Equal Employment Opportunity Commission, Cross-Appellee v. Sears, Roebuck & Company, Cross-Appellant

spe inquiry all into the actual existence v. Sibron consequences.”

cific collateral York, 40, 55, 1889,

New S.Ct. (1968).

1898, It is an 20 L.Ed.2d 917 “obvi most criminal convic

ous fact of life that collateral

tions do in fact entail adverse thus the mere

legal consequences,” and

possibility consequences that such ex preserve

ist is sufficient to a live controver Dallman, Id.; see also

sy. 686 F.2d at petitioner argues

422. The mere fact that

that this matter is moot is not control now Mimms, Pennsylvania v.

ling. See 106, 108 3,

U.S. n. 332 n. (1977).

L.Ed.2d 331 appearing Supreme

It va- Court prior judgment

cated this court’s in order question

for us to consider the of moot-

ness, having considered the briefs sub- parties

mitted on this issue and the law,

applicable it is ORDERED that

judgment of this court be reinstated be- petitioner’s

cause we conclude that release custody parole has not mooted corpus

her habeas action.

EQUAL OPPORTUNI EMPLOYMENT COMMISSION, Plaintiff-Appellant,

TY

Cross-Appellee, COMPANY,

SEARS, & ROEBUCK

Defendant-Appellee,

Cross-Appellant. 86-1519,

Nos. 86-1621. Appeals,

United States Court

Seventh Circuit.

Argued May 1987.

Decided Jan. Rehearing En Banc

Rehearing and 15, 1988.

Denied March investigation

who headed the Sears had a conflict of interest because he served on the Board of Directors of the National Or ganization Legal for Women Defense and (LDEF) prior Education Fund to and at the charges against time NOW filed Sears with Sears, the EEOC.1 EEOC v. Roebuck & Co., F.Supp. (N.D.Ill.1980) (Sears I). During the ten-month trial which be gan September 13, 1984, and consumed 135 days,2 sought prove trial engaged pattern Sears practice in a nationwide against of discrimination women 31, 1980, from March by failing 1973 to December Jeffrey Scanlan, Bannon and James P. promote to hire and females into E.E.O.C., Washington, D.C., plaintiff- positions commission sales on the same ba appellant, cross-appellee. by paying sis as males and female checklist Morgan Jr., Morgan management employees similarly Charles Associates less than Chtd., Washington, D.C., employees. Pamela S. Horo- situated male The district witz, defendant-appellee, cross-appel- January 31, 1986, court on held for Sears lant. on all claims and also denied the EEOC's outstanding partial summary motion for CUDAHY, Before WOOD and Circuit judgment.3 Sears, EEOC v. Roebuck & Judges, ESCHBACH, Senior Co., (N.D.Ill.1986)(Sears Judge. Circuit II). appeals WOOD, Jr., the district court's HARLINGTON Circuit judgment disparate Judge. on the treatment partial summary claims and its denial of appeals outgrowth pro These are the judgment regarding provision that had litigation stemming tracted from an EEOC existed in the Sears Personnel Manual until charge against Sears, commissioner's filed allowing employee day 1974 a male a off Company (Sears) August 30, Roebuck & on pay gave when his wife birth.4 Sears 1973. After efforts at settlement and con appeals cross the district court's refusal to failed, brought ciliation suit alleged ground dismiss the case on the of against 22, 1979, alleging Sears on October conflict of interest. several claims of nationwide discrimination against (and minorities, women but those I. DISPARATE withdrawn) employ claims were later TREATMENT-LEGAL PRINCIPLES practices. ment Before trial the district dismiss, A. Standards of Review court denied Sears' motion to grounds challenge which was based on several includ The EEOC does not ing attorney holding an assertion that an EEOC district court's that this case in- underlying brought partialsummaryjudg- 1. Wediscussthefacts theconflictof 3. TheEEOC claim, February3, 1982, subsequent- interest whichalsoilluminatethe investi mentmotion gative,settlement, history ly and conciliation informedthe districtcourt that it wouldnot pursue interestingpart the issuesin that motionat trial if the the caseandare themost of this case, against otherwisestatistical in our discussioninfra court ruled the EEOCon that motion. appealbrought of the conflictof interestcross by partial summaryjudgment Sears. 4. The motion for containedfour other claims. The EEOCwith- drewtwoof thoseclaimsbeforethe courtruled 20,000pages transcripts. 2. Thetrial involved motion, pursue During trial, testimony on the appeal and decidednot to the court heard the 2,172 of 49 witnessesand admitted (1,080 exhibits the court'sdenialof relief totaling9,377pages claims, pregnancy for the EEOC two other policies. which involved 1,092 totaling12,858pages). for Sears Bazemore, disparate preme explained treatment under Court volves claims of 703(a)(1), 3008, 42 U.S.C. Title VII 5.Ct. at § 2000e-2(a)(l).5 To in a claim succeed § responded have if the defendants ... treatment, disparate the EEOC ul alleging offering plaintiffs’ proof by proving by a timately had the burden own, then of their the factfinder must the evidence that Sears preponderance of plaintiffs have dem- decide whether the practice” of engaged “pattern in a or dis practice pattern a or of dis- onstrated against employees. female crimination preponderance of crimination a 385, 106 Friday, v. Bazemore 478 U.S. This is because the issue evidence. 3000, 3009, (1986) (per L.Ed.2d 315 S.Ct. point is to be decided at that whether the curiam) (Brennan, J., writing majority, plaintiffs actually proved have discrimi- proof This concurring part). includes nation. discriminatory intent, employer’s a See also United States Postal Service required disparate impact in the factor Aikens, Board Governors 460 U.S. International Brotherhood of analysis. 711, 715, 1478, 1482, 103 S.Ct. 75 L.Ed.2d 324, v. United States. Teamsters (1983). plaintiff Whether the has done 1843, 1854 n. 14, 52 335 n. 97 S.Ct. depend given so in a on the “will case (1977). EEOC, Initially L.Ed.2d light factual context of each case in of all plaintiff, establishing had burden a presented by plaintiff the evidence both prima case that facie Sears followed an Bazemore, and the defendant.” pattern practice or unlawful discrimina essence, therefore, at 3009. In the fact- tion. Id. “The plaintiffs’ prima facie case evidence, finder looks at all the and does usually will thus consist of statistical evi any special significance not accord to the demonstrating disparities dence substantial plaintiff initially fact that the met its bur- employment application actions as proving facie case. See prima den of a unprotected group, to minorities and the Aikens, 1482; 460 U.S. at 103 S.Ct. at general policies buttressed evidence of Co., Inc., Magnet Pollard v. Rea Wire specific or instances of discrimination.” (7th Cir.1987). F.2d Johnson, Coates v. Johnson & (7th Cir.1985) (footnote omitted). The EEOC has not structured its argument burden, princi fit within these plaintiff

If a meets that initial shifts, ples, gives impression that it has temporarily, then burden stronger “employer prima actually case than it does. to defeat the facie show- claims, ing pattern practice by presenting hiring promotion of a demonstrat- ing [plaintiffs’] proof argues is either the EEOC first inac- has made out case, Id. The burden insignificant.” prima curate or facie and then contends that because, only temporarily shifts as the Su- Sears failed to rebut that case.6 In addi- recognized responding plaintiffs 5. The district court that the EEOC substantial burden of to disparate impact theory had also alluded to a discrimination. The court of reject suggestion evidence." We the EEOC's held, however, that unimportant among that characterization is — disparate the EEOC had failed to make out a impact proof other distinctions in the of methods (because any case it had not identified theories, showing disparate those a of treatment facially specific, policy neutral Sears that dis- intent, requires proof discriminatory of while proportionately excluded women from the proof disparate impact of does not. See Team indeed, issue), right at had waived its to sters, 14, 1843, 431 U.S. 324 at 335 n. 97 S.Ct. pursue disparate impact theory a of discrimina- 14, (1977); 1854 n. 52 L.Ed.2d 396 v. Griffin challenge tion. The EEOC does not ing. that hold- 1281, (7th Regents, Board 795 F.2d 1287 Cir. of does, however, urge The EEOC us to con- 1986). involving disparate impact sider cases theories arguing and methods of that Sears example, headings appel- 6. For the EEOC’s in its against hiring has discriminated promotion, women in hiring late brief state the claim "The because, EEOC, according to the Compelling EEOC Presented a Case of Discrimi- inquiry disparate impact "how the disparate [whether Rebutted," Hiring nation in That Not Was ought is characterized treatment] importantly "Sears Failed to Rebut the EEOC’sStatistical to affect the outcome of a case such this, claim, plaintiff Regarding promotion as since Case.” the when the makes a the sound case, prima headings Compel- statistical the defendant has a read “The EEOC Presented a facie

309 consequence. continuing is of no case as such makes statements tion, the the persuading of “The burden ultimate statisti a sound present plaintiffs “where intentional defendant of trier fact case, ais substantial there prima cal facie re against plaintiff the ly ... discriminated respond to defendant on the burden plaintiff.” Bur Although at all times with mains analysis.” probative more with a 1093; see dine, 253,101 at S.Ct. U.S. at 450 that did find court Bazemore, 3008. case, at 106 S.Ct. whether prima a facie had presented sta sound a actually presented as determination court’s The district district questionable. case is tistical proved discrimi has plaintiff whether had, and we it find did the evidence of preponderance a nation by im later. issue address stan clearly erroneous subject “is presenta of its through the structure plies Bazemore, 106 review.” appellate on dard quoted as the one such statements tion factual district court’s 3008. The at S.Ct. heavy burden had that Sears above that ultimate determi underlying findings more com evidence rebut clearly erro subject to the also nation are had that Sears and even evidence pelling standard, Under neous standard. sup no We find “ persuasion. of the burden “clearly erroneous” when finding is ‘[a] for those contentions. case law port it, support is evidence although there burden, had a rebuttal that Sears It is true is evidence the entire reviewing burden, needed meet that conviction and firm the definite left with genu ” that “raise[d] evidence produce been committed.’ has a mistake it discrimi whether fact as to of ine issue City, 470 City Bessemer v. Anderson of Communi Department Texas nated.” 1511, 1504, 84 573, 564, S.Ct. 105 U.S. 254, Burdine, 248, U.S. 450 v. ty Affairs States (1985) (quoting United L.Ed.2d (1981). 1094, 67 L.Ed.2d 101 S.Ct. Co., 333 U.S. Gypsum States v. United producing a burden not have did 525, 542, 92 L.Ed. 364, 395, 68 S.Ct. did the than evidence compelling more if such a conviction (1948)). We have Teamsters, Supreme Court In EEOC. of the facts interpretation judge’s “the trial must defense “employer’s that an stated internally inconsist illogical, implausible, is prima facie to meet designed ... be documentary or oth ent or contradicted 360 n. U.S. at [plaintiff].” case City evidence.” er extrinsic Ratliff We noted n. 46. at 97 S.Ct. Cir. Milwaukee, “ evi strength of the Coates ‘[t]he account court’s 1986). “If the pre produce to must the defendant dence light plausible in the evidence carrying the bur plaintiff[s] vent the court entirety, in its viewed record inas depends, ... persuasion den of though even it not reverse appeals may plaintiffs’ strength of case, on the sitting it had been convinced Segar v. (quoting at 532 F.2d proof.’" 756 weighed the fact, have would trier (D.C.Cir.1984), 1249, 1268 Smith, F.2d Anderson, 470 U.S. differently.” 1115, 105 S.Ct. denied, cert. addition 573-74, na (1985)). Whatever L.Ed.2d 258 permissible views two there are “[w]here evidence, de “the defendant’s ture of be evidence, choice factfinder’s per carry burden need not fendant clearly erroneous.” cannot be them tween dispari of a nonexistence as to the suasion Muskego Gun Big Id.; see Boxhom’s 1268. The F.2d at ty.” Segar, 738 Local Club, Electrical Workers v. Inc. showing of its necessarily rest on could Cir.1986). 1016, 1018 494, F.2d 798 above, noted we case. As facie prima case made in this The district plain “respond[s] a defendant once of findings. Some of factual multitude offering evidence by proof [its] tiffs' findings rested on determinations 3008, those Bazemore, 106 own,” S.Ct. witnesses, and of various credibility prima facie proved a plaintiff fact that the EEOC’s that Could Rebut Evidence That Reliable Promotions Discrimination ling Case of Claim.” Rebutted," Promotion Not Offer “Sears Did Was Not 310 demonstrations, and diametrically opposed on the district court’s

some were based events,” evidence. ac evaluation of statistical We accounts crucial the court cord even more deference to necessary “wholly found defer to the findings court’s based on evaluations of credibility district court’s assessment of the Anderson, types those of evidence. lay expert of both witnesses and [had 575, 1512, Supreme U.S. at 105 S.Ct. at only disturb resolution of dissonant to] findings Court noted that “[w]hen if manifestly that resolution is regarding based on determinations erroneous.” Id. We follow suit in this witnesses, 52(a) credibility of Rule demands case, as the district court had resolve the greater even deference to the trial court’s “contradictory statistical demonstrations” findings; judge for the trial can be proffered by primary experts statistical aware variations demeanor and for each side—Dr. Bernard R. Siskin heavily tone of voice that bear so on the the EEOC and Dr. Joan G. Haworth for understanding listener’s of and belief in together pro- Sears. These two witnesses finding A “purport what said.” that is 5,275 pages testimony. duced of trial edly credibility on a based determination” judge specifically regarding stated this tes- erroneous, may however, be if doc- timony credibility of statistical “[t]he objective ments or evidence “contradict the experts weight given and the to be their story; story witness’ or the itself testimony great importance.”7 were ... of internally implausible so inconsistent or II, F.Supp. Sears at 1279 n.& 2. its face that a reasonable factfinder would 575, 105 1512; not credit it.” Id. at S.Ct. B. Absence Individual Victim Testi- Buick, Inc., Snyder see Wheeler mony 1228, (7th Cir.1986). F.2d Regarding major all claims at issue We must also defer to the district —hiring, promotion compensation —the court to a certain extent district court found that EEOC’s failure to findings court’s factual based on statistical present testimony witnesses who noted, evidence. As the district court claimed had been victims of dis “[virtually proof all the offered crimination Sears confirmed the weak nature, EEOC this case is statistical in nesses of the statistical evidence. or related to the statistical evidence.” EEOC, conceding present it did not II, F.Supp. at 1285. We stated any witnesses who testified to individual Ozinga Brothers, Inc., in Soria v. discrimination,8argues acts of that the dis (7th Cir.1983), 995 n. 6 “[ejspecial gave weight trict court undue to the ab ly involved, where statistical evidence is sence of testimony. individual victim great deference is due the district court’s determination EEOC cites of whether Bazemore resultant v. Friday, 478 U.S. numbers are sufficiently probative S.Ct. (1986), L.Ed.2d 315 Soria, ultimate fact in issue.” In International which Brotherhood Teamsters of 324, involved, States, as did this case “a of com v. United Babel peting experts, contradictory (1977), statistical 52 L.Ed.2d 396 and Hazelwood testimony 7. The against court also heard of various criminated women. Each witness was expert nonstatistical witnesses—ten for EEOC testimony called to rebut of a Sears witness Sears, lay and nine for and heard four witnesses EEOC, regarding coding applications. twenty-four lay for EEOC and witnesses for matter, does not claim that witness either Sears. testimony The court stated that "the testimony offered credible that Sears discrimi- important nonstatistical witnesses was as as the against applied nated her—one at a store in testimony experts.” of the statistical applications for full-time commission at 1279 & n. 2. usually processed sales were not because the hiring store had done so little outside of Sears Although the EEOC claims that two of its sales, for commission while Sears showed that [they witnesses "testified in rebuttal that were] applied the store to which the other witness had definitely posi- interested a commission sales qualifi- hired two other women who had better [they] applied,” tion at the time found, cations correctly, than the witness. believe that neither provided any witness evidence that Sears dis-

311 States, properly recognized The district court 433 U.S. v. United District School of anecdotal evidence when it (1977), the value 2736, 768 299, 53 L.Ed.2d 97 S.Ct. determined that lack of individual victim individual victim proposition that for testimony reinforced its conclusions re agree that unnecessary. We testimony is in garding the deficiencies the EEOC’s sta individual did not hold cases those Smith, Segar v. 738 F.2d tistical evidence. necessary support a testimony is to victim denied, 1249, (D.C.Cir.1984), cert. 1278 471 in vio intentional discrimination finding of 2357, 1115, 105 S.Ct. 86 L.Ed.2d 258 U.S. Hazelwood, Title VII. See lation of Segar, (1985), contrary. is not Team 2741; 307-08, at at S.Ct. U.S. the District of Columbia Circuit stated that 339, 97 sters, at 1856. at S.Ct. 431 U.S. plaintiff’s methodology “when a statistical court in this case. did the district Neither appropriate pool focuses on the labor court accorded We believe generates evidence of discrimination at a proper weight. We lack of evidence level, statistically significant poli no sound School Hazelwood in acknowledge that cy subjecting plaintiff reason exists for 307-08, District, at at 97 S.Ct. 433 U.S. requirement pro to the additional of either 2741, held that Supreme Court viding showing gross or anecdotal evidence disparities can be gross statistical “[w]here Id. disparities.” at 1278. In this case the shown, proper in case they may alone the lack of anec district court considered proof pattern of a prima facie constitute finding only dotal evidence after there had The Court practice of discrimination.” major problems with the la Teamsters, recognized in 431 U.S. earlier pool determining bor that the EEOC’s 1856, however, 339, 97 S.Ct. severely statistical evidence was flawed. testimony victim is useful individual This in with District of accord Colum bring convincingly to life.” “cold numbers v. Valentino statement bia Circuit’s “examples recognized has This court Service, 674 F.2d 56, United States Postal always discrimination are not of individual (D.C.Cir.1982), the statistical that “when required, we think that the lack of such but adequately for evidence does not account arising proof reinforces the doubt from the specialized qualifications ‘the diverse and validity of the statistical questions about necessary positions question],’ [the Regents, v. Board evidence.” Griffin strong evidence of individual instances of 1281, 1292 (7th Cir.1986). Similar 795 F.2d plain vital to the discrimination becomes Rush-Presbyterian-St. v. Garcia ly, (bracketed origi tiff’s case.” material Center, 660 F.2d 1217, Luke’s Medical University nal) (quoting Wilkins (7th Cir.1981), we stated: Houston, Unit A 654 F.2d Cir. plaintiffs’ 1981), vacated, very damaging We find Dec.

