually every" minority applicant may be admitted even in the absence of any preference for race. Justice Souter made this point in his dissent in Gratz. Gratz, 539 U.S. at 296 (Souter, J., dissenting) ("[T]he fact that the university admits virtually every qualified under-represented minority applicant, may reflect nothing more than . . . the possibility that self-selection results in a strong minority applicant pool." (citations and internal quotation marks omitted) (quoting Petition for Writ of Certiorari, supra note
209. app. at 111a)).
210. Peter Schuck has observed that the Grutter Court approved of the Law School's affirmative action program only because "[t]he program is sufficiently opaque by design and allows enough scope for subjectivity and discretion in the arbitrary and undisclosed weighting of the 'soft variables' in individual cases that a skeptic cannot prove unmistakably that race-ethnicity is the predominant factor in the admission of preferred minorities." SCHUCK, supra note 157, at 16. We agree that Grutter applied a lower level of scrutiny to the Law School program because of its opacity. We disagree, however, that the Court was not capable of determining the weight assigned to race in the Law School program. Indeed, Part III demonstrated one method the Court could have used to determine the weight the Law School assigned to race.
211. Circuit courts have varied in the degree to which they have adhered to what we argue is Grutter and Gratz's "Don't Tell, Don't Ask" approach. Only one circuit court has reached the question whether an admissions program in the higher education context was narrowly tailored to the diversity interest. Smith v. Univ. of Wash., 392 F.3d 367 (9th Cir. 2004). That court faithfully adhered to Grutter and Gratz's "Don't Tell, Don't Ask" approach, subjecting a law school admissions program that did not "tell" to little scrutiny. See id. at 375-82 (holding that the admissions program was narrowly tailored to the diversity interest). Several circuit courts have reached the question whether affirmative action programs in contexts other than higher education admissions are narrowly tailored to particular interests. Two circuit courts have avoided the "Don't Tell, Don't Ask" individualized consideration regime by expressly limiting the circumstances in which Grutter and Gratz require individualized consideration, holding that the individualized consideration requirement does not apply to K-12 non-merit-based school-choice plans. See Parents Involved in Cmty. Sch. v. Seattle Sch. Dist., No. 1, 426 F.3d 1162, 1180-84 (9th Cir. 2005) (en banc), cert. granted, 126 S. Ct. 2351 (2006); Comfort v. Lynn Sch. Comm., 418 F.3d 1, 17-19 (1st Cir. 2005) (en banc), cert. denied, 126 S. Ct. 798 (2005); cf. McFarland v. Jefferson County Pub. Sch., 416 F.3d 513 (6th Cir. 2005) (per curiam) (affirming district court opinion upholding a race-based school-choice plan but not conducting any independent analysis of the constitutionality of the plan), cert. granted sub nom. Meredith v. Jefferson County Bd. of Educ., 126 S. Ct. 2351 (2006); Goodwin Liu, Seattle and Louisville, 95 CAL. L. REV. (forthcoming Feb. 2007) (manuscript at 24-26, on file with authors) (arguing that the individualized consideration requirement of Grutter and Gratz should not apply to K-12 school-choice plans not based on merit);
212. James E. Ryan, Voluntary Integration: Asking the Right Questions, 67 OHIO ST. L.J. 327, 341-42 (2006) (same). The Supreme Court will soon decide the extent to which Grutter and Gratz apply to race-based K-12 school-choice plans when it decides Parents Involved and Meredith. Three circuit courts have implicitly limited the circumstances in which Grutter and Gratz require individualized consideration by upholding affirmative action programs without considering whether the programs provided individualized consideration. See W. States Paving Co. v. Wash. State Dep't of Transp., 407 F.3d 983, 995 (9th Cir. 2005) (upholding the Transportation Equity Act for the 21st Century, which authorizes the use of race- and sex-based preferences in federally funded transportation contracts), cert. denied, 126 S. Ct. 1332 (2006); Petit v. City of Chicago, 352 F.3d 1111, 1114-18 (7th Cir. 2003) (upholding an affirmative action program used to promote police officers); Sherbrooke Turf, Inc. v. Minn. Dep't of Transp., 345 F.3d 964, 969-74 (8th Cir. 2003) (upholding the Transportation Equity Act for the 21st Century). Nonetheless, Petit did address one of the weight requirements in Gratz. See supra note 142.
