EEOC v. Sears, Roebuck & Co., 628 F. Supp. 1264 (N.D. Ill. 1986)
January 31, 1986
SEARS, ROEBUCK & CO., Defendant.
United States District Court, N.D. Illinois, E.D.
*1265 *1266 *1267 *1268 *1269 *1270 Karen H. Baker, Gerald D. Letwin and James P. Scanlan, E.E.O.C., Chicago, Ill., for plaintiff.
Charles Morgan, Jr., Pamela S. Horowitz, Howard T. Anderson and Paul F. Colarulli, Morgan Associates, Chartered, Washington, D.C., for defendant.
ON MOTION FOR PARTIAL SUMMARY JUDGMENT
NORDBERG, District Judge.
This matter is before the court on the motion of plaintiff Equal Employment Opportunity Commission ("EEOC") for partial summary judgment. For reasons set forth below, the EEOC's motion is denied.
In its motion, the EEOC moved the court to grant partial summary judgment against defendant Sears, Roebuck and Co. ("Sears") as to liability on the following claims:
1. Sears violated Title VII of the Civil Rights Act of 1964, § 701 et seq., as amended, 42 U.S.C. § 2000e et seq. ("Title VII"), by limiting the working *1271 hours of women pursuant to state protective laws.
2. Sears violated Title VII by maintaining a policy of refusing to hire pregnant applicants.
3. Sears violated Title VII by maintaining a policy in its Personnel Manual whereby women absent due to pregnancy-related disabilities were subject to lay-off during such absenses, but persons absent because of other temporary disabilities were not subject to lay-off during such absences.
4. Sears violated Title VII by maintaining a policy in its Personnel Manual of involuntary transfer of pregnant women when appearance was deemed to be a factor in the performance of the job.
5. Sears violated Title VII by maintaining a policy in its Personnel Manual of granting to a male employee a day's paid absence when the employee's wife gave birth, but not granting a day's paid absence to a female employee when she gave birth.
The period of liability for all of these claims commenced on August 30, 1971, two years prior to the filing of EEOC's charge against Sears. See infra n. 10.
The EEOC contended in its motion that there was no genuine issue of material fact as to Sears' liability on each of the above five claims. The EEOC submitted, in support of its motion, a summary of Sears' responses to the EEOC's First Request for Information and the affidavit of Jane L. Dolkart, the EEOC's Director of the Litigation Enforcement Division of the Office of Systematic Programs, with attached Exhibits.
In its Response to the EEOC's motion for partial summary judgment, Sears asserted that: (1) each of the EEOC's five claims was based upon a long-discontinued policy; (2) material issues of fact existed with respect to each claim; (3) the EEOC was not entitled to summary judgment as a matter of law, because it failed to meet its initial burden of demonstrating that unlawful discrimination had been a regular procedure or policy followed by Sears; and (4) the injunctive relief sought by the EEOC was moot and the monetary relief barred by the doctrine of laches. Sears attached to its Response various documents refuting the *1272 EEOC's claims that Sears violated Title VII by limiting the hours of female employees pursuant to state protective laws, by maintaining a policy of refusing to hire pregnant applicants, by maintaining a policy whereby women absent due to pregnancy-related disabilities were guaranteed less protection from reductions in force than employees receiving disability benefits, and by maintaining a policy of involuntary transfer of pregnant women.
The EEOC, in its Reply to Sears' Response, asserted that the fact that all of the alleged discriminatory policies had ceased was not a defense. The EEOC also asserted that it need not produce the number and identity of the persons affected by the policies in the Personnel Manual because, absent evidence by Sears that the policies were not implemented, the necessary presumption is that the policies were implemented and did affect individuals. In addition, the EEOC contended that both injunctive and monetary relief were appropriate, and that there was no genuine issue of material fact as to Sears' liability on each of the claims, except for the claim that Sears violated Title VII by limiting the working hours of women pursuant to state protective laws. The EEOC attached to its Reply a Sears Directive which ordered that the policy of allowing a day off with pay for male employees whose wives gave birth be discontinued effective October 1, 1974.
After the EEOC filed its Reply Brief, on April 7, 1982, the EEOC produced a Statement of Issues. All five of the claims on which the EEOC sought a partial summary judgment were included in the April 7, 1982 Statement of Issues. The EEOC submitted another Statement of Issues on October 15, 1982. In the October 15, 1982 Statement of Issues, the EEOC omitted the claims that Sears violated Title VII by limiting the working hours of women pursuant to state protective laws and by maintaining a policy of refusing to hire pregnant applicants. Therefore, the EEOC had withdrawn these claims against Sears. This is confirmed in the transcript.
*1273 Subsequently, in an order dated January 27, 1984, the court stated the following:
The April 7, 1982 statement was the clarification of the definitive statement of issues of December, 1981 ordered by Judge Grady to be the final statement of issues in the case. EEOC will not be permitted to expand its case beyond the specific job categories listed in the April 7, 1982 statement.
Memorandum Opinion and Order of January 27, 1984, p. 3. Thus, the court invalidated those portions of the October 15, 1982 Statement of Issues which added claims not included in the April 7, 1982 Statement. However, the court did not invalidate the portions of the October 15, 1982 Statement of Issues which withdrew claims included in the April 7, 1982 Statement. Therefore, the EEOC's withdrawal stands, and the court denies the EEOC's motion for partial summary judgment as to these two claims, because they have been withdrawn by the EEOC.
The remaining claims on which the EEOC seeks partial summary judgment are that Sears violated Title VII by: (1) maintaining a policy whereby women absent due to pregnancy-related disabilities were guaranteed less protection from reductions in force than employees receiving disability benefits; (2) maintaining a policy of involuntary transfer of pregnant women; and (3) maintaining a policy of granting to a male employee a day of paid absence when the employee's wife gave birth, but not granting a day of paid absence to a female employee when she gave birth.
Summary Judgment In a Title VII Disparate Treatment Action
The EEOC has moved for partial summary judgment as to Sears' liability on these three claims. On a motion for summary judgment, the moving party has the burden of establishing that there is no genuine issue of material fact and that it is entitled to judgment as a matter of law. Cedillo v. International Association of Bridge and Structural Iron Workers, 603 F.2d 7, 10 (7th Cir. 1979). The non-moving party is entitled to all reasonable inferences that can be made in its favor. United States v. Diebold, Inc., 369 U.S. 654, 655, 82 S. Ct. 993, 994, 8 L. Ed. 2d 176 (1962) (per curiam); Matthews v. Allis-Chalmers, 769 F.2d 1215 (7th Cir.1985) (per curiam). However, the non-moving party may not merely rely on conclusory pleadings to withstand summary judgment. In responding to a motion for summary judgment, a non-moving party must set forth specific facts in affidavits or otherwise show that there are genuine issues that must be decided at trial. First National Bank of Arizona v. Cities Service Co., 391 U.S. 253, 88 S. Ct. 1575, 20 L. Ed. 2d 569 (1968), reh. denied, 393 U.S. 901, 89 S. Ct. 63, 21 L. Ed. 2d 188 (1968); Posey v. Skyline Corp., 702 F.2d 102, 105 (7th Cir.1983), cert. denied, 464 U.S. 960, 104 S. Ct. 392, 78 L. Ed. 2d 336 (1983); Fed. R.Civ.P. 56(e).
On a motion for summary judgment in a Title VII case, consideration of whether the moving party has shown that there are no genuine issues of material fact and that it is entitled to judgment as a matter of law necessarily involves an examination and application of the Title VII burden of proof analytical framework. In a recent case, Coates v. Johnson & Johnson, 756 F.2d 524 (7th Cir.1985), the Seventh Circuit outlined the Title VII burden of proof framework for a government or class action alleging a pattern or practice of disparate treatment. The Seventh Circuit first stated that the general Title VII burden of proof framework in government or class disparate treatment cases is "essentially comparable" to the framework for individual disparate treatment actions, as outlined in McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S. Ct. 1817, 36 L. Ed. 2d 668 (1973) and Texas Department of Community Affairs v. Burdine, 450 U.S. 248, 101 S. Ct. 1089, 67 L. Ed. 2d 207 (1981), but "the content of the specific stages of that framework will be different." Coates, 756 F.2d at 532.
The general Title VII burden of proof framework for disparate treatment cases, as set out in McDonnell and Burdine, consists of the following stages: (1) *1274 the plaintiff has the burden of proving by the preponderance of the evidence a prima facie case of discrimination; (2) if the plaintiff succeeds in proving the prima facie case, the burden shifts to the defendant to articulate some legitimate, nondiscriminatory reason for its action; and (3) should the defendant carry this burden, the plaintiff must then have an opportunity to prove by a preponderance of the evidence that the legitimate reasons offered by the defendant are merely a pretext for discrimination. Burdine, 450 U.S. at 252-253, 101 S. Ct. at 1093. In private, nonclass Title VII disparate treatment actions such as McDonnell and Burdine, in order to establish a prima facie case of discrimination, the plaintiff must prove by a preponderance of the evidence that he or she applied and was qualified for an available position, but he or she was rejected under circumstances which give rise to an inference of unlawful discrimination. Burdine, 450 U.S. at 253, 101 S. Ct. at 1094; McDonnell, 411 U.S. at 802, 93 S. Ct. at 1824. In a government or class disparate treatment action, on the other hand, in order to establish a prima facie case of discrimination, the plaintiffs must demonstrate "that unlawful discrimination has been the regular policy of the employer, i.e., that `discrimination was the company's standard operating procedurethe regular rather than the unusual practice.'" Coates, 756 F.2d at 532 (quoting International Brotherhood of Teamsters v. United States, 431 U.S. 324, 336, 360, 97 S. Ct. 1843, 1855, 1867, 52 L. Ed. 2d 396 (1977)).
