Introduction

In 1997, the graduate and professional schools of the University of California implemented a resolution passed by the system’s Board of Regents in the Summer of 1995 and ceased to consider the race of applicants, thereby beginning the end of affirmative action in higher education in the state. Just months before, the voters of California had followed the Regents’ lead and voted to end all affirmative action in public hiring, contracting and education. The results at most of UC’s professional and graduate programs were somewhat disheartening, including at Boalt Hall, Berkeley’s law school. Boalt made 860 offers of admission: 122 to Asians, 46 to Latinos, 15 to African-Americans, 2 to Native Americans, and 675 to White applicants. Of those offered admission, 268 applicants opted to enroll; 215 of them were White, 38 were Asian, 14 were Latino, one was African American, and none were Native American. Those numbers prompted a complaint to the Department of Education’s Office for Civil Rights, which began an investigation over the Summer of 1997. The complaint alleged, in part, that Boalt had used a facially neutral policy that had disparate racial impacts in violation of Title VI of the 1964 Civil Rights Act, which prohibits discrimination in federally funded programs, including the UC schools. This paper investigates that same question, reaching the conclusion that Boalt did practice impermissible discrimination.

Part I briefly summarizes the path by which the UC Regents and the voters of California decided to end the state’s public affirmative action programs. Part II describes the history and present state of disparate impact theory under Title VI. Because the courts have held that a case of disparate impact under Title VI should follow the framework of such a case under Title VII, Part III summarizes the current jurisprudence of Title VII disparate impact theory. Part IV applies the facts of the situation to that framework, concluding that Boalt’s facially neutral policy had impermissibly discriminatory effects on racial minorities.

The California Constitution no longer allows the consideration of race as part of affirmative action programs in public education, public employment, and public contracting; at least, that is the simple interpretation of a law that is poorly worded, unclear, and that has not yet been tested in court or in legislatures as to its specific implications and applications. How the law got changed is a political tale. Like most political tales, it is the story of small men, personal and partisan agendas that have nothing to do with the law itself, short-run goals, the absence of meaningful dialog as prelude to policy formulation, ill-conceived policies, and products that, though facially clear, lack polish, direction, substance, or detail. Before the state electorate voted to amend the California Constitution, the Regents of the University of California (UC), led by Ward Connerly, a political appointee of a Republican governor with presidential aspirations (i.e. a governor looking to increase his national notoriety), and a man with no pedagogical knowledge to speak of but any number of personal opinions without philosophical framework or factual support, voted to end the consideration of race, religion, sex, color, ethnicity, or national origin as criteria in admission to UC schools. As a result of that vote, all UC schools, including UC Berkeley’s School of Law (Boalt Hall), reformulated their admission policies to eliminate all consideration of race. In 1997, the first year in which Boalt applied its reformulated policies, as many predicted, the number of African-Americans, Latinos, and Native Americans admitted to the entering class dropped dramatically. The number of those students who decided to enroll at Boalt was even lower. The Mexican American Legal Defense and Educational Fund (MALDEF) subsequently filed a complaint with the United States Department of Education’s Office for Civil Rights alleging that Boalt’s admission policy had had a disparate racial impact in violation of Title VI of the 1964 Civil Rights Act (CRA). That is the issue at the heart of this paper.

When California Governor Pete Wilson appointed Ward Connerly to the UC Board of Regents in 1993, the two had known each other for twenty-five years. They had met when Wilson, then a young legislator, was appointed to the state Assembly’s Committee on Urban Affairs and Housing and Connerly was working in the state Department of Housing and Community Development. After working together for several years, Wilson urged Connerly to leave government and enter private enterprise, which Connerly did, becoming a successful land use consultant. Their relationship did not end, however; as part of his business, Connerly has taken advantage of parts of the state’s housing law that Wilson wrote, and with his business contacts, Connerly has helped to raise millions of dollars for Wilson’s various campaigns. Considering their personal and professional relationship, then, one could hardly be surprised that Wilson appointed Connerly to the Board, a veritable club of the monied, powerful, and politically connected. What might nevertheless give one pause is the appointment of a man with a personal agenda but no expertise in the area in which he has been given power, a man known to play fast and loose with facts, who resorts to the tools of illogic and bias--appeals to emotion, popular appeals, poisoning the well, and ad hominem attacks--when he has no grounds for his arguments, and who passes such illogic off as a sound philosophical framework.

When Connerly joined the Board, he almost immediately carved out a unique place with what he himself has called a dogmatically rhetorical style that usually involved a near-harassing campaign of faxes and letters to other Regents expressing his opinions on the issues on the Board’s agenda. But on no issue was that style used to greater effect than on his resolutions to eliminate affirmative action in UC’s hiring, contracting, and admissions. Whether SP-1 and SP-2 were the result of personal inspiration or a concocted scheme between him and Wilson, the Regents passed them both in July of 1995, and the vote garnered Wilson’s fledgling, but ultimately abortive, presidential campaign the media attention it needed if he was to be able to solicit donations from outside the state. The requirement that UC’s graduate schools cease considering race and sex in their admissions processes did not become effective until January 1, 1997, but in the interim, the California electorate, much at the urging of Connerly and Wilson, followed the Regents lead.

Enter Tom Wood, Connerly’s and Wilson’s partner in the campaign for the Proposition with the doublespeak name--the California Civil Rights Initiative (CCRI)--and a veritable poster child for the hypothetical White men much abused by the supposedly "reverse discriminatory" effects of affirmative action. Wood and Glynn Custred, another frustrated academic, purportedly had the idea for something like Proposition 209 independently, but they joined to write the actual initiative in 1991. Together they tried and failed twice to gather sufficient signatures to get the measure on the state’s ballot. By 1993, they realized that they needed help, so they took their idea to well connected members of the Republican Party, who recognized that an initiative (or federal statute) ending affirmative action but framed as a civil rights bill would make a fantastically powerful wedge issue to divide Democratic voters. From the GOP they received financing, national public support and press, and general aid. What the campaign really needed, however, was an event to bring the issue of affirmative action from the background of societal dialog, where conservatives themselves had labored for years to put it, into the forefront of American politics; of course, the event had to be one that would help them frame and deliberately mischaracterize the elimination of affirmative action as a continuation of the country’s evolving civil rights struggle. The Regents’ decision to end affirmative action in the UC system provided just the needed event.

And so it made sense that Connerly became a leading figure in the campaign to pass the CCRI as well. Connerly, while making the colorblind society argument and saying that he had succeeded without any special help because of his race, nevertheless used his race to give the campaign a credibility it otherwise might have lacked. At the same time, Wood argued for the Initiative in terms of its fairness to White men long and unjustly burdened by affirmative action, using himself as an example. A wannabe academic, Wood had not received teaching positions, he claimed when he first started to receive media attention, because they had gone instead to less qualified minorities. It later became known that he had not even applied for one of the positions, and those hired might have been at least as qualified, if not more so, than he. Apparently never considering the possibility that it was he who was unqualified to teach in any classroom anywhere in the country, or perhaps realizing that exactly that point, he turned his energy and time to the campaign for CCRI, using his perhaps fallacious experiences as examples of the dangers of affirmative action, adding to the anecdotal evidence of such cases, which is the only kind affirmative action foes ever seem to have.

Despite the perhaps purposefully and ultimately unhelpfully muddled language of the Proposition that the state’s Office of the Legislative Counsel allowed onto the ballot, despite the almost transparent agendas at work behind the measure, and perhaps because of a poorly run and sickly underfunded opposition campaign, in November of 1996, 54% of those Californians who bothered to cast ballots voted to approve the CCRI. Other numbers, however, show that much of the state, at least by the end, had managed to see through the well orchestrated charade: 74% of African-American, 76% of Latinos, and 61% of Asian-Pacific-American residents voted against the measure, as did 59% of those with incomes under $20,000. Similarly, three large counties (Los Angeles, Alameda, and Santa Clara), the City of Los Angeles, San Francisco, and the Bay Area generally all voted against CCRI. Their votes were not enough, though, so opponents had to turn to the courts.

By December 23, 1996, Judge Thelton Henderson had ruled that plaintiff Coalition for Economic Equity had demonstrated a probability of success on its claim that Proposition 209 violates the Equal Protection Clause of the Fourteenth Amendment of the United States Constitution under the Hunter v. Erickson and Washington v. Seattle Sch. Dist. No. 1 line of cases and granted a preliminary injunction against enforcement of the measure.

Defendants, of course, appealed and pulled perhaps the most mind-bogglingly reactionary panel of circuit judges one could hope/fear to encounter. That panel, unsurprisingly, overruled Judge Henderson; though it attempted (poorly) to find a way around the Hunter and Seattle jurisprudence, its decision is rather transparently grounded in the shockingly simpleminded belief that a popular enactment could not possibly violate the Equal Protection Clause and that a federal court should not interfere in the decisions of the voters of a state. By the end of the summer of 1997, the Ninth Circuit voted not to grant an en banc rehearing of the panel decision. And the Supreme Court put the matter of 209’s constitutionality to rest temporarily by refusing to grant cert and letting the panel’s decision stand.