213. The Gratz opinion never directly cited precedent to argue against quantification, but there is one place in which the Gratz Court cited precedent in what may have been an effort to argue against quantification. Chief Justice Rehnquist followed the statement that "[t]he admissions program Justice Powell described . . . did not contemplate that any single characteristic automatically ensured a specific and identifiable contribution to a university's diversity," Gratz, 539 U.S. at 271, with citations to two cases, Bakke and Metro Broadcasting. To the extent that the Gratz Court meant for this statement to assert an anti-quantification position, the cases cited to support that position fail to do so. The page of Bakke that Gratz cited made two points. First, it stated that the diversity interest that supports an affirmative action program is not an interest in racial diversity alone, but rather "encompasses a far broader array of qualifications and characteristics of which racial or ethnic origin is but a single though important element." Regents of the Univ. of Cal. v. Bakke, 438 U.S. 265, 315 (1978) (Powell, J.). Second, it noted that the state's diversity interest would not be served "by expanding petitioner's two-track system into a multitrack program with a prescribed number of seats set aside for each identifiable category of applicants." Id. Thus, Justice Powell did not imply that in a plus-factor plan race cannot be assigned a "specific and identifiable" weight. The part of Justice O'Connor's dissenting opinion in Metro Broadcasting that Gratz cited stated that policies that "presume that persons think in a manner associated with their race" are constitutionally impermissible. Metro Broad., Inc. v. FCC, 497 U.S. 547, 618 (1990) (O'Connor, J., dissenting). Unqualified systems that take account of race in making admissions decisions, however, make assumptions about applicants that are based on race just as much as quantified systems. So, while this critique may provide a basis for arguing against racial preferences, it does not provide a basis for distinguishing between quantified and unquantified systems. Furthermore, a system that uses race to make admissions decisions - whether it is quantified or not - does not necessarily embody the notion that "persons think in a manner associated with their race." Rather, a race-conscious admissions system can embody the notion that in our society, race often affects people's experiences, which in turn shape how people think. Grutter v. Bollinger, 539 U.S. 306, 333 (2003) ("Just as growing up in a particular region or having particular professional experiences is likely to affect an individual's views, so too is one's own, unique experience of being a racial minority in a society, like our own, in which race unfortunately still matters."). Therefore, the only citations to precedent that may have been intended to support a no quantification requirement fail to support the position.
214. See Spann, supra note 157, at 243-44 (noting that "for any individual applicant, race is either dispositive or it is not," that this "is true whether race is used holistically in connection with a flexible admissions process, or mechanically in connection with a mathematical score" and that "[t]herefore, the differences that exist between the ways in which race was used in Grutter and in Gratz are simply irrelevant to any constitutionally protected individual right").
215. See infra notes 215-19 and accompanying text (describing the diversity interest).
216. See Bakke, 438 U.S. at 315 (Powell, J.) (stating that the diversity interest "encompasses a far broader array of qualifications and characteristics of which racial or ethnic origin is but a single though important element").
217. Gratz, 539 U.S. at 295 (Souter, J., dissenting) ("Since college admission is not left entirely to inarticulate intuition, it is hard to see what is inappropriate in assigning some stated value to a relevant characteristic, whether it be reasoning ability, writing style, running speed, or minority race. Justice Powell's plus factors necessarily are assigned some values. The college simply does by a numbered scale what the law school accomplishes in its 'holistic review'; the distinction does not imply that applicants to the undergraduate college are denied individualized consideration or a fair chance to compete on the basis of all the various merits their applications may disclose." (citation omitted)). Robert George makes a similar point when he states that the "question that the remaining Justices, particularly Justice O'Connor, fail to address adequately is: How can it be unconstitutional to do honestly and above board what it is constitutionally permissible to do 'through winks, nods, and disguises'?" George, supra note 109, at 1634-35 (quoting Gratz, 539 U.S. at 305 (Ginsburg, J., dissenting)).
218. Several scholars have noted that after Grutter and Gratz, courts sanction admissions systems that are lacking in transparency. See, e.g., Guinier, supra note 135, at 194-95 ("[T]he likely immediate consequence of Grutter is that trusted admissions officials are now freer to make their decisions without a great deal of transparency."); Post, supra note 20, at 74 ("Although transparency is ordinarily prized in the law, the Court in Grutter and Gratz constructs doctrine that in effect demands obscurity."); Sunstein, supra note 109, at 1905-07 (discussing the lack of transparency in systems not using points).
