Rudebusch v. Hughes

KLEINFELD, Circuit Judge,

concurring in part and dissenting in part:

I.

I concur in the majority’s conclusion that President Hughes, acting on behalf of the State of Arizona, denied equal protection of the law to whites by categorically excluding them on account of their race from pay raises given to minorities. Appellants’ brief puts at issue whether female professors were denied equal protection with respect to minority males based on gender. Though the arguments relating to race and sex are commingled, appellants’s brief puts at issue whether white males were denied equal protection with respect to females as well as minorities. The first and second issues presented for review address “non-minority males” and “the non-minority male Class Action Plaintiffs.” I understand the majority decision to imply that, if the majority so read the briefs, then we would also be in agreement that President Hughes denied equal protection of the law to males on the basis of sex and white females on the basis of race.

I also concur in the majority’s conclusion that pay raises limited to minorities and females would not deny equal protection of the law, if they were narrowly tailored to remedy past discrimination and bring their salaries into parity with comparable persons of previously favored classes. I agree that a university president could (and should) identify and remedy any past discrimination based on race or sex by bringing the victims’s salaries into line with what they would have been in the absence of past discrimination. Such “catch-up” raises would properly be limited to individ *526 uals in the classes that had been discriminated against.

But, as the majority correctly concludes, the raises here weren’t just for parity or a “catch-up.” The Chambers study, on which President Hughes relied, said that females made, on average, $751 less than males, and minorities made, on average, $87 less than whites. But President Hughes didn’t raise average female and minority salaries by $751 and $87 respectively. Instead he raised female and minority salaries by averages of approximately $2400 and $3000 respectively. 1 Thus, being female was worth an average of about $1,649 more than being white male as a university instructor (a $2,400 raise less a $751 catch-up yields about $1,649 excess value of females over males in an average case). Likewise, being a male member of an ethnic minority group was worth an average of approximately $2,913 more than being a white male. That’s not “parity” or “equity” or a “catchup.” Plain as day, it’s just paying minority males more than whites and females and females more than white males based on their race and sex.

Though I have put it more starkly, so far the majority opinion and I agree. President Hughes did indeed violate the constitutional rights of those disfavored on account of their sex and race. I respectfully dissent, however, from the majority’s holding that President Hughes is entitled to qualified immunity. The question on qualified immunity is “whether it would be clear to a reasonable officer that his conduct was unlawful in the situation he confronted.” 2 In my opinion, it would. Any competent university president would know that he can’t pay people more or less than others based on their sex and race. The scheme here was straightforward: minorities are gold, women are silver, white men are bronze. It’s been a long time in America since anyone thought the Constitution allowed governmental discrimination based on sex and race. The law has been clearly established on this point for many years. 3 President Hughes wasn’t entitled to pretend ignorance on this point because there was no case directly on point specifically on pay raises to “predicted” salaries based on regression analyses with the peculiar flaws of the Chambers study.

The majority justifies qualified immunity by saying “how tricky it is to measure whether sex and ethnicity are a significant *527 determinant of salary.” 4 Sometimes it is. But here, it wasn’t tricky. President Hughes’s own Chambers study 5 said that minorities made an average of $87 less than whites and females made an average of $751 less than males. Moreover, it conceded that the disparities were not statistically significant between minority males and white males and that the largest individual case of disparity for a female professor and a predicted salary for a similarly situated white male was 1.77 standard deviations. But it wasn’t just that Hughes awarded differential pay raises based on the statistically insignificant pay disparity. Instead he instituted raises for females and minorities to levels much higher than the average amounts of $87 and $751 necessary to achieve parity.

This was because the stated theory of the Chambers study was to award raises to females and minorities to a “predicted” salary level based on rank, experience and tenure. This was mere obscurantism, but it was obvious which shell covered the pea. The “predicted” salaries aren’t “parity,” as they don’t bring women’s and minorities’s salaries up to the salary levels of similarly situated males and whites. Because of some crudities in Chambers’s analysis, such as assuming that the' difference in pay between assistant and associate professors would be the same as between associate and full professors, the “predicted” salaries were a theoretically ideal — not an empirically derived — number. That is what accounts for the resulting raises being far in excess of what was needed to achieve parity. Over half of the white males, about seventy percent of the females, and over half of the minority males were paid less than their “predicted” salaries, but Hughes only raised the females and minorities to the “predicted” levels, while white males under the predicted level got no raise. White males were not even allowed to apply for raises to bring their salaries up to “predicted” levels.