In the interim, however, Boalt Hall had proceeded to implement the Regents’ decision by adjusting its admission policy. Part IV.A describes the change and its results in detail, but put simply, the school’s admission policy in 1997 contained no consideration of race itself as a criterion, though it did allow candidates to describe adversities they had faced and overcome. The result of the new policy was a radical drop in the number of African-Americans, Latinos, and Native Americans admitted; the number of individuals who identified themselves as part of those groups who decided to attend Boalt was, of course, even lower. In part on the basis of those numbers, MALDEF filed a complaint with OCR raising the question that this paper explores: whether in complying with SP-1 and Proposition 209, Boalt implemented a facially neutral policy that nevertheless had discriminatory effects in violation of Title VI.

The Supreme Court has only rarely addressed the scope of disparate impact theory under Title VI of the Civil Rights Act of 1964, and its jurisprudence in the area, though relatively clear in holding that one may bring a disparate impact claim against those who receive federal funding, has not, perhaps because of the rarity of pronouncements, provided a detailed approach to how to frame such a claim. Indeed, the Court has, since the passage of the Act, changed its mind several times about whether one may bring such a claim, whether one may bring it under the statute itself, or whether one may only bring it under Title VI’s implementing regulations. And considering the number of justices who have left the bench since the last decision on the issue, the Court may well change its mind again about whether one may bring such a claim at all, if given the chance. Thus, although the following is a fair summary of the law as it is, the Court may rewrite this chapter of civil rights law if the case against Boalt proceeds that far.

Under its most recent decisions, Guardians Ass’n v. Civil Service Comm’n of the City of New York and Alexander v. Choate, both already more than a decade old, the Court said that the prohibitory language of Title VI reaches only intentional discrimination, but that plaintiffs can press disparate-impact-based claims of discrimination under the regulations that the Department of Education has issued pursuant to Title VI. How the agency then had the statutory authority to create such regulations is difficult to understand. On the other hand, when the Court decided Guardians, it still claimed to give "great deference" to agencies’ interpretations of the statutes Congress empowered them to implement; and the Chevron doctrine should have led the Court to defer to the agency’s interpretation of the ambiguous term "discrimination" in Title VI when it decided Alexander v. Choate. Thus, it is equally as troubling that the Court would ignore those principles and give the somewhat ambiguous language of the statute a different meaning, even though agencies promulgated the regulations years before the Court first spoke on the issue, and Congress had never acted to overturn the agencies’ interpretation of Title VI. The following is a summary of the current status of disparate impact theory under Title VI and the twisted path by which the Court got there.

In the course of five years and in just two decisions, the Court managed to make and then change the law under Title VI. The statute states, briefly and ambiguously, "No person in the United States shall, on the ground of race, color, or national origin, be excluded from participation in, be denied the benefits of, or be subjected to discrimination under any program or activity receiving Federal financial assistance." The next section states that

In conformance with Congress’ command, the then Department of Health, Education, Education, and Welfare, issued general regulations regarding federally funded schools as early as 1968. By 1970, it had made them more specific. They now read, in relevant part, as follows:

The relevant portions of the regulations had substantially the same form when the Court decided its first Title VI case, Lau v. Nichols, in 1974. In the decision, the majority did not treat the regulations and the statute as conflicting in any way: the regulations prohibited policies that had discriminatory effect, even when no purposeful design was present, and a violation of the regulations was a violation of the law. Those who wrote separately wondered "whether the regulations and guidelines promulgated by HEW go beyond the authority of § 601." However, they concluded that the regulations were reasonably related to the purposes of the Act, implying that the purpose of Title VI was to eliminate both intentional discrimination and policies with discriminatory effect. Indeed, although somewhat circular, they gave the regulations "great weight" in determining what the purpose of the legislation was.

That deference and perception of consistency did not last more than five years, as the Court’s decision in Regents of the University of California v. Bakke cast serious doubt on Lau’s interpretation of Title VI. The votes and opinions in Bakke are as strange as they are messy. Justice Powell, writing for himself and delivering the opinion of the Court, noted but ignored both Lau and the regulations and held, "In view of the clear legislative intent, Title VI must be held to proscribe only those racial classifications that would violate the Equal Protection Clause or the Fifth Amendment." Two years earlier, the Court had held that only purposeful discrimination by state and federal actors was subject to strict scrutiny under the Fourteenth and Fifth Amendments, respectively; in other words, according to the Court, the Constitution and, after Bakke, Title VI prohibit only intentional discrimination. Although Powell had voted with the majority in Washington v. Davis, he had also voted with the majority in Lau.

Even stranger, Brennan and Marshall, who voted with the majority in Lau and dissented in Davis, agreed with Powell on the proper interpretation of Title VI. The decision, however, does not seem so strange, only short-sighted, when one recalls that Bakke was a "reverse discrimination" case in which a two-time applicant turned down by many other medical schools claimed to have been the victim of racial discrimination because UC Davis Medical School did not consider him for its special admissions program (thereby demonstrating an audacious refusal to admit only white men). The bloc of justices formed by Stevens, Burger, Stewart, and Rehnquist wanted to decide the case on statutory grounds by holding that the school had violated the law by excluding Bakke on the basis of race. For them, "the meaning of Title VI’s ban on exclusion is crystal clear: Race cannot be the basis of excluding anyone from participation in a federally funded program." Their reading of the law is understandable, if simpleminded, but such statements are ultimately unhelpful, since they do not say anything about how one proves such exclusion. At issue in the case was a program that, at least according to the Court, facially considered race, so proving exclusion on that basis was not difficult. However, one cannot deny the possibility that a recipient of federal funds could effectively deny someone participation in the program on the basis of race, and the Stevens opinion, despite its survey of legislative history, provides no convincing reason why such discrimination should not be legally cognizable.

More difficult to understand, considering their defense of affirmative action in other cases, is Brennan and Marshall voting with Powell to interpret Title VI as coextensive with the Constitution and, therefore, prohibiting only intentional discrimination in federally funded programs. Perhaps they believed that only by adopting that position of Powell’s could they get the votes they needed to assure that affirmative action would remain a viable option for public schools in the future to remedy past societal discrimination. Thus, they were able to prevent the outcome desired by Stevens et al. only by effectively disempowering Title VI by reading the statute to prohibit only intentional discrimination in federally funded programs. As Brennan’s opinion acknowledged, the position that he, his bloc, and Powell staked out called Lau seriously into question. Only by later importing the principles of disparate impact under Title VII into the DoE’s regulations were they able to return some power to Title VI and make it an effective means of fulfilling the one Congressional purpose on which all agree: the elimination of segregation in federally funded programs.

Having left the proper interpretation of Title VI somewhat unclear in its absurdly messy opinion(s) in Bakke, the Court again addressed the issue a few years later in Guardians Ass’n v. Civil Service Comm’n of New York. The votes in the latter case were equally messy, but those justices with an interest in actually preventing discrimination in federally funded programs managed to restore some remedial power to Title VI by cobbling together five votes for the proposition that the regulations promulgated under the Title prohibit policies with discriminatory effects, and that the regulations are valid, even if the Title itself does not prohibit the use of such policies. Justice White, who delivered the judgment of the Court and was one of those five, began his analysis by distinguishing any holding regarding Title VI in Bakke on the ground that the issue there was

Put somewhat more clearly, in White’s opinion, the prohibitory reach of Title VI and the Constitution are coextensive when applied to an affirmative action policy: both disallow only intentional racial discrimination; unlike the Constitution, however, Title VI itself also prohibits facially neutral policies that have discriminatory effects. Thus, Bakke did not actually overrule Lau.

From there, Justice White went on to reason that

So, Justice White believed that Bakke’s ruling on Title VI applied only to affirmative action policies, that Title VI otherwise prohibited unintentional discrimination, and that the regulations under Title VI were valid and reached disparate impact, even if the statute did not. Unfortunately, the majority that agreed on the last point did not also agree on the first.

Indeed, of the nine justices, seven agreed that discrimination under Title VI does not encompass disparate impact, five agreed that the regulations promulgated under the Act validly do, and only Marshall agreed with White, presenting the most compelling argument. In his opinion, Justice Marshall summarized the history of administrative regulations issued pursuant to Title VI. He found that they had "uniformly and consistently interpreted the statute to prohibit programs that have a discriminatory impact and that cannot be justified on nondiscriminatory grounds." He also found that that interpretation originated in the same year as the Act, 1964; that Congress had never altered that interpretation, though it had been aware of it; that it had modeled other statutes on it without requiring an intent test; and that it had even rebuffed attempts to overturn that interpretation. He therefore concluded as follows:

Justices Stevens, Brennan, and Blackmun, who, considering the tone of their opinion, must have found Marshall’s arguments persuasive, nevertheless did not join his opinion, relying on stare decisis to justify their decision. "If a statute is to be amended after it has been authoritatively construed by this Court, that task should almost always be performed by Congress. Title VI must therefore mean what this Court has said it means, regardless of what some of us may have thought it meant before this Court spoke." Thus, they concluded that Bakke must govern, and "proof of invidious purpose is a necessary component of a valid Title VI claim." Apparently unwilling to eviscerate the statute completely, however, they went on to hold that the regulations implementing the statute are valid, even though they incorporate an effect test: "By prohibiting grant recipients from adopting procedures that deny program benefits to members of any racial group, the administrative agencies have acted in a reasonable manner to further the purposes of Title VI." They, with White and Marshall, thus generated the interpretation of Title VI that governs today and that would control any challenge to Boalt’s admission procedures.