219. Gratz, 539 U.S. at 298 (Souter, J., dissenting).
220. Id. at 305 (Ginsburg, J., dissenting).
221. Id. at 280 (O'Connor, J., concurring). Justice O'Connor concluded that the review provided by the committee did not satisfy the requirement of individualized consideration, citing several grounds in addition to the one described in the text. See id. (noting two further problems with the committee's review of applications: (1) it was not clear that the committee reviewed a "meaningful percentage" of applications, and (2) qualification for review by the committee was still based in part on the selection index score).
222. In Grutter, the Court cited Law School policies that assert that each application is assessed in a manner that provides individualized consideration. See Grutter v. Bollinger, 539 U.S. 306, 337-39 (2003); see also supra notes 132-33 and accompanying text. But, the Grutter Court did not examine any evidence about whether decisions were actually made in accordance with these policies, except to examine some statistics suggesting that race was not always dispositive. See supra notes 132-33 and accompanying text. So, citation to policies asserting there was individual review was sufficient for the Grutter Court to support the proposition that the system provided individualized review. Yet, in Gratz, Justice O'Connor was not satisfied that the admissions review committee provided individualized review, despite the fact that according to the brief for the University of Michigan, this review was perhaps just as individualized: "every member of the [admissions review committee] closely reviews each applicant's entire file; the whole committee discusses the applicant's strengths and weaknesses." Brief for Respondents, Gratz, supra note 49, at 9.
223. Cf. Peter H. Schuck, Response to the Symposium Participants, 23 YALE L. & POL'Y REV. 75, 81 (2005) (arguing that affirmative action programs should be transparent because "transparency . . . is designed to discipline the granting of preferences by forcing institutions to be more candid about their value choices and by triggering reputational, market, and other informal mechanisms that make the entity bear more of the costs of adopting preferences instead of shifting them to innocent third parties" and noting that "[c]ustomers, students, alumni, investors, journalists, and other interests to which the entity must be attentive can then hold it accountable, rewarding, punishing, or ignoring the preferences, as they see fit" (internal quotation marks omitted)
224. (quoting PETER H. SCHUCK, DIVERSITY IN AMERICA: KEEPING GOVERNMENT AT A SAFE DISTANCE 196-97 (2003))).
225. See Richard Kahlenberg, The Conservative Victory in Grutter and Gratz, JURIST LEGAL INTELLIGENCE F., Sept. 5, 2003, http://jurist.law.pitt.edu/forum/symposium-aa/kahlenberg.php (characterizing the Grutter system as antidemocratic because it permits public universities to make preferential admissions decisions without having to disclose their policies to the public);
226. see also David Crump, The Narrow Tailoring Issue in the Affirmative Action Cases: Reconsidering the Supreme Court's Approval in Gratz and Grutter of Race-Based Decision-Making by Individualized Discretion, 56 FLA. L. REV. 483, 528 (2004) (noting that "the system struck down in Gratz has the advantage of making the level of the preference visible, so that it can be analyzed, critiqued, and reconsidered");
227. Drew S. Days, III, Fullilove, 96 YALE L.J. 453, 458-459 (1987) (observing that programs not "openly adopted and administered" do not "benefit[] from the scrutiny and testing of means to ends assured by public deliberation").
228. Cass Sunstein has elaborated: [I]t is unlikely that any particular officer [making admissions decisions under an unquantified program like the one at issue in Grutter] will be able to give constant, rather than fluctuating, weight to race. It is even less likely that different officers will use the same system, in the sense that they will allocate the same informal points, or weight, to race. The resulting criteria will likely be highly variable across applicants, and they will not be transparent to anyone. It follows that as compared to [a quantified program], [an unqualified program] sacrifices three important values: predictability, transparency, and equal treatment. Sunstein, supra note 109, at 1905; see also Leonard M. Baynes, Michigan's Minority Point System "Compensated" Minority Students for Inferior Public Education, JURIST LEGAL INTELLIGENCE F., Sept. 5, 2003, http://jurist.law.pitt.edu/forum/ symposium-aa/baynes.php (observing that "[r]elying solely on individualized determinations to implement affirmative action programs can cause" results that are "arbitrary" because an "individualized determination is a standardless process").