Arguably, a layman like President Hughes might not know that a statistically insignificant disparity did not mean anything, because he might not know what statistical significance means. A statistically significant difference is one “too large to be plausibly attributed to chance alone” 6 and recognizing such a difference filters out variations due to chance. Thus, if you spin a roulette wheel (with no zero or double zero) twice and it comes out red both times that doesn’t mean it is rigged. On two spins it could come out Red-Red, Red-Black, Black-Red, or Black-Black, so a perfect wheel would come out Red-Red 25% of the time by chance. On two spins, Red-Red means nothing; it doesn’t show any “disparity” that needs correcting.

A common convention among statisticians is to use 0.05 as the measure of statistical significance, so that a result is treated as significant, where the number of cases is sufficient, if it would occur by chance no more than one timé in twenty. 7 Analogously, a result is commonly treated as statistically significant when one can be 95% confident that it didn’t occur by chance. The majority’s reference to “two *528 standard deviations” means the same thing: 95.4% of the cases are within two standard deviations above or below the mean of a normal distribution or curve if the number of cases is sufficiently high and the curve is of a normal shape.

A statistically insignificant difference is one where chance cannot be ruled out as explaining the result, like Red-Red in two spins of a roulette wheel. Chambers admitted that ethnicity was not a statistically significant variable and his study found the largest disparity for a female professor was 1.77 standard deviations from the predicted mean salary. In the case at bar, chance could explain differences in pay between candidates if, for example, one instructor joined the faculty in a fat year for faculty salaries and another in a lean year. “Correcting” a statistically insignificant disparity is like rigging a roulette wheel that came out red twice in a row so that it comes out black more often. A competent university president relying on a statistical study to award raises only to those persons with favored ethnic and sex characteristics would know what “statistically insignificant” meant or would find out. President Hughes claimed to rely on the Chambers study, so he was obligated to read and understand it.

Even if we couldn’t expect this level of statistical understanding from a university president, the Chambers study had another serious failing obvious to anyone affiliated with a university, which no university president could have missed. On the first page of the executive summary of Chambers’s study, which Hughes relied on to award the raises, in conspicuous type set off from the rest, Chambers stated that he did not control for “other factors influencing salary levels, including doctorate and performance.” 8 There is not a university president, or even a university student, who wouldn’t know that a doctorate is the sine qua non for advancement on a university faculty. No competent university president could take seriously a study of pay discrimination among instructors that did not control for whether they had doctorates.

Indeed, in Bazemore v. Friday, the Supreme Court noted that “some regressions may be so incomplete as to be inadmissible” and irrelevant. 9 Although incomplete regressions that leave out some factor will often be admitted and accorded less probative weight, Bazemore plainly states that a regression’s failure to include certain critical variables could lead to its complete irrelevance and inadmissability. 10 A regression of salaries in a university setting that doesn’t include doctorate, merit, or performance as variables is like a regression study that predicts shoe size from weight without considering foot size. Had Chambers included doctorate, performance, and merit ratings (or even just doctorate) it may have even eliminated the statistically insignificant salary disparity.

Finally, even if President Hughes didn’t understand statistical significance, he would have had to see from the practical effect of the raises that he was causing rather than remedying sex and race discrimination. Here are a couple of typical examples. Two Assistant Professors of Criminal Justice, one minority and one “Anglo” (as the study put it) held the identical position, rank, title and duties. Before President Hughes’s raises, both were paid the identical amount, $33,000 *529 per year. Pursuant to the Chambers study President Hughes gave the minority Assistant Professor a raise of $3,353. As Chambers admitted, he hadn’t controlled for “performance,” and this case, like others, highlights the significance of that failure. The Anglo professor had several more publications than the minority professor, including several books, and brought in more money in grants than the minority candidate. President Hughes testified that he was not concerned about merit or performance in awarding the raises. It shows. The additional $3,353 for the minority professor was because of his ethnicity and nothing else. And only ethnicity explains why President Hughes paid the Anglo nine percent less than the minority after the raise.