On motions for summary judgment in Title VII disparate treatment cases, the courts have defined the burdens of the moving and non-moving parties in terms of the plaintiff's establishment of a prima facie case, the defendant's articulation of a nondiscriminatory reason, and the plaintiff's demonstration of pretext. In other words, courts have defined the burdens of parties on motions for summary judgment in terms of the general Title VII disparate treatment burden of proof framework. For example, in cases where defendant-employers have moved for summary judgment, the Seventh Circuit has held that a plaintiff-employee, in order to defeat the summary judgment motion, must present facts sufficient to establish a prima facie case, and, if the defendant-employer presents some legitimate reason for its action, the plaintiff-employee must also present evidence that the legitimate reason is a pretext for discrimination in order to defeat the summary judgment motion. Parker v. Federal National Mortgage Association, 741 F.2d 975 (7th Cir.1984); Kephart v. Institute of Gas Technology, 630 F.2d 1217 (7th Cir.1980), cert. denied, 450 U.S. 959, 101 S. Ct. 1418, 67 L. Ed. 2d 383 (1981). See also Trembath v. St. Regis Paper Company, 753 F.2d 603 (7th Cir. 1985); Herman v. National Broadcasting Company, Inc., 744 F.2d 604 (7th Cir.1984), cert. denied, ___ U.S. ___, 105 S. Ct. 1393, 84 L. Ed. 2d 782 (1985); Huhn v. Koehring Company, 718 F.2d 239 (7th Cir.1983).
In Weahkee v. Perry, 587 F.2d 1256 (D.C. Cir.1978), a Title VII disparate treatment action, the plaintiff-employee moved for summary judgment, and the District of Columbia District Court entered judgment for the plaintiff-employee. However, the Court of Appeals reversed, holding, inter alia, that there were genuine issues of material fact precluding summary judgment. The Court of Appeals stated:
It is elementary that summary judgment is granted properly only where no material fact is genuinely in dispute and only if the movant is entitled to judgment as a matter of law. ... In this case Weahkee bore the initial burden to present a prima facie case of discrimination.
Weahkee, 587 F.2d at 1265 (citations omitted). Thus, the Court of Appeals found that a plaintiff-employee must at least present a prima facie case of discrimination in order to prevail on a motion for summary judgment.
Applying the above analysis to the present case, the court finds that the EEOC must at least establish a prima facie case of discrimination in order to prevail on its motion for summary judgment. Under Coates, that prima facie case consists of a demonstration that discrimination was *1275 Sears' "regular policy" or "standard operating procedure."
Establishment of a Title VII Pattern or Practice Prima Facie Case
The EEOC has established that, as to each of the three claims still remaining in its motion for summary judgment, there were included in Sears' written materials allegedly discriminatory policy statements which were in existence during the relevant time period. However, in its pleadings, EEOC does not identify any victims of the policies, nor does it present any statistical evidence demonstrating discriminatory treatment of pregnant employees. The EEOC contends that it need not identify victims of the policies in this liability phase of the pattern or practice action because the policies are clearly unlawful, and that, "absent evidence by Sears that the policies were not implemented, the necessary presumption is that the policies were implemented and did affect individuals." Plaintiff EEOC's Reply Brief, p. 2.
In Durant v. Owens-Illinois Glass Co., Inc., 517 F. Supp. 710 (E.D.La.1980), aff'd, 656 F.2d 89 (5th Cir.1981) (per curiam), the plaintiffs similarly contended that they need not identify victims of a maternity leave policy in order to prove liability under Title VII, because the policy was explicitly set out in their Production and Maintenance Contract and was clearly discriminatory. In that case, the Special Master had concluded that, even if the maternity leave provision did not comport with Title VII, the plaintiffs still had no claim under Title VII, because there was no evidence that the provision had ever been enforced against or adversely affected anyone in the class. The plaintiffs challenged the Special Master's interpretation of their burden of proof in the District Court. The plaintiffs contended that all they need to prove in the liability stage of the litigation is the existence of a discriminatory policy, and that identification of those harmed by the policy should be postponed until the relief stage. The District Court held:
Plaintiffs' description of their burden is correct, but the fact is they have failed to prove the existence of a policy. They have been unable to show that anyone was ever affected by the attacked provision, which was never enforced and which ceased to be a part of the collective bargaining agreement six years ago. A provision which was never applied and now cannot be applied to a single individual does not constitute a discriminatory policy. Plaintiffs have failed to meet their burden of proving liability in this area.
Durant, 517 F. Supp. at 723.
As in Durant, the EEOC has failed to show that Sears ever enforced *1276 these three policies against pregnant employees or that the policies adversely affected pregnant employees in any way. Although the court agrees with the Durant court that identification of all victims of the policies should be postponed until the relief stage of the litigation, the court also agrees with the Durant court's holding that plaintiff must show the existence of a discriminatory policy that affected actual employees in the liability stage of the case. Here, the EEOC failed to provide any evidence that the policies were ever enforced. While the EEOC need not identify every victim of the policies, it does need to identify some victims or produce some valid direct or circumstantial evidence to show that Sears actually followed the policies in its treatment of employees. Absent some evidence of such action by Sears, the EEOC has failed to establish a prima facie case of discrimination.
The court notes that the holding of the Durant court is in accord with Teamsters, the leading Supreme Court case dealing with the burden of proof framework in a government or class Title VII action alleging a pattern or practice of disparate treatment, and Coates, the recent Seventh Circuit case following Teamsters. In Teamsters, the Supreme Court held:
The plaintiff in a pattern-or-practice action is the Government, and its initial burden is to demonstrate that unlawful discrimination has been a regular procedure or policy followed by an employer or group of employers.... At the initial, "liability" stage of a pattern-or-practice suit the Govermnent is not required to offer evidence that each person for whom it will ultimately seek relief was a victim of the employer's discriminatory policy. Its burden is to establish a prima facie case that such a policy existed.
Teamsters, 431 U.S. at 360, 97 S. Ct. at 1867. The Supreme Court went on to hold that the Government's evidence, statistics and testimony by forty individuals of specific instances of discrimination, was sufficient to establish a prima facie case that a racially discriminatory policy existed in fact.
Although Teamsters and Coates did not involve written discriminatory policies, as did Durant and the present case, the Supreme Court in Teamsters and the Seventh Circuit in Coates did not in any way indicate that the requirement of establishing the existence of a discriminatory policy would differ in a case where such a policy was referred to in an employment contract or a personnel manual. The court finds no basis in Teamsters, Coates, or any other case, for the EEOC's theory that the existence of written discriminatory policy language creates a presumption that the policy was enforced and did affect individuals.
For the reasons stated above, the court finds that the EEOC has failed to establish a prima facie case that the three allegedly discriminatory policies in fact existed during the relevant time period. The EEOC presented no evidence that Sears ever enforced any of the policies. Therefore, the court denies the EEOC's motion for partial summary judgment.
Since the EEOC has previously stated that it wishes to dismiss these claims with prejudice should its motion for summary judgment be denied, the court hereby dismisses these claims with prejudice.
*1277 MEMORANDUM OPINION AND ORDER TABLE OF CONTENTS I. Introduction 1278 A. Jurisdiction 1279 B. Legal Standards 1279 1. Disparate Treatment 1279 2. Disparate Impact 1281 C. Statistics 1285 1. In General 1285 2. Statistical Significance 1286 3. Multiple Regression 1287 II. Hiring and Promotion Into Commission Sales 1288 A. General Background 1288 1. Commission Sales at Sears 1289 2. Qualifications for Commission Sales Positions 1290 3. The Hiring Process 1291 4. Affirmative Action at Sears 1292 B. EEOC's Evidence: Hiring Into Commission Sales 1294 1. Data Bases 1294 2. Statistical Analyses - Unadjusted Analysis 1295 3. Adjusted Analyses 1296 4. Possible Biases 1298 5. Other Statistical Evidence 1299 6. Nonstatistical Evidence 1300 C. EEOC's Evidence: Promotions Into Commission Sales 1300 1. Year-end Store Pool Analysis 1300 2. Year-end Division Pool Analysis 1301 D. Analysis of EEOC's Evidence: Hiring 1301 1. Inadequacies in Data and Variables Analyzed by EEOC 1301 a. Data Analyzed 1301 b. Omission and Inadequate Coding of Important Variables 1302 c. Coding Errors 1304 2. Faulty Basic Assumptions 1305 a. Assumption of Equal Interest 1305 (1) Store Witnesses 1306 (2) Survey Evidence 1308 (3) Comparing Sears to National Data 1312 (4) Other Evidence of Interest 1314 b. Assumption of Equal Qualifications 1315 3. Other Deficiencies in EEOC's Statistical Analysis 1315 4. EEOC's Applicant Interview Guide Analysis 1317 5. EEOC's Nonstatistical Evidence 1317 E. Sears' Commission Sales Analyses: Hiring 1318 1. Hiring and Promotion Figures at Sears 1319 2. Characteristics of Commission Salespersons 1319 3. Sales Performance Data 1320 4. Applicant Interview Guides 1322 F. Conclusions Regarding Commission Sales Claim: Hiring 1324 G. Promotion Claim 1324 1. EEOC's Promotion Analysis 1325 2. Sears' Evidence Regarding Promotions 1326 H. Conclusions Regarding Commission Sales Claim: Promotions 1327 III. Checklist Compensation 1328 A. Legal Standards 1328 1. Elements and Burdens of Proof 1328 2. "Similar" Work Standard 1332
*1278 B. Background 1334 1. Pre-1976 Organization and Compensation 1335 2. New Executive Compensation Program 1336 C. EEOC Evidence 1337 1. Persons and Jobs Analyzed 1338 2. Statistical Analyses 1340 D. Deficiencies in EEOC's Evidence 1340 1. 1973-1975 Analysis 1340 2. 1976-1980: Equality of Jobs 1341 3. Overall Statistical Analyses 1342 a. Data Base 1343 b. Omitted Variables 1343 c. Other Flaws in EEOC Models 1345 E. Sears' Evidence 1346 1. Regression Analysis 1346 2. Cohort Analysis 1350 3. Sears' Witnesses 1352 F. Conclusion: Checklist Compensation 1352 ORDER 1353
This opinion marks the culmination of a lengthy dispute between the Equal Employment Opportunity Commission ("EEOC") and Sears, Roebuck & Co. ("Sears"), the world's largest retail seller of general merchandise. In 1973, an EEOC commissioner's charge was filed. After an extensive investigation and extensive conciliation discussions, EEOC filed this suit in 1979, alleging nationwide discrimination by Sears against women in virtually all aspects of its business, in violation of Title VII of the Civil Rights Act of 1964, as amended, 42 U.S.C. § 2000e et seq. ("Title VII"), and the Equal Pay Act, 29 U.S.C. § 206(d).