229. Indeed, the University of Michigan explained that one of the purposes of the point system was to make decisions consistent and fair: The volume of applications requires procedures and routines to promote fairness and consistency, while preserving counselors' ability to exercise judgment. A single, unitary set of guidelines, which is reviewed annually and altered periodically, governs the admissions process. The aim of the guidelines is to "blend the consistency of a formula with the flexibility of a review that is ultimately a matter of human judgment." Brief for Respondents, Gratz, supra note 49, at 6 (citation and footnotes omitted). And, university admissions officers have reported that quantifying characteristics helps produce more consistent results. See Guinier, supra note 135, at 195 n.318.
230. See Crump, supra note 181, at 528; see also Jessica Bulman-Pozen, Note, Grutter at Work: A Title VII Critique of Constitutional Affirmative Action, 115 YALE L.J. 1408, 1420 (2006) ("Grutter's conception of narrow tailoring depends on the same unchecked subjective decisionmaking that, according to Title VII doctrine, invites bias.").
231. Guinier, supra note 135, at 196.
232. Id. at 196 n.320.
233. For example, Robert Post states that the Gratz Court may be prohibiting point systems because "the twenty-point bonus sends a message to applicants and to the world that being a member of a racial group is worth a certain, named amount, and it therefore invites members of that group to feel entitled to that amount." Post, supra note 20, at 74; see also Sunstein, supra note 109, at 1906 (describing the argument that unquantified systems impose a lesser expressive harm than quantified systems). See generally Primus, supra note 101, at 566-67 (describing literature discussing expressive harm).
234. Cf. Ayres, supra note 21, at 1793-800 (describing - and refuting - arguments that race-neutral means are less socially divisive than race-conscious means).
235. Cf. id. at 1793-96 (criticizing the argument that race-neutral means are less socially divisive than race-conscious means because the citizenry is less likely to be aware of the racial motivation of the former); cf. also Sunstein, supra note 109, at 1906 (persuasively refuting the argument that quantified affirmative programs impose a stigma on minorities while unquantified programs do not).
236. Cf. Ayres, supra note 21, at 1795-96 (noting that the doctrine requiring consideration of race-neutral alternatives may force legislatures to be less open about their use of race, shielding not only the citizenry, but also the courts from information about when legislatures are acting on the basis of race). Of course, if the Court prohibits only ex ante quantification (e.g., points systems) but not ex post quantification (e.g., regressions that capture the weight assigned to race), then we need not worry that the Court is being shielded from information about the amount of racial preferences. We do, however, still need to be worried that the citizenry does not have access to this information.
237. See infra subpart V(B).
238. The test that measured a constitutionally relevant burden was Chief Justice Rehnquist's Gratz test, which compared the number of points assigned to race to the number of points assigned to other characteristics. See infra subsection V(A)(2)(a) (arguing that while Rehnquist's test in Gratz captures the means dimension, it is inadequate because it does not measure outcomes and fails to suggest a theory that determines how much weight assigned to race is excessive).
239. See Gratz v. Bollinger, 539 U.S. 244, 273 (2003); see also supra subsection IV(B)(3)(a).
240. Gratz, 539 U.S. at 270; see also supra subsection IV(B)(3)(a).
241. For a description of the point values assigned to characteristics under the system at issue in Gratz, see supra notes 92-93 and accompanying text.
242. Gratz, 539 U.S. at 272 (omission in original) (quoting Regents of the Univ. of Cal. v. Bakke, 438 U.S. 265, 317 (1978) (Powell, J.)).
243. Bakke, 438 U.S. at 317 (Powell, J.).
244. See supra note 35 and accompanying text.
245. Gratz, 539 U.S. at 272.
246. Grutter v. Bollinger, 539 U.S. 306, 338 (2003).
247. Id.
248. Brief for Respondents, Grutter, supra note 88, at 10.
249. As a reference point, each year the Law School made approximately 1,300 offers of admission and received over 3,500 applications. Id. at 2 n.3.
250. Indeed, perhaps for this reason, the one circuit court that has applied this "test" ignored the word "frequently" altogether, noting that the law school whose admissions program was at issue "also accepted nonminority applicants with grades and test scores lower than underrepresented minority applicants who were rejected, thus showing that the Law School 'seriously weigh[ed] many other diversity factors besides race that [could] make a real and dispositive difference.'" Smith v. Univ. of Wash., 392 F.3d 367, 376 (9th Cir. 2004) (alterations in original) (quoting Grutter, 539 U.S. at 338). The Smith test is, of course, entirely too lenient, as an admissions program would pass muster under that test as long as at least two nonminority applicants were accepted with grades and test scores that were lower than two minority applicants who were rejected.