The remaining justices disagreed, but their reasoning is less than convincing. In the face of Justice Marshall’s lengthy and detailed summary of the development of the administrative regulations implementing Title VI, in which he explicitly points out that the regulations predate even Lau by a decade, Powell wrote, "I know of no precedent whatever for asserting that [the deference earlier opinions had required the Court to give reasonable administrative interpretations] is proper after this Court has issued a definitive--and contrary--construction of its own." One cannot win a debate in which some participants ignore obvious and known facts to effectuate their own purposes.

Regardless of how the Court reached its interpretation of Title VI and its implementing regulations, a majority had agreed that the regulations validly prohibit recipients of federal funds from using policies that have disparate racial impact; it had, to that point, however, provided little guidance about how to structure a disparate impact case under Title VI. Justice White provided some hint in his summary of the case in Guardians. The plaintiffs in the case had initially brought claims under both Title VI and Title VII. The trial court held that they had proven a case under the latter, so it did not reach the former. The appellate court, however, found that an intervening decision by the Supreme Court had made part of the ruling untenable. On remand, the trial court agreed but then reached Title VI, concluding that proof of discriminatory effect was enough to establish a violation. It apparently never reconsidered the finding of disparate impact it had made under Title VII. And since the Court never disputed the findings of disparate impact, it implicitly accepted that a finding of disparate impact under Title VII principles would also make out a disparate impact case under Title VI.

The Court had been somewhat more explicit a few years earlier in Board of Education v. Harris, but that case involved legislation analogous to Title VI and not the Title itself. In Harris, the Court confronted a denial by the then-Department of Health Education, and Welfare for funds under the 1972 Emergency School Aid Act (ESAA); the Department based the denial on statistics it had developed during a compliance investigation under Title VI, showing a disproportionately high assignment of minority teachers to schools with a majority of students of color, as well as a disproportionately low assignment of teachers of color to schools with a minority of students of color. The Court first determined that the ESAA’s ineligibility provisions encompassed school policies with racially discriminatory effects. It then implicitly approved a Title VII-type framework in which the Department could make out a prima facie case of disparate impact using statistics, and though that showing was rebuttable, the "burden is on the party against whom the statistical case has been made." In support of that proposition, the Court cited two Title VII cases and said that that "burden could be carried by proof of ‘educational necessity,’ analogous to the ‘business necessity’ justification applied under Title VII . . . ." The decision had Title VI implications because one of the eligibility criteria under the ESAA was implementation of a desegregation plan approved by HEW as adequate under the Title. And though the Court did not reach the issue of whether the ESAA and Title VI were coextensive, since a violation of the latter could result in ineligibility under the former, the Title VII-type framework implied as appropriate under the ESAA would also be appropriate for the resolution of a disparate impact claim under Title VI.

The federal circuits and district courts have dealt much more with Title VI cases, allowing private plaintiffs to proceed under the regulations, and making explicit what the Supreme Court had only implied by importing Title VII’s burdens and defenses for disparate impact claims into the Title VI setting. Larry P. v. Riles is the Ninth Circuit’s primary statement on the outline of a Title VI disparate impact claim, and since any challenge to Boalt’s admissions practices would be heard there, a brief review of the court’s treatment of the issue seems worthwhile. The case dealt with the state’s use of an IQ test for the placement of students in special classes for the mentally retarded and the disproportionate placement of African-American students in such programs as a result of the practice. The court held, analogizing specifically to Title VII cases, that a plaintiff establishes a prima facie case by showing that the tests have a discriminatory impact on a protected group. The burden then shifts to the defendant to demonstrate that the requirement that caused the disproportionate impact was required by educational necessity. However, that is not even close to a full description of disparate impact as developed under Title VII; the courts on all levels seem to have had little opportunity to consider disparate impact claims under Title VI, so the jurisprudence remains underdeveloped, saying that Title VII principles apply but only occasionally discussing their adaptation in any amount of detail. Therefore, Part III sketches an outline of the issues, burdens, and defenses involved in Title VII disparate impact cases, particularly in light of the changes worked by the Civil Rights Act of 1991.

B. The Future of Disparate Impact under Title VI.

Perhaps the most important--and most unanswerable--question surrounding a disparate impact case against Boalt for its 1997 admissions process is whether the Court will continue to hold that one can even claim disparate impact under Title VI’s regulations. Considering the litany of decisions that have come out of the current Court that are hostile to affirmative action, not to mention claims of racial discrimination generally, the survival of disparate impact claims under those regulations is very much in doubt. When the Court decided Guardians and Alexander v. Choate, its membership included Justices Powell, Rehnquist, O’Connor, Marshall, Stevens, Brennan, Blackmun, White, and Chief Justice Burger. Of them, only Rehnquist, O’Connor, and Stevens remain on the bench. In Guardians, Rehnquist and O’Connor both opined that Title VI prohibits only intentional discrimination. Stevens voted that Title VI’s implementing regulations are valid, and therefore, that they permissibly prohibit facially neutral policies that have discriminatory effects. In any challenge to Boalt’s 1997 admission process, Scalia and Thomas can almost certainly be expected to vote with O’Connor and Rehnquist, despite stare decisis, on the basis of textualism alone, if not on the basis of their personal politics. Of the remaining current Justices, Kennedy, Breyer, Souter, and Ginsburg, Kennedy at least seems likely to vote with the others to overturn the Guardians Court’s interpretation of Title VI and its regulations. Thus, it certainly seems likely that the legal basis for this paper has been eliminated by changes on the Court. It nevertheless proceeds on the assumption that Guardians will remain good law.

The ‘91 CRA adds an interesting wrinkle to any rumination on the survival of disparate impact under Title VI. Recall that Brennan and those who joined with him in Bakke probably agreed with the others that Title VI is no broader than the Constitution, and therefore prohibits only intentional discrimination, because they believed that any other holding would permit White applicants to bring disparate impact claims against schools using affirmative action plans. Setting aside that under a Title VII framework, such plaintiffs could probably not even make out a prima facie case, § 116 of the Civil Rights Act of 1991 provides that none of the amendments to Title VII shall affect affirmative action that is in accordance with the law. Since a Title VI disparate impact case is supposed to follow the Title VII framework, one could infer from that part of the ‘91 CRA that as long as an affirmative action program is otherwise in compliance with federal law, it is not subject to a disparate impact attack on the basis of Title VI. That means that on a practical level, those who would, like Brennan, Marshall, Powell, and others, vote to restrict Title VI’s reach to intentional discrimination as a means of protecting affirmative action programs no longer have to take that position, as the general disparate impact framework immunizes otherwise valid affirmative action programs.

Since the federal courts now consistently use Title VII’s framework of burdens and defenses for claims of disparate impact in Title VI cases, an overview of that framework is necessary to understand how the courts and the OCR would evaluate the challenge to Boalt’s admission processes. The most important recent development in the operation of disparate impact analysis under Title VII is the 1991 Civil Rights Act. The ‘91 Act overturned certain portions of Supreme Court jurisprudence in the area, reaffirmed some, and brought others into relief. At its most basic, the form of a disparate impact claim today is essentially as it was when the Court first dealt with the issue in Griggs. The plaintiff must first make out a prima facie case of discrimination by showing that the employment practice or policy challenged "selects applicants for hire or promotion in a racial pattern significantly different from that of the pool of applicants." If the plaintiff sustains that burden, the defendant must then show that the policy or practice is job related and consistent with business necessity. Even if the employer makes that showing, the plaintiff can still prevail by showing that another practice, policy, selection device, etc. would satisfy the employer’s legitimate business concerns but would not have as much or any discriminatory effect. The model is simple; the devil, of course, is in the details, as well as in converting the principles to the setting of Title VI.

The heart of plaintiff’s prima facie case is the identification of a selection device or employment policy or practice that has the effect of discriminating against a protected group and the provision of proof that that device, policy, or practice actually causes the discrimination. The central prohibition of Title VII in § 703(a) states "It shall be an unlawful employment practice for an employer--

Obviously, that prohibition does not itself explicitly refer to discriminatory effects. Thus, the development of disparate impact theory resided in the courts until the ‘91 Act. In that legislation, Congress reacted to developments in the area, particularly to Wards Cove Packing Co., Inc. v. Atonio, by enacting the following language:

Although the codification in the ‘91 Act was helpful in some ways, it provided no definition of causation, leaving intact the jurisprudence of the federal courts on the issue. The following section addresses that issue. Those that follow discuss the particularity requirement codified in §§ 703(k)(1)(A)(i) and (B)(i) and introduced in Wards Cove, as well as the problems presented by the various kinds of criteria employers use in making decisions.

The first and most difficult step in a disparate impact case is proving that a specific employment practice had a significant adverse impact in the selection of the plaintiff’s protected group. The Court has characterized a disparate impact as both a "significantly different" selection rate and a "substantially disproportionate" disqualification rate. Plaintiffs normally proceed by introducing statistics that they believe show such a disparity. They may introduce such statistics as selection rates, pass/fail comparisons, population/workforce comparisons, and regression analyses, among others.

The Equal Employment Opportunity Commission (EEOC) and other enforcement agencies have developed their own test for when a when disparity is statistically significant in the Uniform Guidelines on Employee Selection Procedures. The so-called "80 percent rule" states that

Courts have criticized the 80 percent rule as being not as probative of statistical significance as other techniques and as prone to mislead when sample sizes are small.