This is a case of claimed statistical disparities. As originally filed, this suit involved 42 distinct claims of nationwide sex discrimination by Sears. By the time of trial, EEOC had abandoned all of its Equal Pay Act claims, and all but two of its claims under Title VII. The two allegations EEOC sought to prove at trial were that Sears engaged in a nationwide pattern or practice of sex discrimination: (1) by failing to hire female applicants for commission selling on the same basis as male applicants, and by failing to promote female noncommission salespersons into commission sales on the same basis as it promoted male noncommission salespersons into commission sales (commission sales claim); and (2) by paying female checklist management employees in certain job categories lower compensation than similarly situated male checklist management employees (checklist compensation claim). The allegations are limited to the time period beginning March 3, 1973 and ending December 31, 1980.
A trial before the court was held lasting 10 months. The court observed each witness, made extensive contemporaneous trial notes of the testimony of each witness, and made specific determinations of the credibility of each witness and the weight to be given to the testimony as the witness testified. The court has drawn what it *1279 believes to be the correct reasonable inferences from this evidence and has evaluated the legal principles presented by the parties and developed through legal research of the court.
Based on all of the testimony presented, the credibility of the witnesses and the weight to be given their testimony, the exhibits received in evidence, and the law governing this case, the court concludes that the EEOC has failed to prove its case on either claim of discrimination and finds that Sears has not discriminated against women in hiring, promotion, or pay, as claimed. The court makes the following findings of fact and conclusions of law in accordance with Rule 52(a) of the Federal Rules of Civil Procedure:
The court finds that it has subject matter as well as personal jurisdiction, pursuant to Title VII of the Civil Rights Act of 1964, as amended, 42 U.S.C. § 2000e et seq. The court adheres to the prior decisions in this case, including Equal Employment Opportunity Commission v. Sears, Roebuck & Co., 504 F. Supp. 241 (D.C.Ill.1980). The EEOC has presented sufficient evidence to support this court's finding that EEOC has met all the minimum statutory preconditions to this suit.
B. Legal Standards
The court must first determine the legal standards to apply to this case. Two separate legal analyses are applied to Title VII cases, disparate treatment and disparate impact. Each has its distinct elements and burdens of proof.
In this case, the EEOC relies principally on the disparate treatment theory. (Plaintiff's Final Argument.) However, the EEOC has also made a number of vague references to the disparate impact theory. The EEOC failed to specify this legal theory of its case prior to trial. Even at the close of trial, the extent of its reliance on the disparate impact theory was unclear. EEOC stated in its trial brief and on several other occasions that, although it contends that the disparate impact model should be applied to "subjective" hiring systems such as Sears', "the characterization of the alleged discrimination as disparate impact or disparate treatment ought not affect the outcome here." (Plaintiff's Final Argument at 2.) However, the choice of theory can make a difference, since intent to discriminate need be proved only under the disparate treatment theory, not under the disparate impact theory. Despite EEOC's contention that the choice of theory is unimportant, and its decision to discuss all the evidence in terms of disparate treatment analysis only, EEOC apparently now seeks application of both theories to its case. Therefore, both the disparate treatment and disparate impact theories must be discussed.
1. Disparate Treatment
The disparate treatment theory evolved under § 703(a) (1) of Title VII, 42 U.S.C. § 2000e-2(a) (1). Under this theory, *1280 employers are prohibited from treating an employee less favorably than the employee's peers because of the employee's sex, race, color, religion, or national origin. McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S. Ct. 1817, 36 L. Ed. 2d 668 (1973). In a disparate treatment case, a plaintiff must prove discriminatory intent. Texas Department of Community Affairs v. Burdine, 450 U.S. 248, 256, 101 S. Ct. 1089, 1095, 67 L. Ed. 2d 207 (1981). This is the most common type of discrimination claim. It may be brought by an individual, or as a "pattern and practice" claim alleging systemic disparate treatment of a protected group. See, e.g., Hazelwood School District v. United States, 433 U.S. 299, 97 S. Ct. 2736, 53 L. Ed. 2d 768 (1977); Coates v. Johnson & Johnson, 756 F.2d 524 (7th Cir.1985).
The Supreme Court has provided an analytical framework for disparate treatment cases "progressively to sharpen" the inquiry into the defendant's intent. Burdine, 450 U.S. at 255 n. 8, 101 S. Ct. at 1094 n. 8. To establish a prima facie case of intentional discrimination, the plaintiff must prove that he is a member of the protected class, has substantially the same qualifications as those not in protected classes, but was not treated on an equal basis by the employer. Id.; Coates, 756 F.2d at 531. A plaintiff's prima facie case creates a rebuttable presumption that the defendant unlawfully discriminated against the plaintiff. Burdine, 450 U.S. at 254-55, 101 S. Ct. at 1094.
The defendant must then offer evidence that "raises a genuine issue of fact as to whether the defendant discriminated against the plaintiff." Id. This burden may be met by articulating a legitimate, nondiscriminatory reason for the defendant's employment action. Burdine, 450 U.S. at 254, 101 S. Ct. at 1094; McDonnell Douglas, 411 U.S. at 802, 93 S. Ct. at 1824. If the explanation provided by the defendant is "legally sufficient to justify a judgment for the defendant," then "the presumption raised by the prima facie case is rebutted, and the factual inquiry proceeds to a new level of specificity." Burdine, 450 U.S. at 255, 101 S. Ct. at 1094-95 (footnote omitted). The plaintiff then must persuade the court that the defendant's explanation is unworthy of credence, or that a discriminatory reason more likely motivated the defendant. Id. at 256, 101 S. Ct. at 1095. Importantly, the plaintiff always retains the ultimate burden of persuasion. Id.
This analytical framework is applied, in modified form, to government or class actions alleging a pattern or practice of discrimination. The burden of proof is "essentially comparable" to that applied in individual disparate treatment cases, but "the content of the specific stages of that framework will be different." Coates, 758 F.2d at 532. In a pattern or practice class claim, the plaintiff has the initial burden of demonstrating that unlawful discrimination has been the regular policy of the employer, that "discrimination was the company's standard operating procedurethe regular rather than the unusual practice." International Brotherhood of Teamsters v. United States, 431 U.S. 324, 336, 360, 97 S. Ct. 1843, 1854, 1867, 52 L. Ed. 2d 396 (1977); Coates, 756 F.2d at 532. The focus often is on a pattern of discriminatory decisionmaking, not on individual employment decisions. Teamsters, 431 U.S. at 360 n. 46, 97 S. Ct. at 1867 n. 46. The plaintiff's prima facie case in a pattern or practice case usually consists of statistical evidence of substantial disparities between the protected group and the unprotected group, buttressed by evidence of general discriminatory policies or specific instances of discrimination. Coates, 756 F.2d at 532.
*1281 Once the plaintiff establishes a prima facie case in a pattern and practice case, the burden shifts to the employer to defeat the showing by demonstrating that the plaintiff's proof is inaccurate or insignificant, or by providing a nondiscriminatory explanation for the apparently discriminatory result. Teamsters, 431 U.S. at 360 n. 46, 97 S. Ct. at 1867 n. 46. See also Dothard v. Rawlinson, 433 U.S. 321, 338-39, 97 S. Ct. 2720, 2731-32, 53 L. Ed. 2d 786 (1977) (Rehnquist, J., concurring). The strength of the evidence the defendant must produce depends on the strength of the plaintiff's proof. Coates, 756 F.2d at 532; Segar v. Smith, 738 F.2d 1249, 1268 (D.C.Cir.1984), cert. denied sub nom., Meese v. Segar, ___ U.S. ___, 105 S. Ct. 2357, 86 L. Ed. 2d 258 (1985). However, at the liability stage of a pattern or practice suit, the plaintiff always bears the burden of persuasion with respect to discrimination. Coates, 756 F.2d at 532.
2. Disparate Impact
The disparate impact theory evolved under Section 703(a) (2) of Title VII, 42 U.S.C. § 2000e-2(a) (2). The seminal case interpreting this provision is Griggs v. Duke Power Co., 401 U.S. 424, 91 S. Ct. 849, 28 L. Ed. 2d 158 (1971). In Griggs, the Court held that this provision prohibits employers from applying policies or job requirements which, although neutral on their face, disproportionately disqualify members of a protected class from employment opportunities. Proof of discriminatory motive is not required under this theory. If a plaintiff can prove that a job requirement or policy impacts disproportionately on a protected group, and the requirement or policy is not related to job performance, it violates Title VII. Examples of policies found to violate § 703(a) (2) are educational requirements, Griggs, supra; intelligence tests, id.; and height and weight requirements, Dothard, supra.
To establish a prima facie case under the disparate impact theory, the plaintiff must show that a facially neutral standard for hiring excludes a disproportionate number of members of a protected class. The burden of proof then shifts to the defendant to show that the job requirement has a manifest relationship to the job in question. If the employer demonstrates that the challenged requirements are job related, the plaintiff may then show that other selection devices would also serve the employer's legitimate interest in efficient and trustworthy workmanship. Dothard, 433 U.S. at 329, 97 S. Ct. at 2727.