position (1982)). the fact that not their experts When 74 L.Ed.2d insufficient, here, may statistical disagree, they did the court completely failed to come forward help witnesses to relate need the of live any direct or anecdotal evidence of experiences. their actual discriminatory employment practices not The EEOC’s reasons present defendants. Plaintiffs did not presenting testimony are such individual specific evidence even one instance argues satisfying. discrimination. There was no individual “inappropriate” such evidence would be be testify how defendant discriminated 47,000 promotions cause “where hires and against him. impos were at issue ... it would have been Mather, Inc., Ogilvy Rossini v. See also & present enough individual demon sible to (2d Cir.1986) (“In meaning evaluat- strations of discrimination [sic] ing all of the evidence a discrimination fully reflect on the statistics.” We do case, properly agree district court consider instances examples of individual quality anecdotal to be evidence or the of discrimination must be numerous EEOC, examples would (citing meaningful. absence such evidence.” Even a few 2)). bring convine- helped at 1278 n. have “cold numbers here, considering variety 339, Teamsters, sary to do so at 431 U.S. ingly to life.” and nonstatistical Furthermore, agree amount of statistical we and 1856. 97 S.Ct. at summary of all considering the detailed evidence and the judge the district with and the district court in its by lawsuit length of the the evidence ten-year Although and investigation by the EEOC we consid- on the merits. opinion amount EEOC, by to the evidence, Sears passed we will review the evi- information er all the could how the EEOC to see necessary it is difficult the extent here dence some members “identify at least fail to contentions that the EEOC’s consider pur- huge class of victims alleged reaching clearly erred district court has II, F.Supp. Sears ports represent.” findings. various factual argues also The EEOC 1325 & n. 82. at presented by The nature of the evidence applicant would not know that an individual significantly party in case was each this against. had been discriminated if she helpful briefly overview different. It is argument speculative While this presented by par- types of evidence situation, hiring in which apt more for the presented, almost exclu- ties. EEOC know whether there applicant may not in the form of sively, statistical evidence of other qualifications vacancy or the regression analyses on information based hired, agree we persons who were employment applications rejected pro- in the area judge applicants computerized and Sears’ sales least, compensation at motions and through payroll records from 1973 by filed individu- of Title YII suits number regression analyses based other general against employers in “seems to als Applicant on information from Interview Id. fairly refute EEOC’s contention.” had administered at various Guides Sears through 1980 at two times from 1978 Sears II. DISPARATE Territory. stores in its Southwestern TREATMENT-HIRING attempted this statistical to bolster AND PROMOTIONS through nonstatistical evidence re- argues that the district court garding subjective nature of Sears’ se- determining that it did not show a erred process allegedly discriminatory and lection by pattern practice or of discrimination aspects testing practices. of Sears’ hiring pro against women positions.9 respond regres- into commission sales Sears did not with like motions analyses employment applica- decision on the well-reasoned sion based on We base our Instead, comprehensive opinion payroll tion and records. most of .of court, see Sears Sears’ evidence was directed at undermin- principal dispositive ing assumptions claimed and address the two parties. Although faulty validity and fatal issues raised Johnson, & analysis assump- statistical Coates Johnson 756 F.2d EEOC’s —the (7th Cir.1985), briefly equal qualifications tions of interests and disparate applicants posi- in that reviewed all the evidence for commission sales case, expedient testimony treatment it is not or neces- tions. This evidence consisted of burdens, proved 9. The EEOC contends that it discrimina- dered the EEOC’s statement of the showing, tion because it met through its burden of produce which is that if Sears did not more evidence, prima facie EEOC, statistical cases compelling evidence than did the promotions hiring discrimination in automatically proven EEOC has a case of dis prima Sears failed to rebut those facie cases. crimination. As we noted in our discussion of above, analysis Based on the noted Bazemore review, plaintiff the EEOC as standard disposi- appropriate believe it is more to cast persuasion. had the ultimate burden of dine, Bur tive issue as whether the district court 101 S.Ct. at 1093. As this determining erred in EEOC did not Co., Magnet court stated in Pollard v. Rea Wire prove by preponderance of the evidence a Inc., (7th Cir.1987), F.2d ‘‘[o]nce pattern practice against or of discrimination disparate case is over ... the trial in a treatment hiring promotion. women in See either cases, governing prima the rules facie order of Bazemore, 106 S.Ct. at 3008. This method of proof, responses, longer matter.” and so on no engen- eliminates the false inference manag- implies EEOC that Sears had the burden of managers, personnel store Sears statisti officials, responding probative with a more study a based ers, store and other Team in cal analysis. Supreme Court in nontraditional of women on interviews stated, however, sters polls specifically Sears, surveys national jobs at suggest in do not that there are women ... changing status of regarding “[w]e particular type limits on the of evidence an surveys of Sears society, morale American 46, employer may 360 n. use.” 431 U.S. at job interest and 1976 and employees employer may n. 46. An labor 97 S.Ct. at 1867 employees, national surveys of Sears proof attempt plaintiffs’ to show that data, analysis Applicant force id. insignificant,” “either inaccurate or attempts to measure Interview Guides 360, 1867, plaintiff may 97 S.Ct. at or the among men and in interest differences “nondiscriminatory ex attempt provide re- presented evidence Sears also women. discriminatory planation apparently for the hiring figures, general evidence garding its 46, result.” Id. at 361 n. n. 97 S.Ct. at 1867 of commission regarding the characteristics Then-Justice, study all now Chief Justice Rehn including a case salespersons Rawlinson, in Dothard v. stores, quist, concurring in hires all based commission sales 321, 338-39, 2720, 2731, appli- 433 U.S. 97 S.Ct. personnel files of on information (1977), defend perform- 53 L.Ed.2d 786 stated that hired and sales cants who were “may endeav- data, regarding employ- ants in a discrimination case evidence ance Sears, impeach reliability of especially relating to or milieu at ment [in rebuttal] evidence, they may offer re selling the structure of the statistical evidence, butting they may disparage or Sears, affirma- evidence its arguments probative briefs tive action efforts. weight plaintiffs’ Regarding the claim of discrimination also Catlett v. See should accorded.” promotions, again relied almost Transportation Highway & Missouri regression analyses based on fe- totally on Commission, (8th 1260, 1266 Cir. 828 F.2d proportions of noncommission sales- male (defendant 1987) may introduce evidence year persons in each store at the end of a part on that a lesser interest certain compared proportion of women with applicants explains statistical dis of female actually hired into commission sales. Smith, 1249, parity); Segar v. relating again relied on the evidence (D.C.Cir.1984) (defense “[t]ypically sales, interest, the nature of commission integrity plain ... will focus on the salespeople, characteristics of commission sig methodology and the tiffs’ statistical In addi- and its affirmative action efforts. shown”), cert. de nificance of the results tion, introduced further evidence of nied, 105 S.Ct. responses to Career interest form (1985). “The defendants’ re L.Ed.2d 258 Questionnaires Aspirations it had adminis- genuine must ... at least raise buttal salespersons, to noncommission tered concerning accu issue of material fact performed adjustments various plain racy picture painted analysis. EEOC’s statistical Segar, 738 F.2d at 1268. tiffs’ statistics.” primary expert Each side featured one support by the The cases cited EEOC on statistical evidence —the EEOC starred argument that Sears had the burden role, Dr. Bernard R. in this while Siskin more rebutting its statistical represented by Ha- Sears was Dr. Joan G. “refined, valid” statistical evi accurate and worth. dence did not state that the defendant must “gen produce argues that Sears’ such evidence to succeed Instead, inadequate rebutting plaintiffs’ case. eralized interest evidence” is defendant sta those cases indicated that a a matter of law to refute the EEOC’s use such a presentation. already par have could or “was entitled to” tistical We in the con means of rebuttal. See Shidaker Car tially argument discussed lin, Cir.1986), vacat (7th proof, F.2d text of the burdens of but there is — ed, —, aspect argument. U.S. another of this In particular proposition. Oppor supporting (1987); Movement L.Ed.2d 195 most, hold, stead, Motors the cases we believe Equality v. General tunity & Cir.1980). employer’s corrective action does 1235, 1245 that an Corp., F.2d impact employer liability disparate not absolve These cases involve occurring deciding practices these discriminatory whether claims—without based disparate in a applicable the corrective principles employer took before the Teamsters, case, say 341-42, can action. See treatment 431 U.S. at These cases principles. Mar 1857-58; Rich v. Martin misconstrues 97 S.Ct. at have cited above suggest, and the cases 333, (10th Corp., 522 F.2d Cir. ietta confirm, that statistical Bell Tele v. Southwestern 1975); Parham case. rebutting a statistical one method Co., 421, (8th Cir. phone 433 F.2d 425-26 340, 97 S.Ct. at Teamsters, 431 U.S. Youngstown Patterson v. Sheet & 1970); Cf. irrefutable; they (“Statistics are not 409, (N.D. Co., 412-13 Tube and, any other variety like in infinite come (defendant Ind.1977) employer permit not In evidence, they rebutted. kind of precharge ted to combine statistics short, depends on all usefulness their “[ajctions tak post-charge periods, because and circumstanc surrounding facts subsequent filing of en to the E.E.O.C. con es.”). reject the EEOC’s therefore We charges moot issues of discrimina cannot evidence, con interest tention that Sears’ aff'd, time”), occurring prior that tion testimony of Sears’ store wit sisting of denied, Cir.), cert. 659 F.2d data, nesses, force national external labor 1100, 674, 102 S.Ct. 70 L.Ed.2d 641 U.S. data, surveys of data from survey Inc., (1981); Capad Besthoff, v. Katz & cf. mat is insufficient as a employees, Sears’ (5th Cir.1983), cert. 655-57 the EEOC’s statis ter of law to undermine denied, tical evidence. (1984). are L.Ed.2d 182 These cases all specifics Before we discuss by assumption em characterized that challenges district court’s EEOC’s practices in ployer existence before some hiring and findings regarding the EEOC’s (usually form of “corrective action” occur claims, the EEOC’s promotion we address ring charges after Title were filed VII erred arguments that the district court as against employer) in fact were discrimi analyzing those claims. matter of law natory. the district court The EEOC contends that case, however, The district court in this failing by erred as a matter of law engaged in discrimi- never found that Sears liability separately period of address natory hiring promotion practices be- Second, early years.” EEOC terms “the charges fore the filed or before by court erred argues the EEOC began affirmative action efforts. considering its claims of discrimination The district court also never held that separately from regarding part-time hires policies programs later absolved regarding full- its claims of discrimination any liability Sears of for earlier discrimina- time hires. Instead, tory practices. the district court found that the EEOC’s statistics failed to Early A. Years prove liability any period at issue in this argues and that action ef- case Sears’ affirmative by failing court erred as a matter of law began early forts as well before as separately the EEOC terms address what August charge. the EEOC’s years” hiring “the earlier at issue its challenge The EEOC’s is thus not promotion into commission sales claims. that the district court erred as a matter of support argument, of this the EEOC cites argument properly char law. Its is more says proposi various cases it for the stand the court incor clearly requires tion that acterized a claim law that a “[t]he findings, employment rectly to make two factual practices prior defendant’s failed taking finding its corrective either of which should have led the action be examined separately.” “early years” We do not read the cases as district court to consider the the district court did not err in period. One of from the later separately statistics, criticisms of the EEOC’s findings claims the district we can- those accept argument making there the EEOC’s in not was that those court erred statistics indicated differences that the significant differences between dis- recognized years by sepa- trict court should have “early years” and the later rating years” “early years. The second from later liability period of 1973to 1980. finding have been would Even if the district court had found the Mandatory adopted the Achievement sound, underlying statistics we are (MAG) plan in re- affirmative action Goals arguments not convinced the EEOC’s charge, or that Sears sponse to the EEOC’s that the district recog- court should have designed program as a corrective the MAG “early years” nized differences between discriminatory earlier action “bless” First, period. and some later we are un- findings, of factual practices. On review sure what EEOC means the term findings uphold the district court’s we will “early years,” because the EEOC never unless erroneous. See specifically defined that term. The district City, 470 City Bessemer Anderson designated liability period as be- 564, 573, 1504, 1511, 84 105 S.Ct. U.S. 3, ginning ending March 1973 and in 1980. (1985); L.Ed.2d 518 Andre v. Bendix points argument, At various the Cir.1985). 786, 788 Corp., 774 F.2d suggests “early years” EEOC is a persuaded that the district We are not (the period year from 1972 before the to make either of the above court’s failure filed) (two charge EEOC years was to 1975 findings error. was clear filed), charge period after the was the be- 30,1973 (when during “early August claims that the fore The EEOC the EEOC filed 1975, years” changes charges), period fundamental “there were from 1973to or the practices regarding hiring period (when and the before the summer of 1974 promotion implemented program).10 of women into commission sales Sears the MAG positions.” The EEOC then cites statistics Because the ultimately EEOC wants to ar- indicating depart- gue that various sales that it can show that Sears discrimi- furniture), (e.g., implementation ments auto accessories or nated before the of its af- percentages program by separating of commission saleswomen firmative action the doubled, tripled, quadrupled, figures implementation or increased before and after up eightfold during “early” action, peri- various affirmative we the believe EEOC Initially, note the “early years” beginning ods. that district means the as the all liability period, court found that of the EEOC statistics plagued by arbitrary 30, 1973, were and false as- court found was March sumptions. implementation We discuss program these statistical de- MAG fects in more detail Assuming later. Because we find summer of 1974.11 that this fif- Initially, suggests "early 10. the EEOC the that of discrimination and corrective action use as years” period fully point charges is the "before Sears had the cutoff against the date when are filed implemented Mandatory employer, the Achievement of which in this case was 30, See, August Program.” point, e.g., Goals At another 1973. International Broth- the EEOC States, striking states that "there erhood 324, 341-42, Teamsters v. United increases in the 1843, 1857-58, proportion 52 female of full time hires from L.Ed.2d 1975, (1977); years Corp., to implementation before Rich v. Martin Marietta and after the mid-1974 333, (10th Cir.1975). suggests program,” F.2d "early years” This of the MAG acknowl- edging period filing is the before that that the district court found that MAG charges. implemented by was the summer of 1974. citing support statistics in of its "fundamental changes” argument, charges Although by the EEOC cites 11. the EEOC indi cases cited periods ranging year point from before the cate that the relevant for discrimi cutoff filed, filed, charge years charges natory was two after action is the time the filed, charge generally focusing the changes on the all assume that corrective action is taken suggesting point, between 1972 to after that and so we use the EEOC’s period “early years.” point implementation this is the The cases the cutoff which is the date support argument (what program EEOC cites in of its of the MAG EEOC would action"). See, distinguish early periods e.g., courts should between term "corrective Internation- division, Sewing Ma- figures by constitutes period sixteen-month teen- to period, the three-year chines, during this by not convinced arewe “early years,” almost full-time hires proportion of female statistics and arguments the EEOC’s 47.9%) in Furni- (from and recog- failing tripled 13.5% erred district between quadrupled ture, than this it more between differences significant nize 33.8%) (8.1% and 1974 alone following it. and years period and 43.0%) (to fivefold more than had increased period emphasize this did in- obfuscate These statistics by 1976. data present it complaint, nor did in its “early during the place that took creases to sixteen- fifteen- to this corresponding implementation years,” before claims should it now period that month claimed Although the EEOC plan. MAG Neither separately. considered have been sales hires percentage of female pro- data for distinguish attempt to did tripled” between than “more during period hires motions nearly doubled percentage the same during the hires promotions data Analysis divi- 1972 and between end- “early years” beginning time after misleading— similarly products is sion of to us on Instead, it cites the statistics ed. hires that Sewing Machines percentages presented to (which were also appeal tripled” says between “almost court) indicating “dramatic than doubled between 1975 more 1972 and encom- hiring promotion changes” for the percentages 1973 and 1972 and “early after pass periods before quad- “more than division Furniture pro- presented data years.” tripled be- and 1974 rupled” between 1972 by year, down hires broken motions and and 1973.13 tween 1972 figures Its 1972.12 beginning with pro- hires and of two months thus included cites are of the statistics Some liability period the relevant approximates motions before period that for time five 1974 included its data for began, argu- support and “early years,” fail to promotions that hires and six months to before there were differences that ment imple- plan was the MAG plan. occurred after of the MAG implementation and after mented. 1973 and that The between EEOC claimed alone, during the which would 1974 this presented to compiled and The data percent- period, the “early years” female misleading well are as by EEOC tripled Hardware more than in age hires EEOC, citing in statis- imprecise. The as times in twenty increased almost and increases” dur- says “dramatic tics it show (the figure Heating Plumbing latter and carefully “early years,” does ing the year). Other trebling the next by almost For period. time recognize refer to that mis- compilations statistics EEOC 1972, the that between example, states The EEOC leading for reasons. filed, different charge was year before percentage of com- claims that the female charge 1975, was years after two part-time outside for mission sales hires filed, proportion of full-time the female Territory than dou- more tripled than Midwestern hires more sales close look and 1975. A 31.1%) breaking between 1972 (from down bled to 9.97® States, recognizes and coun- estimate.” v. United Teamsters al Brotherhood of 1843, 1857-58, 324, 341-42, contrary provided estimate. no U.S. that Sears 431 L.Ed.2d 396 ters (1977); ques- Rich v. Martin Marietta Sears did not EEOC also claims 333, Cir.1975). trial, Corp., F.2d figures not seri- at and did the 1973 tion hiring question figures. ously the 1972 reliability parties quibble 12. The over hiring figures. Sears and 1973 states Stoves, tripled" figures be- that "almost 13. In computerized personnel/payroll because Sears’ be- than doubled more tween 1972 process being implemented system was Coverings, what In Floor tween 1972and 1973. simple during 1972 to "even counts quadrupled "almost between had promotions were less reliable hires and by eightfold almost addition, 1974” and increased years." years In than for later those doubled, tripled, and almost more than had expert de- Sears notes EEOC’s Siskin 1972 and 1973. figures for hires as "at between scribed best his nothing in percentage supports the record the data indicates increasing steadily from 1972 to findings “attracting court’s was women to 1973-1974, including during fairly which is commission sales at Sears has been an im- “early years” period at issue. close to the portant priority in Sears’ affirmative action support to While some of the data seems program since the first affirmative action by the the “differences” claimed EEOC— questionnaire was circulated in 1968” and example, percentage in Auto the female engaged pre-MAG that Sears had efforts eightfold Accessories increased more than promote to hire or women into commission between 1972 and 1974 and more than acknowledges sales. Yet the EEOC that in twelvefold between 1972 and 1975—the dif- 1972 Sears’ plan affirmative action men- percentage ference increases between tioned the movement of women into com- great 1972-1974 and 1974-1975 is not as as mission positions. year sales 1972 is compilation suggests. the EEOC’s of data pre-MAG, because, found, as the court short, attempts the EEOC’s to show implemented MAG was in mid-1974. The significant pre-MAG differences between EEOC also fails to refute the district percentages post-MAG female finding court’s long-term Sears set a persuade workers are insufficient us goal women at all 38% clearly that the district court erred in fail- presumably jobs” “all includes com- 1969— ing recognize alleged by increases jobs. mission sales considering “early years” separately.14 attempts to characterize the aspect The second of the EEOC’s claim plan response charge MAG as a to its sepa- the district court should have against by quoting the remarks of rately “early years” considered the is that Sears Chairman Arthur M. Wood at a meet- failing the district court erred in ing top to discuss executives affirmative implementation find that Sears’ action. Wood stated: plan MAG corrective action taken The fact is that Sears has been identi- response Sears in charges target fied [Equal Employ- as a escape liability for its earlier discrimina- Opportunity] ment Commission and we found, however, tion. The district court improve must our quickly situation that Sears’ commitment to affirmative ac- order to position be in as defensible began tion Sears in 1969 set a possible urge do this I want to our long-term goal of women at all 38% —to lawyers working Sears, and others on discrimi- April 1970, “Sears instituted centralized, every nation company-wide possible cases to conciliate affirmative ac- claim; short, program minorities,” stay tion for women and we should out of which was revised in 1972. and out of direct confrontation F.Supp. at argues 1293. The EEOC agencies long with enforcement as as argues pensation 14. The changed EEOC that the district court rec- for commission sales from ognized representa- some salary plus increases in female draw versus commission to commis- sion; among tion promotions, female availability commission sales hires and and that there increased was clearly attributing day but erred in care—did not occur until at least 1977 willingness these increases to the apply increased and so could not to increased interest of accept positions during "early years.” women to Regarding commission sales women court, post-charge implementa- by rather than to Sears’ final reason noted which was that plan. tion above, serving of the MAG As we have indicated there were more incumbent women models, compilation the EEOC’s of statistics does role the EEOC states this factor “can support hardly its claim that increases in female serve to account the dramatic promotions imple- changes hires and years.” were due to the of one to three The district plan. say mentation of the MAG The EEOC attacks court did not that these reasons were exclu- validity sive, event, any of the reasons the court mentioned however. In even if the court’s support finding finding of its regarding willingness of women’s increased increased willingness accept posi- supported commission sales women to work could not be for the period tions. The EEOC "early years," claims three of the four it would not show failing reasons clearly mentioned the court—that commis- that the court erred in to hold jobs changed being sion sales percentages from almost ex- that the increases in female of com- clusively largely part-time promotions full-time to and more mission hires and were due to preferred part-time; women implementation plan. to work that com- MAG B. Part-Time possible, giving program our of volun- tary meaningful action time to show argues The EEOC also that the dis progress for minorities and women. trict court erred as a matter of law in remark, however, proof This is not failing to consider the EEOC's claims of implemented response MAG in to the regarding part-time discrimination separately hires charge. EEOC's The remark was made from its claims full- May 21, 1973, more than three months be- authority time hires. The EEOC cites no charge fore the was filed. On its face the proposition, argues sepa for this nothing remark indicates more than that part-time rate consideration of from full- potential action, Sears was aware of compelled by "significant time is differ although the EEOC claims that Wood had part ences between time and full time by Ray Graham, been informed J. Sears' question law; hires." This is not a rath Equal Opportunity Director of from 1968 er, arguing the EEOC must be that the through 1980, of the likelihood of the failing district court erred in to find bringing Programs EEOC's a National alleged fact that differences between charge against Sears. Other evidence is part-time signifi and full-time hires were finding consistent with the court's enough separate cant to warrant discussion began Sears' affirmative action efforts part-time figures. difference, testified, emphasize of the One 1968. Graham "I can't according EEOC, strongly is that the mix of too that it was not the threat of products [adopt sold was different between full- EEOC action that caused Sears to part-time, plan]," and significant which the EEOC claims is the MAG and stated that at the meeting knowledge because "the court based much time of the he had no "targeting" opinion products of Sears. Gra- of its on the nature of the pro- ham further testified that the MAG sold." The other difference mentioned gram designed [Sears'] to "accelerate the EEOC is that Siskin's multicell produced larger disparities progress" "leadership" part-time and continue Sears' action, role in affirmative and that Sears hires than for full-time hires. plan [affirmative action] had "one very from the not, The district court did as the EEOC Furthermore, [in 1968]." outset appears suggest, ignore or never consid policies EEOC never identified at trial the part-time figures. er pletely Neither did it com practices plan allegedly of the MAG part-time subsume the within the changed disparities generated by points full-time case. At various opinion in its analyses. Apart saying statistical part-time the court referred to data "materially that MAG was different" from disparities as distinct from full-time programs, earlier affirmative action disparities. appears data and court did refer to the full-time case more It changed EEOC does not state what the practices per-, were. The EEOC has not part-time case, clearly than it did the but that suaded us that the district court explained by finding the fact that the full-time erred in that Sears' affirmative ac- emphasized began case was more at trial.15 In tion commitment in 1968. We hold considering deed, explicitly "[f]ull that the court did not err in not the court stated that "early years" separately. part positions time and time ... were Also, course, Although territories. these full time 15. the EEOCclaims that more time money people, was devotedto the full-timecase because"so mean a lot more be- pertainedonly time, peoplestay much of Sears' defense time," to full cause are full duringclosing longer. So, important, EEOC'scounselstated ar- there fulltimeis more guments: important people butthisis as wellto the who are affected. perhaps given Andwe havenot that much So, ought lose-entirely part not to focuson attentionto the time case as it should case,although by large given,given remaining thefulltime when I will and have been enor- talking disparities particularly about this case refer to these mous regard edgedly we have with figures purposes simplicity. [full-timel to these four territories. Acknowl- reason, talking Whateverthe the fact remainsthat the we are about more hires in time, compared appeared full at least to these four focus to lie with the full-timecase. selling usually ‘big involved separately [by parties]. analyzed “Commission items, analysis according meaning high-cost ticket’ merchan- court will structure II, dise, F.Supp. furnaces, ly.” major appliances, at such as conditioners, tires, roofing, sewing air ma- analyzed part-time Although the EEOC chines, etc.,” at figures, it figures separately from full-time product per- and “the lines into which 95 employed the same mode cent Sears’ full time commission sales- Thus the part-time and full-time data. persons were hired included such items as subject to the part-time statistics were hardware, building supplies, paint ap- court as same criticisms the district pliances.” Id. at 1313 n. 58. These state- address were the full-time statistics. We product ments do not exclude lines in which recognized by the court these flaws per- the EEOC claims there awas more point in more detail later. At this it is Second, part-time suasive case. these part-time analy- to note that the sufficient statements are insufficient to convince us ses, analyses, fraught full-time are like the products the nature of the sold was design as flaws well as a failure to reasoning. crucial to the district court’s in- capture differences male and female seventy-six- These are two statements a addition, qualifications. terests page published opinion. We do not believe applied defense of affirmative action district court failed to address the part-time analyses. full- and To to both against (cid:127)EEOC’s claims of discrimination part-time figures the extent as part-time employees. figures misleading well as the full-time imprecise, the EEOC’s claim that dif- Hiring C. part-time ferences between and full-time

disparities significant enough were to war- judge plethora The district found a separate part-time rant consideration of the problems in analyses the statistical disparities is without merit. support EEOC had offered to the claim against that Sears discriminated women argues The EEOC also the district hiring positions into commission sales from impermissibly recognize failed to addressing 1973 to 1980. Before part-time distinction between and full-time specific challenges to the district in product hires based on differences lines. evidence, court’s criticisms of its statistical The EEOC criticizes two of the district helpful key findings it is to discuss three give impres- court’s statements court, made the district which we be- sion, EEOC, according to the that the dis- clearly lieve are not erroneous. Those selling trict court believed commission in- findings during period are that issue solely “big volved ticket” items such as (1) (1973-1980): in this case commission major appliances. impression Such an selling significantly different according would lead to false conclusions Sears; (2) selling noncommission women EEOC, selling because commission not as sell- interested involved smaller items like shoes as well as men; (3) ing as were were not women furnaces, big “big ticket items like qualified selling for commission as were product part ticket” lines were a small men. part-time compared hires with smaller product lines like shoes. finding that colors the district

We are unconvinced that court’s hir entire treatment of the EEOC’s failing recognize ing erred in promotion as well as its claims is that part-time selling very distinction between and full-time on is a commission at Sears figures product based on certain in job “regular,” lines different or noncom- First, selling say commission sales. we do not mission believe at Sears. We cannot claims, finding clearly as the EEOC “the court The based erroneous. opinion description much of its on the nature of the and non- court’s of commission products selling sold.” The two statements cited commission indicates that at Sears support selling the EEOC in of its claim are the two forms of differed major find next two involved, The court’s sold, the risk merchandise type of interests different ings, there were com- the manner in was reflected among men and women technical, qualifications and knowledge, ex- and pensation, grounded selling, in were commission dis- involved. and motivation pertise, recognition of differ the court’s part on the differences describes trict and commis noncommission ences between 1289- F.Supp. at length, see Sears The court based selling at Sears. sion major differ- only need mention thus we large amount of evi findings on the these found, commis- district theAs ences. on these issues. presented by Sears dence usually sell- involved selling at Sears sion extensively this evi discusses The court items, high-cost which are ing “big ticket” Again, cannot Id. at 1305-15. dence. major appliances, fur- such as merchandise clearly errone findings are say that these sewing Mer- machines. naces, roofing, and ous. basis a noncommission on sold chandise differing in- Regarding question and generally low-cost understandably was among and general men women terests in and cosmetics. paint, apparel, included selling, already we have in commission risk, es- selling some involved Commission types of evidence briefly reviewed the During period 1977. before pecially district court on by Sears presented generally received salespersons commission highlight some only this issue. We need per- ranging from to 9% commission 6% a support significant findings the court The draw week. plus a “draw” each cent that the court’s find- of our determination average or not usually did exceed 70% The court ing not erroneous. was re- earnings, subject to was estimated and con- credible found that most “[t]he employee’s commission did if the duction at trial vincing evidence offered There of the draw. equal the amount sales at commission women’s interest employee could always a risk that was detailed, uncontradicted tes- Sears was if the commissions some draw lose timony numerous men and women who of the draw.16 equal the amount did managers, personnel man- store were Sears salespersons no commission After officials, regarding their agers other and the court In what longer faced deficits. women into commission efforts to recruit reduce the finan- an effort “to was noted sales.” Id. at 1306. These witnesses testi- in an selling commission risk of on cial their success affirma- fied to limited more women to commis- to attract effort persuade women tive action efforts to sales,” paid commission sales- sion commission, sell on testified wom- commis- salary plus a persons a nominal 3% prod- interested in generally en were more sales- Id. at 1289. Noncommission sion. jewelry, clothing, uct lines like cosmet- straight hourly paid were persons usually on a ics that were sold noncommis- rate, salespersons received and full-time 1% basis, they product sion than were lines January until on all commission sales selling like involving commission automo- practice discontinued. when tives, roofing, contrary and furnaces. selling commission court found that applied were also to men. less Women salespersons required to be available often sales which often re- interested outside working hours of 8:00 the normal after night than quired calls on customers required that p.m., to 5:00 a.m. sometimes men, selling exception of custom with the homes, might require a people’s sell in draperies. Various reasons women’s sold, products depending on selling license of interest in commission in- lack necessary required qualities usually they per- not as what cluded a fear or dislike of selling, including high degree in- regular competition, ceived as cut-throat knowledge, expertise, pressure and mo- creased risk associated with technical selling, commission sales. Noncommission tivation. equal draw salesperson’s eliminated with an excess If did not was not week, draw, Failing carry next it would cleared. meet the amount of the he or she would firing. period time was cause for week. If the deficit draw over a over a deficit to the next hand, mostly largely part-time was associated with full-time and on the other friendship, (more less preferred part-time), social contact and women change more less risk. This evidence was pressure compensation plus and salary commission study surveys confirmed a national (which risk), eliminated a lot of increased through polls the mid-1930’s care, availability day group and a changing women in regarding the status of successful saleswomen who served as role society, American from which a Sears’ ex- models. pert regarding conclusions women’s made findings EEOC attacks the court’s selling; sur- interest in commission morale grounds. on interest on various We have veys employees, which the court Sears already argu- refuted the EEOC’s basic that noncommission found “demonstrate[ ] ment that Sears’ nonstatistical interest evi- generally happier saleswomen were dence is insufficient as a matter of law to Sears, present jobs at much their successfully attack the EEOC’s statistical counterparts likely less than their male argues case. The further positions, in other such as be interested analysis Applicant Interview Guides sales,” 1310; job commission id. at inter- (AIGs) quantified provided interest and 1976;17 survey est taken at Sears persuasive more evidence of interest than survey taken in 1982 of commission and According Sears’ evidence on interest. salespeople at noncommission Sears re- EEOC, analysis the AIG showed that attitudes, interests, garding their and the while women were over of full-time 60% personal lifestyles em- beliefs and applicants sales approximately 40% ployees, which the court concluded showed persons who considered themselves that noncommission salesmen were “far ready, willing most and able to sell install- more interested” in commission sales than improvements Sears, ed jobs home wom- saleswomen, were noncommission id. only comprised en of full-time commis- 1.7% 1312; and national labor force data. sion sales hires in 1973 and between 10.6% recognized expert The court the EEOC’s argues thereafter. in re- 5.3% testimony regarding gen- witness women’s sponse that some of its evidence employment, eral interests in which essen- interest, survey in the form of and labor tially significant was that there were no data, quantified. force was indeed The dis- differences between women and men re- gave trict weight little to the EEOC’s garding aspirations. interests and career analysis AIG discuss our decision that —we We cannot determine the district court finding the district court’s clearly was not clearly finding erred the evidence not erroneous in point. more detail at a later credible, persuasive probative. These short, the district court found that al- expert samples witnesses used small though neither nor Sears’ women who had taken traditional analysis of the AIGs was entitled to much opportunities Larger samples when arose. weight, helpful was more persuasive. would have more been In addi- question on the of differences in interest found, tion as the court of these “[n]one selling among men commission any specific knowledge witnesses had women. Sears.” Id. at 1314. The court found argues next Sears main- persuasive Sears’ evidence more part pro- tained as of its the issue of different affirmative action interest in commis- gram selling refusing data on women sion between men and women. The positions significant court also found sales and that this would Sears’ evidence have provided increasingly willing easy way substantiating that women became accept positions arguments, commission sales Sears’ interest between but Sears did to, among 1970 and 1980 due things, other not introduce of this evidence. Sears changes positions 25, however, in commission sales from had introduced exhibit recognized 17. The court validity that the EEOC attacked "affect the essential of the results.” See bases, job survey interest on several but we at 1310 & n. 49. agree with the court that these criticisms do not an there was instead argument on refusals of documents collection was a oppor- existing interest lack of interest com- actual lack of and indications problem of tunities, Territory. we are faced with the Eastern mission sales first, opportunity, interest or comes this contains which notes that The EEOC again This is applicants chicken-egg problem. outside eight of female instances rep- period might at have called a throughout refusing positions area where group disappointed found that witness- nonetheless issue. The court resentative selling, Sears’ witnesses’ commission preferred exhibit corroborated who es ex- were sincere and testimony that there It did not. were rebuffed. into com- to recruit women efforts tensive short, the district court we hold that Furthermore, we find the mission sales. finding that women clearly err did not crediting clearly err court did sales in commission were not as interested management testimony of Sears’ store men. positions as were encourage regarding efforts to witnesses similarly the district court find that We positions sales to take commission women concluding clearly err in that wom- did not posi- in those women’s lack interest qualifications applicants had different en tions.18 applicants. men The court noted than did argument another on The EEOC bases Report the EEOC’s Commission Sales although finding that the sur- the court’s appli- average, female indicated that “on re- testimony preserved by Sears veys and pool younger, were less cants the ‘sales’ the interests of women primarily to lated educated, likely commission less to have Sears, already employed by likely experience, and less than male sales good indication of the interest also “a was experience applicants prior work to have positions at applying for sales of women products sold on commission at with the II, 1312 n. Sears.” Sears F.Supp. at 1315. Sears.” argues that there is no 56. The EEOC finding. challenge this The EEOC does not in the record for this inference that basis findings dis- All three of the court’s employees’ sales interests