Two of the statistical techniques to which courts have given greater approval are the "Z score" test and the "chi-square" test. The formula for the former is the following:

In the equation, %PC is the percentage of members of plaintiff’s protected class in the relevant labor market or selection pool. Conversely, %NPC is the percentage of others in the relevant labor market or selection pool. To calculate the value of ExpectedSelections_pc, multiply the total number of selections made (TotalSelections) by the percentage of the protected class members in the relevant labor market or selection pool. If the resulting Z score is less than 1.96, a court could conclude that there is not enough evidence to believe the disproportion is discriminatory because the relationship is not statistically significant at the 0.05 level. In other words, if the Z score is less than 1.96, there is a greater than 5 percent chance that the disproportion occurred randomly.

The chi square test also compares differences in proportions, though somewhat differently. The simplest version compares just two classes, though one can construct more complex models. To perform a two-class analysis, one begins by setting up a four-cell table as follows:

The variables a, b, c, and d represent the number of people in each group. The model assumes that if class status is unrelated to performance on the test, the proportion of each class selected and not selected will be the same, which expressed mathematically, would be: . The chi-square formula is this:

If X² is greater than 3.84, then at the 95 percent confidence level, one can reject the theory that class status had no statistically significant role in the outcome; if X² is greater than 6.635, then at the 99 percent confidence level, a court can conclude that plaintiff has made out a prima facie case because class status played a statistically significant role in the outcome. But in using the statistical techniques, plaintiffs still have to identify the proper numbers to plug into the formulae.

As part of the ‘91 Act’s failure to discuss the methods of proof of causation, it also does not identify the appropriate statistical samples to use in making out a prima facie case. Thus, one has to look to the cases that preceded the Act. In its early disparate impact cases, the Court focused on the comparison between the actual applicant pool and those selected. In making out a prima facie case, the plaintiff must show that "the tests in question select applicants for hire or promotion in a racial pattern significantly different from that of the pool of applicants." In later cases, however, the Court began to express concern about the nature of the applicant pool, especially its qualifications. For example, in Watson v. Fort Worth Bank and Trust, in making clear that an employer could undermine a prima facie case by attacking weaknesses in plaintiff’s statistics, it stated that "statistics based on an applicant pool containing individuals lacking minimal qualifications for the job would be of little probative value."

In Wards Cove, the Court expanded on that aside in Watson, turning it into a more substantive alteration of the requirements of a prima facie showing in a disparate impact case.

Since plaintiffs in disparate impact cases usually challenge the qualifications about which the Court is so concerned that applicants meet, the decision is usually understood by emphasizing the "otherwise" in the Court’s language in Wards Cove, implying a permissible focus only on applicants who satisfy qualifications other than those challenged. Such a focus would be consistent with the particularity requirement that the Court imposed in Wards Cove and that Congress seemingly codified in the ‘91 Act, since plaintiffs could dispute the effects of one qualification but still have to base the statistical sample on those who meet the others (if there are any). However, so parsing the applicant pool may be a prohibitively difficult task for plaintiffs. Additionally, according to the Wards Cove Court, before plaintiffs may use that alternative showing, they first must be able to demonstrate that ascertaining labor market figures would be difficult or impossible, and the Court has provided no guidance in how plaintiffs may make such a showing. Moreover the ‘91 Act admits that there may be situations in which identifying the offending qualification is impossible, permitting plaintiff to deal with the employer’s decisionmaking process as a whole, and making unnecessary a focus on qualifications at all in the statistical samples.

As touched on briefly in the preceding section, both the Court in Wards Cove and the Congress in the ‘91 Act required plaintiffs in disparate impact cases to identify the particular business practice that had the discriminatory effect. In the former, the Court stated that "a plaintiff must demonstrate that it is the application of a specific or particular employment practice that has created the disparate impact under attack." In the latter, Congress required plaintiffs alleging disparate impact to demonstrate that an employer "uses a particular employment practice that causes a disparate impact." In clarifying that requirement, the Wards Cove Court held that the plaintiffs in the case could not use so-called "bottom line statistics" offensively. Bottom line statistics "look at the results of an employer’s selection process rather than at a single component . . . ." Thus, plaintiffs in Wards Cove could not make out a prima facie case by showing that two job categories had starkly different racial compositions; rather, they had to show that some particular employment practice operated to cause a statistically significant disproportion. That the job categories themselves were de facto segregated was not enough.

In an earlier case, Connecticut v. Teal, the Court similarly held that an employer could not use bottom line statistics defensively. There, the employer used a test in making promotion decisions, promoting those who passed and not promoting those who failed. The test therefore operated as a pass/fail barrier. It also had a statistically proven disparate impact, and the employer had not validated it. However, at the bottom line, the employer eventually promoted disproportionately more African-Americans than Whites. The Court nevertheless held that the employer could not use bottom line statistics indicating racial balance in its work force to insulate itself from claims that it used a test that had proven adverse impacts, at least when the employment practice is a test used as a pass/fail barrier.

The Court and Congress seem, however, to have carved out exceptions to both the offensive and defensive use of bottom line statistics. The dissenters in Teal, who included Rehnquist and O’Connor, argued that employers could, within the reasoning of the majority in the case, still defend with bottom line statistics a multicomponent decisionmaking process that includes consideration of a scored test, if the test does not constitute a pass/fail barrier. Presumably, their compatriots Scalia, Thomas, and Kennedy might agree with Rehnquist and O’Connor, making theirs the position of a majority of the Court. And though those five might attempt to find a way to avoid the result, consistency would seem to require allowing plaintiffs to use bottom line statistics offensively when employers use such a decisionmaking process, and it results in a disproportionately constituted work force.

Congress also carved out an exception for the offensive use of bottom line statistics in the ‘91 Act. Section 703(k)(1)(B)(i) states

Apparently attempting to give meaning to that provision of the statute, the legislative history provides the following guidance:

Federal courts have added little meaningful construction to that section, so what makes the elements of a decisionmaking process acceptably difficult to differentiate for the purposes of the ‘91 Act remains unclear. Two cases do provide some guidance in the area. In Caron v. Scott Paper Co., plaintiffs challenged their employer’s method of deciding whom to lay off.

The court decided to treat the combined subjective and objective criteria as a single subjective process. In dealing with the same process in a later case, the same court held that the components were not separable and that limiting "the disparate impact model to a discrete component of the selection process would completely exempt the situation where the adverse impact is caused by the interaction of two or more components of the process." Thus, the court was willing to allow the use of bottom-line statistics to make out a prima facie case in the proper circumstances.

Many of the principles in the preceding discussion are more well adapted to and many of the Court’s most important disparate impact cases deal with the use of so-called "objective criteria" in an employment decisionmaking process. Scored tests, diploma requirements, and height and weight minima are generally considered objective criteria. However, many employers use an interview, essays, or the general impressions of supervisors in their decisionmaking processes and base their decisions on their subjective impressions of applicants. The Supreme Court has said that such subjective criteria can cause adverse impacts cognizable as actionable discrimination under Title VII. In so holding, it added that "[h]owever one might distinguish ‘subjective’ from ‘objective’ criteria, it is apparent that selection systems that combine both types would generally have to be considered subjective in nature."

In Wards Cove, O’Connor and three other justices believed that the extension of disparate impact theory to the use of subjective employment criteria could encourage employers to adopt quotas or preferences to counterbalance the discriminatory effects of such practices. To "safeguard" against that possibility, the plurality attempted to alter the "evidentiary standards." First, they redefined "business necessity" as "legitimate business reasons." Second, they made the employer’s burden only one of producing evidence of the existence of legitimate business purposes for the use of the criteria under attack. Third, they rejected any requirement that employers validate subjective criteria to satisfy their burden of production. The majority in Wards Cove essentially adopted all of those principles, and the ‘91 CRA overruled that decision, eliminating the necessity but allowing the possibility of distinguishing between subjective and objective criteria for the purpose of making out a prima facie case. But even if all of the preceding and following principles of a disparate impact case apply to an attack on the use of subjective criteria, some questions remain.

Among the questions the courts and Congress have yet to answer are how a plaintiff must establish the disparate impact of subjective criteria, what the proper groups for comparison are, and whether the focus should be on how the employer applies the criteria or on the impact of applying those criteria at all. Perhaps the only answer currently available to such questions is to conform as closely as possible to disparate impact principles as articulated, to the extent that such articulations are consistent or capable of harmonization; and that is the approach in Part IV to adapting Title VII’s disparate impact principles to a Title VI case against Boalt.

Employers can defend themselves against claims of using employment practices that have discriminatory effects in two ways. In cases in which plaintiffs use statistics, employers can attack the sufficiency of those statistics, thereby preventing plaintiffs from even making out a prima facie case. Otherwise, employers must demonstrate the business necessity and job relatedness of the practice under attack. A common means of making that demonstration is through statistical validation. Of course, as in other areas of Title VII jurisprudence, the meanings of many of those terms and how to make those demonstrations are less than clear.

An employer can prevent a plaintiff from even making out a prima facie case by challenging the propriety of plaintiff’s statistics (when plaintiff attempts to prove disparate impact through the use of statistics). The most successful criticism of a plaintiff’s statistics is that they are based on too small a sample size; if justified, that criticism minimizes both the statistical and practical significance of a plaintiff’s statistics. Additionally, a defendant can rebut any inference of discrimination by providing the court more sophisticated statistics that show that the apparent disproportions are not statistically significant.