In this case, EEOC has admitted that it cannot identify any specific employment practice of Sears which has a discriminatory impact on women. Instead, EEOC contends that there is "something in the process" at Sears which causes the disparities shown in EEOC's statistical evidence. It asserts that the hiring and promotion "process" at Sears is highly subjective, and that the disparate impact analysis should be applied to such subjective systems.
EEOC relies on a number of cases in which courts applied the Griggs disparate impact analysis to overall hiring processes which are subjective in nature. For example, in Rowe v. General Motors Corp., 457 F.2d 348 (5th Cir.1972), decided shortly after the Griggs decision, the court applied the impact analysis to a "subjective" promotion system. However, the court made no mention of the disparate treatment theory, did not compare the two theories or discuss which should apply, and did not discuss why the disparate impact analysis would apply when a specific facially neutral employment policy had not been identified.
*1282 The Seventh Circuit, relying on Rowe, applied the Griggs analysis in similar circumstances in Stewart v. General Motors Corp., 542 F.2d 445 (7th Cir.1976), cert. denied, 433 U.S. 919, 97 S. Ct. 2995, 53 L. Ed. 2d 1105 (1977), reh. denied, 434 U.S. 881, 98 S. Ct. 244, 54 L. Ed. 2d 165 (1977). Without mentioning the disparate treatment theory or any reasons for applying either theory to the facts of the case, the court analyzed a "highly subjective and loosely structured" promotion system under the Griggs analysis. 542 F.2d at 450. Other cases applying the Griggs analysis to "subjective" systems have been equally unilluminating. E.g., Rowe v. Cleveland Pneumatic Co., 690 F.2d 88, 93 n. 10 (6th Cir.1982); Hung Ping Wang v. Hoffman, 694 F.2d 1146, 1148 (9th Cir.1982). None of these decisions analyze the United States Supreme Court cases applying the disparate impact analysis, or provide any justification for extending the application of the Griggs analysis beyond cases in which plaintiffs attack specific facially neutral employment practices which impact disparately on protected groups.
In all of its decisions under the disparate impact theory, the Supreme Court has applied the theory only to specific facially neutral employment policies, and has implicitly indicated that the theory should be applied only to specific neutral policies.
The Supreme Court first recognized the disparate impact theory in Griggs, supra. In Griggs, the Court found that the defendant's policy of requiring a high school diploma or a passing score on a standardized intelligence test, although neutral on its face, had the effect of disproportionately excluding black employees from promotions. The court held that proof of the disparate impact of the test or diploma requirement was enough alone, without proof of the defendant's motive, to establish a violation of Title VII under § 703(a) (2), 42 U.S.C. § 2000e-2(a) (2). The court reasoned that Congress intended to eliminate such "artificial, arbitrary and unnecessary barriers to employment." 401 U.S. at 431, 91 S. Ct. at 853.
As the Court later noted in Connecticut v. Teal, 457 U.S. 440, 102 S. Ct. 2525, 73 L. Ed. 2d 130 (1982), the focus of the Griggs analysis is on "procedures and testing mechanisms that operate as `built-in headwinds' for minority groups." 457 U.S. at 448, 102 S. Ct. at 2531 (quoting Griggs, 401 U.S. at 432, 91 S.Ct. at 854). When summarizing the elements and burdens of proof of a Griggs disparate impact case in Dothard, the Court stated, "[Griggs and Albermarle Paper Co. v. Moody, 422 U.S. 405, 95 S. Ct. 2362, 45 L. Ed. 2d 280 (1975) ] make clear that to establish a prima facie case of discrimination, a plaintiff need only show that the facially neutral standards in question select applicants for hire in a significantly discriminatory pattern." 433 U.S. at 329, 97 S. Ct. at 2726-27 (emphasis added). By specifically referring to "facially neutral standards," the Court indicated that the impact theory would apply only to such facially neutral policies.
Similarly, in Teamsters, supra, the Court defined the two theories as follows:
[d]isparate treatment ... is the most easily understood type of discrimination. The employer simply treats some people less favorably than others because of their race, color, religion, sex, or national origin. Proof of discriminatory motive is critical, although it can in some situations be inferred from the mere fact of differences in treatment.... Undoubtedly disparate treatment was the most obvious evil Congress had in mind when it enacted Title VII.
* * * * * *
Claims of disparate treatment may be distinguished from claims that stress "disparate impact." The latter involve employment practices that are facially neutral in their treatment of different groups but that in fact fall more harshly on one group than another and cannot be justified by business necessity.... Proof of discriminatory motive, we have held, is not required under a disparate-impact theory.
431 U.S. at 335 n. 15, 97 S. Ct. at 1854 n. 15. By once again defining disparate impact *1283 only in terms of facially neutral employment policies, the court indicated that other types of discrimination should be analyzed under the disparate treatment analysis.
In Griggs, Dothard, and indeed in all cases decided by the Supreme Court under the disparate impact theory, the Court has applied the theory only to specific employment practices that were neutral on their face, but had the effect of disproportionately excluding members of a protected class from employment opportunities. E.g., Albermarle, supra (employment test); Connecticut, supra (ability test); Dothard, supra (height and weight requirements); New York City Transit Authority v. Beazer, 440 U.S. 568, 99 S. Ct. 1355, 59 L. Ed. 2d 587 (1979) (employer's refusal to employ persons who use methadone).
Moreover, in McDonnell Douglas, supra, the Court refused to apply the disparate impact analysis to the plaintiff's claim because he could not identify a specific employment practice of the defendant which disproportionately excluded blacks. The Court distinguished the plaintiff's claim from Griggs because the employer did not seek to exclude the plaintiff "on the basis of a testing device which overstates what is necessary for competent performance," or through some "sweeping disqualification" of all applicants on the basis of a particular characteristic. 411 U.S. at 806, 93 S. Ct. at 1826. The Court concluded that the plaintiff had not presented the kind of "artificial, arbitrary and unnecessary barriers to employment" that violated Title VII under Griggs. The Court therefore refused to apply the impact theory when no specific policy had been identified.
All of the Supreme Court's discussions of the disparate impact theory indicate that it should be applied only to specific employment practices which are neutral on their face, but disproportionately exclude members of protected groups from employment opportunities. The Court has applied the theory only in these circumstances, and has never stated or suggested that the theory should be applied in other circumstances. Thus, there is no basis in Supreme Court decisions for applying the disparate impact theory to cases where the employer merely uses a "subjective" hiring system, as most employers do.
Several Courts of Appeals have specifically rejected the extension of the Griggs analysis to "subjective" hiring decisions. For example, in Pouncy v. Prudential Insurance Company of America, 668 F.2d 795 (5th Cir.1982), the same court which ten years earlier decided Rowe v. General Motors Corp., supra, rejected application of the disparate impact theory to a wide-ranged attack on a company's hiring system. In Pouncy, the plaintiff alleged that the defendant systematically failed to promote blacks within its workforce and otherwise afford them the same conditions of employment given to whites. He pointed to a number of employment policies which he alleged resulted in the discrimination, including the failure to post job openings, the selection of employees for promotion using minimal objective criteria, and the use of subjective criteria in employee performance evaluations. 688 F.2d at 799. He sought to apply the disparate impact analysis of Griggs to his challenge of the defendant's company-wide system of "subjective" employment decisions.
The court rejected the plaintiff's attempt to fit his case within the disparate impact model by listing specific "practices" of the defendant. It reviewed a number of Supreme Court decisions on disparate impact. Relying on Supreme Court precedent, the Court concluded that a prima facie case under the disparate impact model is shown by "identification of a neutral employment practice coupled with proof of its discriminatory impact on the employer's work force." 668 F.2d at 800. The court rejected use of the disparate impact model of proof for a wide-ranging attack on the cumulative effects of a company's employment practices, stating that, "The disparate impact model applies only when an employer has instituted a specific procedure, usually a selection criterion for employment, that can be shown to have a causal *1284 connection to a class based imbalance in the workforce." Id.
The court reasoned that proof of such a specific employment policy is required under the disparate impact model to fairly allocate the parties' respective burdens of proof at trial. Id. The plaintiff must first prove a disparate impact due to the selection procedure. Id. The employer then has the burden of proving that the selection procedure is justified by a legitimate business reason. Id. Identification of the specific practice is necessary so that the employer can respond with proof of its legitimacy. 668 F.2d at 801. A plaintiff cannot challenge an entire range of employment practices under the impact model merely because the employer's workforce reflects a racial imbalance that might be related to any one or more of several practices, because to do so "would allow the disparate impact of one element to require validation of other elements having no adverse effects." Id. (quoting Rivera v. City of Wichita Falls, 665 F.2d 531, 539 (5th Cir.1982)).
The court found that none of the employment practices singled out by the plaintifffailure to post job openings, use of a "level" system, and evaluating employees with subjective criteriawere akin to the facially neutral employment practices the disparate impact model was designed to test. Id. The court therefore applied the disparate treatment model to the plaintiff's claims.
A number of other Courts of Appeals have reached the same conclusion. For example, in EEOC v. Federal Reserve Bank of Richmond, 698 F.2d 633, 638-39 (4th Cir.1983), rev'd on other grounds sub nom., Cooper v. Federal Reserve Bank of Richmond, 467 U.S. 867, 104 S. Ct. 2794, 81 L. Ed. 2d 718 (1984), the court refused to apply the disparate impact model where the plaintiff offered no evidence of any "objective standard, applied evenly and automatically." 698 F.2d at 639 (quoting Statsny v. Southern Bell Tel. & Tel. Co., 628 F.2d 267, 274 n. 10 (4th Cir.1980)). The plaintiff alleged only that the employer practiced discrimination against an entire group. The court found the disparate treatment model was appropriate for "subjectively based practices." Id. (quoting Statsny, 628 F.2d at 274 n. 10). See also Talley v. U.S. Postal Service, 720 F.2d 505, 507 (8th Cir.1983), cert. denied, 466 U.S. 952, 104 S. Ct. 2155, 80 L. Ed. 2d 541 (1984); Mortensen v. Callaway, 672 F.2d 822, 824 (10th Cir.1982); Heagney v. University of Washington, 642 F.2d 1157 (9th Cir.1981). But see Hung Ping Wang v. Hoffman, 694 F.2d 1146, 1148 (9th Cir.1982) (court held without citation of any authority, that to prevail on disparate impact claim, plaintiff need only demonstrate lack of objective criteria and a disparity); and Moore v. Hughes Helicopters, Inc., 708 F.2d 475, 481-82 (9th Cir.1983) (court noted division within circuit on this issue, and then assumed without deciding that impact analysis applies to subjective hiring process, because plaintiff failed to make case under either analysis).