noncommission selling above —that commission cussed who the same as those of women significantly different from noncommission positions. But as we later applied for sales equally selling, that women were not inter- applicant pool, appli- regarding the discuss selling with men in commission ested interests were not clear and the cants’ Sears, applicants among and that women were not distinguish applicants EEOC did op- equally qualified in commission as with men for commission who were interested posed selling. Further- selling to noncommission the bases for the at Sears —form more, expert the EEOC’s own Siskin testi- court’s criticisms the EEOC’s statistics ap- qualified fied fewer interested and regarding hiring at Sears.

plicants sales were to be for commission Applicant Pool

expected applicant pool. challenges argued If it is that there was no interest applicant selling only court’s characterization of its in commission because there opportunities, disputes pool, were no which was the source of the EEOC’s acceptance techni- 18. The EEOC attacks the court’s conducted- interviews of women service testimony regarding cians and auto mechanics at Sears and their Sears' store witness inter- est, arguing supervisors. argues that Sears’ store witnesses were that the court as- identify single weight specific signed to her “unable to instance an inordinate amount of applicant testimony regarding toward where an outside refused the offer of women’s attitudes aspect position." a commission sales There is no basis sales because that of her commission notes, testimony with women for the EEOC’sstatement. As Sears was based on interviews state, testified to more The court did not how- witnesses than instances of outside of Sears. ever, testimony expressions aspect refusals of lack of interest of her Furthermore, selling. specifically The EEOC also based on the Sears interviews. expert acceptance testimony wit- attacks the court's relied on several the court interest, Brudney, and did not of Sears’ witness Juliet who is a writer nesses rely exclusively women’s employment Brudney’s testimony. on women’s issues and in 1980 statistics, hiring “inflated.” kin analyzed sample applications expert ap- Siskin constructed this applicants nonhired and counted as commis- plicant employment applica- from pool applicants sion applicants sales all had who 33,000 rejected applicants sales tions of applied any job for at except Sears those stores, randomly from selected 33 Sears persons requested only who a job. nonsales 1,920 applications approximately and the application, however, The did not distin- persons part-time hired into full-time and guish between commission and noncommis- positions sales commission between 1973 jobs. sion sales Siskin then compared the approximately and 1980 at stores. 210 estimated percentages of women commis- attempted With the data Siskin to deter- (“actual sion percent sales hires female”) mine characteristics of male and female percent with the of women in the “sales” applicants by coding from the information applicant pool (“expected percent female”), applications computer tapes; onto in addi- on a nationwide and territorial basis for tion, attempted to estimate year each from 1973 through 1980. proportion applicants female of sales at all The court found that com- “these Sears stores on a nationwide and territorial parisons in large disparities resulted be- identify To basis. all commission sales tween EEOC's actual expected percent promotions through hires and for 1973 of female commission sales hires aon na- 1980, analyzed the EEOC also Sears’ com- tional and territorial basis for all years,” puterized payroll records for all Sears em- values, with the z or number of standard ployees years.19 II, for those See Sears deviations expected between actual and F.Supp. 628 at 1294. figures, “highly a statistically signifi- records, Using payroll Siskin estimat- range cant” nationwide as well as in indi- proportion ed the female of full-time and II, vidual territories.20 Sears F.Supp. part-time commission sales hires in the na- at 1296-96. territory tion each year for each through initially 1980. The figures To arrive at court criticized the EEOC’s proportion applicant for the female pool full-time sales as “inflated” because part-time applicants, commission sales Sis- pool Siskin included in the applicants although 19. The court noted that Supreme EEOC and results. The Court noted in Castane disagreed Sears over the actual Partida, numbers of da v. promotions, commission sales hires and it need (1977), L.Ed.2d 498 that under a "two-tail" test not resolve experts that issue because both ad- significance, general of statistical "[a]s rule percentage mitted that the of female hires and large samples, for ... if the difference be promotions vary identified each side did not expected tween the value and the observed significantly. greater number is than two or three standard deviations, hypothesis [dispari then the that a 20. The court noted that the z values for the ty] suspect was random would be to a social comparisons very large, nationwide were “all scientist.” Id. at 497 n. 97 S.Ct. at 1281 n. ranging from 11.9 to 45.1.” The z for the values lower, general applied 17. The rule should be territories were the “somewhat but still in caution, however, ‘highly statistically significant’ range.” particular applica and its II, F.Supp. bility at may vary 1295-96. also from case to D. case. See performed found product when Cole, Baldus & J. Statistical Proof of Discrimi division, z estimated values were 9.03, (1980), (Supp. nation § 293-97 108-12 still, lower "but most had values 5.0.” Id. above 1984); see also [EEOC] American National at 1296. Bank, [1176,] [(4th 652 F.2d at 1190-93 Cir. As this court noted in Coates v. Johnson & 1981)]. Johnson, 1985), n. Cir. principles, with the accordance above [t]he "standard deviation” is a number district court assumed that “differences between quantifies degree disparities to which expected actual and values that exceed 3 stan- spread out above and below mean of statistically significant." dard deviations distribution, describing probability thus 1287. As the district responsible is any chance difference noted, "standard deviation and other mea- expected between served outcome in a outcome the ob- significance merely attempt sures of statistical sample consisting of two chance as the reason for eliminate the results. (a distribution). groups greater binomial They prove do not what in fact caused the deviations, the number of standard the less results.” Id. likely it is that chance is the cause expected difference between the and observed pool applicant sales commission EEOC’s checked had not who anyone positions sales were overestimated and thus meant that This only.” “nonsales of women percentage expected applicant for as an EEOC considered appli- hires, on that had had been based anyone who position sales above,” seriously “sales,” does “any pool. cant either checked (possibly conclusion,21 job and we do type challenge and another or “sales” Furthermore, jobs). erroneous. including “nonsales” find it “arbitrarily” assuming the court found Analyses Adjusted pool the “sales” members that all positions sales commission for all applying that the argument is basic The EEOC’s “one of was divisions all at Sears in applicant of its criticism district court’s all of EEOC’s pervading flaws most serious The contends superfluous. EEOC pool is Id. 1301. We analyses.” statistical into ac- regression analyses took conclusion court’s agree applicant interest differences count pool, sales an overinclusive this led to significant still showed qualifications and com- distinguish between it did not actual expected and disparities between the selling noncommission selling and mission sales female commission percentages of in interests or differences account for myriad of found district court a hires. The We also among applicants. qualifications however, and thus analyses, in these flaws conclusion the court’s agree with given they could be determined that sup- no credible “offered EEOC any weight. appli- assumption that all either the port” regres briefly summarize the We in sales any interest indicated cants who performed various analyses. Siskin sion in commission interested specifically were analyses to determine multiple regression all members assumption that or the sales ac expected and disparities if the between for all applying pool were the sales sales of commission percentage female tual in all positions at Sears sales commission explained by could differences hires proved divisions, Sears’ evidence and that app among and female male characteristics Id. false. assumptions were that both that he He chose six factors findings that court’s on the 1305. Based licants.22 applicant’s chance might affect an believed qualified and inter- differently women were for; (2) (1) age; job applied being hired: at Sears than selling ested commission (5) experience; (3) education; (4) job type men, women were the court found were (6) commis- experience; and product line represented disproportionately disparities. Whatev- quantita- been no argues there would have that there is “no The 21. EEOC considering ex- females’ er the reasons for that the court's tive basis” for the conclusion preferences sales on pressed for commission applicant pool A criticism of was sales inflated. forms, statistics, however, most of application the court found that underlying assumptions selling qualities for commission the essential quantified to be valid. have to be does not Furthermore, interview, only from be determined "could quantitative basis there is some II, application.” not from a written accepted conclusion because for court’s F.Supp. at 1303. expert that of of Sears’ Haworth the conclusion prefer- applicants a written who indicated those sales, analyses, designed Multiple regression to de- men. over 75% were ence for commission independent noting varia- did not define of several termine effect It is worth variable, ap- dependent in this case percentage sales bles on a commission female accepted preference hiring, method plicants a for is are an common those who indicated application disparate See proving The treatment claims. on the form. sales commission Friday, because "far less it did not do this stated Bazemore (1986); 3000, 3008-09, Camp- go persons.” positions L.Ed.2d 315 to such than half of the bell, Analysis Mini- group applicants Regression Title VIICases: argues who Worth, Standards, Comparable and Other preference sales a commission mum had for indicated Meet, predominantly 36 Stan.L. writing application Law and Statistics on the Issues Where in male, Finkelstein, (1984); Re- expected The Judicial the EEOC did not base Rev. 1299 Regression in Race ception Multiple Studies percentage sales hires on of female commission Cases, pref- and Sex of females who written Discrimination Colum.L.Rev. the numbers had had, (1980). sales because if it erence commission for product experience. sion sales Siskin cod- of characteristics.” Sears at 1297. ed information each of these character- (from applications) per-

istics for each noted, As the district court logit son in the hired and sample nonhired analyses multivariate reduced the dispari- groups, grouped each characteristic ties expected between and actual female categories. into various per- logit He then hires. analysis sales reduced the expected nationwide proportion formed multiple regression analy- various of female commission sales hires for 1973 designed ses to demonstrate that discrimi- through 1980 from for 61.1% full- 49.5% nation than rather the six character- time and from part-time. 66.2% 63.3% disparities istics accounted for hiring multivariate reduced the ex- applicants. female The EEOC relied for pected percentage of female full-time hires part the most regres- on the results of two years nationwide for all from 61.1% analyses. weighted sion logit regres- (later 37.2%),23 40.3% reduced to and the analysis, sion im- measured the disparity part-time hires was reduced pact independent of each age) variable (e.g., 66.2% 60.5%. court also noted dependent on the variable of whether or that the z values below 3 for a num- hired, person not the controlling for the years ber of in various territories and that impact of independent other variables. In the EEOC had dispari- admitted that those cross-classification, the multivariate or mul- ties were not statistically significant, but analysis, ticell Siskin’s model allowed all recognized also the EEOC’s claim that sta- interact, variables there was a tistically significant disparities existed na- separate possible cell for each combination tionwide and by territory for years several of all the different values for all of the in which the z value exceeded 3.24 independent variables, or characteristics. The EEOC conducted another multivari- As the explained, this method ate analysis for full-time hires for each of analysis expected “calculates per- product ten groupings (e.g., line home fur- cent by comparing female the number of nishings, appliances) on separately a na- any given hires with combination of charac- years tionwide basis for the through 1973 teristics, with number of female non- 1975, and through 1976 1980. The court hires with particular combinations noted that there statistically were signifi- [sic] 23. The proportionately by of may in the six ticell reduce this nated most of the composition to become ties on the bias can bias and greater the hires, possibly masking that Siskin had from the female no cells analysis analysis 1298. The reduced to time commission sales hires at 40% after con- sidering possible biases in estimate of losing precision applicants, underestimate of the occur relative to the number of analysis, commission sales court noted that this occur proxy logit analysis. and characteristics and basis court noted that Siskin when bias 37.2%, number expected of the possibly subject sexual bias, by collapsing explained fragmentation and that Siskin arrived at an sex.” Sears Siskin claimed he had elimi- men, expected higher characteristic expected in the model. either of which could cause female composition cells in which there are hires. Then "the sexual expected proportion such as automobile ex- hires. that the larger size district court noted rerunning disparity hires tends proportion various the multivariate bias at the risk applicants, Fragmentation the number of fragmentation possessed attempted of the actual multivariate any dispari- Proxy proportion categories was later the mul- of full- simply bias dis- Part time: territories in cally significant cess in greater acteristics held favor of held men would more due to discrimination. As the district court explained, Full time: perience The district closely hiring. success of men in men, "if there is in fact Southwestern— Southern— Midwestern Pacific Coast— Four Territo- Southern— 1973-1978, ries— Southwestern— Eastern— automotives, Midwestern— Pacific Coast— Nationwide— Eastern— correlated characteristics more court stated that the This association disparities women.” Id. appear with general, presumably associated are as follows: 1973-1980 1973-1974, 1973-1980 1976-1980 1973-1979 1973-1980 1973-1977 1973-1974 1976-1978, 1973-1977 1973-1977, success discrimination claimed statisti- analysis commonly than char- years 1976-1980 ^[1980] mask the with suc- to be and and in sales in commission interest product of the in several disparities cant con- The EEOC sold.” Id. to be product conducted groupings.25 line characteris- four of its appeal that case, tends ex- part-time for the similar experi- and the three tics, Job-Applied-For product of ten instead fourteen cept that product experience, (job type ence variables and the Midwest- used were groupings line product experience, line separately from analyzed Territory was ern highly correlated experience), “were sales be- territories the other the combination however, nothing, cites interest.” It hiring pat- differences apparent cause record, or outside the record either on terns.26 is thus The case support this statement. of the major criticism court’s The district Friday, unlike that of Bazemore six analyses is that the regression EEOC’s L.Ed.2d 106 S.Ct. U.S. ex- by the EEOC’s chosen characteristics Supreme Court (1986), in which inade- “simplified” were pert Siskin independent varia- type of that the noted differences applicant account quate to analyses regression plaintiffs’ bles used II, 628 qualifications. in interest and testimony discovery based on were selected n. 35a. F.Supp. at certain officials that of defendants’ by one in- did characteristics that these contends depend- of the determinative factors in interest contemplate differences deed Although ratio- salary. ent variable criti- court first qualifications. independent variables nale choice of characteristics. choice cized the compre- preclude questioning did found, the EEOC does The court variables, see Baze- of these hensiveness this, factors chosen six challenge 3008-09, re- more, Court opinion as his result of by Siskin were conclusion appeals’ the court of jected important for would be factors to what unacceptable because analyses were the selling. The court further commission might variables that not include all did weight to Siskin’s give little it could found salary Court stated effect on have an —the “is Siskin regard opinion given some analyses should be economics, has no expert in labor not an We think that probative value. sales, has of retail expertise in the area support its to choice failure EEOC’s knowledge of Id. at Sears.” no direct on the casts a in this case shadow variables to such great deference accord We must analyses regression value of the probative determinations. credibility-based factual those variables. Further- incorporating City, Bessemer City See Anderson indeed more, were if those four variables 1504, 1512, interest,” “highly correlated with (1985). L.Ed.2d 518 satisfactorily, explain, does not EEOC still incorporating varia- analyses those how also determined The court for adequately interest. controlled bles to evidence” presented no credible “EEOC particular six support Siskin’s choice affect- only not factor was Interest selling important commission being factors hired ing applicant’s chance an Instead, found the court “all for at Sears. not court found was controlled that the regression analyses. at indicates ... offered trial EEOC’s challenges were not con- important factors district court’s number analyses not regression did analysis.” in the EEOC’s clusion that included qualifi- adequately adjust for differences in II, The most critical F.Supp. at 1302. found, the As the district by EEOC” “intentionally excluded cations. factor Report itself Sales Commission applicant’s “the EEOC’s to the court according was 3 in 8 z values exceeded court noted period, 26. The noted that the 1973-1975 25. The court product groupings line in the combination groupings product line z exceeded 3 in all values periods 1973-1975 territories for four except apparel, but in 1976-1980 women’s the Midwest- 3 in Z values 1976-1980. exceeded disparities were reduced and were statisti- groupings Territory product line for 8 of ern groupings. cally significant of 10 in 3 1973-1975, product line 7 of 14 and in groupings for 1976-1980. that, ap- average, Supreme “on female cluded. The demonstrated Court later reversed pool younger, issue, plicants in the ‘sales’ the Fourth Circuit on this holding educated, likely commis- Appeals less less to have that the Court of erred deter experience, likely mining regression less than analysis sion sales that a “ ” applicants prior experi- ‘unacceptable’ to have work male because it did not include “ ” products on commission ence with the sold ‘all measurable variables’ that could af at 1315. dependent Bazemore, at Sears.” Sears fect the variable. experience The EEOC counters that its var- 106 (quoting opinion S.Ct. at 3009 of court appeals). iables took those differences into account. of The Court noted that the re experience gression three variables major accounted for the would seem to account for differences in factors and in that situation failure in “experience” aspect qualifications. normally goes clude probative variables characteristics, ness, The court found that other admissibility. Id. The district however, appearance, “physical such as as- this case did not make the Fourth sertiveness, communicate, ability Circuit’s fatal It mistake. is clear from the friendliness, motivation,” and economic opinion district court’s that it did not deter managers factors Sears had identified as analyses mine that the EEOC’s were inad salespersons, desirable missible or irrelevant because of the failure “could instead, be determined from an inter- to include the court de variables— view, application.” not from the written analyses termined that the proba were not recognized Id. at 1303. The court tive because of the Although omissions. easily quanti- these characteristics are not we do not believe the court found the re fied and this data generally gression would not be analyses totally unacceptable, the forms, application available from the court’s criticisms of analyses are so found that to the pervasive extent EEOC’s re- that the court have believed gression analyses incorporate did not these regressions were what the Baze- factors, they weight. were entitled to less incomplete more Court termed “so as to be say finding, going We cannot irrelevant,” that this inadmissible as which would be probative value rather than the relevan- exception general rule established cy regression analyses, Bazemore, in Bazemore.27 S.Ct. Bazemore, erroneous. See S.Ct. 3009 n. 10. 3008-09. The district court’s other criticisms of the regression analyses go coding of the

We believe the district court ac regres- factors the EEOC did include in findings regarding corded its the the failure to sions. coding The court found that the independent include of variables such as inter “clearly inadequate several factors was aspects qualifications est and of in the re gression properly adjust” for analyses those factors. proper evidentiary the weight. The district at 1303. The court court cited the noted Fourth regarding prior Circuit case that the Friday, Bazemore v. characteristic of F.2d (4th Cir.1984), job experience, the part adjust EEOC did not for and aff'd job experience. various part, forms of The court rev’d in (1986), gave 92 L.Ed.2d 315 principle example prior job for the the of experi- that the meaningful regression in a analysis, in ence selling all of commission were —women dependent variables significantly likely more af than men to have been cashiers dependent fect the basis, variable should be in commission-selling on a and thus in salaries, mining 27. The Court failing found some of the dis- Bazemore to mention that one trict regression court’s conclusions omitted vari- included that variable and “inexplicable light addition, ables greater disparity. of the record.” showed an even In Baze- more, 106 S.Ct. at 3011 n. 15. These conclu- the Court noted that while the district had court finding regression sions included fault with a listed nine regressions, many variables not accounted for in the variable, particular because it did not contain a "missing” of the re- variables regressions when in other the included var- lated to other variables which were included. statistically significant dispari- iable there were Id. We find none of these infirmities in this ties; stating key that one variable was to deter- case. likely varia- into account taking without in com- experience have actuality not did applicants at individ- quality at Sears. in the done tions selling as it was mission expert openings Siskin actual- that its whether or even responds ual stores separat- regression stores, did a multicell the “EEOC performed at ly existed been cashiers who had persons all ing out actually con- hiring situations the analyze Al- results.28 on the impact” “trivial at managers.” Id. 1316. by Sears fronted the have been may not though cashiers the statistical accorded court thus experience, we job example of earlier best that on weight. believe We analyses less prior emphasis on the court’s do believe finding plausible. such this record implausible on experience job findings, the district on the above Based in- other found court The district record. z both for the values determined court experience coding of the in the firmities expected comparisons unadjusted the not code did variables; Siskin namely, hires, regres- for the interrup- female and actual experience, prior the amount relevant statistical- history, or other not constitute analyses, did in work sion tions the was on (the Regarding of which the latter experience disparities. ly significant coding inadequacies These application). court found that analysis, the unadjusted led the court experience variables of the in- for never controlled nature of the “[bjecause of conclude no were qualifications, there terest deficiencies, are credit- women coding these actual comparisons meaningful between experience greater amounts of ed with “z hires thus the female expected actually had. The higher levels than nothing.” Id. prove[d] values is to inflate inadequacies of all these effect attempts to found that also the statistical models percent women among applicants, adjust for differences hired.” Id. should estimate qualifications, were mainly interest and some, but not all responds to above, so inadequate as discussed woefully argu- criticisms, on its mainly relying these “fair not achieve esti- the EEOC did data cite actual did not that the court ment hires. expected female of the mates” be- do not criticisms. We support the concluded, z case, such coding court’s criticisms lieve the comparison of with the statistics connected erroneous. experience variables lit- hires “have expected female actual and that one of agree with We credited The court further tle value.” Id. coding was criticisms of court’s into cal- incorporation of “variances” that Siskin court concluded misplaced. The to account z statistics culation varia- Job-Applied-For not recoded had proportion difference between coding, discovering in the an error after ble pool pro- and the women nationwide was thus an overestimation that there pool. After in each of women store portion proportion of com- expected female variances, the court taking into account however, had, Siskin sales hires. mission “significantly noted, z statistics “overlapping variable recorded lower,” nationwide so that EEOC’s reported “[f]or sample” twenty-nine stores analysis of full time impact the multicell was no in that there hires, only in two z exceed analysis. sales values 1317. court then Id. The years.”29 none of the argues that The EEOC z values did concluded that these because regression of its other criticisms court’s in male into differences not take account consequence. The analyses sales, an in commission interest female by creating artificial na- court found that differences incorporating those applicants, analysis pools of and territorial tional put in likely them other traits proportion was had expected female hires cashiers 28. unlikely. selection where their was cells by percentage points for full-time 0.9 reduced part-time. percentage points Siskin 0.2 and explained taking account vari- analysis likely z after into did not show 29. values and 3.86 1974. 4.12 in 1973 persons were impact who were ances of an much produce statistically significant would not from analyses. those The district court disparities. noted that Siskin expect- estimated that the proportion ed of female commission sales challenges findings, the above hires higher based on analysis this arguing regression analyses its did than it was based on the multivariate re- statistically significant dispari- indeed show gression analysis. ties that satisfied its proving burden of by discrimination as articulated the Su- argument EEOC’s main is that its preme Bazemore, Court in S.Ct. analysis AIG supports regression plaintiff 3009: “A in a Title VII suit need analyses. implies, however, It also prove not discrimination with scientific cer- own, even considered on its analy- AIG rather, tainty; prove his or her burden is to proves sis hiring Sears’ intentional discrimi- discrimination preponderance aby reject nation. We both contentions. The evidence.” The Bazemore Court went on district court explicitly stated that he as- note, however, “[w]hether, fact, cribed little weight to the analyses, AIG regression such a analysis carry does analyzed by whether Sears, the EEOC or plaintiffs’ ultimate depend burden will in a agree we with this conclusion. The given case on the factual context of each very AIGs have a scope, coming limited light case in presented of all the evidence only two stores in Sears’ Southwest plaintiff both the and the defendant.” Territory. addition, the Guides at one yet Id. We have not as discussed all of the store were administered while evidence, EEOC’s but we briefly need only the other were administered from consider that other evidence because the 1978 to 1980. The simply Guides cannot primary EEOC’s focus was regression its serve proving as basis for the EEOC’s analyses (the exception with one claim of nationwide discrimination for the AIGs), the EEOC casts that evidence as years 1973 to 1980. supporting regression its analyses alleg- The EEOC nonetheless claims that the edly showing hiring discrimination. We be- AIGs were corroborative of its multicell lieve district court’s treatment analyses because after adjusted regression EEOC’s analyses was consistent pool using “sales” expe- interest and &emdash;the with Bazemore require court did not responses AIGs, rience on the there were the EEOC to disparities show with scien- statistically significant disparities. The tific certainty, but rather all the considered analysis found that this was tainted evidence in determining that it could not because of the flawed pool, “sales” rely on the purported disparities. above, we discussed and because the ad- justments incomplete, taking into