251. If this is the case, then we might be able to demonstrate that the College also "frequently" admitted nonpreferred applicants over preferred applicants with higher scores who were rejected. Indeed, our data analysis shows that if we divide College applicants into two groups, the first of which is applicants with a GPA of 2.6 or higher and an SAT of 900 or higher, and the second of which is applicants with either a GPA of less than 2.6 or an SAT of less than 900, we find that between 1995 and 1999, 595 nonpreferred applicants with the lower credentials in the latter group were accepted, while 165 preferred minority applicants with the higher credentials in the former group were rejected. So, on average, each year 119 nonpreferred applicants with the lower credentials were accepted, while 33 preferred minority applicants with the higher credentials were rejected. As reference points, for the years between 1995 and 1999, each year the university admitted an average of 10,011 applicants and received an average of 14,230 applications. Whether this reference cell is an appropriate one to use for the test and whether these numbers rise to the level of "frequently" is, of course, entirely unclear because the Court did not elaborate on this test.
252. Grutter, 539 U.S. at 320. The Court recited this statistic in the section of the opinion describing the facts and the district court opinion.
253. The University of Michigan's expert statistician computed similar statistics for other years at the Law School and the College, and the Court had access to these statistics. See supra notes 61-72 and accompanying text. Despite its access to such statistics, however, the Court showed no interest in factoring such statistics into its constitutional calculus.
254. The First Circuit has made progress in examining marginal costs and benefits. See Comfort v. Lynn Sch. Comm., 418 F.3d 1, 15, 21 (1st Cir. 2005) (en banc) (finding it significant that the district court concluded that "the benefits of intergroup contact continue to accrue as a school becomes increasingly diverse"), cert. denied, 126 S. Ct. 798 (2005).
255. Courts may need to aggregate the outcome and means dimensions. We describe one possible way to aggregate these dimensions. See supra note 43.
256. If the yield - the percentage of admits who accept offers of admission - is the same for preferred minorities and nonpreferred applicants, then the percentage of admits who are but-for admits will be the same as the percentage of enrollees who are but-for admits. If, however, the yield is different for preferred minorities and nonpreferred applicants, then the two statistics will be different. See supra note 44.
257. Of course, it may be that the number of but-for admits is also independently significant. Indeed, if two admissions programs have the same percentage of enrollees who are but-for admits but different numbers of admits who are but-for admits, the program with the larger percentage of admits who are but-for admits may be more burdensome than the one with the smaller percentage.
258. In some contexts other than higher education, it makes sense to measure costs only along the outcome dimension. For example, when considering a secondary school assignment policy that takes account of race but does not take account of qualifications of the students, there is no need to examine whether there is a differential in qualifications along the means dimension, as qualifications are irrelevant to the admissions decision. Relatedly, the cost of each unit along the outcome dimension will vary depending on the context. See Comfort, 418 F.3d at 20 ("Every child . . . is guaranteed a seat in a district where, as the parties have stipulated, every school provides a comparable education. The denial of a transfer under the Plan is therefore markedly different from the denial of a spot at a unique or selective educational institution."). Finally, the cost calculation should probably take into account not only the extent to which the affirmative action program at issue burdens individuals, but also the extent to which the program burdens each race. All other things being equal, it may make sense that an affirmative action program imposing an equal burden on the races imposes lower net costs than an affirmative action program imposing an unequal burden on the races. See Parents Involved in Cmty. Sch. v. Seattle Sch. Dist., No. 1, 426 F.3d 1162, 1192 (9th Cir. 2005) (en banc) (finding it significant that "the [racial] tiebreaker does not uniformly benefit any race or group of individuals to the detriment of another"), cert. granted, 126 S. Ct. 2351 (2006).
259. When we use the word "qualified," we mean qualified in a holistic sense, taking into consideration all qualifications, including academic and nonacademic qualifications (but not race).
260. Cf. supra note 74 (discussing BOWEN & BOK, supra note 42 and Wightman, supra note
261. This analysis assumes that the marginal costs along the outcome dimension do not depend on the percentage of admits who are but-for admits, an assumption that may not be accurate. For example, the costs of increasing the percentage of but-for admits from 1% to 2% might be lower than the costs of increasing the percentage of but-for admits from 21% to 22%. To the extent that the marginal costs vary depending on the percentage of admits who are but-for admits, these varying marginal costs on the outcome dimension should be considered as well.