In Coates, supra, the Seventh Circuit addressed this issue. The plaintiffs in Coates alleged racial discrimination in the defendant's pattern and practice of employee discharges. No specific neutral policy of the defendant was identified by the plaintiff. The plaintiffs asserted on appeal that, although they tried their case primarily on the theory of disparate treatment, they also alleged disparate impact. They argued that the district court erred in ignoring this theory. 756 F.2d at 530 n. 7.
The Coates court distinguished between the two theories, quoting the Teamster's decision set forth above, which defines the disparate impact model as applying to facially neutral employment policies. The court noted that plaintiffs had alleged facts appropriate for disparate treatment analysis, but that they never argued that there were facially neutral termination rules that had a disproportionate impact on blacks. Rather, they alleged that the rules were more severely applied, through supervisory discretion, to blacks than whites. The court concluded that the case was therefore susceptible to a disparate treatment, *1285 and not a disparate impact, theory. Id. The court then cited EEOC v. Federal Reserve Bank of Richmond, supra; Pouncy v. Prudential Insurance Company of America, supra; and Mortensen v. Callaway, supra, all cases discussed above which reject the application of the disparate impact theory when the plaintiff has not identified a specific, facially neutral practice of the employer which disproportionately excludes members of a protected group.
Thus, without specifically discussing the decision in Stewart v. General Motors, supra, the Seventh Circuit has now aligned itself with those courts which apply the disparate impact theory only in circumstances similar to those in which the Supreme Court has applied the theory, for example, when the employer has a specific practice which is neutral on its face but impacts disproportionately on protected groups.
This court therefore concludes that, under Supreme Court precedent, Coates, and the Pouncy line of cases, the disparate impact theory should be applied only when such a policy has been identified by the plaintiff. Since the EEOC admits that it has not identified any specific, facially neutral policy of Sears which disproportionately excludes women from the jobs at issue in this case, the court will apply only the disparate treatment theory to this case.
1. In General
Virtually all the proof offered by the EEOC in this case is statistical in nature, or related to the statistical evidence. Statistics are an accepted form of circumstantial evidence of discrimination. In some cases, where "gross disparities" are shown, statistics alone may constitute a prima facie case. Coates, 756 F.2d at 532 n. 6; Segar v. Smith, 738 F.2d at 1278. However, as the Court in Teamsters cautioned, "statistics are not irrefutable; they come in an infinite variety and, like any other kind of evidence, they may be rebutted. In short, their usefulness depends on all of the surrounding facts and circumstances." 431 U.S. at 340, 97 S. Ct. at 1856-57 (citations omitted).
Statistical evidence, like other evidence, must not be accepted uncritically. The usefulness of statistics depends to a large extent on "the existence of proper supportive facts and the absence of variables which would undermine the reasonableness of the inference of discrimination which is drawn." White v. City of San Diego, 605 F.2d 455, 460 (9th Cir.1979) (quoting United States v. Ironworkers Local 86, 443 F.2d 544, 551 (9th Cir.1971), cert. denied, 404 U.S. 984, 92 S. Ct. 447, 30 L. Ed. 2d 367 (1971)). Inaccuracies or variations in data or in the formulae used to test *1286 such data may easily lead to different, contradictory, or even misleading conclusions by experts. E.E.O.C. v. Federal Reserve Bank of Richmond, 698 F.2d at 645. Courts therefore must carefully evaluate all the assumptions and data underlying the statistical analyses to determine whether they are sufficiently related to reality to provide any useful information to the court.
As this case will demonstrate, the assumptions made by a statistician in formulating a model can be far more important than the numerical complexities and results of the analysis. Without a sound theoretical basis, which is carefully reasoned and closely tailored to the factual circumstances of the case, the statistical results can be meaningless. Close attention will therefore be paid in this case to the assumptions made by the experts and their relation to reality.
2. Statistical Significance
In view of the highly statistical nature of the EEOC's proof in this case, a short discussion of a few relevant statistical principles is necessary.
An important concept to bear in mind in evaluating any statistical analysis is that no statistical analysis can prove causation per se. Statistical experts on both sides of this case have readily admitted this. Rather than attempting to prove what caused the results obtained from their analyses, statisticians endeavor to estimate the likelihood that the results occurred merely by chance. In statistical terms, they attempt to measure the level of statistical significance of their results.
One commonly used measure of statistical significance, the likelihood that the result occurred by chance, is the standard deviation. The standard deviation measures how much a typical observation varies from the average of all observations. D. Barnes, A Commonsense Approach to Understanding Statistical Evidence, 21 San Diego L.Rev. 809 (1984). A standard deviation of 2 indicates that there is a .045 probability (that is, almost a 5% probability), that the results observed occurred due to chance. A standard deviation of 3 represents a .003 probability, and a standard deviation of 4 represents a .006 probability that the result occurred by chance.
Statisticians generally consider results to be statistically significant at two or three standard deviations. However, statistical significance merely indicates that chance is not likely to have caused the result. As EEOC's expert, Dr. Siskin, testified, statistical significance and practical significance are two completely different concepts. Statistical significance can be determined merely by calculating the standard deviation or some other test statistic. To determine the practical significance of statistical results, a court must look at the theories and assumptions underlying the analysis and apply common sense.
Courts have not blindly adopted any test of statistical or practical significance. In Castaneda v. Partida, 430 U.S. 482, 496 n. 17, 97 S. Ct. 1272, 1281 n. 17, 51 L. Ed. 2d 498 (1977), and Hazelwood School District v. United States, 433 U.S. at 308 n. 14, 311 n. 17, 97 S. Ct. at 2742 n. 14, 2743 n. 17, the Court indicated that, as a general rule, if the observed result is more than 2 or 3 standard deviations from the expected or average value, a hypothesis that the result occurred by chance is suspect. However, the Court has not laid down any hard and fast rules for evaluating statistical or practical significance in every case, and the Court's "2 or 3" standard deviation approach should be applied with caution. See D. Baldus and J. Cole, Statistical Proof of Discrimination 294-95 nn. 12-13 (1980). The Seventh Circuit has recommended using "extreme caution" in drawing any conclusions from statistical significance at a two-to-three standard deviation level. Coates, 756 F.2d at 547 n. 22. See also EEOC v. American National Bank, 652 F.2d 1176, 1198 (4th Cir.1981), cert. denied, 459 U.S. 923, 103 S. Ct. 235, 74 L. Ed. 2d 186 (1982). The Coates court recognized that courts are not required to conclude from low standard deviations that the statistics *1287 conclusively prove discrimination, particularly when the statistical model is not well adapted to the facts. Coates, 756 F.2d at 547 n. 22. See Teamsters, 431 U.S. at 340, 97 S. Ct. at 1856-57 (the usefulness of statistics depends on surrounding facts and circumstances). The court will therefore carefully examine the models underlying the statistical analyses before attributing practical significance to standard deviations in the 2 to 3 range. As a guide for discussion, however, this court will generally consider that differences between actual and expected values that exceed 3 standard deviations may be statistically significant. However, it is important to emphasize that the standard deviation and other measures of statistical significance merely attempt to eliminate chance as the reason for the results. They do not prove what in fact caused the results.
EEOC's statistical expert calculated "z" statistics for the disparities between his estimated actual and expected results for the commission sales claim. These z values represent the number of standard deviations between his actual and expected results. "T" values were calculated to measure the same for the checklist compensation analyses. Accordingly, when z values or t values in the analyses exceed three, the court will assume that the disparities may be statistically significant. The court's view of the practical significance of the results will be discussed at length below.
3. Multiple Regression
The primary statistical analyses performed by the EEOC in this case were multiple regression analyses. Multiple regression is a statistical technique designed to estimate the effect of several independent variables on a single dependent variable. It attempts to measure the effect of factors such as age, education, experience, or sex, on an outcome, such as a hiring decision. It estimates how much each factor affects the outcome.
The most important issue in a regression analysis is what factors should be included as the independent variables. It is important to include all variables that significantly influence the dependent variable. However, if too many factors are added that do not significantly affect the dependent variable, the model can become distorted and then may not accurately estimate how much the independent variables influence the dependent variable. For a regression analysis to be meaningful, it is therefore important to strike a balance by including all factors which significantly affect the dependent variable, and excluding those variables which do not significantly affect the dependent variable.
As discussed above, the assumptions underlying any statistics must be critically examined by the court. In a regression analysis, this requires a careful evaluation of the variables included in the model, the ability to accurately measure these variables, and the importance of variables not included in the model, including variables which cannot be measured on a mathematical scale. In short, the court must evaluate "the fit" of the model to the reality of the facts and circumstances. See T.J. Campbell, Regression Analysis In Title VII Cases: Minimum Standards, Comparable *1288 Worth, and Other Issues Where Law and Statistics Meet, 36 Stan.L.Rev. 1299, 1305-09 (1984). Measures of statistical significance estimate the likelihood that the results occurred by chance, assuming that the model accurately reflects the decision making process it evaluates. Id. Measures of statistical significance will not have meaning where the model does not accurately reflect the process it seeks to replicate. As discussed above, the court in Coates, 756 F.2d at 547 n. 22, recognized this in advocating caution in relying on statistics in the standard deviation of 2 or 3 as proof of discrimination, particularly when the model does not have a good "fit." Courts need not accept the results of any regression model as proof of a prima facie case merely because the results are statistically significant, unless the model has a sufficiently good "fit" to justify an inference of discrimination.