3. AIGs age account or more than education support To the results of the re just product a few groupings. line While gression analyses, the statistically agree age factors such as analyzed responses to Applicant In were crucial to analysis, we cannot (AIGs) terview Guides that had been used find that the court’s conclusion Denver, in certain Colorado, stores in the evidence, the value of this especially con- 1978-1980, area in Waco, Texas, and in sidering scope, its limited was clearly erro- analyzed Siskin these applicants’ neous. ratings, Guides, found in the on a scale of Furthermore, one to five of apparent their own it is expe interests and that the dis- regard rience with to four trict forty-five court found Sears’ relatively categories helpful regarding chosen Siskin more closely ap separate issue proximating product four groups line differences in at male and female interest appliances, automotive, Sears: in home build commission sales. The court found that ing materials, improvements. evidence, and home this which showed that men were analysis This incorporate did not any of the more jobs interested in usually associated characteristics in used the other regression with positions commission sales at Sears analyses, completely independent while disproportionately women were inter- fig- the smaller pares figures those jobs, corrobo- noncommission-type in ested responds that these Sears dif- ures at Sears. survey evidence Sears’ rated the na- comparisons are inaccurate interests. female male ferences in the break- not reflect had ana- does tionwide data Haworth expert Dr. Joan Sears’ commission “normalizing” the scores versus of noncommission by down lyzed AIGs interest, product in line ex- depending on gave salespersons themselves applicants Sears, considering According that some to experience, isting at Sears. skill in- scores numbers might inflate their “included vast applicants product line data being hired. Ha- sell their chances who did not crease store clerks department categories roughly Siskin, worth, chose way in like no and whose commission on lines product with Sears’ corresponding salespeople commission those of resembled percent- expected the EEOC’s adjusted commis- Sears,” data for retail and other hires based sales commission age of women product salespersons failed indicate sion in Guides. responses “normalized” on it sales- so that included line differences percentage expected adjusted the She also and furs selling like fashions items women indicating that one figures on based on sold on commission which are “seldom for applicants is selected job every 250 unclear whether It is Sears.” adjust- these After sales. commission argument this actually presented reduced for ments, greatly disparities Indeed, expert Siskin district court. The EEOC period of 1976 force external labor rejected on reliance validity of Haworth’s criticizes constructing analyses. Issues his data in argues that grounds and several on court are waived presented much analysis too gave the district court Mercury Lincoln DeValk See appeal. court’s light weight, especially Co., Motor v. Ford analysis. EEOC’s AIG criticisms event, Cir.1987). considering the defi- disagree. The court various noted We comparison likely present in a inaccuracies explicitly analysis, Haworth’s ciencies in data on data with Sears of the nationwide given it “substantial had not it stated question, cannot find particular this weight the it is clear weight,” and conclud- court erred that the district not in evidence was did this accord indicated force ing that external labor data prod- specific sense of Haworth’s the more hiring discrimination Sears. coun- comparisons, which was the line uct the court’s conclusion The EEOC attacks rather analysis, but terpoint to the EEOC’s force external labor that certain nationwide lack of interest regarding women’s relative salesper commission employed data for all Sears. selling at commission availability a lower female sons reflected say cannot at 1323-24. We during positions sales for commission attributing to clearly erred in district court estimat than the years 1976 to 1980 significance limited evidence the pool. on its based “sales” ed were available did. “inappropriate argues that it is The EEOC Data External Labor Force static work force compare a nationwide A hiring figures for vari figure with Sears’ the external claims that compari agree that years.” such ous We Sears on data introduced labor force misleading. by this As son can be noted and female male issue differences Opportunity in Movement & interest, “corrob the court found Corp., Motors Equality v. General strong of female orate[d] (7th Cir.1980), 1235, 1345 cumula such F.2d and in sales lack of interest hiring de “snapshot” statistics include tive lines,” id. particular product ... twenty thirty years. the past cisions for EEOC’s claim that supports instead past mask discriminato Thus the against in hir statistics women discriminated be com hiring and should not ry decisions nationwide ing. EEOC cites various availability commis pared with female percentages of sales female statistics for during four- positions a recent sion sales product lines and com- people in various

331 id.; See United States v. period. proposition for the year objective that a lack of 415, (7th City Chicago, 549 F.2d regarding standards employment decisions of Cir.1977) (“it charge is no to a of defense can be a discriminatory practice because everyone in discrimination else is such system with its safeguards lack of law”). compliance with the may easily subject to abuse. See Davis Weidner, v. 726, (7th Cir.1979) 596 F.2d Nonetheless, we believe that the main (court noted generali or “absence extreme purpose according of this evidence to the ty any predetermined of standards district court not to for se was show overall fe- lecting employees, availability posi- male the failure commission sales to record tions, but job rather further corroborate dif- evaluations of applicants, and the ab ferences male and female interest sence employment of minorities on the deci- product line. The court found that sionmaking bodies add can credence of the national labor force did data corroborate aggrieved claim of an employee, especially selling female lack of particular interest when employer rely objec does not lines, product say and we cannot that this tive, easily measured criteria for his em finding is erroneous. To the extent ployment decision”); Stewart v. General the usefulness of this static labor Corp., Motors 445, (7th 542 F.2d 450-51 subject force is to the same criti- Cir.1976) (court found discriminatory pro cism incorporating past discrimination, practices motion process where a which we find when doubtful the evidence “highly subjective loosely structured,” purpose is used for the showing differ- the supervisory being recommendations im ences men’s and women’s interest portant but there no safeguards lines, particular product we note that some guidelines no written regarding promotion of the evidence that the court found most denied, criteria), cert. 919, significant comparisons related to of male- 2995, (1977); S.Ct. see also 53 L.Ed.2d 1105 businesses, and female-owned which would Carmichael v. Birmingham Works, Saw not seem to subject be as to discrimination. Cir.1984) F.2d (subjec procedures tive promotion context of in 5. Hiring Practices cluded word of regarding openings mouth The EEOC maintains that the dis procedures); informal review Harris v. determining trict court erred in that none Birmingham Education, Board regarding of its evidence subjective na (11th Cir.1983) (“lack F.2d hiring process ture Sears’ or objective hiring standards contributed to testing practices supported [plaintiff’s] prima establishment facie EEOC’s statistical analyses provided any case” of hiring); racial discrimination in proof credible discrimination Sears. Springs Williams Colorado School Dis respect With subjectivity of Sears’ trict, (10th Cir.1981) (ob 641 F.2d hiring process, argued that there serving subjective promotion and em was no formal training and a lack of writ ployment possibilities criteria afford ten instructions for Sears’ interviewers re abuse); Rowe v. Corp., General Motors garding qualities to look for in commis (5th Cir.1972) (discrim 358-59 applicants. addition, sion sales inatory promotion/transfer procedures pointed what referred to as a where foreman’s recommendation most “highly description masculine” of a com important promotion process, factor salesperson mission found in the Retail given foremen are written no instructions Manual, Testing phrases which included regarding qualifications necessary pro “active,” drive,” such “has a lot of pos motion, controlling “vague standards were vigor,” sesses “considerable and “likes subjective,” hourly employees not noti requires work which physical energy.” promotion opportunities fied of and no The EEOC also notes that a Sears official safeguards). admitted that this reflected characteristics average “on possess more men hiring than Sears concedes process women.” The EEOC cites subjective, numerous cases was argues subjectivity Manual, instruc- argues that this EEOC hiring process necessary part of is a in instructions to hiring oral tion reflected that because was follow it does not discriminatory. found that necessarily The district court subjective, it is interviewers.

process is historical Opportu that this “no evidence Movement there was for Cf . Corp., hiring de- any on description had influence Equality v. General nity & Motors Cir.1980) (court (7th in cisions, during years particularly 622 F.2d pro in inherent subjectivity paid was managers “often noted that Sears question,” manual,” evidence established system, but motion to information no attention and balances checks system contained in the record “[tjhere is no basis and that addition, system in promotion and, in description had concluding that this for ac fulfill affirmative procedures cluded of women impact on the selection negative Inc., Jeffboat, obligations); Mozee tion II, 628 sales.” Sears for commission (7th Cir.1984)(“sub 365, 371, 746 F.2d consulting After F.Supp. at 1318. may intangible factors jective and other record, the court did not conclude that we and ... decisions employment influence finding. clearly err in this may not subjective misjudgments even argues also The EEOC Title VII liabili necessarily be the basis finding in that one as court erred district addition, probative “the value ty,” testing practices was not pect of Sears’ may be diminished evidence statistical discriminatory. The EEOC contends that genuinely on sub employer relies where scale, scales on the “vigor” one of seven making decisions and jective factors Schedule, Temperament con Thurstone incomparable” (emphasis where data are likely more be that would questions tained Springs original)); v. Colorado Williams men, such as affirmatively answered District, School you played on a football team?” “Have Cir.1981) employ (subjective definitions of ques agreed these The not promotion ment selection criteria likely answered af tions would more se, subjectivity some per unlawful men, to believe firmatively by but chose realistically making be avoided cannot witnesses, many Sears’ who testified decisions). ar hiring promotions tested, they or if were not applicants impor gues subjectivity especially is tested, after it was not until action efforts. tant in its affirmative Cf. believed Sears’ Equality, were hired. The also Opportunity & Movement for the test had managers, testified that 622 F.2d at who 1277. hire impact any decision to little then, subjec- is how much question, adjusted for women. that test scores were agree tivity permissible. We argues implicit in the that it is that a court must consider Ninth Circuit impact phrasing district court’s subjective employer’s use of criteria “an impact. some that it had at least test facts and ... with the other circumstances that there light In of the court’s conclusion of the case.” Casillas v. States United no credible evidence that a woman’s (9th Cir.1984). Navy, 735 F.2d being prevented vigor ever her score policy been better While have Sears, how hired for commission sales written instructions and Sears to have had ever, clearly err find the court did training for interviewers formal testing determining aspect this Sears’ qualities for in sales to look commission own, impact, enough on its did not have applicants, say on record we cannot prove a case discrimination. subjectivity so much that Sears exercised engage discriminatory practice. as to in a addition, Past Discrimination finding regarding the court’s appears efforts action affirmative the court contends that impermissible subjectivity on the obviate hiring mentioning data for erred

part of Sears’ interviewers. because, although it was before the period began, it constitute Regarding description liability would Testing past salespersons found in the Retail discrimination. support an inference of later discrimina- Friday, Bazemore v. cites 3000, 3010, tion. 92 L.Ed.2d 315 (1986), Supreme stated in which the Court “ engaged ‘[p]roof employer that an in D. Promotions prior to the effective racial discrimination court, According to the district might date of Title VII some circum- exclusively relied “almost on statistical support the inference that such stances analyses” prove its claim that Sears continued, particularly discrimination against promotions discriminated women in aspects relevant of the decisionmak- where ” from noncommission sales ing undergone change.’ to commis- process had little Hazelwood School District v. (quoting jobs.30 sion sales 628 States, United 309-10 n. expert 1300. The EEOC’s Siskin estimated 2742-43 n. 53 L.Ed.2d 768 expected proportion pro- of female (1977)). year motions by looking for each proportion female of noncommission sales- prevail argu- this The EEOC cannot persons in year each store at the end of a ment, however, on because based the dis- (year-end pool). compared store He these findings, trict court’s have we held figures proportion with the of women actu- erroneous, proved are not it has not ally promoted sales, into commission against that Sears discriminated women in up came with z values that exceeded 3 in hiring in 1972. The EEOC cites various year all but one territory. one To ac- products statistics that are of the same possibility count for the persons above, in a analyses we have described and ar- promotion division in which a gues occurred they show discrimination. These greater would however, pro- have chance at the analyses, subject to the same salesperson motion than a outside the defects the district divi- court found fatal to the sion, performed Furthermore, analysis Siskin a second years. statistics of later expected this data. He may likely these statistics estimated the fe- be even less proportion promotions male years. by looking reliable than those of the later expert proportion to the female Siskin testified that these of noncommission figures salespersons 1972 were “at best an on (year- estimate.” the basis of division say Because we cannot pool). that these statistics end division The z values for this provide proof discrimination, past analysis years exceeded 3 in all and territo- need not determine whether the except territory statistics ries for one year.31 one 30. The EEOC 31. The Years All ^{[1980] [1979] [1978] [1976]} Pacific tories from 1977 1099 ^[1973] 1974 1016 ^[1975] female EEOC's Division Pool full-time noncommission sales to full-time com- sion to full-time commission sion sales to mission sales and from Year Total 1973 to 1980 were as follows: disparities ^{[6082] [238] [419] [934] [819] [732] [825]} Coast, Southern, disparities promotions Percent Female Number Female Exp. 71.8 68.8 70.8 70.4 63.7 68.0 68.4 72.8 65.5 part-time only analyzed 42.5 47.5 43.1 51.5 48.2 43.9 47.7 29.1 37.9 Act. between from full-time noncommis- part-time commission sales. and Southwestern terri- ^[4185] Exp. part-time ^{[596] [267] [518] [665] [581] [635] [752] [171]} were expected positions using for the nationwide from movement from ^[2620] Act. Diff. Z as follows: ^{[349] [422] [450] [483] [240] [385] [113] [178]} noncommis- and actual 1565 32.9 269 20.5 280 20.7 341 9.2 ^[169] 185 14.5 174 14.5 ^[89] 58 9.5 Eastern, 15.1 11.2 Territory from 1973 to 1980 were as follows: All Years Years All ^{[1977] [1978] [1979] [1980] [1973] [1975] [1976] [1974] [1974] [1975] [1973] [1978] [1976] [1980] [1979] [1977]} Year Year Total disparities Total ^{[3649] [207] [916] [380] [348] [251] [308] [360] [372] [206] [317] [361] [546] [426] [549] [815] [438]} Percent Female Number Female Percent Female Exp. Exp. 71.5 78.7 83.9 76.0 70.4 75.8 79.9 72.3 78.8 80.1 67.9 67.2 86.9 70.5 73.7 70.0 69.7 72.7 49.8 57.3 60.2 67.9 38.1 51.1 44.6 52.4 24.3 59.2 61.2 45.4 Act. 39.4 50.7 46.2 48.8 Act. 53.4 53.9 part-time Number Female ^[2573] Exp. Exp. ^{[273] [293] [254] [278] [319] [696] [218] [207] [254] [298] [371] [568] [399] [323] [149]} for the Midwestern ^[1683] Act. Diff. Z Act. Diff. ^{[258] [206] [551] [234] [296] [398] [215] [195] [194] [184] [114] [183] [190] [117] [50]} 674 32.6 145 12.4 890 32.3 104 22.6 ^{[103] [108] [156]} 170 13.0 45 7.4 98 13.5 ^{[60] [89] [99] [71]} 90 10.9 15.4 ^[72] 61 9.7 89 9.7 11.0 11.7 11.4 14.3 9.4 9.9 8.2 Z credence the little even entitled was not hiring statis- the EEOC’s As it did with it, however, gave pro- court the EEOC’s tics, criticized subject to the (1) adjustment sta- line product levels: statistics two motion of Sears’ assumptions court made criticisms the false same based tistics were employ- adjustment sales Feeder Pool” those “Division noncommission that all assign this promotions, pool for therefore ees, composed data.32 We promotion who determining were qualified weight equally no adjustment were from non- promotions erred deter- equally interested whether (2) selling; even successfully to commission produced mining that Sears statistics promotion if the EEOC’s to undermine probative sta- those valid, disparities created the disparities. The district promotion EEOC’s or even reduced significantly were tistics analysis tended recognized that court dif- adjustments for after Sears’ eliminated and disparities, the EEOC’s to overreduce interest, adjustments. other and in ferences adjustment the only. gave apparently adjustments that principle the weight for the that burden do not think We disparities. reduce the could adjust to the EEOC initially was on and qualifi for interest statistics promotion re introduced data After Sears cations. Interest how qualifications, interest and

garding regarding general evidence Sears’ litigation careful ever, EEOC used the interest and women’s in differences men’s inter show attempting to that strategy in selling that we discussed in commission little have qualifications would and est hiring applies claim context of the the at least would disparities, or bearing on its also as well. Sears promotion claim the question is wheth them. The not eliminate promotion based EEOC’s data adjusted the disparities remained statisti er the EEOC’s The district measurements. on interest taking into account significant after cally analysis of that Sears’ court concluded in men’s and adjustments for differences that of that evidence “demonstrates some selling, for interest women’s dispari for the alone can account interest and con qualifications, other differences analysis.” EEOC’s computed under ties and (such action as affirmative siderations 1326. The EEOC at testimony). victim of individual the lack however, it did appeal, that contends gave that the court Initially, we note “major adjustments for take into account “Product Line” to a credence some significant dis statistically and interest” analy- promotion the adjustment of EEOC’s part-time for both parities still remained expected reducing the sis resulted period is throughout and full-time promotions from proportions of female sue. from to commission sales noncommission the record analyzing After analysis 1974- pool for the division EEOC’s concerning interest parties’ evidence both period from for full-time 68.6% the dis- conclude that promotions, part-time for full-time to 60.8% 73.8% err determin- trict did part-time. adjustment That and 64.5% analysis weighted proportion of product pool analysis line the division feeder 32. In both promotions expert from divisions within analysis, different product Haworth line line, product the court adjust expected proportion found same because attempted to analysis since the product was valid by determining propor- line promotions female product adjustments in the promotions for divisions same that came from a of female tion particular experience. weight- might Accord- reflect related product line and line ing division or product explanation line according proportion Haworth’s ing proportion to the analysis, was not confined promotees from that "feeder" that came total pro- weighted product line instead product line. The district criti- same division cized feeder portion promotions different pool analysis division feeder respect product it is constructed lines. had not the court found that Sears shown analysis, Pool just Feeder experience like the Division divisions was related feeder subject as selling same criticisms required is analysis. type of in the division thus opening. Apparently the court believed ing substan- in interest can that Sears’ interest evidence differences account for dis- (and tially parities. reduced indeed almost eliminat- ed) alleged dispari- promotion the EEOC’s adjustment EEOC’s other interest ties. quantified least was based on some data first the EEOC’s contention We address regarding different interest levels. The major adjust- that it did take into account responses EEOC as well Sears relied on significant ments for interest and that dis- to the 1982 National Nonsupervi- Timecard parities adjustments. remained after those sory Special Survey (NTNSS), which was points in- purported us to two administered to incumbent noncommission adjustments. adjust- In the terest first salespeople by and commission Sears in ment, picked years, two expert 1982. The analyzed Siskin expected pro- and estimated the responses question 13C of the NTNSS33 portion promotions if of female noncommis- quantified to achieve levels interest and 1.5, 2, sion salesmen or 3 times as up came with a 1.78 ratio male to female interested commission sales than were full-time noncommission interest in com- noncommission saleswomen. For the 1973 selling, mission and a 1.56 ratio of male to figures, expected proportion of female part-time female adjusted interest. Siskin reduced, promotions was but it was still original expected the proportions female greater proportion than the actual at all promotions based on these interest re- figures, three interest levels. For the 1977 sponses disparities and found that the disparities except propor- existed the when reduced, significantly although the EEOC tion of noncommission salesmen three was argues that statistically significant dispari- proportion times the of noncommission ties remained.34 disparities saleswomen—then the between The adjustment EEOC’s NTNSS interest proportions expected and actual fe- disparities expected reduced the between promotions male were eliminated. proportions and actual pro- of female adjustments meaningless These in a motions If adjustment one-half.