If a plaintiff can establish a prima facie case of disparate impact, Title VII now requires the defendant employer "to demonstrate that the challenged practice is job related for the position in question and consistent with business necessity." By defining the term "demonstrate" to mean meeting the burdens of production and persuasion, the ‘91 CRA overturned that portion of Wards Cove that said the employers burden was only one of production. Additionally, as the O’Connor plurality in Watson had characterized the employer burden of producing a "legitimate business reason" for its challenged practice, the Wards Cove Court characterized the burden as requiring the production of a "business justification." The ‘91 CRA amended Title VII by codifying the language of cases before Watson and Wards Cove; it also stated in its purposes that it was intended to restore the pre-Wards Cove meaning of those terms and prohibited courts from examining any legislative history but a specially prepared interpretive memorandum.

Besides being an obviously unusual lawmaking technique, Congress’ approach to overruling Wards Cove’s characterization of the burden of an employer in a disparate impact case is not as helpful as it could be, as the Court’s decisions before Wards Cove, which do include Watson, do not themselves provide consistent construction of the terms business necessity and job related. So, one has to scan the cases themselves. In Wards Cove, in getting to its statement that an employer need only produce a business justification, the Court differentiated that formulation by stating that "a mere insubstantial justification" would not suffice, but at the same time, "there is no requirement that the challenged practice be ‘essential’ or ‘indispensable’ to the employer’s practice." The Court thereby presented three possible interpretations of the employer’s burden: provide any justification for the challenged practice, provide a business justification for the practice, or prove that the practice is essential to the business. Since the ‘91 Act rejects the first two formulations, one could infer from Wards Cove that "consistent with business necessity" means that the employer must prove that the challenged requirement(s) is essential to the functioning of the business. However, such a reading might not be consistent with other pre-Wards Cove cases, nor does it give meaning to the full statutory language, as the employment requirement must also be "job related for the position in question."

In Watson, the Court nowhere mentions business necessity, stating only that a practice is job related if the employer can produce "legitimate business reasons" for it. As extremely tangential support for that position, it cites to New York Transit Authority v. Beazer and Washington v. Davis. In the former, the Transit Authority (TA) prohibited from employment drug users, including those on methadone maintenance. Plaintiffs complained that the policy had a disproportionate impact on Black and Hispanic employees. The Court held that the TA’s goals of safety and efficiency were legitimate and permitted the exclusion from employment of any form of drug user in safety-sensitive positions. It added that the policy was job-related for all other (nonsafety-sensitive) positions because the exclusion "significantly served" the TA’s goals of safety and efficiency. Thus, the Court effectively defined job-relatedness as significantly serving a legitimate business goal. That cannot, however, provide a basis for using Watson’s as the definitive pre-Wards Cove reading of the terms "business necessity" and "job related," as neither Watson nor Beazer give any construction to "business necessity;" both deal only with whether the challenged employment practice is job related. Additionally, only a plurality joined in the part of the Watson decision that says "job related" means producing a legitimate business reason.

One might find more helpful, then, the Court’s decisions in Griggs, Albermarle Paper, and Dothard v. Rawlinson. In Griggs, the first Title VII disparate impact case, the Court held that the "touchstone" of whether an employment practice’s discriminatory effects are illegal "is business necessity. If an employment practice which operates to exclude Negroes cannot be shown to be related to job performance, the practice is prohibited." It then went on to add that the employer must show that the requirement bears "a demonstrable relationship to successful performance of the jobs for which it was used." The Court pointed to two things that undermined Duke Power’s ability to defend its high school diploma requirement and IQ test: the employer had adopted both "without meaningful study of their relationship to job-performance ability," and the evidence available showed that employees who had not taken the test or earned a high school diploma continued to perform adequately and to progress within the organization, implying that neither requirement was necessary to improve the quality of work (i.e. perform the business’ essential functions). It concluded by saying, "What Congress has forbidden is giving these devices and mechanisms controlling force unless they are demonstrably a reasonable measure of job performance. . . . What Congress has commanded is that any tests used must measure the person for the job and not the person in the abstract."

The Court more fully addressed the employer’s burden in Albermarle, in which plaintiffs challenged the use of a pair of standardized tests, concluding that the defendant had failed to meet the burden for several reasons. First, the employer had not validated the tests for the jobs at issue. Second, the paper company had made no effort to analyze the jobs in terms of the specific skills they might require. Third, the validation study that had been done had produced inconsistent results. Fourth, the study compared test scores with subjective supervisorial evaluation. Though a permissible technique, there was "no way of knowing precisely what criteria of job performance the supervisors were considering, whether each of the supervisors was considering the same criteria or whether, indeed, any of the supervisors actually applied a focused and stable body of criteria of any kind." In other words, when validation is based on supervisor rankings, the employer must make sure that the supervisors are all using the same relevant criteria. The Court also noted the EEOC’s concern, as expressed in its Guidelines on test validation, with the problem of bias creeping into subjective evaluations. One can, thus, take from the case that proof of job-relatedness is factual and empirical and not at all an easy burden for an employer. As for business necessity, one can take from the case the same point the Court made in Griggs: if those who have advanced in the company and performed satisfactorily on the job do poorly on the test, it may not be necessary to the business.

Finally, in Dothard, the Court dealt with an employer’s burden of justifying the use of objective criteria that are not standardized tests. There, plaintiffs challenged the use of a height and weight requirement for employment as a prison guard. The employer said that height and weight were related to strength, an unspecified amount of which one would need to perform the job effectively; because of the nature of the position, safety was as important as efficiency. The Court said that the employer had provided no evidence correlating height and weight with the requisite amount of strength necessary to good job performance. It added that when safety is, in addition to efficiency, a necessity of the job, "a discriminatory employment practice must be shown to be necessary to safe and efficient job performance . . . ." One can conclude two things from the Court’s holdings. First, job-relatedness and business necessity are two different things. Not only did the defendant have to show that height and weight were related to strength (job related), it had to show that strength, and thus a practice that evaluated a candidate’s strength, was necessary to the safe and efficient performance of the job (business necessity). Second, both parts of defendant’s burden require empirical demonstration.

Section 703(h) of the Civil Rights act expressly authorizes employers to use "professionally developed tests," "provided that such test, its administration or action upon its results is not designed, intended or used to discriminate because of race, color, religion, sex, or national origin." The EEOC’s Guidelines on employment testing set out "detailed and relatively stringent technical standards for validity studies" for tests found to have significant disparate impacts on groups that Title VII protects. Although the Guidelines do not have the weight of promulgated regulations, the Court does give them deference and does cite to them when dealing with disparate impact cases that it believes require validation of employment practices. The message of the Guidelines and the cases is the same: tests with discriminatory effects "are impermissible unless shown, by professionally acceptable methods, to be ‘predictive of or significantly correlated with important elements of work behavior which comprise or are relevant to the job or jobs for which candidates are being evaluated.’"

There are three general methods of test validation: criterion related, content, and construct. An employer establishes criterion-related validity by showing that relatively successful performance on the test has a positive correlation with relative achievement on some measure of job performance. The three primary issues in establishing criterion validity are the selection of appropriate criteria; the selection, design, and reliability of each criterion’s measure; and the degree of correlativity necessary to establish validity. The criteria are the aspects of job performance the test is designed to predict; thus, an employer should choose skills necessary to the job. The most commonly used criterion measure is a subjective supervisory evaluation. Employers must make sure to examine the actual data to eliminate any distortions that might lead to an erroneous finding of validity. Although industrial psychologists consider a correlation coefficient of 0.20 sufficient to establish validity, some courts demand at least 0.30.

Establishing content validation focuses on whether a test directly measures the knowledge, skills, or abilities a job actually requires a candidate to perform. It does not compare on-the-job results to test results; the test should itself mirror the actions demanded on the job, so a good performance on the test correlates with ability to do the job. An examination is content valid if it "tests a representative sampling of specified job functions or the underlying knowledge or skills necessary to perform those functions." Five factors that the federal courts and the EEOC have identified as critical to good content validation are:

An employer establishes construct validation when the test correlates significantly with a trait necessary to the performance of the job, such as intelligence. Employers rarely use it, so it has not seen sufficient litigation to give it much legal meaning.

4. Training Program Validity.

An occasionally recurring question in the courts has been the requirements for validating employment tests used not for the purposes of hiring or promotion but for entrance into a training program. The Court addressed the issue in Washington v. Davis, in which plaintiffs attacked a test for admission to a police training program as effectively discriminating against African-Americans. Plaintiffs argued, and the Court of Appeals found, that even if success on the test was demonstrably positively correlated with success in the training program, the crucial correlation to prove the validity of the test was, rather, between success on the test and successful job performance. The Supreme Court disagreed.

At least one three-judge panel has followed the broad principle the Court enunciated in Washington v. Davis and held that an employer can validate a test used for entrance into a training program by proving a positive correlation between success on the test and performance in the program. Other district courts and circuits have read the decision more narrowly.

The Ninth Circuit, where any challenge to Boalt’s admission process would be heard, has held that "the skill, characteristic, or ability which the . . . [training program entrance] test purports to measure must be important to the requirements of the job. [An employer] may show the requisite relationship by demonstrating that the test correlates significantly with important elements of [training program] performance and that those elements of . . . performance are important to actual job performance." In reaching that holding, the court limited the decision in Washington v. Davis to its facts: the use of a verbal abilities test is valid to determine whether applicants to a training program possess the minimum skills necessary to understand and succeed in the academic portion of the training. The court read the case in that way because of its fear that if "employers were permitted to validate selection devices without reference to job performance, then non-job-related selection devices could always be validated through the simple expedient of employing them at both the pre-training and training stage."