The court therefore will carefully examine the factors included and excluded from the regression analyses in this case, and the weight given to measures of statistical significance will depend upon the fit of the statistical model to the actual hiring and compensation practices of Sears.
Finally, it is important to recognize at the outset the limitations of statistical analyses when evaluating complex decision making processes. Although mathematical models such as those used in regression analyses can provide useful information in appropriate circumstances, the intricacies of human decision making often cannot be reduced, to mathematical formulae. The more complex the decision making, the less accurate a regression model will generally be. Thus, statistical analyses can be quite accurate when used to analyze relatively simple fact situations, such as when a job requires only a small number of minimum objective job qualifications. However, as will be seen below, as more qualifications and subjective factors are required, mathematical models are less able to accurately analyze the decision making process.
II. Hiring and Promotion Into Commission Sales
EEOC attempted to prove at trial that Sears intentionally discriminated against women in hiring and promotion into commission sales on a nationwide basis from 1973 until 1980. Both parties have analyzed separately those hired directly into commission sales and those promoted to commission sales from other jobs at Sears. Full time and part time positions in each category were also analyzed separately. The court will structure its analysis accordingly.
A. General Background
Sears is the nation's largest retailer of general merchandise, employing approximately 380,000 persons in over 4,000 facilities across the country. During the relevant time period, Sears had approximately 920 retail stores. Its corporate headquarters is located in Chicago, Illinois. Between 1973 and 1980, the relevant time period, Sears was divided into five territories: Eastern, Midwestern, Pacific Coast, Southern, and Southwestern.
General corporate policies at Sears are formulated at the corporate headquarters, communicated to the territorial organizations, and eventually disseminated to individual stores. Although Sears has some very strongly enforced corporate policies, its entire management system overall is highly decentralized. This includes its hiring and compensation practices. Maximum flexibility, within bounds, is given to individual store managers to respond to the surrounding markets. Each store manager's compensation is significantly tied to the profitability of his or her store. Commission sales is a major source of store profits, and having a qualified and successful commission salesforce is of great importance *1289 to individual store managers. All of this has been very important to Sears' success over the years.
1. Commission Sales at Sears
During 1973-1980, Sears retail stores were divided into approximately 55 retail divisions. Salespersons in these divisions were paid either on a commission or noncommission basis. Merchandise sold on commission was usually much more expensive and complex than merchandise not sold on commission. Commission selling usually involved "big ticket" items, meaning high cost merchandise, such as major appliances, furnaces, air conditioners, roofing, tires, sewing machines, etc. Noncommission selling normally involved lower priced "small ticket" items, such as apparel, linens, toys, paint and cosmetics.
Until 1977, commission salespeople were paid on a "draw versus commission" basis. Commissions ranged between six and nine percent. Under this system, salespersons were guaranteed a "draw" each week, usually not exceeding 70% of average or estimated earnings. However, if the sales-person's commissions did not equal the amount of his weekly "draw", he incurred a deficit. This deficit was carried over to the next week. If the salesperson exceeded his draw for the next week, the excess would be applied against the preceding week's deficit. If the deficit was not eliminated in the following week, the deficit would be cleared. Thus, the deficit would be carried over for only one week. However, if the salesperson failed to meet his draw over a period of time, his employment could be terminated. Thus, there was substantial risk attached to commission selling.
After 1977, this risk was ameliorated to some extent when Sears changed its method of compensating commission salespersons to a "salary plus commission" basis. This change was implemented primarily to reduce the financial risk of selling on commission in an effort to attract more women to commission sales. Under this system, the salesperson earned a nominal salary plus a three percent commission. However, the salesperson's income was still substantially dependent on the amount of sales made, and failure to make sufficient sales could result in termination.
Noncommission salespeople were compensated on a straight hourly rate, except that a nominal 1% commission was earned by full time salespersons on all sales until January 1979. Throughout the period from 1973 through 1980, full time and part time commission salespersons on average earned substantially more than full time and part time noncommission sales persons.
Commission selling techniques varied from division to division, depending on the nature and price of the merchandise. Technical skills required to sell also varied widely between divisions. Some divisions required a high level of technical knowledge and expertise, while others required less technical knowledge. For example, selling in the Installed Home Improvements divisions required substantial technical knowledge and a high degree of motivation. The selling of some products was done on an "outside basis," at the customer's home or place of business. Products requiring "outside" selling included central air conditioning, heating, and plumbing systems. Merchandise sold "inside" included water heaters, attic fans, and other big ticket items which did not require complicated estimates or installations.
Products sold "outside" the store generally required a high level of technical expertise. Many jurisdictions required these salespeople to be licensed, which often required salespeople to pass a test. To be successful, the salesperson had to spend significant amounts of his own time and energy educating himself on all aspects of the product, from its technical functioning to installation and customer service. Sales usually would be closed in the customer's home, requiring substantial weekend and night time hours. Other important aspects of the process of selling home improvements include following up on sales, developing leads and referrals, and responding to emergency calls. In short, commission *1290 selling in these Installed Home Improvement divisions required a high degree of technical expertise, motivation, and persistence.
Although other commission sales divisions required lesser degrees of technical skill than Installed Home Improvements divisions, all required some technical knowledge of the product. Equally important, all commission sales positions required a high degree of motivation. The salesperson's income and continued employment was dependent upon his ability to sell. In addition, as noted above, a substantial part of the store's revenues, and the store manager's income, was dependent upon commission sales. Managers therefore endeavored to hire the most highly qualified commission salespeople.
2. Qualifications for Commission Sales Positions
No written document at Sears specifically identifies the qualifications for commission sales positions. Sears' Retail Testing Manual (entitled "Psychological Tests for Use in Sears Retail Stores," Pl.Ex. 108, Def. Ex. 34), contains the only written description of a desirable commission sales candidate. The commission salesperson is described as a "special breed of cat," with a sharper intellect and more powerful personality than most other retail personnel. According to the manual, a good commission salesperson possesses a lot of drive and physical vigor, is socially dominant, and has an outgoing personality and the ability to approach easily persons they do not know. A good commission salesperson needs the ability to react quickly to a customer's verbal suggestions and modify the approach accordingly. Thus, according to the Retail Testing Manual, a higher level of "salesmanship" is required of the commission salesperson than is required of the general salesperson.
However, the Retail Testing Manual generally was not relied on to any significant extent in selecting commission salespersons. Most managers and interviewers relied far more heavily on past experience with commission salespeople. Many managers and personnel employees involved in the hiring process testified about the qualities they sought in commission sales people based on their own experience and the experience of their superiors.
First, in reviewing applications to identify potential commission sales candidates, Sears managers and interviewers look for: prior commission selling experience, prior experience selling a product line sold on commission at Sears, any specific indication of interest in commission sales, knowledge or experience of any kind in a product line, availability during the hours required for a particular job, a work or personal background reflecting a history of achievement, prior experience involving significant public contact, and any special training, education or experience indicating an active, outgoing personality. Previous product line experience was particularly important when there was a commission sales vacancy in an installed home improvements division, the automotive divisions, or custom draperies.
However, most of the essential qualities for commission selling could be determined only from an interview, not from a written application. As will be discussed below, all candidates were interviewed at least twice before being hired. During the interview, managers looked for a number of important qualities, including aggressiveness or assertiveness, competitiveness, the ability to communicate effectively, persuasiveness, an outgoing, social or extraverted personality, self-confidence, personal dominance, a strong desire to earn a substantial income, resilience and the ability to deal with rejection, a high level of motivation and enthusiasm for the job, maturity, and a good personal appearance. The extent that each of these characteristics was required depended on the particular job opening. However, one critical prerequisite for all commission sales jobs was availability during the required hours. An otherwise qualified individual could be rejected if he was not available during the hours required for the particular job. Each manager and interviewer used his best judgment *1291 to choose applicants with the best combination of these qualities for each job. However, as with all hiring decisions at Sears, the qualifications sought in commission sales applicants were often modified or relaxed to comply with Sears' Affirmative Action Plan, described below.
3. The Hiring Process
The hiring process at Sears varied from store to store. Normally, a person who appeared at a Sears store and expressed an interest in employment at Sears would be given an application form to complete. The application was usually a two-sided form provided by Sears corporate headquarters. It provided space for the applicant to indicate the type of employment sought, whether permanent or full time, the hours available for work, education, past work experience, and other background information.
On the application forms used from 1974 to 1980, the applicant checked boxes next to various types of jobs available at Sears. The application contained only one box for "sales"; it did not provide separate boxes for noncommission sales and commission sales. Applications usually remained "active" for approximately 90 days, although the length of time varied somewhat from store to store.
It was Sears' policy to extend an initial interview to each applicant as a courtesy, whether or not a position was available at the time. This initial interview was often conducted at the personnel counter when the applicant handed in a completed application. The interviewers were usually experienced personnel department employees. The great majority of Sears' interviewers were women.
The content of the initial interview varied from store to store, but often applicants were not informed of specific openings during this interview. The interviewer attempted to glean basic information from the applicant, such as availability for work and areas of interest. The interviewer often wrote comments on the applications regarding the applicant's personal interests, qualifications, or suitability for a particular position. In some cases the interviewer would convey her comments to the personnel manager orally.
The second step of the hiring process involved a second interview and sometimes testing. Procedures for moving to this second step varied among stores. At some stores, the initial interviewer screened applications on the basis of factors such as availability, prior job experience, and affirmative action requirements before moving applicants to the second stage. At other stores, the store personnel manager reviewed all applications before deciding whether a second interview or testing was appropriate. At still other stores, applications were screened by various members of the personnel staff.
However, despite this variation in the mechanics of the process, at all stores, the most important part of the hiring process was the second interview. This interview was usually conducted by the store's personnel manager, and normally lasted approximately one-half hour. The interviewer would evaluate the applicant's personal characteristics, such as appearance, manner, assertiveness, friendliness, ability to communicate, motivation, and overall potential. Various job openings might be discussed with the applicant at this time.