262. Lani Guinier characterizes Justice O'Connor's opinion as emphasizing three elements of the diversity interest, the first of which was endorsed by Justice Powell's opinion in Bakke, and the second two of which are new. These three elements are as follows: "diversity is pedagogical and dialogic; it helps challenge stereotypes; and it helps legitimate the democratic mission of higher education." Guinier, supra note 135, at 176 (citations omitted). The Ninth Circuit has characterized Grutter as articulating two elements of the diversity interest, combining our first and second elements into a single element. See Smith v. Univ. of Wash., 392 F.3d 367, 373 (9th Cir. 2004) (characterizing the two primary benefits of diversity articulated in Grutter as the enrichment of the "educational experience itself and ensuring the effective participation of all racial and ethnic groups in the nation's civic life). Several commentators have analyzed the ways Justice O'Connor's diversity interest differs in scope from Justice Powell's. See Guinier, supra note 135, at 173-76; K.G. Jan Pillai, The Defacing Reconstruction of Powellian Diversity, 31 T. MARSHALL L. REV. 1, 40 (2005) (arguing that "Justice O'Connor chose to expand the Powellian concept of diversity specifically to accommodate the integrative and remedial components of all [extant] race-conscious admissions policies"); Post, supra note 20, at 59-61 (comparing Grutter's focus on, inter alia, ensuring the participation of all racial and ethnic groups in the nation's civic life and leadership with Justice Powell's focus in Bakke on the educational benefits of diversity); Siegel, supra note 108, at 1538 (noting that "Grutter does not simply incorporate Justice Powell's diversity rationale for race-conscious admissions practices into the fabric of constitutional law" but rather that "Grutter transforms the diversity rationale in the course of adopting it, expanding the concept of diversity so that it explicitly embraces antisubordination values"); Rosman, supra note 120 (comparing the diversity rationales in Grutter and Bakke); see also Primus, supra note 101, at 560 (characterizing the diversity interest articulated in Grutter as being "in part about mitigating the effects that existing racial hierarchies have on the composition of national leadership cadres" and not just about diversity's educational benefits).
263. Grutter v. Bollinger, 539 U.S. 306, 330 (2003) (quoting Petition for Writ of Certiorari app. at 246a, 244a, Grutter, 539 U.S. 306 (No. 02-241)).
264. Id. (alteration in original) (quoting Petition for Writ of Certiorari app. at 246a, Grutter, 539 U.S. 306 (No. 02-241)).
265. Id. at 331.
266. Id. at 332.
267. Note that the marginal benefits as a function of the number of minority enrollees may not be - and likely is not - a constant function. Many claim that in order to enjoy many of the benefits of a diverse student body, schools must enroll significant numbers of minority students. See, e.g., Patricia Gurin et al., Diversity and Higher Education: Theory and Impact on Educational Outcomes, 72 HARV. EDUC. REV. 330, 360-61 (2002) (explaining that enrolling "significant numbers of students of various groups" is necessary to enable students to "perceive differences both within groups and between groups"); cf. Regents of the Univ. of Cal. v. Bakke, 438 U.S. 265, 316 (1978) (Powell, J.) (noting the "necessity of including more than a token number of black students"). If this is so, then the marginal benefits of admitting preferred minority applicants are smaller when the number of preferred minority students who will be enrolled is less than the number needed to form a "critical mass," and the marginal benefits then jump suddenly when the number of preferred minority students who will be enrolled reaches the "critical mass number."
268. Cf. Kathryn R.L. Rand & Steven Andrew Light, Teaching Race Without a Critical Mass: Reflections on Affirmative Action and the Diversity Rationale, 54 J. LEGAL EDUC. 316, 332-34 (2004) (noting that under a cost-benefit analysis it may be more difficult to uphold an affirmative action program when a university is unable to enroll a critical mass of minority applicants). Whether universities have an interest in enrolling more than a critical mass of minority students should turn on the marginal benefits (and costs) of enrolling those students. Cf. Comfort v. Lynn Sch. Comm., 418 F.3d 1, 20-22 (1st Cir. 2005) (en banc) (upholding an affirmative action program that sought to enroll more than a critical mass of minorities because the evidence suggested that marginal benefits accrued when more than a critical mass of minority applicants was enrolled, but not analyzing the marginal costs), cert. denied, 126 S. Ct. 798 (2005).