Turning to the facts of the case, the court found that the employer had demonstrated a significant, indeed unusually high, correlation coefficient between test scores and academic performance in the training program as measured by a written examination at the end of the program. The employer thereby satisfied the first part of its burden: proving a correlation between the entrance exam and the program. Plaintiffs attacked the sufficiency of that showing on the ground that the validation study was unrepresentative of the applicant pool because the pool included only those who passed the entrance exam; thus, the validation study did not indicate whether those who failed the test could nevertheless successfully complete the training program. In support of the argument, they cited EEOC Guidelines that required that a criterion-related validity study must include a sample of subjects representative of the typical applicant group. The court agreed that such a sample would normally have to include applicants who passed and failed the test. It nevertheless found defendant’s validation study sufficient only because of the unusually high correlation between the entrance exam and performance in the program (0.60); it inferred from those results that those who failed the exam would not have succeeded in the program. However, it remanded to the trial court because the employer had not proven that the training given was manifestly related to the knowledge and skills that employees must have to perform their jobs satisfactorily.

The Fifth Circuit reached essentially the same conclusion in Ensley Branch, NAACP v. Seibels. It too limited Washington v. Davis’ holding on training program validity to cases in which employers use tests to ascertain whether applicants have minimum skills necessary to complete job-relevant training; since the employer in the case was using the results of the entrance exam to rank applicants and admitting only the highest scorers, a study showing a correlation between success on the test and performance in the program would not suffice. Instead, the court held that use of a test for ranking purposes, rather than as a device to screen out candidates without minimum skills, "is justified only if there is evidence showing that those with higher test scores do better on the job than those with lower test scores." The employer in the case had produced no such evidence. Indeed, the district court had found that the test was not a valid predictor of ability to pass the training curriculum. And although the validation studies showed a positive correlation between entrance exam score and grades in the program, the employer could not use grades as the criterion for comparison, as grades did not predict job performance. Thus, under Title VI, to use a positive correlation between performance on an entrance exam and performance in a training program to validate the test itself, an employer must be able to show one of the following: that the employer uses the test only to screen out those applicants without the minimum skills necessary to complete and benefit from the training; that the content of the test, the program, and the job are all the same; that the skills gained in the program are necessary to perform the job; or that performance in the training program predicts performance on the job (if the employer attempts to validate the test by comparing scores on the test to grades in the program).

Having made out a prima facie case, and defendant having demonstrated the business necessity and job relatedness of an employment practice with discriminatory effects, plaintiffs may still prevail by demonstrating the existence of an alternative employment practice that serves the business’ necessities, with less adverse impact, that the employer refuses to adopt. That demonstration must be "in accordance" with the law as it existed prior to Wards Cove, but as in other areas, the pre-Wards Cove law on the issue is not entirely clear.

The Supreme Court has elucidated this part of a disparate impact case very little. Its statement on the issue in Albermarle Paper remains its most significant: "If an employer does then meet the burden of proving that its tests are ‘job related,’ it remains open to the complaining party to show that other tests or selection devices would also serve the employer’s legitimate interest in ‘efficient and trustworthy workmanship.’"

Having laid the legal groundwork to explore an answer, the question nevertheless remains whether the Office of Civil Rights has good reason to believe that the admissions process Boalt used in 1997 produced an adverse racial impact that violated Title VI’s regulations, assuming, of course, that the Court continues to hold that the regulations prohibiting the use of effectively discriminatory polices are a permissible reading of Title VI. In exploring that question, this part first summarizes the admission procedure at issue. It then examines whether the process as a whole produced statistically significant racially disproportionate results. Assuming the ability of a plaintiff to make out or the OCR to find a prima facie case, this part will then examine the school’s ability to prove the "educational necessity" and "educational relatedness" of the identified criteria and the process as a whole. The discussion will include the validity of the LSAT, particularly its validity as a tool for selecting entrants to a training program. Finally, this part examines the existence of other selection methods that would serve the school’s legitimate interests but have less disparate impact on racial minority groups.

In response to SP-1, the faculty of Boalt Hall made three general changes to the school’s admissions process. First, it abandoned the use of enrollment goals for underrepresented racial groups. Second, whereas the school had previously given an applicant’s Index score -- the combined total of an applicant’s LSAT score and UGPA, weighted according to undergraduate institution -- "the greatest weight," the faculty decided to accord that score "substantial weight." The admissions policy also gives "substantial consideration" to such things as "letters of recommendation, graduate training, special academic distinctions and honors, difficulty of the academic program successfully completed, work experience, and significant achievement in nonacademic activities or public service." Third, purportedly "to get a sense of" the applicant "as a person and as a potential student and graduate of Boalt Hall," the faculty expanded the permissible length of the required personal statement. More generally, in evaluating an applicant, the Admissions Committee is supposed to consider "disadvantages that adversely affected his or her performance" and "to select students who will attain the highest standards of professional excellence and integrity, and who will bring vision, creativity, and commitment to their professional endeavors." Essentially, then, the school has invested tremendous discretion in both the Admissions Office and Committee in making admissions decisions, and that discretion may have pushed in one of two directions. Either it may have encouraged improper and unexpected reliance on the numerical data, particularly the LSAT. Or it left decisionmakers free to make such unprincipled decisions that one cannot distinguish what factors led to any individual decision to admit or reject an applicant (other, perhaps, than those almost automatically admitted because of their Index scores).

When an application arrived at Boalt during the 1997 admissions cycle, the school’s first step was to assign it an Index score. In computing index scores, whereas Boalt’s peer schools typically weight an applicant’s LSAT score at about 70%, Boalt gave the LSAT and UGPA equal weight in 1997. However, that year it also adjusted applicants’ UGPAs before computing the Index score. The school adjusted the UGPAs according to the age of the degree (to compensate for grade inflation) and the rank of the undergraduate school (to account for "quality of the student body and the grading patterns at the school"). Boalt used the LSAT to rank schools by averaging the LSAT percentile for all test-takers from the school and the LSAT percentile ranking of students at the school who have a UGPA of at least 3.6. The school then assigned all applications to one of five strata--A, B, C, D, and deny--based on the applicant’s Index score. Moving from stratum A to D, Index scores declined, as did the ratio of admitted to total students per stratum. Indeed, the Admissions Office admitted the majority of applicants in stratum A without further consideration and accorded the same treatment to some students in stratum B.

The Admissions Director read all applications of those whom the Admissions Office had not administratively admitted or rejected and recommended which should move on to the "Committee read" stage of the process. Although the faculty’s general pronouncements on diversity controlled the Director’s decisions about which applications to send to the Committee, since as indicated, the faculty did not provide detailed instructions about how to evaluate different types of claims of disadvantage and types of diversity, the Director exercised a great deal of discretion in deciding which files to send to Committee. Committee members received some instructions explaining the Index number and providing guidance on evaluating grades by considering such factors as exceptionally high UGPA, grade trends, discrepancies with in a transcript, graduate school grades, and disadvantage. If an applicant experienced such performance-affecting disadvantage as language barriers, low socioeconomic status (SES), or level of parental education, and an application shows the effects of the disadvantage lessening markedly over time, a Committee member could expect improved performance in law school. However, applications did not identify the applicant’s race, and the Committee received no instructions about how to evaluate claims of disadvantage attributable to an applicant’s race. Also faculty-student teams within the Committee disagreed internally and among themselves about how to substantiate claims of racial disadvantage and how to make decisions generally.

As touched on above, for the purposes of analyzing Boalt’s 1997 admissions process for disparate impact, one might conclude that the process the faculty and administration adopted pushed those who made admissions decisions in one of two directions. Either, as the students who prepared the report called New Directions in Diversity argue, the unstructured discretion led those making decisions to rely inappropriately on the hard numbers. Or considering the multiple additional factors reviewers were supposed to take into account and the lack of instruction about how to take them into account, one could conclude that , with the possible exception of students in the A and B strata admitted almost solely on the basis of their numbers, the process was so discretionary that one cannot be sure what factors led to any individual decision to admit or deny, so one should and may evaluate the process as a whole.

The authors of New Directions give a number of reasons why the unprincipled discretion granted by the 1997 admissions process might push decisionmakers toward overreliance on the numbers. First, and perhaps most obviously, the numbers provide a convenient way to make decisions. Especially when one considers the level of turmoil surrounding admissions decisions after SP-1 and Proposition 209, as well as the highly pitched (and poorly rationalized) rhetoric about admitting only the "most qualified" applicants, one might reasonably expect decisionmakers to do the seemingly least controversial (i.e. convenient) thing: rely almost exclusively on the purportedly objective factors. Second, the school told those making admissions decisions that one of the goals of the admissions process was to admit an experientially diverse class with a proven history of overcoming disadvantage, but it did not provide them clear criteria or procedures for identifying and evaluating those factors or for weighing them against other factors, such as the Index score. Without structure for discerning, measuring, or weighing those subjective factors, evaluators are more likely to place greater reliance on data they perceive as objective. Third, the New Directions authors point to the lack of a standardized policy for evaluating claims of disadvantage caused by race as pushing decisionmakers toward reliance on numbers. Although the application encouraged students to discuss disadvantages that may have inhibited past performance, those who wrote about experiences with racial prejudice may have been subjected to higher scrutiny. For instance, since some readers required substantiation of such claims, if an essay lacked support for the claims of racial disadvantage, those readers may have disregarded the essay altogether, leaving them with even less material to weigh against an applicant’s Index score. Finally, since the school established no numerical goals for the admission of diverse applicants, indeed it seemingly did not differentiate among different types of diversity, readers lacked a firm mandate for admitting such students, likely causing them to place greater emphasis on Index scores to differentiate.