vacuum, The however. EEOC does not valid, were considered the EEOC is correct point any to evidence on it based its arguing in that there would nonetheless assumption proportion of interest- disparities remain that are above the statis- 1.5, ed noncommission salesmen was tically significant z level of 3 for both full- great proportion or 3 times as part-time promotions. time and For a vari- interested noncommission saleswomen. reasons, ety however, think the dis- we might The district court well have conclud- trict court did not in err not credit- ed, based on Sears’ evidence dif- ing adjustment by this interest interest, ferences in male and female that determining statistically and thus not that more than three times the proportion significant disparities remained even after noncommission salesmen as of noncommis- taking interest into account. sion saleswomen was interested in commis- selling. Indeed, First, adjustment sion was NTNSS administered probative argument seems of Sears’ applied that but the EEOC that 1982inter- Question you original expected proportion 13C of the NTNSS asked: “If of female salesper- asked were son would to become a commission promotions part-time from noncommission to you accept?" part-time commission sales based the divi- on pool analysis sion was 70.5% in all territories original pool analysis 34. Siskin’s division Territory, but the Midwestern and 78.7% in the promotions full-time from noncommission to Territory, compared Midwestern with the actual period commission sales for the to percentages of 46.2% in all territories but the yielded expected percentage pro- of female (z 32.3), Midwestern level of 57.3% 68.8%, compared motions of with the actual (z 32.6). Territory Midwestern question level of After the 43.1%, percentage of a z value of 32.9. After adjustment, expected propor- 13C adjusting responses for interest on based to 64.9%, part-time promotions tion of reduced to 13C, question expected percentage of female for a z level of promotions 23.5. years, was reduced to 55.4% for all for a z value of 19.3. on these relying is if the EEOC years all year to each est ratio statistics. found, to 1980. deter- previously however, have problem with biggest Perhaps the clearly errone- finding was it relied is that adjustment mined that interest changed sub- 13C of ous, question interests responses that women’s on It found NTNSS, the 1970’s. the district court stantially the decade whereas over survey,35 the re- inter- 42 of question the EEOC’s after significant that in its relied on sponses to which promotions, for full-time adjustment est indicator a better adjustment, was interest disparities declined significance of selling. The in commission of real interest dispar- in 1979 the during so that the 1970’s question 42 was determined signifi- statistically (less than ity 2.9 in an- of interest because measure better 3.6 z level was cant), respondent had swering question, dis- statistically significant). (barely selling commission affirmatively choose changing inter- noted women’s trict court selling. Ques- forms of than other rather interest addressing some ests hand, merely re- 13C, other on the tion important finding it adjustments, whether respondent to answer quired the sales- 1982-[19]83, noncommission “even sales accept a commission he or she would likely as noncom- three times as men were Sears’ interest if offered. position it were interested to be mission saleswomen data promotion of the EEOC’s adjustment sales.” Sears moving 42 reduced question responses on based added). (emphasis at 1326 pro- proportion of expected female incongruous a bit It is also virtually eliminate so as to motions inter- rely the NTNSS on EEOC chooses fe- expected and actual disparities between dis- appeal, because in adjustment est promotions both proportion male court, the reliabili- the EEOC criticized trict the “all part-time cases for and full-time survey. The that responses to ty of the 1980.36 period of 1974 to years” sales- argued that noncommission question the argues that The EEOC adequately be identified not people could in statisti- results adjustment still interest the survey, and that responses and disparities full-time cally significant was to reduce imprecision effect of that (the z levels promotions part-time 1973 promotions proportion expected female part-time and 11.5 for 6.1 for full-time were The district adjustments for interest. after the court 1973). think that We do not although recognized that there court sug- reaching conclusion in not erred design and administra- above, some “flaws As we gested by noted EEOC. NTNSS, significant are not they tion of the changed the 1970’s. over interests women’s validity the essential enough apply undermine to questionable to It is therefore II, results,” F.Supp. statistics. promotion levels to 1973 interest specifically that Sears was stated expert determined also Haworth Sears’ as one identify noncommission interest levels changing to women’s adequately able applied the 1982 why im- she seem that had It reason salespersons. would earliest for the promotion against the data to statistics to cut also tend precision would compared actual part-time. with When listed 26 Question different the NTNSS 35. EEOC’s respondents: promotions from the types proportion and asked of timecard headquarters Ti- "Considering analysis, all the field for full-time 43.1% below, assignments during mark listed part-time mecard case or 57.3% for the 46.2% you assignments to be that want 1973-1980, three disparities Timecard years it is clear that the original). (emphasis in for next." considered year- not conduct Sears did are eliminated. with analysis, done by-year the EEOC had adjustment question re- Sears’ interest 13C, not think question did because Haworth pro- proportions expected duced the female early years with 1982 compare appropriate to (1973 figures years to 1980 motions for changing levels interest there were when data expert Ha- of Sears’ included because were not during the 1970’s. reliability pro- of those as to the worth’s doubts figures) for full-time 53.6% to 40.1% motion Furthermore, percentages presented Ha- sulted in new issue. both years court, may expert Siskin ac- we not consider and the EEOC’s that worth analysis appeal. on knowledged underlying the 1973 that data unreliable.37 promotion statistics were Perhaps recog- because the district court imprecision nized some in the NTNSS sur- also contends that the vey responses, rely the court chose to on noting implications erred in not court expressed the evidence of interest in re- analysis” responses based on a “combined sponses Aspiration Question- to Career question 42. The questions 13C and (CAQs) naires administered Sears at approved EEOC notes that applicant stores in the EEOC’s nonhired suggestion perhaps Sears’ witness’s sample in 1982-1983. The court concluded of interest would the best measurement analysis question- “Sears’ of these by combining responses ques achieved naires demonstrates that interest alone can presents 42. The then tions 13C and disparities computed account for the under results adjustment of its Division Pool analysis.” analysis.” Al on this “combined based at 1326. The EEOC attacks the court’s may helpful though it have been to have reliance on conclusions from these analysis, this had the benefit of combined questionnaires respects. Al- several may not consider these statistics on though troubling aspects we find some appeal presented were not data,38 agree reliance on this we do not the district court. The EEOC contends adjust- the district court’s conclusion that adjustment this that we consider be virtually ments for interest can eliminate accepted cause the district Sears’ disparities the EEOC’s is erroneous. testimony witness’s method, and that the would be best EEOC contends that results analysis through EEOC has done the “sim CAQs from the are unreliable because ple fully explained.” mathematics expert Haworth combined full-time simple inap Even the results of math are part-time responses regarding interest ap propriate for this court consider in commission sales to achieve the results however, peal, if the not ini results were proportion of men interested tially presented to the district court. approximately See commission sales was three Ohio-Sealy Manufacturing Mattress Co. proportion times of women interested Inc., 650 n. 1 Sealy The district court commission sales. rec- Cir.1985) plaintiff’s argument (rejecting ognized this criticism determined that “simply that because summaries were tools part the “there is no indication that time understanding in the to aid court responses fairly reflect female do not inter- [the] record,” appeals court of could consider est in commission sales.” Id. We do not “summaries of voluminous materials” al reasoning faulty. this The believe that though presented they argues had not to the aspirations been EEOC that career of court); Vogel, part-time salespersons district Rebuck v. 713 F.2d full-time and differ (8th Cir.1983); 10(a). persons Fed.R.App.P. product in those lines which usually the re- are distinct. Commission Because EEOC’s combined work expert questionnaires were not adminis- 37. Sears’ Haworth testified that there states that the stores, promotion argues were no counts for 1973 that she at and the EEOC that tered those expert The found to accurate. EEOC’s Siskin gave explanation because Sears no for not dis- (on figures that stated 1972 hires which the stores, tributing CAQs the those Sears must year-end part) promotions counts for were based in doing improper in not so. have had motives "at best an were estimate.” the reason Sears have had for Whatever facilities, administering surveys the at those not complains CAQs 38. The EEOC were CAQs unreliable we will not assume that the are among distributed to two facilities 29 of the included. Twen- because two facilities were not Chicago sample that or were included, ty-seven and the EEOC facilities were groups, traditionally Detroit where there had possible speculate im- would have us higher representation female in commis- been (because pact that were of the from the two facilities part-time data sales there more sion commission sales hires in those were stores). not included. full-time sales, from of which were four not available are at Sears positions sales Although salespersons.40 however, noncommission salespersons, only to full-time had helpful to have have been it would in interest commission of question thus the not con- analyze, we are responses to more the line between across to cut sales seems responses number the small vinced part-time. full-time inter- not indicative question to this the suggests that also light sales, especially est CAQs on the results relied court respondents significant numbers of the because from the NTNSS results not the men) who (63.4% of women and 50.1% totally did not of the NTNSS results question of whether “no” answered while disparities promotion eliminate pro- transfer interested were As noted CAQs did. we of the the results immediately preced- question This motion. reliability questioned the earlier, regarding inter- question headed the ed and grounds, on several of the NTNSS See D. Baldus & commission sales. est in reliability of questioned even Discrimina- Statistical Proof of Cole, J. Furthermore, did the court the NTNSS. (“Sample 4.123, (Supp.1986) tion § NTNSS; the of the the results not discount course, validity of actu- size, affects the results found that court instead data, not lead applicant should al corrobo- to and similar were NTNSS altogether. Even disregard it the court CAQs. clear It is results of the rated the applicant flow sheds sample of data a small on sole place did not reliance that the court oper- [employer’s] actual light on the some survey that one interest CAQs as the ation.”). disparities. promotion eliminated CAQ argues applying surveys, other two court found still ratio 3.19 to female interest NTNSS, male Survey and the 1976 Job Interest As noted disparities in 1973. we results reliability some although they involved survey regarding applying NTNSS above results on problems, yielded similar promotion to the 1973 from 1982 results Furthermore, question of interest.39 apply results of data, questionable it is factors aside expressed several 1973 statistics survey data to or 1983 to determine from interest caused it changed interests had women's promotion disparities did the EEOC's addition, the 1973 during the 1970’s. discrimination. prove a case of intentional based unrelia- were promotion statistics factors, in more discuss Those which we promotion data. ble later, qualifica- include differences detail women, Sears’ af- men and tions between reliability problems considering the Even efforts, and a lack of indi- action data, firmative survey interest of some testimony. determining vidual victim not err the court did think if significantly, the interest reliability questions the also totally, promotion reduced not almost CAQs sample on the size based disparities. CAQs dis- responses. The CAQ 1,220 357 male full- female and tributed Qualifications salesper- part-time time and noncommission of differences number, Regarding issue fifty there were sons. Of that among male and female qualifications in commission responses of an interest language as “twice precise more such alleges evi- court used The that court 39. Sears misunderstanding of the ... as percentage of female male.” denced mathematical survey surveys language other data. is not al- morale results of The point, be, notes that one it is clear ways precise could but as as it were salesmen that noncommission concluded two judge understood the context that than noncommission more interested times significance percentages. of the take into account but did not saleswomen noncommission sales women were 75% of the pool, respons- percentages male and female 40. The numbers, two more than so that in actual 6.7% of constituted es on these numbers based many men were women interested times as in as responses responses of female 2.1% male expressing nitpicking argu- is a sales. This commission selling. in commission an interest opinion, district other times in the ment. At expert testimony Sears cites witness salespersons, 10%. noncommission regard- claiming expected that it is not “all court concluded males at performance shows have lower ing qualifications would women commission qualified for more were Sears rates than men. Id. females.” than positions sales [were] year It in their first after is true court did not complains that the outperformed men women being promoted, rely purported to what evidence state year from 1974 to 1980 the full-time each conclusion, and as- reaching this upon overall, top salesper- and in the case 10% that the court was along sumes 10%, performance In sons. the bottom salespersons’ referring to noncommission equal, relatively of men and women was entering background before experience and outperformed women although men still par- Both at Sears. sales noncommission case, part-time there was a slightly. In the for evidence dig into the record ties clearly outper- sample, men smaller prior proposition that support the refute or top usually in the formed women 10% impact on chances experience can have per- outperformed overall. The women sales to promotion from noncommission part-time in the case were formance rates commission sales. data, Based on this the same. about aspect of only one experience is Prior judge clearly erred in say cannot however, and we believe qualifications, usually possessed determining that men perform- referring to the district court qualifications for commission more salesper- noncommission aspect of a ance (even argues, selling though, as the EEOC promotion to com- qualifications for a son’s in this data wheth- Sears has not identified discussing data on mission sales. come from noncom- promotions er the have context, hiring performance in the sales sales). go criticisms mission The EEOC’s on sales to Sears’ data also referred hiring figures, they if to the and even salespersons by commission performance data, we do not applied promotion to the promotion. during year first after their the essential validi- among they undermine full-time believe The court found that salespeople, figures. the sales ty We conclude that part-time commission of those males exceeded inferring rates of the performance from these judge did not err The court deter- of the females. the rates usually pos- figures that men performance pro- Sears hired and mined that the men for com- qualifications more sessed more commission sales were moted into jobs than did women. mission sales inferred that salespeople, and successful qualifications more of the true had S. Conclusion-Promotions did the women. sales than re hires and The district court’s conclusions true for all new court found this incorpo com- among top also promotions, garding promotion 10% claim saleswomen, and missions salesmen and regarding the absence rated its decisions of commission among bottom 10% af testimony and Sears’ individual victim salespersons. previ We have action efforts.41 firmative these considerations ously determined that EEOC, perform- response of the claim justified in the context context, hiring states that in the ance issue discrimination, no reason not hiring and see pro- of the hires or only the bottom 10% promotion context apply them to the important to a discrimination motions are Indeed, absence we think that the well. if did discriminate claim because more testimony is even individual victim women, expected against it was to be promotion understand difficult performance have lower women still should average top in context. the than men on rates in we note promotion For the reasons claim. incorporated claims The EEOC

41. claims, reject practices those subjective employment alleged our treatment these discrimination, alleged past which we dis- well. promotion context claims in the claim, hiring in its the context of the cussed in 340 wage with- discrimination bring a claim reli- some may have been Although there of the meeting equal work standard out adjust- interest in Sears’ problems ability argues that the district EPA. The in- not whether ments, question is case was mistakenly held that this court totally eliminate alone adjustments terest Gunther, purview of not within question The disparities. promotion prove that requiring in it to erred therefore in clearly erred court rather whether substantially equal had and women men adjustments determining interest that pay discrimination in its Title VII jobs such as along considerations with other court accu- believe the district claim. We lack of qualifications, in differences Gun- scope of recognized the rately limited af- testimony, and Sears’ victim individual correctly ther determined any negate infer- efforts firmative action scope. fell outside case against discrimination of intentional ences Gunther, sell- courts determined to commission Before promotions in women harmoniously Title VII to construe ing. conclude We wage addressing discrimination EPA in clearly err. did not alleging claims, plaintiffs discrim VII Title equal to meet the pay would have ination DISPARATE III. v. Plemer See EPA. of the work standard TREATMENT —WAGE Parsons-Gilbane, 1127, (5th 713 F.2d DISCRIMINATION Cir.1983); see, e.g., v. Chamber DiSalvo again al- attempted prove, The EEOC Commerce, 593, (8th F.2d Cir. (multiple regres- solely by statistical most Son, & v. Frank R. MacNeill 1978); Orr sion) discriminated analyses, that Sears denied, (5th Cir.), cert. 166, 511 F.2d in checklist employed against women L.Ed.2d 94 S.Ct. less by paying management positions them Co., v. Zia (1975); Ammons jobs. em- similar Checklist men with than Cir.1971). 119-20 are salaried