Following this interview, the applicant might also be interviewed by the appropriate division manager or merchandise manager, and in some cases by the store manager. However, the personnel manager's *1292 recommendation was followed in most cases.
Another part of the "second phase" of the hiring process was testing. Testing procedures varied from store to store even more widely than interview procedures. However, most persons hired by Sears were tested at some point in the process. At some stores, tests would be given before the second interview. At other stores, they were given after the second interview, while at still other stores tests were given only after an applicant was hired.
If an applicant was tested, he would usually be given two tests, Sears' Test of Mental Alertness and the Thurstone Temperament Schedule. Although the extent to which the test results were relied on varied, in most cases test scores did not play a dispositive role in the decision-making process. Sears' testing process will be discussed more fully below.
Finally, all aspects of the hiring process, from the initial interview to the final selection of a candidate for commission sales, were highly influenced by the requirements of Sears' stringent affirmative action programs.
4. Affirmative Action At Sears
Sears has for many years had a far ranging and effectively enforced affirmative action program. Sears was the first major retail employer in the nation to institute an affirmative action program. Sears' program became a model plan followed by other corporations in the retail industry and in other industries as well.
Sears' commitment to affirmative action began in 1968, when it appointed its first Equal Opportunity Director, Ray J. Graham. Also in 1968, at the urging of Mr. Graham, Sears voluntarily declared itself to be a government contractor, thereby submitting itself to the stringent affirmative action and recordkeeping requirements of the Office of Federal Contract Compliance Programs ("OFCCP"). As a result, Sears submitted itself to approximately 2,000 compliance reviews by the OFCCP.
In 1968, Sears began to develop its affirmative action program by distributing a questionnaire to obtain information about minorities and women at Sears. One of the pages of the questionnaire was specifically devoted to the movement of women into commission sales. (Sears Ex. 247, Tab. 1268, at 7). Sears then began to formulate its affirmative action program, which required Sears managers to consider the race, sex, and national origin of applicants and employees in making every employment decision. Because Sears was decentralized and each Sears unit hired independently, each unit had its own specially tailored plan.
In 1969, Sears set a long-term goal of 38 percent women in all jobs at Sears. This figure was derived from the government's estimates of women in the labor force. Because the female percentage of Sears employees substantially exceeded 38 percent, Sears concentrated on promoting women into non-traditional areas. Minority goals were also developed based on local labor force percentages.
*1293 In April, 1970, Sears instituted a centralized, company-wide affirmative action program for women and minorities. In 1971, full time executives were hired to administer the program from the corporate headquarters and territorial offices. Sears held meetings with officials of the EEOC and its major manufacturers and suppliers to discuss and promote affirmative action programs. Sears' plan was revised in 1972. The stated long term goal of the plan was to obtain parity with the proportion of women and minorities in the total work force. Hiring goals were set for each unit to reflect the local populations. Managers were required to report efforts to contact minority and women's groups and other recruiting efforts. They were also required to identify and develop promotable women and minority employees, and to encourage minorities and women to participate in training programs. Community activities relating to affirmative action were also reported.
In February, 1973, A. Dean Swift became president of Sears. One of his first actions as president was to call a conference of Sears' top 250 executives, for the first time in 23 years, to discuss affirmative action. The topic of the conference was the personal accountability of managers for affirmative action. Out of this conference grew the Mandatory Achievement of Goals ("MAG") Plan. The MAG Plan set long term goals for both timecard and checklist women. To meet these goals for timecard employees (including commission sales), the Plan required that managers fill one out of every two positions with women or minorities. When jobs were vacated by women or minorities, the Plan required replacement-in-kind.
The one-of-two rule was not applied for women in technical service, automotive, and other craft-type jobs, or for men in clerical jobs, because of the lack of interested and qualified candidates. The rule for hiring into these positions was set at one in three, and then reset to one in five.
The MAG Plan was implemented in all Sears facilities by the summer of 1974. Management at all levels were involved in the Plan. Individual store managers were held primarily responsible for meeting Plan requirements. Managers were informed that their performance ratings and compensation would be tied in part to their affirmative action performance, and that they would be held accountable for failure to comply with the Plan. Some individual managers were disciplined for insufficient efforts.
Each store's compliance was monitored at the group and territorial level. Regularly scheduled personnel meetings always included discussions of affirmative action. Managers seeking to deviate from the Plan were required to submit Good Faith Efforts forms to the territorial office to obtain written permission to deviate from MAG requirements for a particular job. The manager was required to state the reason for the deviation, what efforts had been made to locate an individual of the required race or sex, why such a candidate could not be located, and why it was essential that the opening be filled at that time. Deviation requests were rarely granted.
In 1977, Sears raised its long range goal for women from 38 percent to 40 percent *1294 for all jobs at Sears except craft-type jobs, for which the long range goal was 20 percent. In 1979, Sears added to its checklist affirmative action program the requirement that 50 percent of all openings be filled by women and minorities. Prior to that time, there were too few qualified female and minority candidates in line for checklist promotions to make this a realistic goal.
In addition to imposing these hiring and promotion goals, in 1972, Sears implemented an Equal Pay Affirmative Action Plan. Under the plan, a nationwide formula was established for setting timecard salary to ensure that male and female timecard employees received equal pay. Task forces, a majority of whose members were women, conducted audits in every store and had complete authority to make on-the-spot pay changes. In 1974, Sears also developed its "Equitable Pay Package," which consisted of administrative procedures designed to ensure uniform application of wage scales to male and female employees.
The sincere dedication and commitment of Sears management at all levels to affirmative action was evident from the testimony of the Sears' officials and employees, whom the court found to be highly credible witnesses. Sears' program exceeded the requirements of Title VII or any other governmental regulation. Sears' specific efforts to place women in commission sales will be discussed below.
B. EEOC's Evidence: Hiring Into Commission Sales
EEOC has relied almost exclusively on statistics to prove its case. Not a single witness was called by EEOC to testify that she had been discriminated against by Sears, or had witnessed discrimination at Sears. In the absence of any direct evidence of any sort of discrimination against women at Sears, EEOC's statistical evidence must be carefully scrutinized.
1. Data Bases
EEOC's statistical expert, Dr. Bernard Siskin, performed several analyses of Sears' hiring decisions in commission sales, in which he compared rejected applicants with persons hired into commission sales. To compile a data base for these analyses, he first obtained the employment applications of 33,000 rejected sales applicants from 33 randomly selected Sears stores across the country. Some of the information from these applications was coded onto computer tapes. From this sample, EEOC attempted to estimate the female proportion of sales applicants at all Sears stores on a nationwide and territorial basis, and also to determine the characteristics of the male and female applicants. To analyze the characteristics of persons hired into commission sales, EEOC took a second sample of applications of approximately 1920 persons hired into full time and part time commission sales positions between 1973 and 1980 at approximately 210 stores.
EEOC also attempted to identify all commission sales hires and promotions for the years 1973 through 1980, to determine the composition of these groups. To do this, *1295 EEOC used computer tapes compiled by Sears from its computerized payroll records for all employees of Sears during the relevant period. EEOC attempted to identify from those tapes all persons hired or promoted into commission sales. Considerable trial time was expended in debate over the most accurate means of identifying commission sales hires and promotions. Substantial disputes remain as to whether certain groups of these commission salespeople should be excluded because the data on the tapes relating to them is likely to be unreliable, or for other reasons. The EEOC finally adhered to a figure of almost 34,000 commission sales hires and promotions for the eight year period, while Sears identified approximately 20,000. However, the court need not belabour here the nuances of their disagreements, because both experts admit that the percent of female hires and promotions identified by each side does not vary significantly.
2. Statistical AnalysesUnadjusted Analysis
EEOC performed several analyses of this data. First, EEOC's expert, Dr. Siskin, attempted to estimate the female proportion of full time and part time commission sales hires in the nation and each territory for each year from 1973 through 1980. This estimate was made simply by determining the percent of women in the group he identified as commission sales hires as described above.
Dr. Siskin next attempted to estimate the female proportion of full time and part time applicants for all sales jobs at Sears by year and territory. He analyzed the applications of the sample of non-hired applicants. He counted as "sales applicants" any person who applied for any job at Sears, except those persons who requested only a non-sales job. His definition of a "sales" applicant included not only persons who checked "sales" on their application, but also persons whose interest could not be determined, and those who checked "sales" in addition to other job categories. Dr. Siskin then determined the percent of women in this group of "sales" applicants.
Dr. Siskin also made the very critical assumption that all of these "sales" applicants were applicants for commission sales jobs at Sears. He further assumed that all of these "sales" applicants were equally qualified for and interested in all commission sales jobs at Sears. (Subsequent evidence proved that these assumptions were not correct.)
Based on these two arbitrary foundational assumptions, Dr. Siskin postulated that, in the absence of discrimination, Sears would be expected to hire into commission sales approximately the same percent of women as the percentage of women in his pool of "sales" applicants. He compared his estimated percentage of women commission sales hires, called the "actual percent female", with the percent of women in his "sales" applicant pool, the "expected percent female." These comparisons were made on a nationwide and territorial basis for each year from 1973 through 1980.
As Table 5 of the EEOC's Commission Sales Report demonstrates, these comparisons resulted in large disparities between EEOC's actual and expected percent of female commission sales hires on a national and territorial basis for all years. Table 5 also provides the "z" values for the differences between the actual and expected figures. The "z" value equals the number of standard deviations between the actual and expected figures. The "z" values estimated by Dr. Siskin for his nationwide comparisons are all very large, ranging from 11.9 to 45.1. Estimated z values for the individual territories are somewhat lower, but still *1296 in the "highly statistically significant" range.
The same "unadjusted" analysis was performed by division, with the results appearing in Table 7 of EEOC's Commission Sales Report. The estimated z values were somewhat lower than the nationwide and territorial figures, but most had values above 5.0.