However, the school did obligate application readers to include subjective criteria in their considerations. So, if the Admissions Director, the committees, and the committee members did not rely solely on Index scores and did take the other, subjective criteria into account, then in doing a disparate impact analysis, one must consider the effects of those subjective criteria on the outcome of the decisionmaking process. I certainly would not argue that Boalt or other schools should not consider such factors as performance-affecting disadvantage, socio-economic status, character, letters of recommendation, work and other nonacademic experience, etc., since I believe such factors probably more important than grades and LSAT scores in discerning which applicants will contribute to society and the profession, but to the extent the decisionmakers considered such factors, those factors impacted the outcome of the admission process and are, thus, subject to scrutiny as part of a disparate impact analysis. As discussed in Part III.A.3, the areas of uncertainty are even larger in a case of subjective-criteria disparate impact than in cases involving claims that a particular objective criteria caused the discriminatory effects. But if one could discern the impact of the consideration of some of those criteria and show that the impact was statistically significant, the fact that the school did not even attempt to validate their use would weigh heavily in a finding that Boalt’s admission process had impermissible discriminatory effects.

For the purpose of evaluating possible disparate impact in Boalt’s 1997 admissions process, one might not need or want to demonstrate either that a particular objective factor(s), the LSAT and/or GPA, or one of the subjective criteria caused the statistically significant impact. Thus, those of the New Directions authors are not the only or the necessary conclusions to draw from Boalt’s 1997 admissions process. One could, just as reasonably, conclude that the unprincipled discretion delegated to application readers makes it impossible to discern a single factor of the process that may have caused the disparate impact. Since it may be unclear how the Admissions Director decided which applications to send to committee or how the committees as a whole or their individual members decided which applicants to admit, identifying the impact-causing factor may be impossible, meaning one may and should examine the process as a whole. Thus, in doing statistical analysis, one may use a "black box" or "bottom line" approach in determining whether a statistically significant disparate racial impact resulted from the process.

The numbers that resulted from the 1997 admissions process at Boalt are, put nicely, not pretty. Boalt received 4171 applications from people seeking to enter the class that will graduate in 2000. Of those, 499 came from Asians, 316 from Hispanics, 256 from African-Americans, 28 from Native Americans, and an overwhelming 3072 from White applicants. Boalt made 860 offers of admission: 122 to Asians, 46 to Latinos, 15 to African-Americans, 2 to Native Americans, and 675 to White applicants. Of those offered admission, 268 applicants opted to enroll; 215 of them were White, 38 were Asian, 14 were Latino, one was African American, and none were Native American. That means that 73.6% of applicants and 78.5% of admittees were White, 11.96% of applicants and 14.2% of admittees were Asian, 7.5% of applicants and 5.3% of admittees were Hispanic, 6.1% of applicants and 1.7% of admittees were Black, and 0.67% of applicants and 0.2% of admittees were Native American.

A decent prima facie case exists against Boalt’s 1997 admissions process. Recalling from Part III.A, a prima facie case under Title VII requires plaintiffs to show as an initial matter that a facially neutral policy had a statistically significant disparate impact; the law primarily contemplates plaintiffs making that showing for a particular decisionmaking criterion, but it also allows, in some circumstances, for plaintiffs to focus on the bottom-line results of the process as a whole. Here, the bottom-line data for Boalt’s admissions process show statistically significant disparate outcomes in offers made to African-Americans and Latinos, and such an analysis is appropriate and permissible in this matter because the elements of the decisionmaking process are not capable of separation. However, even if one focuses on the LSAT as the particular criterion that caused the impact, one can make out a prima facie case in several ways. As explained in greater detail in Part IV.B.2, a statistical analysis of the impact of the LSAT on different racial groups in the admissions process was not possible, but other means of making out a prima facie case based on Boalt’s use of the LSAT exist. OCR’s draft guidelines on Fairness in Testing allow it to find a prima facie case without the complex statistical analysis that would be required of the test itself when use of a test contributes to a statistically significant bottom-line disparity. Also, Boalt made several misuses of the LSAT in its admissions process that constitute per se violations under those guidelines; such per se violations allow OCR to find impermissible discrimination regardless of the validity or educational necessity of the test, or whether alternative practices exist.

Use of bottom-line statistics is appropriate in analyzing Boalt’s admission process for disparate impact, and such statistics show that the process had impermissible discriminatory effects against African-Americans and Latinos. As discussed in Part III.A, the ‘91 Civil Rights Act and, seemingly, the Court have carved out exceptions to the general rule of particularity for the offensive use of bottom-line statistics. In Connecticut v. Teal, though the majority held that an employer may not use bottom- line statistics indicating racial balance in its work force to insulate itself from claims that it used a test that had proven adverse impacts, at least when the employment practice is a test used as a pass/fail barrier, the dissenters, who included Rehnquist and O’Connor, argued that employers could, within the reasoning of the majority, still defend with bottom line statistics a multicomponent decisionmaking process that includes consideration of a scored test, if the test does not constitute a pass/fail barrier. If Scalia, Thomas, and Kennedy, agreed, theirs would be the position of a majority of the Court. Consistency would then seem to require allowing plaintiffs to use bottom line statistics offensively when employers use such a decisionmaking process, and it results in a disproportionately constituted work force. Additionally, Congress also carved out an exception for the offensive use of bottom line statistics in the ‘91 Act for when "the complaining party can demonstrate that the elements of a respondent’s decision making process are not capable of separation for analysis." A decisionmaking process is so constituted when it "includes particular, functionally integrated practices which are components of the same criterion, standard, method of administration, or test."

The decisionmaking process Boalt instituted for its 1997 admissions cycle conforms to the models the Court and Congress have established as permitting bottom-line analysis. Like the one that the dissenters in Teal said would permit a bottom line defense, Boalt’s was a multicomponent process that included a scored test not used as a pass/fail barrier. Also, because of the unprincipled discretion granted to application readers, one could reasonably conclude that, with the possible exception of the administrative admits, the elements of the process are not capable of separation; in other words, it may be impossible to tell which criterion or criteria formed the basis for an individual decision to admit or deny. And all of the elements were a functionally integrated part of a single method of administration that attempted to determine which applicants were "best qualified." Finally, like the process at issue in Graffam v. Scott Paper Co., in which the court held that plaintiffs could use bottom line statistics offensively, Boalt’s admissions process combined objective and subjective criteria to form a single subjective process in which multiple criteria interacted to lead readers to their decisions, thereby causing the disparate impact. For those reasons, the admissions process is subject to a bottom-line attack.

The following are the Z-test results for Boalt’s 1997 admission process at the bottom line. A Z score of more than 1.96 is enough to conclude that discrimination occurred at the 95% confidence level. The Z-scores for African-American (5.39) and Latinos (2.45) are enough to make out a prima facie case of disparate impact at the bottom line for those groups because the results show that there is a less than 5% chance that the results for those groups are random.

The results of a chi square test of the bottom-line of Boalt’s 1997 admission process also show a likelihood of discrimination sufficient to make out a prima facie case. Using a two-by-two chi square test, one may conclude that a decisionmaking process resulted in impermissible discrimination at the 95% confidence level if X² is greater than 3.841 and at the 99% confidence level if X² is greater than 6.635. For African-American applicants, the result was 35.34; for Latinos, it was 7.28; for Native Americans, it was 2.35. Thus, a prima facie case of disparate impact against African-Americans and Latinos exists at the bottom line.

Assuming that a court agreed that it is appropriate to examine the bottom line results of Boalt’s admission process, rather than a particular selection criterion, either of the foregoing would be more than sufficient to make out a prima facie case against that process. However, it is possible to level an attack against Boalt’s use of the LSAT as the particular impact-causing criterion without performing the same kind of statistical analysis of the effects of its use on the racial composition of the group of applicants it admitted in 1997.

If this were a perfect world, or if, as the economists like to say, people operated with perfect information, this section would begin with a statistical analysis of the LSAT scores of those who applied to Boalt in 1997. But this is not a perfect world (if it were, I would not be writing this paper), and information flow is not perfect. Thus, this section does not contain such an analysis, mostly because it is simply beyond the statistical skills I could master in the course of preparing this paper. Also because, in the wake of Proposition 209 and SP-1, I am not sure that Boalt has done such a racialized categorization of LSAT scores, nor am I sure that it should be preparing such records under those laws. This part of the paper nevertheless contains several ways to make out a prima facie case against Boalt on the basis of its use of the LSAT. According to the Office for Civil Rights’ draft guidelines on Fairness in Testing, it may/should conclude that there was discrimination in violation of Title VI if use of a test caused or contributed to a bottom-line disparate impact. And if the test so contributed to a disparate impact and was either the sole or principal criterion used in decisionmaking, or the decisionmaking process used the test for purposes for which it was not designed, then the use constituted a per se violation of Title VI. Boalt’s use of the LSAT in 1997 falls within the circumstances under which OCR may find a prima facie case of impermissible discrimination.