ployees at executive Gunther, guards prison could female and administra- management, professional, of the equal work standard not meet the employees. tive guarding jobs involved EPA their originally Although the EEOC prisoners than substantially female fewer under wage discrimination claims brought guarding prison- male prison guards male VII, it Equal Pay Act and Title danger- ers, prisoners both less female were Equal under dropped its claims Supreme later male inmates. ous than however, issue is whether Pay preliminary significant, Act. A Court found correctly held that compara- guards alleged the district court that a the female equal Washing- had to meet the in this case study commissioned ble worth Equal Pay guards Act42 to of the County work showed female standard ton guards Title VII sex discrimination male prove paid a case of of what should be 95% alleged guards claims that also pay. paid. The female were incorrectly interpreted County paid of what the 70% Gunther, 161, 101 received, this differ- Washington guards and that U.S. male (1981), constituted intentional in which ence in treatment 68 L.Ed.2d S.Ct. held that in a VII sex discrimination. Court Supreme held that Title Court case, guards did not have the female such a plaintiff in certain circumstances could opposite Pay ployees sex such establish- Equal equal standard work 206(d)(1) performance equal work on § is set forth at 29 U.S.C. ment Act of 1963 effort, skill, part: requires equal and re- provides of which performed sponsibility, are under and which (d)(1) subject employer having employees No conditions, except working where similar any provisions of this section discrim- shall (i) pursuant inate, payment a se- is made any such niority such establishment which within (ii) (iii) system; system; a employees merit employees employed, between earnings by quantity system by paying which measures wages em- on the basis of sex (iv) production; quality or a differential ployees less establishment at rate such factor other than sex. pays other wages to em- based than the at which he rate Gun- analyzed This court the contours of equal work standards to meet the ther American Nurses’ v. in Association sex discrimination prove EPA to intentional Gunther, Illinois, (7th State pay Title VII. 720-22 under of 180-81, Cir.1986). emphasized at 2253. at We the narrowness holding, stating of that that court argues that the district mean, that seems to [Gunther ] [a]ll equal refusing relax the EPA erred in to out, dissenting pointed Justices is Gunther applying to by not work standard showing equal “that even absent a of Gunther reading of this case. Our work, there is a cause of action under that interpretations of case subsequent Title VII when there is direct evidence us, however, convinces intentionally employer that an has de- correctly determined that this Title pressed salary a woman’s because she is does not fall wage claim VII discrimination a today woman. decision does not Gunther. scope of within the approve a cause of action based on a not, as the EEOC The district court did comparison wage rates of dissimi- by Gunther suggests, limit to its facts jobs.” lar Gunther recognizing court was Gunther, Id. at 721 (quoting 452 U.S. at scope of its own deci- careful to limit the (emphasis origi 101 S.Ct. at 2265 passages several sion. This is evident at nal)). We also discussed American Feder Gunther opinion. The Gunther State, County Municipal ation outset the narrowness “emphasize[d] at the Employees Washington, 770 F.2d 1401 case,” question before [that] [it] (AFSCME), Cir.1985) example as an prison female which was whether claim, a case in which there was insufficient evi guards’ based on direct evidence that depressed wages “their dence to establish Title VII sex discrimina discrimination,” (not com- wages intentional sex tion different work. claim), precluded pro- parable worth their AFSCME, paid prevailing the state ceeding Title VII because could under wages employees market rates as to its equal work standard of the not meet the commissioning comparative after worth Gunther, 8,101 at 166 & n. EPA. U.S. studies, employees jobs discovered n. 8. That intentional sex S.Ct. at & mainly by paid approxi- women were held Court, discrimination, according to the con- mately wages by employ- received 80% “setting wage scale for female sisted of predominantly jobs ees in male considered guards, guards, for male at a level but not comparable worth. The AFSCME survey of outside mar- lower than its own state, alleging that the state dis- sued kets and the worth of the warranted.” by having against criminated women Id. at 101 S.Ct. at 2246. The Court comparative compensation system based on opinion by emphasizing ended its “job worth. The Ninth Circuit held that wage decision did not threaten struc- comparable studies and worth evaluation allowing every employer by Title ture of statistics alone are insufficient to establish plaintiffs “ compare job content and VII requisite discriminatory inference of any job predominantly per- pay ‘between motive,” discriminatory it was any job predomi- formed women rates, wage pay prevailing salaries at ”men,’ nantly performed by because the “committed to and that the state was not require suit in that case did “not a court to compensation implement system a new subjective assessment of the make its own comparable worth as defined based on guard jobs, or of the male and female value Id. study” it had commissioned. attempt by technique or other statistical 1406-08. quantify the effect of sex dis- method key distinc- contends that the Id. wage crimination on the rates.” and Gunther tion between AFSCME 180-81, (quoting petition- at 2253 S.Ct. AFSCME, brief). employer had not ers’ Hay job after the Program Compensation study,43but worth comparable the adopted alleged or has not evaluations, it the adopted had Gunther, employer the in found, implementation evidenced paid that this proved Court (and, as the Gunther the motive or intent. discriminatory less than a rate at employees Sears’ women the argued valued district court study had in comparable worth The EEOC never case is this Ex- argues that implementation of The EEOC job). that Sears’ a had made dis- Program because was like Gunther Compensation ecutive form of the study in the comparable worth that court found The district criminatory. adopted the re- had Hay evaluations convincing- very “testified Sears’ witnesses Compen- its Executive study in sults of the establishing in factor was not a ly that sex the dis- do not think Program. We sation Program with Compensation Executive employer has on whether an turns tinction determining 1976, or in in Hay Associates in re- program compensation adopted a employ- pay of checklist at time study. That comparable worth sponse to a 1352. Fur- at ees.” Sears consideration, but part of be factor thermore, found no evidence court imple- is whether key distinction discriminatory intent. in program compensation mentation of argues also that study discriminato- reflects response to the show direct required it to improperly in Nurs- noted American we ry intent. As intent, because discriminatory evidence lacking in thing AFSCME es’, critical “[t]he provided by evidence the circumstantial decided [employer] that the was evidence to indicate sufficient statistics was work- wages particular not to raise in against women intent discriminate were those workers most of ers because later, not find do discuss we pay. As we orig- (emphasis in 783 F.2d at female.” court’s deci- clearly erroneous Parsons-Gilbane, inal); also Plemer see pay did statistics that the sion Cir.1983)(Fifth (5th 1127, 1133-34 agreed Even if we prove discrimination. significant that Gunther Circuit found discrimination, showed statistics employers unex- shown “it was type all sure that are not at objective pay crite- plainedly departed indirect, of discrimina- evidence statistical comparable a adopted”). had Thus ria sufficient under Gun- tory would dis- intent proving sex study is relevant worth was clear Gunther may provide the ther. evidence “it because crimination comparable straightforward. employer is forced occasion on which guard study his female valued women’s intentions results his toward worth declare Nurses’, 783 F.2d but wom- guard jobs, employees.” jobs American at men’s 95% guards’ Meth- 721; paid Merrill v. Southern men guards see also en were 70% F.2d University, emphasized odist The Gunther salaries. intentional Cir.1986)(court to find hesitated “direct such holding was limited to that its was not because evidence discrimination a case it was not evidence” and Gunther, i.e., plaintiff did like that subjectively would have to which a court to abide employer suggest failed and female of male the value determine (or adopted) objec- otherwise preannounced attempt by statistical tech- jobs44 “or to criteria). pay tive quantify the ef- nique or method other wage on the sex discrimination fect of not fit that this case does We believe 180-81, Gunther, 452 U.S. at rates.” Gunther within contours dissenting Justices S.Ct. at 2253. its Executive although implemented differing jobs. found that The district court legislation implement- passed The state later that, requesting just and we comparable ing wage system worth was based EEOC’s claim a AFSCME, finding say effect over a decade. erroneous. that would take cannot Furthermore, 770 F.2d at 1403. major this claim issue indeed sim- the checklist was whether argues comparable worth 44. The later, eventually the court we discuss ilar. As implicated and that there concerns are not here they were not similar. decided subjectively the value of is no need to assess great ex- depended bonuses to a creating a cause aries and saw the case Gunther generosity the discretion and tent on there is Title VII “when under of action Regarding managers at the local levels. in- has employer that an direct salaries, although salary there were some salary tentionally depressed woman’s managers usually negoti- guidelines, store decision to- she is a woman. because employees, ated with and made recommen- of action approve a cause day does not regional managers usually dations to who comparison of the wage rates on a based accepted those recommendations.47 Be- at jobs.” Id. of dissimilar policy paying person of cause of Sears’ J., dissenting) (emphasis (Rehnquist, on position, and not the salaries were based added). original and in including employ- of factors number correct- district court We believe salary, poten- prior history ee’s work did not show that the EEOC ly determined performance, company, job tial with of intentional direct evidence type of employee relo- and the number of times contemplated by pay in discrimination sex (the cated latter of which involved substan- Court, thus the EEOC the Gunther increases). salary Virtually all check- tial of the equal pay standard had to meet eligible employees list were for annual bo- Title VII sex discrimina- prove EPA to nuses, depended greatly the size which wages in claim.45 tion regional managers generosity on the compensation of Sears’ A overview widely,” brief as well as on the which “varied pay employees earnings. in checklist practices performance for and total employee’s help- is during period at issue of the positions The court found that all above influencing compensation of whether the EEOC ful to our discussion factors employees “caused substantial in its claim checklist equal met the work standard among compensation check- had differences wages.46 Sears discrimination sex employees of both sexes.” list systems for compensation two different period 1973 during the employees checklist Execu- introduced a new 1976 Sears content, 1975, job From 1973 to to 1980. Compensation Program, which the dis- tive salaries, significantly varied and bonuses response check- found was in to trict court regions and geographic among different perceptions system employees’ list high decentralization facilities due to Sears’ unwieldy” and re- “inequitable autonomy. store emphasis on individual system management’s desire for a flected job called “C” Initially, Sears had titles competitive in the market- that “would be codes, job not reflect those codes did in- equity, and reward place, have internal II, during because content 1973-1975 Sears performance.” dividual first, from the imprecise Program were was the F.Supp. codes at 1336. This managers assigned coopera- differ- years unit had two of Sears’ culmination of years evaluating Hay over the Associates and additional duties tion with ent used compensation practices. em- Sears the checklist Sears’ depending on work and Method of Hay sal- Guide Chart-Profile employees’ ployees’ talents. Checklist Gunther, 101 S.Ct. at 2253 U.S. at 179 n. the EEOC’s court believed that 45.The Furthermore, agree claim of un- that the EEOC’s fell within the traditional claim equal pay n. 19. equal work. The court went so wage more on the claim was discrimination attempted suggest that the EEOC far as avoid the standards equal unequal pay for work classic order of a proof in the traditional type exceptional/un- it was the claim than by relying because the on Gunther EPA claim EEOCwas usual situation Gunther. having difficulty establishing the sub- equality jobs at issue. We do stantial compensa- court discusses 46. The district court’s conclusions comment on the practices in more detail. See tion rely attempting the EEOC’s motives F.Supp. at 1334-37. agree general the court’s We do Gunther. Court, by that the Gunther allow- observations ing op- employees in credit for checklist 47. Salaries jobs remedy for women had a Title VII who exception. decisions Those were an erations men, appeared particularly to be never held (group of generally territorial made at the remedy providing a for claims concerned with outside the traditional regions) level. equal pay analysis. See change company formally did not actu- jobs. The manner Evaluation to evaluate actually job immediately upon imple- assessed al structures which evaluators determining menting program. whether The court found important standard, equal gradually work so we communicated EEOC met the “Sears point. on this go changes employees period into more detail over a will years.” at 1337. Job Hay-trained Sears Hay consultants and through an in- activities were standardized descriptions position employees wrote changes formal communication of to em- “average” per- of an on the duties based changes job ployees that involved evalu- job, position each former of the job categories. ations and creation of new members, drawn from Sears. Committee Sears, analyzed posi- various levels at The EEOC had the burden of meet descriptions systematically and evalu- tion ing equal Equal the work standard of the positions, of those at- ated the content Mercy Hospital Act. EEOC v. and Pay incorporate changes in tempting to Sears’ Center, 1195, (7th Medical 709 F.2d 1197 part job and structure which were content Cir.1983). provides That standard that an along way, system of the new so that employer may not discriminate on the basis actually analyzed positions were not by paying unequal wages “equal of sex for they performed or at the time were before jobs performance work on of which assigned of The committees evaluation. effort, requires skill, equal responsibili and know-how, position points each for ty, performed and are under which similar problem solving, accountability it en- and 206(d)(1). working conditions.” 29 U.S.C. § salary ranges structured tailed. Sears succeed, establish, To the EEOC “must compensated positions were accord- upon job performance based ‘actual and ing points. to relative evaluation Sears set titles, job content —not classifications or de ranges minimum, salary midpoint, with a scriptions’ performed that the work ... (going and maximum of 75% 120% Hospital, Mercy substantially equal.” 1976, midpoint salary) position in for each County (quoting Gunther v. F.2d at 1197 annually. and modified Washington, of The court found that even after the im- Cir.1979), aff'd, plementation Compensa- of the Executive (1981)). 68 L.Ed.2d 751 Program salary ranges, tion set and its performance The district court found that other factors such as evalua- EEOC, tions, attempting previous job history, prove of substantial number relo- cations, issue, location, seniority, geographic equality jobs and of the did not intro managers any job perform the individual discretion of re- duce evidence of credible garding timing salary Instead, size and of increases ance content.48 the EEOC re still influenced the salaries job descriptions of individual lied on the that emanated employees. checklist The Hay court also found from Sears’ Guide Chart-Profile Meth process phasing of job prove the new od of evaluations to substantial program years took several equality. because of contended policy reducing of employee’s Hay not factors measured Method— pay. Salary employees know-how, increases for problem solving, above and accounta they the maximum were cut until bility equivalent came EPA —were range. skill, effort, within the The court found responsibility, that not factors and only were salaries not standardized if jobs equal Sears received an number years” (for implementation points know-how, “several after problem solving, Compensation Program, the Executive accountability), place which would then 48. The jobs district court found that the evi- tween “C" codes for before was enti- regarding job dence introduced any weight tled if to little and was not credible testimony partner content was the of a former experts rely because the EEOC’s not on this did Associates, Hay Laking. of clearly Jon We do not find testimony Laking had reason to be biased findings erroneous the district court’s against Sears. testimony, discussing relationships that this be- Hay Method are such as range, evaluations salary in the same jobs those equality. The EEOC job probative of under “substantially equal” jobs be would court. has mischaracterized The dis- equal work standard. EPA’s actually found that relies found, therefore “EEOC trict categories po- job Hay [although evaluation own on Sears’ developed by commit- proof description in 1976 as completed and codes sition determining helpful were job code tees persons in each would substantially equal, substantially work.” are similar whether performing proof not a substitute II, F.Supp. at proof Without performance.... actual comparing jobs method of The EEOC’s content, job performance of actual different. was somewhat from 1973-1975 descrip- Hay position evaluations rely recognized that it could The EEOC skill, equal inadequate prove tions are jobs, assigned to various codes on the “C” responsibility. effort indica- codes were not valid those at 1342. thus job content. tors sup codes, cases in cites job numerous from the worked backward job contention that titles and port of its “correspondences” be- develop trying to probative job equality descriptions are job codes. and 1976 job codes tween “C” even without corroborative many job codes Because there recog performance. We job job or content code, process one “C" associated with sug language in the cases cited nize that involved determining correspondences of descriptions job are gests that titles EEOC, by the based subjective judgments “may, some instanc probative and even in among with- jobs patterns of movement es, equal work be highly probative of be Many any knowledge job of content. out respon expect actual cause one would that left out of the 1973-1975 employees were extent, would, conform to to some sibilities analyses the EEOC worked back- description.” Epstein, 739 F.2d at job wards, job of comparing the 1976 codes Coun 6; see also Pearce v. Wichita 277 n. or promoted been persons who had not (“Al 128, (5th Cir.1979) ty, 590 F.2d 133 codes of job “C” transferred in 1976 with to some though are entitled job titles To be year-end in 1975. persons those comparative in assessment of weight 1974, example, employee an counted comparison of actual responsibility,” job code in the same “C” had to have been controlling.); Brennan v. Owens duties is in 1974 and in 1975. County Hospital, 523 F.2d boro-Daviess court found that The district Cir.1975) (“Although (6th job 1013, 1017 methods sub- prove by the above failed to accorded as descriptions should not be any jobs equality of the at issue stantial duties given to the weight much as that finding may not disturb this year. We in deter by employees actually performed that it is erro- we unless determine substantially mining jobs two whether United Epstein Secretary, v. See neous. nevertheless, that equal, we believe Treasury, 739 Department States of the de helpful, particularly where may be Mercy Hospi- Cir.1984); 274, F.2d 277 compared are jobs to be scriptions of tal, Because we have F.2d at 1198. 709 very em by the and were written similar and are not “left evidence reviewed all the wage differentials ployer who claims that that a and firm conviction the definite criterion.”), impermissible based on are not committed,” United has been mistake 973, denied, 96 S.Ct. rt. 425 U.S. ce 2170, Co., Gypsum States 333 v. United States Brennan (1976); L.Ed.2d 542, 92 L.Ed.2d U.S. Co., F.2d Trust Bank & Victoria (1948), court’s affirm the district (“the controlling (5th Cir.1974) factor not meet the finding EEOC did that content, job job be ... has equal work standard. employer, prepared by description [but] rating not] argues employer’s [should The EEOC employer ignored simply because job as a matter law improperly held where, here, argues producing individ- the rat- The EEOC it, especially made study comparisons job performance of com- extensive ual ing came after an degrees responsi- parative qualifications, at trial was “not feasible” consider- content bility, of loss Bank.” ing and chances record and that the trial the size original)). (emphasis in court’s dock- took six months of the district excuse, unconvincing time. This is an et cases, however, supports None of these body especially light of the substantial job titles the EEOC’s contention involving equal work stan- of case law alone, any without corroborat descriptions invariably gone courts have dard which job job perform or ing evidence of content analyze, often beyond job descriptions to case, ance, equality. In each prove job can job job perform- in-depth, actual duties and recognized job content was the court implies might ance. The EEOC controlling it was factor and believed the court’s time introduc- have wasted necessary go beyond job titles and/or might actually ing The case such evidence. descriptions compare and contrast actual effective, how- Epstein, have been shorter and more See job content. duties and 278-80; Pearce, 133-34; ever, right if the EEOC had introduced the 590 F.2d at F.2d at Owensboro-Daviess, 1015-28; kind of evidence or tried the case accord- Bank, out, precedent. 493 F.2d at 899-900. As it turns Victoria ance with cases, appeals went some the court of failure to introduce evidence through the differ job job performance an extended is of actual content job con wages ences and similarities duties and fatal to its sex discrimination Owensboro-Daviess, 523 F.2d at See tent. light claim in 1025-28; hospital (comparing duties of job differences in content. Bank, orderlies); Victoria aides appears suggest that Sears (contrasting F.2d at 899-900 duties of ex showing inequality had the burden of see also tellers); change note tellers and job argument line of content. This v. Kenosha District School Unified Ep- recognized in similar to that which we 1,No. 1220, 1225-26 (7th Cir.1980) 620 F.2d stein, would, it 739 F.2d at 278: “Plaintiff predominantly (contrasting duties of fe seems, equal us infer work from the have custodians); predominantly male cleaners and male prove defendants’ failure to otherwise.” Institute, Mary ner v. 613 Hor argument responded We that (8th Cir.1980) (contrasting F.2d 713-15 ignores elementary the fact that the bur- physical duties of male and female edu proving prima den for the facie case is on in Epstein, teachers). cation As we noted plaintiff. plaintiff point- the The has not 6, “[sjimilar job descrip 739 F.2d at 277 n. evidence, any except job ed to record require finding not of tions alone ... do descriptions, tend to show which would equality jobs substantial of under work, equal and on our review Indeed, Equal Pay Act.” identical or sim agree court record we with the district job deceiving ilar titles can because be any that the record is devoid of such underlying do not reflect the actual duties evidence. Horner, the title. See 613 F.2d at 714-15 Epstein, Id. Unlike the defendant in how- identical,” (although “superficially male ever, of differ- Sears did introduce evidence physical jobs and female education not sub job ences in The found that content. court stantially equal respon in terms of skill or through “Sears has credible testimo- shown Angelo see also v. Bacharach In sibility); job Co., ny that there were differences in con- (3d strument 1172 Cir. employees job code.” 1977) (“ tent of with the same ‘[mjechanical and surface similari F.Supp. 1342. The court inadequate equali ties’ are to establish the Hay also found that Sears showed the ty positions” (quoting Cong.Rec. of two (remarks position de- Frelinghu Method evaluated “idealized (1963) Rep. of Bank, changes actuality job Victoria ysen))); scriptions,” but in 493 F.2d at 899- (duties phased period years in of after dissimilar between bank’s note were over a tellers). Compen- exchange implementation tellers and of the Executive job de- whether determine not needWe in indeed, differences and Program, sation such as joba evaluation on scriptions in based ear- existed performance and content job alone, any without can findings Hay Method these Neither of ly 1986. con- job of actual corroborating evidence clearly erroneous. equality, be- job tent, substantial establish that court concluded The district unrebutted light of Sears’ hold we cause evi- to Sears’ respond not did content, the district job unequal of evidence job content differences of dence sus- had not finding that or content court’s performance job of evidence substantial proving signifi- of burden not its tained differences show burden erroneous. not is not sustain equality did cant, the EEOC jobs of equality substantial proving of descriptions job Assuming in the support ample findWe analyzed. face of in the value persuasive rule is have could conclusion. for this record of differences evidence application credible “the “well established” agree with content, actual job on depends standard equal pay resulting descriptions job performance.” job and requirements job Co., 555 job evaluations Hay Method Instrument v. Bacharach Angelo probative Cir.1977); 51 Fed. lacked (3d see this case circumstances F.2d at 29 equality” (to codified (1986) of “substantial 29,822 as evidence Reg. value 1620.13(e)) ("Application standard. equal work EPA's C.F.R. under § job on dependent is standard equal pay jobs Method, evaluates Hay depends rather or titles classifications “accountability,” “know- on basis perform- requirements job actual solving,” “does how,” “problem test EEOC asserts ance.”). The statutory in terms jobs measure argues that we equality, substantial effort, responsi skill, standards [EPA] distinctions “overly fine not make should working conditions.”49 equal bility, and v. South Brennan issue.” the tasks F.Supp. Co., 464 Penney v. J.C. Marshall F.2d Hospital, Community Davis 1979); Wheeler (N.D.Ohio see 1166, 1191 produc- Cir.1976). EEOC has 1050,1052 Corp., 471 Armco Steel perform- job content evidence ed no evi “is not (Hay system (S.D.Tex.1979) us to draw permit even would ance skill, effort being equal job of a dence Such jobs. between distinctions regulations Federal responsibility”). have allowed would may have ... jobs fact that “the state Pay Equal terms compare evalua under value point total same effort, responsibili- skill, Act factors does employer by the in use system tion job no introduced the EEOC ty. Because concerned jobs the mean that in itself not tous asks evidence, in essence it content terms the according to equal are job differences the speculate whether (em (1986) 29,822 Fed.Reg. 51 statute.” minor or are by Sears out brought content 29 C.F.R. added) codified (to be phasis the do so when may not significant. We as appeal, the 1620.13(e)). On § proving the of burden has the “virtually are Hay factors the that serts of equality the the substantial factors,” cites but statutory identical would evidence Job content issue. jobs at attempted where record nothing in the Hay Meth- whether indicated also have factors, showing sets of equate the two solv- problem accountability, of od factors Hay evaluation point values how the actually equivalents ing, and know-how analysis equivalent be could skill, system ef- of Pay Act factors Equal of statutory factors.50 using the fort, responsibility. equali- job however, suggested it establishes that Hay acknowledged Method have 49. We ty- job evaluation. accepted "widely method" is a 458, Corp., F.2d Radio Zenith McGrath record portions of the points to various denied, Sears Cir.), (7th cert. Sears] [and not, witnesses in which (1981). have We 70 L.Ed.2d S.Ct. addition, similar.”); district court found that only Horner v. Mary Insti- cf. Hay tute, Method was an inaccurate means (8th 613 F.2d Cir.1980) (if measuring job equality “in a plaintiff number of has failed to establish substantial instances” in this case because of equality, any wage various disparity is irrelevant by errors Sears in the process. evaluation Equal Act). under the Pay note, We also II, F.Supp. however, at 1342. These er- that the district finding court’s rors, employees which resulted in classified that the EEOC had proved not sex discrimi- into job codes that did not correspond to wages nation in through its multiple flawed duties, their led the court to conclude regression analyses is clearly not errone- EEOC, who relied on faulty Sears’ ous. analysis of these jobs, proved had not sub- Although expert EEOC’s Siskin equality jobs stantial in four of the performed multiple two regression analy fifty-one job say codes. We cannot ses incorporating different sets of varia finding clearly erroneous. bles, he regression relied analy one The failure of EEOC’s job reliance on sis to support his signifi conclusion that descriptions emanating from Hay Eval- wage cant disparities existed. The district uation Method in proving equal- substantial held “EEOC’s analy statistical ity is years exacerbated in the 1973to 1975. ses so flawed that [were] lack[ed] Sears’ label workforce,” “artificial used to persuasive value.” describe the results of the tracing EEOC’s at 1342. We reiterate that the court’s find method, back altogether is not inappropri- ings regarding statistical evidence are ac ate. The record amply supports the dis- corded even more deference typical than trict findings court’s “correspon- findings fact, subject which are dences” between various “C” code and clearly erroneous standard of review. So 1976 job codes patterns were based on Ozinga Brothers, ria v. Inc., movement taking job without content into 995 n. 6 Cir.1983). consideration. The EEOC introduced no appeal EEOC states on that the dis- other credible job con- trict court’s conclusions regarding its re- years, tent for those and because of the gression analyses “an exhibit irrational be- “tracing method, back” substantial lief that proof statistical is not adequate to (those numbers of employees who transfer- model ‘complex’ making decision red, Sears, left or promoted, even to process.” disagree. We We believe that “corresponding” jobs), were excluded from the correctly district court recognized the analysis. The court did err regression limits of analysis. Our review in concluding EEOC failed to of the court’s careful and considered dis- prove substantial equality during 1973- cussion EEOC’s statistical evidence 1975, and even if it prove could substantial reveals that the court overawed equality, analysis failed to create rea- the results of the analysis, statistical sonable inference of a pattern nationwide instead had a healthy distrust of ability practice of sex discrimination Sears. regression analyses explain Our of the EEOC’s claim Title factors management involved in salary de- YII sex wages discrimination for check cisions. skepticism Such “may justi- list employees at Sears can end here be fied.” Johnson, Coates v. Johnson & 756 cause the EEOC has failed satisfy to 524, (7th F.2d Cir.1985); 539 see Lewis v. proof burden of regarding equal the work Mills, Inc., Bloomsburg 561, 773 F.2d 578 standard. See Spaulding (4th v. University Cir.1985)(remanding case for a more Washington, 686, (9th 740 Cir.), F.2d 700 complete factual finding, court stated denied, 1036, cert. 469 511, U.S. was “not suggesting] that there may not (1984) (“We 83 L.Ed.2d 401 will not plausible ... be a basis for rejecting ... statis- infer intent merely from the existence of tical evidence or finding ... it not suffi- wage differences between that are ciently probative claim”); of the Mozee v. "accountability” that

testified aspects measured differ- job ent of a "responsibility.” did than

349 a court in that (7th Cir. 371 F.2d Inc., 746 Jeffboat, regression a hold analysis, regression subjec “frequently (court that 1984) noted solely evidence” “unacceptable as analysis influ factors intangible other tive and ‘all measurable include not even it “did that decisions employment ence effect on sala an to have thought neces may not variables misjudgments subjective ” court liability” The district 3009. VII Id. at ry for Title level.’ basis sarily be the defects, we dis Prince which v. Allen several original)); found here (emphasis (4th the below, “affect[ed] that County, briefly cuss George’s always admit not not its admissibili (“Courts have Cir.1984) probativeness, analysis’ analyses, [salary] regression Regents, Board multiple v. Id.; ted see ty.” Griffin Cir.1986) sometimes admitted, they have (7th n. and when F.2d understanding of the likely an aid as analysis used been regression (“Multiple proba as than rather position, in parties’ the the model unless ... inappropriate show.”); Fish purport what likely to tive of major variables the all of cludes Proceed Legal Regression er, Multiple variable.”); dependent the on effect have an (mul (1980) 702, 714 80 Colum.L.Rev. ings, 613, 623 TVA, F.2d v. Eastland helpful may be analysis regression tiple 1066, 104 denied, 465 U.S. Cir.1983), cert. sub not circumstances, “it is a but certain (1984). theAs 79 L.Ed.2d S.Ct. definitely court The thought”). stitute fact, noted, “[w]hether, in Court Supreme hostility toward outright show did not carry the analysis does regression such suggests. EEOC statistics, the in a depend will burden plaintiffs’ ultimate limita- specific recognized two each court The context factual on the case given illus- others, we believe that tions, among presented the evidence light of all case consideration careful court’s the trate the defendant.” plaintiff and the by both analyses regression the limitations add (emphasis Bazemore, S.Ct. clearly erroneous. say are we cannot the evidence all consider ed). did The court important if noted First, court the limits the account into took properly model, “the omitted are variables analy- statistical of the probative value compensation estimated of sex effect ses. be- artificially inflated” bewill the model findings court’s holdWe re- will sex variable dependent cause EEOC’s its conclusions underlying sex, also effect only any not flect not flawed analysis was regression variables omitted of the effect the residual briefly address We clearly erroneous. F.Supp. at II, 628 salary. Sears that affect those criticisms serious most significant Second, found court 1344. First, court found findings. sala- compared study that case Sears regression underlying EEOC’s base with data employees checklist male two ries of chal- EEOC was inaccurate. histories. employment very similar asserting finding by as lenges this differed employees of these salaries showed, Sears multi- month, “unavailable.” but Sears’ were data per as $600 much indeed explain the data however, of the some could model regression ple noted, (on paper records “[i]f salary. The available difference in- employees sex- of different been individual employees had these would personnel salary disparity computerized es, of this of on much stead exclusively re- discrimina- sex attributed tapes, have been on which unex- for some lied) Id. at tion.” and that paper analyze did plained reason appro also believe We addition, demonstrated records. analyses statistical considered priately difference amade indeed the data Supreme spirit according to analyses. regression results Fri Bazemore opinion recent Court’s in deter- clearly err not did 3000, L.Ed. 385, 106 S.Ct. U.S. day, 478 analy- regression mining that did here The district (1986). 2d weight because less entitled sis was deficient found Court not, the Bazemore incomplete underly- the inaccurate and data because the variable diminishes the effect ing analysis. that the sex variable would otherwise have had. However, part the EEOC cites no argues The EEOC that certain variables support record that would that idea. It the court found the EEOC should have seems to us that even if a factor was regression analyses in its included were not gender-correlated, neutral, legit- if it is a job salary. related or correlated Al- imate factor job such as the number of though go say we not will so far as to that (which proportionately relocations more the EEOC should have included the factor had), men than women and evidence indi- pre-Sears experience, because on data cated that that factor had an effect on available to either Sears or salary, that factor should be included in the EEOC, agree with the district court regression. prior responsibility time-card and the employee number relocations were varia- attempts The EEOC sidestep the dis- bles EEOC should have included in trict finding regarding court’s defects in analysis. evidence, There ample in- regression model, the EEOC’s arguing that cluding testimony record, and also Sears’ model greater had similar and even law, indicating case these omitted defects. only aspect We need address one bearing factors salary. have a The of the court’s criticisms of the EEOC’s EEOC refers to one of its exhibits that model to affirm the finding. court’s purports to show that certain variables do district impor- court found that the most not correlate with salaries they do tant flaw in the EEOC’s model is that it not, individually, bearing have a on the aggregated (because data nationwide analysis. statistical This accept- be an employees occupied a few some on a testing able method of the correlation of a level), more local salary when decisions at salary. variable with See Coates John- Sears were made at the local level and Johnson, son & decisionmakers at that level had considera- Cir.1985)(“ size and signifi- statistical ‘[t]he ble discretion. analyses Sears’ showed cance of the qualification controversial will wage disparities by territory. differed question illuminate the of whether the argues every salary deci- ” qualification was in fact considered’ approved sion had to be by the territorial (quoting Cole, D. Baldus & J. Statistical office, but Sears refuted this contention. 8.23, Discrimination at 100 § Proof of The court did clearly err in this deter- (Supp.1984)); Asher, Gwartney, Haworth & mination. The analyzed EEOC could have Haworth, Statistics, the Law and Title salary local through decisions study a case View, VII: An Economist’s 54 Notre approach, but it chose not to do so. (1979)(“If Dame Law. potential Briefly, we note that we also do not skill factor is not employ- related to the consider erroneous the district opportunities ment employ- offered finding court’s significant again that it was er, the variable will statistically un- wage produced claim that the EEOC earnings. contrast, related to when sta- no testimony individual wage acts of analysis tistical illustrates a consistent link discrimination. Neither did the court clear- between the skill higher factor and earn- ly err in finding important the Sears’ ings, this constitutes evidence that the skill management testimony witness that Sears job-related.”). factor is Sears has identi- did not compensa- discriminate problem fied a reasoning, with this how- period tion at issue. ever, and we have our own doubts. Sears misleading contends that it is We need not discuss the identify court’s treat- the effect of regression individual ment of salary, analyses, variables on Sears’ except because some jointly variables to the extent problems contributed that it identified the salary. Furthermore, to influence (for the EEOC’s example, appears argue stepwise regressions that because a demonstrated that in- job related, factor is not gender- it must be disparities clusion of variables reduced correlated and therefore skews point the results statistically were not significant). regression analyses pro- The court also used Sears' EEOC's were not analyses wages evidence from its cohort to con- bative of sex discrimination in regression analyses employees elude that EEOC's did checklist at Sears. "accurately complex, not discriminatory decision-making processes." reflect Sears' non- IV. AFFIRMATIVE ACTION at 1352. In the co- analyses, compared hort the increas- The district court concluded that salary among es in checklist male and fe- Sears' affirmative action evidence demon employees, male with the result that wom- strated that Sears had no intent to discrimi favorably en were treated as or more so against hiring, pro nate women in either argues than men. The EEOC that these motions, pay. argues or analyses cohort are invalid because "re- the district court erred in this determina given higher sults showed that men were tion, clearly finding erred in that Sears job- initial checklist salaries in 74% of the program had an affirmative action since years analyzed," "disparities in start- 1968 that made the movement of women ing statistically sig- checklist salaries were priority.51 into commission sales a A dis years nificant in two of the five when test- finding regarding trict court's intent to dis individually." analy- ed sis, however, Sears did a cohort finding subject criminate is a of fact salary that indexed for men clearly erroneous standard. See Pullman- through and women in cohorts from 1973 Swint, 273, 289, Standard v. 456 U.S. 1979 and set the index for women at a 1781, 1790, (1982). S.Ct. 72 L.Ed.2d 66 value that reflected the initial difference in finding "strong This is thus clothed with a pay multiple regression estimated Sears' presumption validity." rough Hilisbo analysis. Men were indexed at and County v. Automated Medical Laborato 1974-by women were indexed at 91 in ries, Inc., 707, 721, 1979,women were indexed at more than 97. 2379, (1985). 85 L.Ed.2d 714 again The results showed that women were say just favorably favorably We cannot that the district court's or more treated finding salary that Sears had no intent to discrimi than were men in increases. We do against hiring promotions clearly nate women in or not find erroneous the court's con- clearly analyses into commission sales is erroneous. clusion that the cohort were "the The EEOC contends that the existence of complete regard- most ing evidence in this case program similarly an affirmative action is not Sears' treatment of situated enough support finding Although of no intent to checklist men and women." Id. argument appealing discriminate. Cf. EEOC v. Keco Indus the EEOC's is at first tries, Inc., glance, Cir. it assumes that sex is the accounting 1984)("Compliance factor for the differences in ini- with the affirmative ac pay requirements agency tial checklist The mere fact that EEOC felt the need to between men and women. tion of one federal automatically does not translate into com multiple regression analysis, pliance VII."). do a acknowl- with Title The Second Cir edging noted, however, that at least some variables other cuit has salary, than sex affect shows that comprehensive [t]he existence of a af- EEOC knows better than to claim that dif- program firmative action is the antithesis pay ferences in initial checklist between pattern practice of a of discrimina- solely men and women are due to sex. The program tion based on sex. Such a challenge indexing EEOC does not Sears' gender evidence of an intent to eliminate of the initial salaries. employment [sic] as an criteria and to We conclude that the district court did root out subtle forms of discrimination. determining directly err in It thus controverts a claim that Although rulings findings necessary, our that the EEOCfailedto tive action theless those less we none- prove against briefly challenges that Searsdiscriminated women considertheEEOC's employees hiring,promotions, compensa- tion makethisdiscussionof the court'saffirma- findings. discrimination is operating the “standard record support evidences for the court’s procedure.” finding. For example, Equal Sears’ first Moore, Opportunity Director, Coser v. 746, Ray (2d Graham, J. Cir. 1984) (quoting International Brotherhood testimony whose the court found “highly States, Teamsters v. United 431 U.S. great credible and id. weight,” entitled 324, 336, 1843, 1855, 97 S.Ct. 52 L.Ed.2d 11, at 1292 n. “[sjince 1968, testified that (1977)). increasing the number of in women com- essentially argues that the mission sales part has been of Sears’ af- evidence of affirmative action is weak and firmative program,” action and at another support cannot a finding of no intent point, that “Sears was aggressively seek- discriminate. EEOC maintains “in that a ing women in commission sales in the ear- typical affirmative program, action em- an ly ’70s adoption ... before the of the MAG ployer long will set range goal for a plan.” testimony This was corroborated group’s representation in a job certain cate- other evidence.52 The may argu- be e.g., gory, 40%, by hiring then group at ing that there is no documentary evidence considerably a rate above enable itself 40% policy, the initiation of this but to reach goal sooner if 40% than the district court found testimony credi- simply hired at It cites as an 40%.” exam- ble, and we must accord finding that even ple United Weber, Steelworkers v. 443 U.S. deference, more unless internally it is in- 193, 199, 2721, 2725, 99 S.Ct. 61 L.Ed.2d consistent, we do noncredibility than factu- (1979), company which the hiring set findings. al City See Anderson v. Bes- goals equal percentage of blacks in City, semer 564, 575, forces, the local labor the work force in one 1504, 1512, (1985). 84 L.Ed.2d 518 39%, area was company mandated that of new openings 50% trainees in craft Regarding pro- Sears’ affirmative action would be black until the percentage of gram during period charges of the at blacks in that approximated division 39%. issue (roughly 1980), 1972 to the court re- The EEOC argues then although ac- lied on description of the Mandatory cording to Sears it long had set a range Achievement (MAG) of Goals affirmative goal 1969, as early as “at 38% no time program, action fully implemented 1974, prior implementation the full and “the testimony of the Sears’ officials program (nor MAG for some time there- employees, whom the court found to be after) did full hiring time rate for wom- highly witnesses,” II, credible Sears en 38%, even approach much less exceed 1294, at who testified based on it.” their personal own experiences how Sears We nonetheless find the district court’s tried to recruit women into commission finding Sears actually long had “a positions. sales We do not find the court’s serious commitment to affirmative action” findings requirements of Sears’ was clearly erroneous. Sears programs affirmative action “highly influ- F.Supp. at 1324. The type of affirmative enced” all aspects hiring process or plan action Weber laudable, but testimony of Sears’ witnesses re- it is not plan. valid form of flected the “sincere dedication and commit- this case found that “at- ment of management Sears at all levels to tracting women commission sales at action,” id. affirmative to be clear- Sears has important been priority in ly erroneous. Sears’ affirmative action programs since ... 1968.” Id. short, 1306. argues finding the district court’s of no finding error, is clear intent to discriminate based findings on its “[sjince Another Sears witness testified that sales." Sears also introduced written memoran- enunciated, universally un- February da of August indicating 1973 and derstood, regularly policy followed was to emphasis recruiting women to commission undertake encourage affirmative action to more sales. women to enter and to succeed claim54 on its mary judgment53 affirmative committed in violation women against discriminated clearly erroneous. is not action eighteen-month during an Title VII PAY provision WITH LEAVE ending DAY’S V. October period-