Since females were silver and minorities were gold, injustice prevailed even among those receiving raises. Two assistant professors of criminal justice both made $33,000 before the raises, one an Anglo female, the other a minority male. The female got a $2,146 raise, the minority male $3,353. Like the previous pair, they were both paid equally before the so-called “equity” raises. The only basis for why the minority male got $1,207 more than the Anglo woman, and she got three percent less than the minority after the raise, was his ethnicity. 11 And all that explained why President Hughes paid the white male six percent less than the female was his sex. As in the previous example, considering performance dramatizes the kind of “equity” at play. Here, just like her white male counterpart, the female professor had more publications and brought in more grant money than the minority professor. But even disregarding performance, the female Anglo was paid less than the minority male because of her ethnicity, and the Anglo male less than her because of his sex.

The table showing these comparisons is in Chambers’s own executive summary. It includes names, so it would be easy for President Hughes' to make the comparisons, and it’s likely that he knew something about these individuals. There is no way to mistake these raises for “equity” raises bringing minorities and females into parity with white males. The record is replete with similar comparisons. And it is perfectly plain, and uncontradicted, that the raises were awarded purely for being of the preferred sex and the preferred ethnicity.

True, the Gantz-Miller study showing precisely why the Chambers study didn’t prove race or sex discrimination or justify the raises wasn’t prepared until after Hughes awarded the raises. But it didn’t rely on new data, it just penetrated the obscurantism and flaws in the Chambers study. And even without it, President Hughes could easily see . what he was doing, even if he couldn’t be expected to see exactly why the Chambers study didn’t justify doing it.

Though the facts were unmistakably obvious, President Hughes would still enjoy qualified immunity if the law were insufficiently clear. The right at issue has to have been “sufficiently clear that a reasonable official would understand that what he is doing violates that right.” 12 Sometimes sufficient specificity requires a fair amount of development of the case law to inform the reasonable officer. For example, a policeman should know that he can’t use excessive force to arrest a suspect, but may need a case to tell him whether hurry *530 ing a suspect away from a location where the Vice President was speaking and shoving him into a van was excessive. 13 But in this case, the law was clearly established at a sufficient level of specificity.

At the general level, everyone in America knows that the constitution prohibits the government from treating some people better than others because they are of a preferred sex or ethnicity. True, no case at the time specifically outlawed what Hughes did under the circumstances. But, we don’t require a case on point at that level of specificity in order to “clearly establish” the law. 14 The qualified immunity standard of clarity is met provided that the “unlawfulness is apparent in light of preexisting law....” 15 Even in the absence of an analogous case a right can be clearly established “on the basis of common sense.” 16 Numerous cases detail the limitations on government to distribute a variety of benefits where the criteria considered for distributing those benefits is being the governmentally preferred race or sex. 17 From these cases, a reasonable university president should have been able to conclude that awarding raises to individuals beyond what was necessary for parity because of their race or sex alone was unconstitutional. Even without cases on point, a reasonable university president would know that he could not constitutionally pay some people more than others based on a preference for their ethnicity or sex.

II.

As for the Title VII claim, I concur in the majority’s remand to determine whether the pay raises were “designed to attain a balance.” 18 With the record as it stands now, there seems to me to be only one correct answer to that question. Chambers himself said that when he controlled for experience, rank, discipline and tenure status (not doctorate or performance), males made $751 more than females and “Anglos” made $87 more than ethnic minorities. I cannot see how, even if, disregarding statistical significance, raises of $751 and $87 could be “designed to attain balance,” that the female and minority raises averaging approximately $2400 and $3000 could be to “attain balance.”