The OCR guidelines described in the following section are subject to a number of attacks, most deriving from the deviation of those guidelines from the format for making out a disparate impact case that the Court has constructed for Title VII. Education is, however, different from employment, the Court has not given as much detailed attention to the nature of a disparate impact case under Title VI in the educational setting, and when and if it devotes such attention to the matter, it may find that the differences between education and employment justify some departures from the Title VII format. Thus, I use the OCR guidelines not because they conclusively make out either a prima facie case of disparate impact or a per se violation of the law, but because they provide a starting point for thinking about how disparate impact law in the educational setting should be constructed.

Focusing on the LSAT as the particular criterion that caused the disparate impact, as the New Directions authors suggest is appropriate, one could make out a prima facie case of disparate impact against Boalt’s use of the test in its 1997 admissions process under the OCR’s guidelines. Those guidelines provide that when a test is the sole or principal criterion in making educational decisions, investigators should determine whether the test caused or contributed to the statistical disparity in the denial of educational benefits or opportunities. The first step is to do a bottom-line analysis using a "Z test," as I did above. If that analysis indicates disparate impact at the bottom line, an investigator should examine the average (mean) scores for all relevant populations. If the mean scores are significantly different for the groups, and the difference is "in the same direction" as the disparity at the bottom line, the test is contributing to the disparity. The contribution of a test to a bottom-line disparity is enough to make out a prima facie case against the use of the test.

The bottom-line results of Boalt’s 1997 admissions process are provided above, so the first question is whether the average scores on the LSAT of the relevant populations are in the same direction as the bottom-line disparity. In 1990-91, the mean LSAT score for Native Americans was 30.27 (on the then-used 40-point scale), 33.22 for Asian Americans, 25 for African-Americans, 30.13 for Hispanics, 29.70 for Mexican-Americans, and 34.35 for Whites. Thus, the average scores on the relevant test are different for each racial groups in the same direction as the bottom-line results of the admission process, meaning that under OCR guidelines, Boalt’s use of the test violated Title VI.

Having established a prima facie case under OCR standards, however, OCR also instructs that if the use of a test caused or contributed to a disparate impact, OCR should conclude that there was discrimination in violation of Title VI either if "1) the test is being used as the sole or principal criterion for an educational decision and the test was clearly not designed to be so used[, or if] 2) the test is clearly not being used for the purposes for which it was designed . . . ." Thus, OCR defines such uses of a test contributing to a disparate impact as per se violations of Title VI; users of such tests do not receive the opportunity to show the educational necessity of that use of the test, and those who challenge such use do not have to identify less discriminatory alternatives. So, the next question is whether Boalt’s use of the LSAT also constituted a per se violation of Title VI.

First, use of the LSAT is a per se violation if Boalt used it as the sole or principal criterion in its decisionmaking process, and the LSAT was not designed to be so used. This particular line of inquiry has proceeded on the assumption that the LSAT was the principal or particular criterion that caused the impact, since as the New Directions authors suggest, the unprincipled delegation of discretion in decisionmaking that Boalt gave its admissions decisionmakers pushed them toward almost exclusive reliance on the LSAT. And the Law School Admissions Council explicitly informs law schools that they should not place principal reliance on the LSAT in determining the qualifications of applicants. Therefore, if consideration of the LSAT did overwhelm all other factors in the admission process, Boalt committed a per se violation of Title VI.

Second, Boalt committed a per se violation if its use of the LSAT, a test that contributed to a disparate impact, was for purposes for which it was not designed. As just discussed, using the LSAT as the principal criterion for decisionmaking would itself be enough to constitute a per se violation under that OCR standard. Boalt also used the LSAT improperly by employing cut-off scores in assigning applicants to one of the five strata discussed in Part IV.A.1; those assignments in turn affected whether an applicant would be admitted. In part because the LSAT has a standard error in measurement (SEM), the LSAC warns law schools against using cut-off scores and making decisions based on fine distinctions in scores. Boalt nevertheless did both in assigning applicants to the five admission strata, since applicants with scores below a certain point would not be placed in certain strata, and the stratum in which the Admissions Office placed applicants influenced outcomes. Most placed in the A stratum were administratively admitted, many in the B stratum were administratively admitted, and the stratum in which an applicant was placed was likely to influence the decisions of application readers. Yet the Admissions office assigned applicants to the strata on the basis of differences in LSAT scores of as little as one point. That was a misuse of the LSAT.

Additionally, Boalt misused the LSAT in the way it used the test to adjust applicants’ UGPAs. Boalt determined the quality of an undergraduate institution by averaging the LSAT percentile for all test-takers from the school and the LSAT percentile ranking of students at the school who have a UGPA of at least 3.6. Boalt then ranked schools on the basis of the resulting score and adjusted applicants’ grades based on the undergraduate institution they attended. That simply is not a use of the LSAT that LSAC contemplated or explicitly sanctioned as proper. And Boalt says it makes the adjustment to account for "both the quality of the student body and the grading patterns at the school," but that is not a use of the LSAT for which it has been validated. Thus, Boalt misused the LSAT in three ways, making possible the conclusion that Boalt’s 1997 admission process constituted a per se violation of Title VI under OCR’s guidelines; that conclusion precludes an educational necessity defense, which in turn makes unnecessary any demonstration of the existence of alternative practices that have less disparate impact but nevertheless serve the school’s legitimate interests.

Assuming the establishment of a prima facie case of impermissible discriminatory effects by Boalt’s 1997 admission process, the next step would be to require Boalt to justify both/either that process as a whole and/or its disparate-impact-causing components. Adapting the requirements of Title VII, as federal courts have done, Boalt would have to prove the "educational necessity" and "law school relatedness" of the process or impact-causing criterion. Focusing on the process as a whole, Boalt would have to prove that the characteristics it instructed the admission decisionmakers to seek are related to some set of skills or characteristics necessary to performance as a law student, and that the process is a valid predictor of performance as a law student. Focusing on the LSAT, Boalt would have to do more than simply show that the test is a valid predictor of first-year grades, it would have to show that first-year grades are the result of professors using the same set of criteria to judge performance, and that those criteria are actually necessary to perform well as a law student. Additionally, looking at law school as a training program, Boalt would have to do more than show that the LSAT has a statistically significant correlation with first year grades, it would have to show that those grades are, in turn, valid predictors of graduation from law school and of performance as an attorney. And more generally, the LSAT is subject to criticism both for the way in which it is constructed and for the limited skill set it tests.

Consistent with the rest of the paper, this part examines the defensibility of both the admission process as a whole and the LSAT individually. Boalt has not validated its overall admission process, and that process is not sufficiently educationally necessary or related to the requirements of legal education. The LSAT, though valid for predicting first-year grades, is not necessarily educationally necessary or related to the requirements of legal education.

Those federal courts that have addressed educational necessity have required defendants to prove that their practices have a manifest relationship to their purposes. Thus, in Georgia State Conference of Branches of NAACP v. State of Georgia, the Eleventh Circuit found that Georgia’s grouping of students by ability was educationally necessary because the practice had a manifest relationship to classroom education in that the technique had pedagogical acceptance and was meeting students educational needs. Similarly, though with a different outcome, a New York court held in Sharif v. New York State Educ. Dept. that use of the SAT in awarding merit scholarships to high school students was not educationally necessary, as the SAT purportedly measured ability to succeed in college and, thus, did not have a manifest relationship to measuring past success.

Neither court, however, seems to have given the idea of educational necessity detailed consideration, and from their use of "relationship" and "relatedness," they seem to have considered relatedness and necessity as synonymous. In developing its "manifest relationship" standard, the Eleventh Circuit cursorily examined the language in Griggs and Dothard to pull out the manifest relationship language, so it provides a good starting point. Recall, though, from Part III.B.2 that the Court in Griggs found that the tests and the high school diploma requirements at issue there did not bear a manifest relationship to the jobs at issue because the employer had adopted the requirements without consideration of their relationship to job performance (and thus with no justification for believing that they were so related), and because employees who had neither received a high school diploma nor been able to succeed on the tests had nevertheless performed their jobs adequately and been promoted. Additionally, in Albermarle, the Court held that though the tests at issue there had been validated, and the validation study had shown that test scores correlated positively with good supervisorial evaluations, it was unclear what criteria supervisors were using, whether they were all using the same criteria, and whether those criteria were actually relevant to job performance. And in Dothard, though the Court was willing to accept defendant’s assertion that strength was necessary to the job of prison guard, it required some proof that height and weight, the requirements at issue, were actually related to strength. So, defendants must identify just what quality or qualities are supposedly necessary to perform satisfactorily in the position in question and show that the criteria they examine in making decisions are related to those qualities.

Unfortunately, academics and courts have devoted essentially no time to the meaning of educational necessity and academic program relatedness. I cannot, in this paper, provide full understandings of the terms, but the following may at least provide a starting point for a dialog about whether the selection criteria that educational institutions use are necessary or related to the programs in the way that the law requires. Recall from Part III.B.2 that business necessity and job relatedness mean more than that the reason or business may justify the practices; those terms may, however, mean less than that the practices must be absolutely necessary or related to the employment in question in an exactingly precise way.

To translate the terms that the Court used in Griggs, Albermarle, and Dothard into the educational setting, one might first want to take at least a couple of steps back and try to understand some aspects of the nature of education, since it is not the same as employment. If nothing else differentiates them, education is not a boss giving an employee a job, it is a student, in some sense, hiring teachers to provide a service. Conceding that, one might i