PROVISION allowed a Manual Personnel the Sears his when paid leave day a of employee male the contends EEOC equivalent no was There gave birth.55 wife sum- partial denying court erred district other this and summary judgment on deny to character- court the district parties and 53. The 54), would not (see supra note summary claims partial for as one motion ized this the court would ask and confusing try issues those may because be label judgment. This prejudice. summary claims with those "partial dismiss phrase parlance, the legal de- used to typically a misnomer judgment” is EEOC the becomes whether question then 56(d) under Fed.R.Civ.P. pretrial order a scribe addition of claim appeal this has waived and are controverted facts specifying which try We think right waiving this claim. defi- is an order Such are established. interlocutory orders other As with not. C. nonappealable. See interlocutory and nitely case, of denials of a during course arise the Kane, Federal Practice Wright, & M. A. Miller merged a final into summary judgment are 1983). (2d 2737, ed. at 463 § and Procedure appealed. Moore 9 be then judgment and for Instead, one the motion treated the court summary judg (denials 110.07, of at 108-09 ¶ claims. to some summary judgment as appeal after on "subject to review ment are however, employ not, expressly did The court action, terminating the judgment entry of a the as a 54(b) motion treat the and Fed.R.Civ.P. interlocutory rulings are the into which all but fewer than one or more judgment on Kane, Wright, & M. A. Miller merged”); C. see and express determination and make claims 2715, at 636- § Procedure Practice Federal effect. to that direction Guaranty Morgan 1983); (2d also see ed. 38 593, Martin, & n. 9 599-600 F.2d 466 Trust Co. summary partial had moved 54. The (dictum); Bell Cir.1972) v. Illinois (7th Smith cf. of discrimination claims judgment on five 588-89, 587, S.Ct. Co., 46 Telephone regarding provisions certain based on were (interlocutory (1926) 408-09, 408, 747 L.Ed. 70 Manual, on a Sears’ Personnel pregnancy in perma of merged final decree injunction into hiring preg- of policy geneial Sears’ This appealable). injunction thus nent alleged adher- applicants, and on nant avoiding piecemeal policy of comports the in- The EEOC laws. protective to state ence policy consideration is the main appeals, which try those not it would the court formed appellate against general rule underlying the summary judg- denied it were even if issues judgment. See summary of review of denials those claims of two withdrew ment. Horne's v. E. Association Cheese Switzerland ruled, de- the court and after the court before 193, 24-25, 23, Inc., Market, 87 S.Ct. U.S. 385 other three on summary judgment the nied Matterhorn, (1966); 194-95, Inc. 23 17 L.Ed.2d claims. those appeal two of chose not claims Cir.1984). 629, (7th 633 Corp., 727 F.2d v. NCR the impression in that of first a situation This is question of the party addressed 55. Neither summa of underlying the denial specific claim judgment is summary of this denial whether not thus was judgment ry tried goes to our question appealable. the particular Because terminating the ac judgment part of the final appeal, we be this jurisdiction over appellate allowing believe We do not tion. is briefly this necessary to consider it is lieve policy the instance undermines in this review against piecemeal summary judgment Generally, of denials sue. of sum denial appeals. This nonappealable. See interlocutory and thus an interlocu in essence judgment mary was not 656, 656, Florian, 61 U.S. 312 States v. United ending effect had the tory it because order curiam) (1941) (per 713, 713, 1105 85 L.Ed. S.Ct. court. district claim in consideration judgment, a court (reversing, of a final want adju- pretrial denial also differed appeal a deni hearing appeals decision released court the district in that dication judgment after the summary al of same on summary judgment denying order decision); final as a the denial had disclosed opinion the merits day released the 277, Co., 641 Ray & F.2d McDermott v. J. Ardoin outstanding no were Thus there case. & Cir.1981); Bank National (5th American action with the claims connection Lloyd’s Underwriters Co. v. Certain Trust judgment. by a final terminated had been action Cir.1971); 640, (7th 6 J. London, 444 F.2d addition, appropriate to it is we believe In Lucas, Moore, Federal Moore's J. Ward & B. deny if we were to appeal because this 1987). hear (2d 56.21[2], ed. at 56-1276 ¶ Practice in effect we would jurisdiction on this claim try case, proceed to parties will typical unless, of appeal, foreclosing avenues all validity of the question of the the issue as an claim course, certify the judge becomes summary judgment effect denial U.S.C. interlocutory under order appealable case, This, however, a rare moot. strategy risky trial pursues a 1292. Counsel § chose court that if informed allowing (quoting 90 S.Ct. at 1609-10 employees Fed.R.Civ.P. provision for female 56(e) (1963)) (em gave advisory birth. committee note day’s pay them a when Court); Big phasis by Supreme O added summary a district court’s Our review of Dealers, Warehouse, Big Tire Inc. v. O novo, is de and we judgment determination (7th Cir.1984). F.2d applied apply standards as those the same Roman v. Unit See court. The district court held that Service, ed States Postal 821 F.2d EEOC failed to meet its burden because it (7th Cir.1987); In re Colonial Discount prima facie case that did not establish a Cir.1986), Corp., day’s-leave-with-pay provision amounted to — U.S. —, denied, cert. discriminatory policy. can an enforced We (1987); J. 6 J. Moore & 95 L.Ed.2d 526 erred in say that the district court Wicker, Moore’s Federal Practice conclusion. (1987 Cum.Supp.1986-1987). & ¶56.27[1] dispute The EEOC does not first determine whether there are We must plaintiff in court’s determination that a a fact, any genuine of material draw issues disparate Title VII treatment action must ing inferences in of the all reasonable favor present prima facie case of a discrimina genuine If find no non-movant Sears. tion, attempting in whether to defeat a fact, must deter issues of material we then summary judgment defendant’s motion for is entitled to sum mine whether the EEOC prevail summary judgment mo or to on a mary judgment as a matter of law. See Parker v. Federal National See tion. Mercury, Ford Mo DeValk Lincoln Inc. v. Association, 975, Mortgage 741 F.2d 979- Co., 326, (7th Cir.1987); tor 329 811 F.2d (7th Cir.1984); Kephart v. Institute 80 56(c). Fed.R.Civ.P. 1217, (7th Technology, Gas 630 F.2d 1219 Furthermore, the we must consider bur- denied, Cir.1980), 959, 101 cert. 450 U.S. 56(c) pro- den on the EEOC. Rule movant 1418, (1981); Weahkee S.Ct. 67 L.Ed.2d 383 moving summary party vides Perry, 1256, (D.C.Cir. v. F.2d 1265 587 showing judgment has the “burden of the 1978). disputes The EEOC instead what genuine any of a issue as to mate- absence showing the district court stated such a Co., fact,” Adickes v. Kress & S.H. rial 398 prima a facie case entails. 144, 157, 1598, 1608, 90 26 U.S. S.Ct. relied district court on the standard (1970), L.Ed.2d 142 and that he or she is International Brotherhood established see judgment law, entitled to as a matter of “ States, Teamsters v. United 56(c). 431 U.S. Fed.R.Civ.P. ‘Where the evidentia- 324, 360, 1867, 1843, 97 S.Ct. 52 L.Ed.2d ry support matter in does motion not (1977) issue, by 396 this court genuine establish the of a followed absence Johnson, must be denied even Coates v. Johnson & summary judgment 756 F.2d if 524, (7th Cir.1985), opposing evidentiary no matter is plaintiffs in 532 a ”56 Adickes, presented.’ 160, pattern-or-practice government at 398 U.S. or class dis- submitting by determining summary versing summary judg- before a trial court's denial ment, judgment opt try noting summary judgment an motion that will that denial of appealable issue if the motion is denied. Our research has was not as such but there was no employing strategy. disclosed no cases such a material issue of relevant fact and both is, however, tactic, legitimate especially It a in a parties ample develop had "more than time to large underly- such as this case when the claim issues"). pleadings and fact ing summary judgment the motion is minor and Catrett, attempting prove might Corp. it at trial confuse 56. In Celotex 2548, 2552-53, (1986), expenditure other issues and cause extensive S.Ct. 91 L.Ed.2d 265 56(c) strategy Supreme resources. That the is novel does not Court held that Rule mandates party entry summary judgment “against party mean that a should be denied all avenues a appeal. showing The EEOC made a conscious deci- who fails to make a sufficient to estab- by sion to stand its contention that Sears’ Per- lish the existence of an element essential to that case, provision discriminatory party’s party sonnel Manual was which that will bear Sears, summary judgment proof and that it was entitled the burden of Because at trial.” defendant, appeal. proof on that issue. We will hear this did not have the burden of Cf. claim, Jorge, Tabacalera Severiano S.A. v. Standard on this discrimination this statement Co., Cigar (5th Cir.1968) (re- apply. Celotex does not (E.D.La.1980), Co., F.Supp. 710 of Glass initial burden claim have crimination Cir.1981) (per cu F.2d aff'd, discrimination unlawful showing that In Du holding. of this riam), support A policy. regular employer’s was allegedly rant, produced an plaintiffs facie case prima a must “establish plaintiff provision discriminatory maternity leave policy existed.” [discriminatory] that such employer’s at out in the 97 S.Ct. set Teamsters, at had been U.S. words, plaintiff would Maintenance Contract” In other “Production 1867. preponderance employer discrimi “by a to show the prove effort have to was discrimination The district court against women. evidence nated [the] operating procedure— their standard had not met company’s plaintiffs held prac- the unusual than regular rather of a showing existence dis burden at tice.” Id. were “un criminatory policy because anyone affected to show able met this burden that it argues was never provision, which attacked discriminato- allegedly showing that the part of to be a and which ceased enforced existed provision day’s-leave-with-pay ry agreement six bargaining the collective eigh- for an Manual Personnel the Sears Tuma v. Amer 723; Id. ago.” years court re- The district cf. period. teen-month “ (D.N.J. 218, 230 Co., Can ican ‘absent EEOC’s statement jected job to fe policy 1974) (employer’s barred policies were by Sears that had not bid males, employees female necessary presumption implemented, any impact on not could show job did so implemented and for policies were that the ” illegal policy); stemming from 628 them individuals.’ affect Inc., 39, Photo, 85 Berkey brief). F.R.D. Kuck v. The reply (quoting 1275 showing (“absence of a provi- (S.D.N.Y.1979) existence held court gener has prima plaintiff a enough injury to show in fact not alone was sion policy judicial ex- all interven discriminatory to ally held bar that a been facie case some prima case”). facie case is Durant a isted, such in a that to show tion provide weak) authority to for the (although perhaps have had EEOC would case the discriminatory en- allegedly was provision that an proposition that some evidence court, equiva the EEOC automatically the According to the provision forced. is identifying purposes policy discriminatory either for done this could have of a lent “some val- by producing case dis establishing prima or facie victims” a “some show showing of en or circumstantial some id direct absent crimination policfy] provision. followed the actually application that Sears forcement Id. at 1276. employees.” con cases its treatment has cited no no has disclosed our research trary, and a fair applied court think We Furthermore, language of such cases. properly law and of the case interpretation suggests that and Coates Teamsters con- summary judgment account took into of discrimina facie case prima a establish determining siderations just more than tion, must show plaintiff a of discrimi- case facie prima a not show did (which was of discrimination one instance to summa- not entitled thus nation here) plaintiff must shown not even deter- properly —the judge judgment. The ry operat “standard policy was show that the more to show had mined regular rather than ing procedure discriminatory provision allegedly than —the Cooper Feder practice.”57 unusual facie case prima Cf. establish alone to 878, 867, Bank, 467 U.S. Reserve al policy rather regular provision was (in (1984) 2794, 2800, 81 L.Ed.2d S.Ct. practice. unusual an than liability, ob- pattern-or-practice determining v. Owens-Illinois Durant cited obligations domestic their would have fulfilled for motion response to the EEOC’s 57. Sears’ rear- work or for day’s not scheduled suggests leave when judgment summary off practice— so that time may ranged an unusual work schedules have been pay their required to necessary.” were not that ‘‘[m]en stated was not many instances provision, and in utilize Cir.1987); see 130, plaintiffs attempt 131-32 serving that “a class 810 F.2d States, 346, v. United companywide poli- also Watts prove the existence of a cy, practice (9th Cir.1983) (“summary judgment consistent within a or even a though given department, avoiding fail even necessity neither a method of against one or two individu- discrimination proce- nor proving one’s case a clever Telephone General proved”); als been has whereby gambit a claimant can shift dural Falcon, Co. v. 159 n. U.S. adversary proof to an the burden of on one (1982) n. 72 L.Ed.2d 740 S.Ct. issues”). agree or more We with the dis- employee’s (finding that an individual claim trict court that the EEOC failed to meet discriminatory treatment did not neces- this burden and thus was not entitled to sarily employee adequate qualify the summary judgment on its contention that representative class of individuals who al- day’s-leave-with-pay provision violated legedly stating had similar claims that Title VII. prohibits Title noteworthy “it is VII discriminatory employment practices, APPEAL VI. CROSS (em- policy an abstract of discrimination” appeals Sears cross from the dis phasis original)). The district court was trict court’s failure to dismiss the EEOC’s deciding faced with whether the EEOC had determining suit after statutorily shown the absence of a genuine issue as to suit, preconditions mandated were con fact. See Adickes v. S.H. any material alleged ducted amidst conflicts of interest. Co., 144, 157, Kress & In view of our affirmance of this case on 1598, 1608, (1970). 26 L.Ed.2d 142 the merits there is less need to resolve this legitimately could have determined appeal, cross but it nevertheless deserves

that an issue of fact remained consideration. day’s-leave-with-pay provision whether the was ever enforced. The EEOC did not part This the case was decided examples ap- show of enforcement or Judge Grady before the balance of the case plication provision, and thus showed reassigned decision on the merits was injury type no evidence of actual of the Judge Nordberg to in May 1982. EEOC v. designed remedy. Title to VII was Sears, Co., Roebuck & (N.D.Ill.1980). Judge Grady dealt with this Du- attempts distinguish to extensively conflicts issue and resolved it rant, noting that in Durant the defendant in favor of the by denying the mo presented testimony that allegedly the dis- so, doing however, tion to In dismiss. he criminatory policy enforced, was never but alleged found that certain abuses of the this case Sears offered no such controverted, EEOC were not and that “the point, evidence. At another the EEOC appearance partiality of part on the of the complains that the district improperly palpable.” Commission Id. at 251. required [was] prove that Sears expressed He also misgivings some as to complied provision personnel with a in its possible the ruling denying effect his the manual had when Sears not contested that might motion have on the future conduct of complied, it had not genu- even raised a employees. the EEOC and its He therefore ine argu- issue of material fact. These made it clear that condoning he was not the reflect ments the EEOC’s confusion re- of conduct the EEOC and certain of garding responsibility its its party moving as a personnel. One of his reasons summary judgment. for for not dis The EEOC im- Sears, missing plies anticipated was that he nonmoving party, de novo had during merits, raising genuine review the trial on the burden of a issue of which he actually preliminary it considered would render material fact when was the improprieties Again genu- Judge EEOC’s burden to establish that no harmless. Gra warned, however, dy ine as to issue material fact existed that de novo review always and that the EEOC was entitled to could summa- be counted on to render ry judgment as a matter of See preliminary law. join errors harmless. We Service, v. United States Wilmes Postal in those admonitions. Since then the Su personal relationship a Young v. United also was added decided Court has preme — S.A., problem. et Fils We will compounded rel. Vuitton which ex States only briefly. story, L.Ed.2d but U.S. —, tell the dismissal argues makes (1987), NOW set begins in story when appropriate.58 even more “major action goals its a national one of of the Grady’s denial Judge affirm NOW’s large We chain retailer.” against a been, have would dismiss which motion to Task Force Enforcement Compliance and sanction,” but it, “extreme as he called considering time what the same was misgiv- considerable do so with likewise a nationwide Sears on done about could be Young, which we consider will ings. convention, 1973 national At NOW’s basis. tip the detail, to not sufficient is more Force, rec- of that Copus, a member Task circumstances particular in the balance settle- to a an effort “demand” ommended Young reached this case. results corpora- highly large, visible ment from a different, the mes- but in this case image of its tion, corporation conscious as- have must the same. sages are “[W]e female women traditional employed wield those who would surance conducted At the convention NOW jobs. solely by their sense guided power will be to demonstrate workshop how about attainment responsibility public target, against “Snears-Doebuck.” Young, It at 2141. S.Ct. justice.” urgent message ap- spring an Later With that. any better than be said cannot District in the newsletter peared trial us after hindsight afforded calling vol- chapter of NOW Columbia greater perhaps even misgivings are our calling campaign by join the unteers Judge Grady Grady’s. Judge than were home, giving or at his Copus at the EEOC harm that rightly concerned was advising keep trying numbers both in inter- parties the “real might result es- him time reach was it Sears, employees. As est,” Copus, who spring of 1973 sence. however, par- “real out, none turned Chairman, directly to the reported witness- produced as in interest” ties heading the responsibility given the parties The “real by the EEOC. es Programs Division National EEOC’s glimpsed only be could interest” charge put in (NPD), particular was and in analysis of the of the statistical shadows August investigation. By the Sears Before consider- expert witness. underway, assisted well Copus was import Grady’s opinion Judge ing he had members whom by other NOW Young briefly review necessary to produced the Com- Copus and staff hired. problem to background of this factual Sears, Co- charge against which missioner’s effect, an ef- insidious appreciate fully investigation then led the signed. He pus in weaknesses may evident fect Copus be- November 1973 of Sears. In it has any In event case. the EEOC’s Directors of the Board a came member as system impact judicial on the had an Education Legal Defense and NOW permit such parties. To as on the well legal (LDEF), arm. NOW’s Fund good can harm to exist conflicts promote. EEOC seeks to Early causes in the Chairman Assistant, Whitney M. Special hired a case do statistical evidence a Seldom the Board Adams, member of a also one this similar to problem find a of NOW. Directors in the lawyer high-ranking Here a case. little indicate was found that there motivated EEOC, Copus, was David A. Sears in the directly involved was Adams devotion to more his duties his official feelings it strong about action, she had Organization for Women National re- In undoubtedly some influence. calling and high by the was (NOW), than he questionnaire sponding to conflict this position. To government his supplemental briefs. party filed thereafter each argument allowed the court oral After Young, briefing on for additional motion support program jointly presented her on the she indicated were NOW convention “compliance” investigation employment law work- of the NOW Sears two position. Copus shops. her EEOC revealed only had an interest Adams not motto, seeking T-shirt Copus outdid the case, Judge in each other. but also “predetermination settlement” from Sears alleged conjugal Grady regarded the rela- argued He in excess of million. $600 tionship Copus and Adams as “rel- between important the settlement was to “outside by.59 atively pass also it trivial.” We will groups.” When Sears declined that offer relationship them Whatever between Copus supervised participated prejudicial and whatever have been production of the Commission’s reasonable impact may had on the conduct of have decision, cause decision. The handed down case, prejudice than it adds no more 19, 1977, April eighteen days on after Co- Copus’ does conduct alone. EEOC, pus adopted left the a to 1

During NOW’s national convention vote. Sears endeavored to have the Com- Chicago chapter alleging justification spring of 1974 NOW’s mission reconsider interest, charge request filed with the EEOC a formal the conflicts of but the against incorporated Copus that was denied. Sears. charge charge. with the EEOC At procedure following The next the deci- adopted a convention NOW formal resolu- statutorily presuit sion was mandated con- condemning

tion Sears. NOW also issued Copus regulations ciliation.60 drafted the Litigation Strategy a document entitled “A governed attempt. that conciliation NOW,” Copus which bore the names of Conciliation efforts failed as the Commis- and Adams. sixty-nine sion would not discuss race, days individually involving