269. Interestingly, it follows from this analysis that universities can assign more weight to race under Justice O'Connor's diversity rationale than they could under Justice Powell's diversity rationale. This is because Justice Powell focused only on the benefits of diversity that were intrinsic to the university experience and were not race-specific. In his view, a diverse student body helps create an "atmosphere . . . conducive to speculation, experiment and creation," and promotes the "robust exchange of ideas." Bakke, 438 U.S. at 312 (Powell, J.) (internal quotation marks omitted). In contrast, in Grutter, Justice O'Connor focused not only on benefits of holistic diversity that were intrinsic to the university setting, but also on race-specific benefits enjoyed both on and off campus. See supra text accompanying notes 215-19. In particular, the Grutter Court stated that racial diversity "promotes cross-racial understanding, helps to break down racial stereotypes, and enables [students] to better understand persons of different races"; helps ensure that higher education is "accessible to all individuals regardless of race or ethnicity"; and also makes the "path to leadership . . . visibly open to talented and qualified individuals of every race and ethnicity." Grutter, 539 U.S. at 330-32 (internal quotation marks omitted). Unlike the benefits of diversity described by Justice Powell, these benefits are particular to racial diversity, and the latter benefits extend beyond the university campus. Thus, when conducting marginal cost-benefit analysis, the marginal benefit to admitting a preferred minority applicant under Justice O'Connor's diversity rationale is greater than or equal to the marginal benefit to admitting a minority applicant under Justice Powell's diversity rationale because there are more marginal benefits to admitting a minority applicant under Justice O'Connor's rationale. Cf. Post, supra note 20, at 62 (noting the "far-reaching implications" of the Grutter diversity rationale and finding it to be in conflict both with the Court's assertion that "'outright racial balancing' would be 'patently unconstitutional'" and also with the individualized consideration requirement (quoting Grutter, 539 U.S. at 330)); see also supra note 215.
270. Two post-Grutter and Gratz circuit court opinions addressing affirmative action programs in contexts other than higher education have made some progress towards engaging in this cost-benefit inquiry. In Parents Involved in Community Schools v. Seattle School District, No. 1, the Ninth Circuit considered the constitutionality of a secondary school assignment plan. 426 F.3d 1162 (9th Cir. 2005) (en banc), cert. granted, 126 S. Ct. 2351 (2006). The court computed the percentage of admits who were but-for admits, id. at 1170-71, and thus computed all relevant costs, see supra note 212. The court did not, however, go on to quantify the benefits and conduct cost-benefit analysis. It did, though, attempt to justify the particular amount of weight placed on race by noting the similar weight assigned to race in desegregation plans. See Parents Involved, 426 F.3d at 1186 (noting that, like the plan at issue in the case, most desegregation programs had a goal of matching the surrounding community within fifteen percentage points). In Petit v. City of Chicago, the Seventh Circuit considered the constitutionality of an affirmative action program that gave minorities a boost on a test used to determine which employees to promote. 352 F.3d 1111 (7th Cir. 2003). The opinion discussed data from which the number of but-for promoted employees could be calculated and thus came close to measuring the weight assigned to race along the outcome dimension. See id. at 1116-17 (stating the number of minorities who would have been promoted in the absence of the affirmative action program and the number of minorities actually promoted). In addition, the opinion cited data that attempted to measure the test score boost minority applicants received and thus came close to measuring the weight assigned to race along the means dimension. See id. at 1117 (comparing the pre-boost and post-boost test scores for the least-qualified white, African American, and Hispanic applicants in the top 500 applicants). The opinion did not, however, attempt to quantify the benefits, and it did not compare the costs and the benefits. It may be that the opinion did not do so because the test score adjustment scheme at issue was justified, in part, on the ground that it was necessary to compensate for a discriminatory test that unfairly favored whites. See id. ("[S]tandardizing the scores can be seen not as an arbitrary advantage given to the minority officers, but rather as eliminating an advantage the white officers had on the test.").
271. Moreover, courts should examine the extent to which preferences are differentiated in fact by looking at actual practices, not by examining only university policies that merely assert that preferences are differentiated. The extreme deference that Justice O'Connor showed to state officials is deeply inconsistent with the whole idea of strict scrutiny as an attempt to smoke out unjustified governmental racial preferences. That is, we would look at actual differentiation, not potential differentiation.
272. See supra text accompanying note 212.
273. See supra notes 213-14 and accompanying text.
274. For a description of the Grutter diversity rationale, see supra notes 215-19 and accompanying text.