The University admits, and the majority notes, that “over half of its white male faculty were making less than the predicted salary of a similarly situated white male” 19 under the Chambers study, but answers their demand that solicitude be nondiscriminatory by saying, “[s]uch is the law of averages: some will be above the mean and some will be below. Such is also the reality of the teaching profession ....” 20 The law of the Constitution requires that people not be discriminated against because they are the wrong color or the wrong sex. Though the law of averages ordinarily leaves some people below average, the government can’t constitutionally bring only some employees up, above average — to “predicted” levels— based on whether they have the right sex *531 and ethnicity. But there’s no harm in leaving this to be fleshed out on remand.

III.

Finally, I dissent from the majority opinion’s holding that admitting the letter from Joseph Franco, the Regional Director of the Department of Labor’s Office of Contract Compliance, was harmless error.

President Hughes did not call Director Franco as a witness, so the evidence came in without any opportunity for plaintiffs to cross examine him, and see what he meant by the letter, whether it was an official position of his office, what Director Franco or his staff had examined prior to taking the position, whether the position was taken as a friendly accommodation to President Hughes rather than as an evaluation of the legal obligations of the university, and so forth. Plaintiffs’ hearsay objection was overruled. As the majority opinion correctly notes, the business and public records exceptions to the hearsay rule do not appear to allow admission of the letter.

The significance of the letter was to show the jury that President Hughes had no choice, the federal government required him to do what he did. The letter says that it is the position of the federal Contract Compliance Office that the female and minority raises at issue were a legal “obligation” of the university under a federal executive order and under the affirmative action plan that the university had an “obligation” to maintain. That is tremendously powerful evidence. A fair jury could hardly impose personal liability on a college president for doing what he was obligated to do by federal law. In the eyes of the jury, the handcuffs on the president were especially tight, because the Office of Contract Compliance could affect the university’s revenue stream.

Consequently, I respectfully dissent from the majority’s conclusion that the error was harmless. The test for harmlessness is whether “the jury more probably than not would have reached the same result absent the error.” 21 Considering how weak the Chambers study was and the egregiousness of the sex and race discrimination, I question whether it would have, but for this letter.

Nor does the case the majority opinion relies on, Gilchrist v. Jim Slemons Imports, Inc., 22 save the verdict from this devastating letter. The letter in Gilchrist was an agency determination that the employer had discriminated. We held that the district court should have exercised its discretion whether to admit it and erroneously admitted it as a matter of right. However, we held that the admission was harmless (though a “close call” 23 ) because it was only introduced on damages, not liability, and the district court told the jury that the jury and not the EEOC was the sole judge of whether there was a violation.

There are several differences between the Gilchrist letter and the letter in the case at bar. First, Gilchrist had been fired about two years before the letter, and a month after the letter, he was offered a position. The employer claimed he had failed to mitigate his damages by refusing the offer, and Gilchrist introduced the letter to show that the offer of a job was part of a package to make him waive damages the EEOC had already determined he was entitled to. In the case at *532 bar, however, the letter did not affect any decision President Hughes made. President Hughes requested it from Director Franco after he awarded the raises, probably after the white males complained. And Hughes introduced it to show absence of liability — the core issue for the jury— not merely on a point of mitigation of damages. Second, the Gilchrist letter was relevant to whether Gilchrist ought to have taken the job offer, but the Franco letter, because it was subsequent to all of the relevant conduct, was irrelevant to any conduct at issue. It is impossible that a letter written after Hughes gave the pay raises was something that he relied on or obeyed. Because of this temporal impossibility, the letter can’t be probative on the qualified immunity issue, and can’t establish that President Hughes was merely obeying a government directive in the letter.

Finally, it is also less likely that cross examination would have mattered in Gilchrist, because of the nature of the letter there. The Gilchrist letter was an agency determination resulting from an adversarial proceeding before an administrative law judge, who reached a conclusion based on admitted evidence and produced a record to evaluate this decision. 24 In contrast, the Franco letter said that it was “to confirm conversations and understandings” between Franco and someone on President Hughes’s staff, based on unspecified “information furnished to this office.” The letter does not even say whether Franco ever saw the Chambers report or any other data, or just had it described to him over the phone by someone on President Hughes’s staff. The indeterminateness of Director Franco’s basis for the letter and the one-sidedness of the process that led to its issuance magnifies the risk of unfair prejudice here.