A the Chairman of the few later claims national EEOC, Powell, Jr., origin, charges. John H. asked for the and sex discrimination It counsel, general did, however, opinion of his William Car- reduce its demand from $600 ey, concerning Copus propriety also million to million. The excuse for the $54.5 serving continuing on LDEF Board. large monetary the NOW Both demand was Carey knowing groups Powell and denied at trial that “outside interest had to be campaign against of NOW’s Sears al- satisfied.” Sears declined the conciliation though Adams, Copus and badly both offer. The EEOC therefore abused Powell, investigation, predetermination EEOC staff and direct lines to settle- ment, actively July statutory prerequi- involved. after and conciliation receiving Carey’s opinion, suit, the Chairman sites to all of “intimately which are Copus give up position. asked his NOW related” in Title VII’s enforcement struc- so, Airlines, Copus however, Allegheny ture. EEOC v. delayed doing until September significant (W.D.Pa.1977); F.Supp. 1974. It is that Co- see pus Corp., EEOC v. leadership was not relieved of his role Container investigation. (M.D.Fla.1972). in the EEOC That prejudice was where the conflict and the The EEOC then filed this lawsuit. The exposed. would been have charges individual were trimmed from six- engaging began ty-nine NOW thirty-five. charges demonstrations incorporated. and other activities at Sears stores vari- NOW had filed were country. responded ous locations in the At alleg- NOW’s with a motion to dismiss interest, ing 1975 convention there was a “Sears Action these conflicts of as Sears it, victory expressed by Copus’ ... celebration.” T-shirts were created and Adams’ NOW, popular on the which front carried the mes- the simultaneous with activities EEOC, other, sage “N.O.W. Sears Action Task Force.” each and the case. We motto, Judge published On the back was the “100 Grady’s Million need not restate Nothing Copus analysis, upon partially and Less.” and Adams our which we rest Copus subsequently 2000e-5(b). and Adams 59. married. 60. 42 U.S.C. § would joined, O’Connor in Justice Justice Young consider decision, need to but we rule and instead categorical adopt a factual not particular against the detail more to deter- the case remanded have would this case. background of was harm- not the error or whether mine lawsuit, but later a began as civil Young White dissented at 2148. Justice Id. less. Ini- proceeding. criminal a developed into conducted, trial was that a fair basis the goods leather French well-known a tially, appoint- the not embrace although he did injunc- permanent sought a manufacturer to party interested of an of ment counsel operat- individuals certain prohibiting tion Id. contempt case. a prosecute manufac- goods leather family-owned ing a in- country from in this turing business although the Justices apparent It is re- A settlement its trademark. fringing the conflicts analyses in their varied paid to the were damages in which sulted disapproved unanimously problem, compa- American and the company, French of an the counsel appointing practice of the permanent entry of a to the ny consented prosecute criminal to party interested it prohibiting generally injunction con- similarly do not We contempt case. products selling leather manufacturing and case, present in the happened what done product. French imitating the Young, a it, not read do far from firm, sus- French year later the About to reach contempt prosecution, criminal being injunction was the pecting in this require reversal far as to quite so to investigation firm observed, an retained case. “sting” operation, an undercover conduct Young and between distinctions Some Two very effective. was was Young are obvious. present case requested attorneys firm’s French involving public proceeding a criminal coun- special appoint them own the court’s in vindication interest developing criminal prosecute to sel such as In a case authority. at 2136. Id. appointed They contempt action. stake. Id. liberty could be itself Young, investiga- proceed to authorized that under considered Court at 2139. The evidence, incriminating tion. Substantial error the harmless circumstances those gath- tapes, was and video including audio assuring the task equal rule followed. conviction Trial ered. official’s conduct. public confidence is relevant appeal only issue on acknowledge the Although we Id. is whether case present importance of private great public attorneys for the appointing erred involved issues discrimination action, as firm, in the civil plaintiff French without case, civil case it was a present contempt criminal prosecutors liberty inter- adversely affect potential adversary, their civil against charge authority. the court’s ests firm. American trial, although before conflicts occurred re- Supreme Court plurality A at least influence their possible that is prohibit- categorical rule versed, adopting a Co- trial. into the carried over extent some prose- an interested appointment of ing the All the case. however, try the did not pus, Justice at 2141. Young, S.Ct. cutor. entirety fairly their aired issues were gone have Blackmun, would concurring, judges, and impartial two before process violation it a due held further lost on early conflicts spite counsel party’s interested appoint an If there. matter leave the We merits. Jus- contempt. Id. criminal prosecute nature arise issues serious concurred, have but would Scalia tice peculiarities, future, its own each with holding result plurality’s reached discretion sound his or her judge in trial power to no “have courts the federal “extreme consider left to nevertheless derivative ... and no contemnors prosecute appropriate as the dismissal sanction” attorney to conduct appoint power to to have found conflict remedy if the Id. contempt prosecutions.” criminal process. judicial whole infected the Powell, the Chief with whom 2146. Justice disap- investigatory stage by presuit performance of EEOC was ed at the pointing only to conciliation. the result of that and did disservice Whatever been, Sears, public might party but also to the and even to have each has now had up and its causes. the chance to offer at trial whatever it NOW had, lacking. has found been merely requiring remedy The EEOC hope We this determination will not be Copus resign showed a from NOW total considered all loss and will instead assist appreciation of the lack of serious conflicts ways the EEOC in various in the future problem. making If the EEOC officials pursuit worthy objectives. of its truly that decision were unaware of the problem existing extent of the conflicts AFFIRMED. office, the Chairman’s own then that too is CUDAHY, Judge, concurring Circuit only Copus indefensible. Not should have in part dissenting part: required relinquish been his NOW con- nections, totally but he should have been extremely It is distinguish difficult to case, separated from the Sears and new superficial blemishes from structural de- appointed. EEOC counsel New counsel confusing case. in this oversized and fects carefully should then have reviewed the Although impressive and, its efforts are development Copus. case under commendable, many respects, majority Copus Sears, conflict was between only partially has been successful. Its Copus Copus, although and NOW. without opinion properly important identifies some office, charge his NOW left in was of the shortcomings case; in the EEOC’s but it partiality Sears case with his conflicts and significance overstates the of others and remaining. remedy was facial entirely seems to overlook equally certain only government unworthy agency of a Thus, serious argument. flaws in Sears’ apparent pursuit which in important of its is true that the EEOC’s internal machina- public responsibilities rendered a disservice initiating tions in this case deserve condem- doing so unfairly. seeking to elimi- nation. equally It is true that the EEOC as alleged nate discrimination at Sears the gave much as away by failing the case guilty EEOC was itself of discrimination produce any flesh and blood victims of dis- against Sears. Regression crimination. by statistics them- selves demonstrate Judge Grady

We affirm correlations be- on the issue of variables; tween appeal. this cross move from correlation causation, there indepen- must be some VII. CONCLUSION theory dent about the relationships causal See, e.g., variables. D. Baldus & J. appears It from the record that this has Cole, Statistical Discrimination been a Proof of parties war between two that both (1980). case, 9.4 at 320 In this much § profess equality commitment to for em- dispute centers on whether crucial inde- ployees regardless of sex. If Sears in fact pendent variables have been omitted or did not have during such a commitment misspecified. Therefore, parties’ caus- 1973-1980, then the EEOC simply has theories, al convincing- substantiated most prove commitment, failed to that lack of ly through accounts, par- first-hand take on which was the burden it had assumed. We importance. Hence, ticular the EEOC’s defer, must, as we to the trial court’s factu- present failure to first-hand evidence findings al unless errone- lopsided makes Sears’ victory far easier to ous. We also defer to the trial court’s reiterate, understand. I however, would judgment credibility of the witnesses key elements in Sears’ case have es- it, heard and by observed and not us. caped critical examination. We view this as a difficult case which was as well-handled the trial court as Perhaps questionable aspect the most reasonably possible. Had it not been for majority opinion acceptance is its Copus early unfortunate bias alleged women’s qualifica- low interest stages, might the case have been terminat- tions for selling complete as a for women’s lack of inter- dispari- reasons huge statistical explanation [the] for the selling a fear in commission included est adoption by the favoring men. ties they perceived of what as cut- or dislike by majority of Sears’ and competition, pressure increased throat arguments strikes me as analysis of these risk associated with commission and has indeed extremely uncritical. selling, on the Noncommission sales. gen- that these varied evidence presented hand, more other was associated with exist, in our so- both der-based differences friendship, pres- social contact and less particular labor and ciety as a whole and less risk. sure virtually insupera- a pool. But it remains weight of the to overcome ble task Supra at 321. by the evidence marshalled statistical court found The district ought courts skepticism that or the noncommission saleswomen were that ac- defenses to Title VII

to show toward present jobs generally happier their with as- unquantifiable traits rely on tions Sears, likely and were much less than groups. protected cribed to counterparts interested their male to be positions, commission in other such as commis- expectation that I start with the sales.... generally see themselves salespeople sion They make of the sales force. as the elite at 1313. peo- noncommission sales money

more than me, conclusions, These it seems to are through positions ple their and obtain are piece proposition with the that women a conse- hiring process. As selective more doing by happier cooking, nature the laun- greater prestige. They quence, they enjoy dry chauffeuring the children to soft- ability people confidence their are with selling games arguing appeals than ball merchan- and move captivate customers stereotype of women as less stocks. expect them to look with dise. I would daring greedy and than men is one that the condescension, contempt, upon retail if not sex discrimination were intended laws working as order-takers for a clerks disturbing It is that this sort address. quarrel I straight wage. do not therefore thinking accepted uncritically is so majority’s proposition that retail by majority. Perhaps district court and order-taking is a “different” task from they forgotten been have that women have expect selling. But I would commission hugely fields as resi- successful such most often in a jobs are seen estate, sales door-to-door dential real alignment selling vertical with commission merchandising. There outside direct other perspective top. Whether or not the lack that women indications are abundant ranking gives rise to this hierarchical strenuously compete desire to neither the commendable, I certainly pervasive. it is is capacity to take gain nor the financial my think differs from that of the view risks. majority, who seem to believe Sears, majority the district court and the as oc- coequally not perceived are tasks hang much of their refutation of pecking in a vertical tiers distinct cupying hiring promotion claims on the order. putative men’s and difference between undertaking commis- perspectives important These be- women’s interest disparities in benign Huge sion sales. statistical majority’s cause the more view selling significance participation in tends to minimizethe of Sears’ various in “inter- that women lack the interest are ascribed to differences contentions recognition scarcely any qualifications to sell on commission. est.” Yet there is Sears, Women, shaping “in- employer’s role as described See, applicants. e.g., majority, very court and the exhibit the terests” of Catlett Comm’n, qualities Transp. stereotypical Highway same for which Missouri Cir.1987). assigned positions 1265-66 been low-status have willing throughout history. majority majority Even the to concede states inter- opportunity may drive lack of lack of 362 1296-99) initial esti- adjusted F.Supp. at a “chicken- est, matter as but dismisses proportion of female expected I mates of Supra at concede egg” problem. might that for six factors hires to account would be government’s case that applicant’s an odds of legitimately affect handful even a stronger produced if had factors, “job ap- Four of these testify that Sears had selection. willing to of women for,” “product “job type experience,” plied of dreams be-

frustrated their childhood “commission sales ex- experience” and roofing, sew- line sellers of coming commission imperfect prox- perience,” can be viewed as How- conditioners. ing machines or air interest, and, extent, for victims, to a lesser I ies for ever, and blood even absent flesh adjust- subjective qualifications. These willingness of the district find the expected propor- brought down the accept the interest ments majority to and the The cross-classifica- recogni- tions of female hires. uncritically, and without defense method, example, ex- for reduced the stereotypes tion parallel tion of its close into full- eradicate, pected proportion of women hired perplex- to that Title YII seeks for the time commission sales nationwide unacceptable. ing and liability period entire from 61.1% 37.2%. event, to wrestle with In if we are statistically significant disparities be- But interest, ambiguous quality of hiring rates re- expected and actual tween the EEOC’s claim honestly assess must majority years every in a of mained disparities re- significant gender-based II Sears region. 1297-98.1 at appropriate adjustments main after findings indepen- analy- adjusted These derive interest. made to control for data, logit support cross- dent from the EEOC’s of hiring sis of II, Applicant This 628 Sears’ Interview Guides. analyses {see classification (discussing one-tailed and two-tailed tests sta- similar results when it of 1. The EEOC obtained Cole, significance); figures D. and J. adjustments ex- tistical Baldus performed these (1980 § Statistical Discrimination 9.221 disaggregated by product pected line. See hires Proof of and, (describing 41A, Supp.1986) & differences Table 15. Even if one EEOC Exhibit 1 advocating supplement, use less restrictive assumption of accepts overly restrictive possibility one-tailed test "if the of intentional only disparities greater than three standard favoring protected group discrimination can significant, 628 deviations are out”). be ruled The Z statistics calculated 1287, figures signifi- adjusted still reveal the EEOC are based on a binominal distribu- tion, disparities product cant for most lines in most approaches a normal distribution years. involving large processes for selection Thus, numbers. mean, however, imply accept I I do not probability obtaining a Z of 3.0 employed by the three standard deviation cutoff greater study in the EEOC due to chance events 1287, court, implicitly id. at very randomly probability is close to the approved by majority, supra at 323 n. 20. selecting an observation three standard devia- significance Statistical convention ascribes tions or more above the mean from a normal probabili- findings percent that have a five substantially distribution —about This is .001. ty having or lower of resulted from chance probability obtaining lower than the 2.0, a Z of five-percent factors. The translation of this very probability which is close to the erroneously ascribing significance to chance of selecting an devia- observation two standard chance results into standard deviation terms tions above the mean of a normal distribution— (also determining ninety-five per- known "the about .023. interval”) depends upon the cent confidence authority applying There is some for courts’ type of distribution at issue. For a normal significance stricter standard of percent than the five- distribution, five-percent corresponds level scientists, favored social convention for a to a 1.96 standard deviation threshold see, States, e.g., School Dist. v. United Hazelwood test, designed hypothesis "two-tailed" to test the 299, 2736, 307-08, 2741, U.S. 433 97 S.Ct. 53 explanatory power (e.g., variable has no (1978); L.Ed.2d v. & Coates Johnson John- being nor female neither increases reduces son, 524, (7th Cir.1985), 756 F.2d n. 22 but I applicant’s landing odds of a per any disparity find none for a se rule that job), sales and a 1.65 standard deviation thresh- less three than standard deviations to be dis- test, designed old for a “one-tailed” to test the See, Partida, regarded. e.g., Castaneda v. hypothesis particular 1272, 17, that a variable does not U.S. n. 1281n. (e.g., influence outcomes in one (1977) (differences direction "greater L.Ed.2d 498 than being female does not reduce the odds of significant land- two or three standard deviations Palmer, ing job). generally scientists’’); (sur- a commission sales See social 815 F.2d at 92 Schultz, (1987) levels). veying Palmer v. 92-97 case law on confidence Caviale Wis measurable variables. more than that women Cf. analysis shows Serv., Health & Social Dep’t appli- consin sales the full-time percent of sixty A) forty (“rule (7th Cir.1984) requir (EEOC Exhibit surveyed F.2d cants who considered persons comparison pool to percent ing members willing and able ready, most themselves ren qualifications would subjective possess improvements at Sears home installed sell production near plaintiff’s burden der a 30). com- But women A (EEOC Exhibit insurmountable”). ly of full-time only prised 1.7% acceptance majority’s uncritical The in 1973 and between 5.3% sales hires power explanatory the inordinate EEOC Exhibit See thereafter. 10.5% subjectively to interest Sears ascribes the may be “Interest” Supp. Table appar- be most qualifications assessed assign- in the crucial —factor valid —even “early rejection the EEOC’s ent in its that Sears positions to applicants ment of EEOC contends argument. The years” for be, hardly accounts but it it to asserts period early years liability of the that the discrepancies involved statistical the vast March (from beginning period the the the ma- Sears, court and district the here. implementation of the through the adjustments made that the contend jority (“MAG”) Mandatory of Goals rea- Achievement for various EEOC understate the 1974) ana- July in inter- should be plan or in June the true effects differences sons (from criticisms, however, ser- years are not later separately from the lyzed These est. explain away plan the enormous to the enough implementation of MAG ious the adjust- the after remain disparities 1980). majority declines to through The made. pri- ments are periods separately, two consider the EEOC’s “arbi- marily it claims the because the also district majority approves The assumptions” deprive its false trary and regression finding that Dr. Siskin’s court’s to respect probative value with statistics of for adjust adequately to analyses failed Supra at 310. In alter- the period. either Su- qualifications. assessed subjectively if the native, majority holds that even goes the even majority pra at 325-27. analysis were court’s approve basic statistical the EEOC’s so far as early take regressions sound, failed to finding separate consideration “ ‘phys- unnecessary such into account characteristics would be years assertiveness, ability appearance, ical the two EEOC has established friendliness, communicate, and economic Nei- Id. at 315-16. distinct. periods ” (quoting Id. at 327 motivation.’ declining analyze for ther rationale 1303), had been F.Supp. at scrutiny. separately withstands early years traits for commis- as “desirable” mentioned the EEOC’s failure majority, For the critique I think a based salespersons. sion differences adequately claimed account pe- obviously unquantifiable and on such quali- subjectively assessed in interest and inordinately crit- is ripheral considerations consider unnecessary to it fications makes being statistical ical of insisting In early years separately. Supreme Bazemore In presented. validity of its EEOC establish rejected the Fourth Circuit’s determi- Court proceeding with findings before basic “un- regression was nation that of the differences between discussion “all not include acceptable” because did however, periods, pre-MAG post-MAG Fri- Bazemore v. variables.” measurable signifi- majority misunderstands 3000, 3009, day, S.Ct. Separate argument. cance of the EEOC’s Here, (1986). is no there claim L.Ed.2d 315 early period not consideration include rele- failed to largest disparities, attention on focuses los- vant, variable. measurable dramatic difference underscores it also because, eyes in the es in commis- representation of women still it has not carried the majority, and the pro- the MAG and after sion before sales disproving the enormous burden heavier gender-based sharp un- gram. The decline that Sears attributes significance disparities suggestive significant more of a ment disparities seems far efforts change hiring practices actually discrete than exist between earlier and later gradual changes aspira- more in women’s periods. I find the discussion of motives qualifications. logical impli- tions and peripheral and the effort to minimize the plan cation is that the MAG removed barri- unconvincing. over difference time hiring practices.2 ers to Sears’ women view, my precise motives light, analytic Viewed this distinction implementing plan the MAG have little periods between the earlier and later is not bearing appropriateness on the separate some extension of the elaborate plan, treatment. The MAG whether argument, only if the valid EEOC’s basic prompted by the threat of enforcement or presumed In- statistical model valid. by Sears’ disinterested dedication to the stead, this distinction is an additional equal opportunity, ideal of in- assessing means of the coherence of Sears’ tended to increase the numbers of women conflicting interpretations and the EEOC’s jobs. II, in male-dominated See Sears of slipperiness the numbers. Given the F.Supp. at 1293. objective provides This qualifications the interest and variables on looking reasonable basis for at the differ- heavily, which Sears relies so the court hiring ences in peri- rates the between two employ every should available means of sharp improvement ods. A in the success Instead, assessing explanation. Sears’ the applicants implementa- women after the majority uses its conclusions as blinders to change hiring practices tion of a in conveys arguments block inconvenient from view. information, valuable whether or not Sears unpersuaded I am by majority’s also the left evidence guilty that a conscience moti- argument declining alternative for to con- change.3 vated the sider the early period separately. The ma- assessing In significance the of the MAG jority temporarily suspends its disbelief in plan, I find that the numbers show the approach EEOC’s overall statistical sharp increase in applicants’ women disparity consider whether the com- between pre-MAG statistics, mission sales post-MAG after 1974. if For exam- valid- measured, ple, ly represented women justify separate would consid- full-time 9.9% eration of early years. the commission respect, sales hires nationwide in majority approves 1974; in court’s the corresponding 17.6% annu- finding that the figures EEOC failed to period establish al for through from 1974 either adopted that Sears pro- ranged the MAG from 30.7% 40.5%. EEOC gram response EEOC’s enforce- Exhibit 1 at Table 1. Similar differ- course, possible, 2. It is plan of discrimination, that MAG did knew per- which candidates would become low not eliminate instead re- perfect formers. Sears assumed an almost cor- placed process an evenhanded with one that qualifications relation of .99 between assessed affirmatively strongest favored women. Sears' performance. Its results do not hold if interpretation evidence for this is its data com- more realistic correlations are assumed. EEOC paring job performance for male and female Exhibit, Testimony Siskind’s Rebuttal at 21-22 people. commission sales Sears Exhibit 6-3-2. (written testimony Siskind); of Dr. Bernard R. majority rely Both the district court and the on 17,407-09 (trial Siskind). testimony Tr. at of Dr. figures acceptance these quali- in their of Sears’ argument. fications Nothing 3. majority the cases cited 1320-22; supra at 338-39. separate periods suggests plain- issue that a unpersuasive. I find these expert data seeking distinguish periods tiff before and only comparisons conceded that between men after a corrective action must demonstrate that percent and women in the bottom ten of candi- employer acknowledged past illegality when provide persuasive dates who are hired implemented See, changes. e.g., Team- 14,616-19, 18,494 of affirmative action. Tr. States, 324, 341-42, sters v. United (testimony Wise). of Dr. David A. Sears’ data 1843, 1857-58, (1977) (cor- S.Ct. 52 L.Ed.2d 396 compare percent performers, the bottom ten however, prior rective action does not excuse discrimina- percent not the bottom ten of candi- tion); Corp., dates Rich hiring. as assessed at v. Martin the time of Marietta Inferior performance among Cir.1975) percent (improvement bottom ten after fil- women ing imply charges sellers prove would pre-filing affirm- does not lack of hiring process ative action in the discrimination). if Sears explained by these part-time over time cannot data on reflected enees are (three appear of which did not regions. five four factors for four of Sears’ hires 1977) changes social byor the broad until questions the EEOC’s com- majority affected women’s role in that Sears claims peri- the earlier and later parisons between during the 1970’s.4 The force work First, majority grounds. od on two the unmeasured differ- majority has vested years the data for calendar observes subjectively ences in interest assessed correspond to the 1974 do not 1973 and power qualifications explain with the from March “early period,” which ran disparities sharp swings both vast July avail- through June or best disparities. majority those thus seems “early years” there- numbers for the able having to credit Sears with devised an all- that ante- fore include two months machina, purpose deus ex invulnerable to liability period or six dated the and five by any argument quantifia- refutation imple- full months of 1974 that followed factors. ble num- plan. of the MAG Precise mentation month for the sixteen to seventeen bers respect promotions, majority With strength- no period question would doubt again attempts to dismiss inexorable argument separate en the EEOC’s by falling highly numbers back on dubious see, however, I how this treatment. do not assumptions disparate about levels imagi- by any stretch of the mismatch can of men “interest” and women commis- negate sharp differences be- nation selling. sion The EEOC tried to meet the periods, particularly the two since tween objection adjusting great- “interest” early including the months in the late-1974 1.5, 2, er male interest to the extent of disparities years should reduce the majority even 3 times female interest. The periods. the earlier and later between dénigrate proof emerging to seeks by arguing majority adjustments some of from these also observes that comparisons by may specific presented the men be more than three times as selling EEOC tend to overstate the differences interested in commission women. least, periods. puts Supra majority the earlier and later at 335. At between Su- agree I pra pre- at 316. that the EEOC’s on the EEOC to that the burden show point on this is more than sentation not a model interest differential three However, clarity. approach vividly differ- to one. This illus- undeniable Id. (differences periods virtually ences between the two trates insurmountable ob- probably by imple- that are stacle differential interest defense understated that the way proof mentation place of MAG Plan well before the can a case of period) appear very end of the 1973-74 to me to this sort. “Interest” its nature is a compel something fuzzy hardly a conclusion quality funda- lends itself say mental occurred. The district court attrib- mathematical demonstration. To during uted the increase in hires women men are three times as interested as wom- liability period changes selling rough attempt that made en in commission is a women, jobs dynamic phenomenon; these more attractive to in- social describe a cluding part-time precise an increase in hardly susceptible commis- it is a matter jobs, sion sales the introduction of base A three seems to delineation. factor of force, plausibility salaries for the commission sales in- near the the ma- limits of creasing availability jority’s of female “role mod- demand that the EEOC rule out the possibility multipliers greater els” commission sales and the in- than three availability day greater proving creased care. rather than Sears these Sears F.Supp. sharp disproportions at 1309. The me In differences strikes as absurd. "refuting place 4. Sears accuses the EEOC of the entire ward home life and the work and the by refusing accept availability Impor- women’s movement” so- increased of birth control. been, change during complete cial I do not 1970’s as a tant as these factors have improved prospects years for account women’s at think that three account two to Appellee’s sharp improvements by gradual Sears. Brief at lev- 98. The factors followed changes eling include cites to- off. women’s attitudes addition, exception admittedly changing sharp majority’s levels of view that during regression feminine the 1970’s—not- “interest” failed to ac- EEOC’s take majority— enough and the independent ed the district court count of variables. majority’s objections make the to the Assuming comparability jobs, basic promotion nitpick- statistics seem took every account of reason- ing passed If unrealistic. what arguable ably variable and the still results early lack of interest 1970’s was decisively showed women the losers. opportunity, really subsequent lack of summary, this case was tainted hardly shift in the numbers should redound origins and was tried without recourse to Catlett, to the benefit of Sears. See flesh and victims. The blood statistical evi- F.2d at 1265-66. quite strong dence is nonetheless aspects majority opin- Two other majority grossly has misconstrued and ion, the treatment of the pregnancy leave overstated its frailties. claim, wage issue and the discrimination I respectfully must therefore dissent as brief comment. warrant The Sears Person- to the matters I have discussed. granted day nel paid Manual leave to employees gave male when their wives

birth denied this benefit female em- giving

ployees birth. The district court

held that the Personnel Manual did not prima

make out a facie case of discrimina-

tion showing without a further that these provisions actually implemented. TEAMSTERS LOCAL 282 PENSION II, 628 majori- F.2d 1275-76. The FUND, Plaintiff-Appellant, TRUST ty reasoning. affirms Supra 353- ANGELOS, Anthony al., G. et majority concedes authority Defendants-Appellees. of Durant v. Fiberglass, Owens-Illinois (E.D.La.1980), principal No. 87-1084. support for the startling result reached United States Appeals, Court of here, “perhaps weak.” I would charac- Seventh Circuit. terize authority virtually Durant’s non- existent since employer the defendant Argued Sept. 1987. presented that case evidence that contract Decided Jan. provisions, appeared to establish guidelines pregnancy stricter leave

than brought for sick by illness, leave routinely subordinated recommen-

dations of personal physicians. women’s

Id. at 723. presented analogous no majority’s

evidence. The insistence that prove implemented that Sears policies

official written illustrates how the

majority heightens its demands on the beyond reason and common sense.

Finally, wage on the discrimination

claim, agree I would existing that under Hay job

case law the Method descriptions probably adequate to show substan- equality

tial for purposes of

Equal Pay analysis. Act Therefore, this would, I however,

claim must fail. take

Case Details

Case Name: Equal Employment Opportunity Commission, Cross-Appellee v. Sears, Roebuck & Company, Cross-Appellant

Court Name: Court of Appeals for the Seventh Circuit

Date Published: Mar 15, 1988

Citation: 839 F.2d 302

Docket Number: 86-1519, 86-1621

Court Abbreviation: 7th Cir.