Because Gilchrist was, as that decision said, “admittedly a close call” 25 as to whether the erroneously admitted evidence was harmless, and the letter in this case was both much more prejudicial and much less relevant, extension of Gilchrist is not justifiable. Admission of the letter alone would be ground for reversal even if the case were not otherwise reversible, as it is.

. Although some minority male and female faculty members received far less than these average values, I note, as the majority does, that most salary adjustments were between $1,001 to $3,000 for the female professors and between $2,001 to $3,000 for minority males. Majority Op. at 511-512.

. Saucier v. Katz, 533 U.S. 194 , 202, 121 S.Ct. 2151 , 150 L.Ed.2d 272 (2001).

. See, e.g., Richmond v. J.A. Croson Co., 488 U.S. 469 , 109 S.Ct. 706 , 102 L.Ed.2d 854 (1989) (striking down a city’s minority set aside plan in the contracting context); Regents of Univ. of Cal. v. Bakke, 438 U.S. 265 , 98 S.Ct. 2733 , 57 L.Ed.2d 750 (1978) (striking down a rigid set aside plan for allocating student spots in state-supported universities as violated the Equal Protection Clause); Mississippi Univ. for Women v. Hogan, 458 U.S. 718 , 102 S.Ct. 3331 , 73 L.Ed.2d 1090 (1982) (holding that a policy of a state-supported university, limiting its enrollment to women, of denying otherwise qualified males right to enroll for credit in its nursing school violated Equal Protection Clause); Reed v. Reed, 404 U.S. 71 , 92 S.Ct. 251 , 30 L.Ed.2d 225 (1971) (holding a provision of the Idaho probate code that gives a preference to men over similarly situated women violates the Equal Protection Clause); McLaurin v. Oklahoma State Regents for Higher Ed., 339 U.S. 637 , 70 S.Ct. 851 , 94 L.Ed. 1149 (1950) (holding that a state sponsored graduate school’s disparate treatment of an admitted black student based on his race violated the Equal Protection Clause).

. Majority Op. at 518.

. It may be that the Chambers study should have been excluded under Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 , 113 S.Ct. 2786 , 125 L.Ed.2d 469 (1993), but that argument is not raised on the appeal, so we do not rule on it. The majority opinion states that the propriety of controlling for particular variables goes to the weight and not the admissibility of a study, but this is dicta, because we are not faced with a dispute about admissibility.

. Gary Smith, Statistical Reasoning § 10.3 at 362 (2nd ed.1988).

. Id. at 397-98.

. Chambers Study, Executive Summary at 1.

. 478 U.S. 385 , 400-01 & n. 10, 106 S.Ct. 3000 , 92 L.Ed.2d 315 (1986) (Brennan, J„ concurring, in a portion of his concurrence joined by the majority).

. Id.

. Similarly, the female professor received a $2146 raise over the white male from the prior example.

. Saucier v. Katz, 533 U.S. 194 , 202, 121 S.Ct. 2151 , 150 L.Ed.2d 272 (2001).

. See id. at 208-09, 121 S.Ct. 2151 (discussing whether this constituted excessive force under the conditions).

. See Giebel v. Sylvester, 244 F.3d 1182, 1189 (9th Cir.2001) ("Precedent directly on point is not necessary to demonstrate that a right is clearly established.").

. See id.

. Id.

. See supra note 2.

. Majority Op. at 524.

. Id. at 523.

. Id.

. Beachy v. Boise Cascade Corp., 191 F.3d 1010, 1016 (9th Cir.1999).

. Gilchrist v. Jim Slemons Imports, Inc., 803 F.2d 1488 (9th Cir.1986).

.Id. at 1500.

. See 29 U.S.C. §§ 211 , 626.; 29 C.F.R. § 1626.15 (a) & (b).

. Gilchrist, 803 F.2d at 1500 .