3. Adjusted Analyses
Dr. Siskin recognized that his unadjusted comparison of the female percent of all those he deemed "sales" applicants with the actual percent of women commission sales hires might not be meaningful, because there may be differences among those in his "sales" applicant pool which could explain some of the observed disparity. He therefore attempted to "adjust" his estimate of the expected percent female by taking into account various characteristics of applicants to determine whether the disparities might be explained by differences in the characteristics among male and female applicants. Three analyses were performed: a univariate analysis, a logit analysis, and a multivariate cross-classification analysis.
The first step for each of these analyses was to identify characteristics which might be associated with selection for commission sales. Dr. Siskin selected six factors which he thought might affect an applicant's chance of selection: (1) job applied for; (2) age; (3) education; (4) job type experience; (5) product line experience; and (6) commission product sales experience. He attempted to code information into a computer file on each of these characteristics for each person in the hired and nonhired samples.
One analysis performed by EEOC to adjust for these characteristics was a weighted logit regression analysis. The logit analysis predicts the effect of a variable on a "yes or no" event, such as whether or not a person is hired. This analysis measures the effect of independent variables (e.g., age, experience, education), on the dependent variable, whether or not the person is hired. The logit coefficient for each independent variable represents the net effect of that variable on selection, after adjusting for the impact of all the other independent variables considered. However, the logit analysis does not allow all the independent variables to interact with each other. It assumes, for example, that experience with a particular product line has the *1297 same effect no matter what the other variables are.
The second analysis, the multivariate cross-classification analysis, is a regression analysis that allows full interaction between all the variables in the model. Under the multivariate (also called multicell) analysis, a theoretical matrix is created with separate cells for each possible combination of all the different values possible for all of the characteristics considered. It calculates the expected percent female by comparing the number of hires with any given combination of characteristics, with the number of female non-hires with that particular combinations of characteristics. Dr. Siskin's multivariate model arbitrarily assumes that all persons within any particular cell are equally qualified for and interested in all commission sales jobs at Sears. In addition, in this and all of Dr. Siskin's analyses, it is arbitrarily assumed that all applicants were applying for all sales jobs in all divisions. (As subsequently explained, the evidence in this case proved that these assumptions were not correct.)
Finally, EEOC conducted a univariate cross-classification analysis, which measures the impact of any one characteristic alone on selection. It calculates the expected percent of female hires with any one characteristic, based on the number of persons hired who have that characteristic, and the number of women in the hiring pool who have that characteristic. This model does not allow for any consideration of the effect of variables other than the one being measured on the hiring decision.
EEOC relied primarily on its logit and multivariate analyses in this case. The results of the logit analysis are set forth in Tables 9 and 23 of EEOC's Commission Sales Report. The logit analysis reduced EEOC's original nationwide expected proportion of female commission sales hires for 1973 through 1980, from 61.1% to 49.5% for full time, and from 66.2% to 63.3% for part time.
Tables 9 and 23 also break the results down for each year, nationwide and for each territory, for full time and part time. In Table 10 of the Report, Dr. Siskin sets forth the relative chance of selection of men and women into full time commission sales positions after adjusting with the logit analysis. According to this table, in the majority of years in each of the territories, men had at least twice the chance of selection as women after adjusting with the logit analysis.
The multivariate cross-classification analysis also reduced the disparity between EEOC's expected proportion of females in commission sales and EEOC's actual proportion of females hired. Tables 12, 13, 25 and 26 of EEOC's Commission Sales Report contain the results of the multivariate analysis for full time and part time positions, nationwide and by territory, for each year. The analysis reduced EEOC's aggregated expected proportion of female full time hires nationwide for all years from 61.1% to 40.3%. This disparity was later reduced to 37.2% (Pl.Ex. 48). The disparity for part time hires was reduced from 66.2% to 60.5%.
The tables also provide the z values for the differences between EEOC's expected and actual figures, which represents the standard deviation of the differences as discussed above. These z values were below 3 for a number of years in various territories. EEOC admitted that these disparities are not statistically significant. EEOC asserts, however, that statistically significant disparities exist nationwide and by territory in years in which the z value exceeds 3. They are:
Full time: Nationwide1973-1977, 1980 Eastern 1973-1974 Midwestern 1973-1977
*1298 Pacific Coast 1976-1978, 1980 Southern 1973-1977 Southwestern 1973-1978, 1980 Part time: Four Territories1973-1980 Eastern 1973-1979 Midwestern 1973-1974, 1976-1980 Pacific Coast 1976-1980 Southern 1973-1980 Southwestern 1973-1980
These are the years and territories in which EEOC presently claims Sears has discriminated against women in hiring into commission sales.
Dr. Siskin conducted a further multivariate analysis by product line groupings. The analysis was performed for each of 10 product line groupings separately, and disparities for each of these product line groupings were calculated. The analysis was conducted on a nationwide basis for the time periods 1973 through 1975, and 1976 through 1980. The results are contained in Table 15 (full time) and Tables 28A and 28B (part time) of EEOC's Commission Sales Report. These tables also provide the estimated z values for the disparities. For the 1973-1975 period for full time, the z values exceeded 3 in all product line groupings except women's apparel. In 1976-1980, the disparities were reduced significantly, and were not statistically significant in three of the ten groupings (home furnishings, appliances, and shoes).
In the part time analysis, 14 product line groupings were used. The Midwest Territory was analyzed separately because of apparent differences in hiring patterns. For both the 1973-1975 and 1976-1980 periods, in the four territories analyzed together, the z values of the disparities exceeded 3, and hence were considered statistically significant by EEOC, in 8 of 14 product line groupings. In the Midwestern Territory, disparities were statistically significant (had z values exceeding 3) in 8 of 14 product line groupings for 1973-1975, and in 7 of 14 product line groupings in 1976-1980.
4. Possible Biases
Dr. Siskin identified two potential problems with the multivariate which he asserts may result in an underestimate of the expected proportion of womenfragmentation bias and proxy bias. Fragmentation bias can occur when there are a large number of cells in the multivariate analysis to allow for all the possible combinations of characteristics. The larger the number of cells relative to the number of applicants, the greater the number of cells in which there are only hires, and no applicants. In this situation, Dr. Siskin asserts that the sexual composition of the expected hires tends simply to become the sexual composition of the actual hires, possibly masking the size of any disparities on the basis of sex.
To reduce this fragmentation bias, Dr. Siskin collapsed a number of the categories in the six characteristics and re-ran the multivariate analysis. He found that the regrouping of variables largely eliminated any fragmentation bias. The regrouping also had the overall effect of increasing the expected proportion of women. However, by collapsing the categories within variables, the model also lost a degree of its "precision," the ability to accurately predict how the hiring decision was affected by variations in the characteristics analyzed.
Dr. Siskin also identified proxy bias as another cause of a possible underestimation of the expected proportion of women in the multivariate analysis. He asserts that the analyses he performed do not reflect the extent to which the characteristics he adjusted for serve as a proxy for sex. In other words, he testified that, although it may appear from the analysis that a particular characteristic possessed disproportionately by males, such as automobile experience in automotives, is associated with success, *1299 in fact it may only reflect that greater success of males in general, presumably due to discrimination. Thus, if there is in fact discrimination in favor of men, characteristics more commonly held by men would appear in the analysis to be more closely correlated with success than characteristics held by women.
As noted above, the final figure arrived at by Dr. Siskin using the multivariate analysis, for all years nationwide, was 37.2% for full time commission sales. However, based on his perception of these possible biases in the multivariate analysis, and in light of the higher expected proportion of women resulting from the logit analysis (estimated at 49.5%), Dr. Siskin estimated the actual expected female proportion of hires, as adjusted for the six characteristics, is 40% for full time commission sales. The figures for each territory by year are set forth in Tables 12, 13, 25 and 26 as discussed above.
5. Other Statistical Evidence
EEOC presented one additional statistical analysis to support the results of its multivariate logit analyses. Dr. Siskin analyzed responses to Sears' Applicant Interview Guides. These guides were used in Sears' Southwestern Territory from 1978 through 1980. In these Guides, applicants rated their skill, interest and experience on a scale of one to five for various activity categories relating to jobs at Sears.
For all years at issue, Dr. Siskin adjusted his original pool of "sales" applicants based on the interest and experience responses to the Applicant Interview Guides he analyzed. In making this adjustment, Dr. Siskin also attempted to factor in the percent of hires with and without product line experience from the EEOC coding of the nationwide samples of full time commission sales hires from 1973 until 1980. This analysis was completely independent from the logit and multivariate analyses. Dr. Siskin controlled only for interest and experience based on responses to the Applicant Interview Guides, not for any of the factors considered in the multivariate or logit analysis.
Dr. Siskin performed his analysis of responses to four categories on the Applicant Interview Guide which he determined were most appropriately matched with four product line groups at Sears: appliances, automotive, home building materials and home improvements. Based on this analysis, he *1300 estimated that the expected proportion of women in each of these categories was higher than the expected proportion of women estimated using the multivariate analysis.
6. Nonstatistical Evidence
Despite the comprehensive, nationwide scope of its lawsuit, EEOC did not produce any victims of discrimination by Sears, or any persons who claimed they witnessed discrimination against women by Sears. EEOC points, instead, to two aspects of Sears' selection process to support its statistical analyses: the subjective nature of Sears' selection process, and its testing practices. First, EEOC asserts that the absence of objective criteria for selection provides a ready mechanism for discrimination. It presented evidence showing a lack of written guidelines for selection of commission salespersons, and the lack of formal training of interviewers, both of which permitted subjective hiring decisions. The interviewers, most of whom were women, were formally trained primarily on nondirective interview techniques designed to encourage the applicant to "open up" during the interview. No formal instruction was provided regarding the qualities to look for in commission sales candidates. Interviewers were expected to learn the desirable characteristics for commission salespersons from observation of those persons presently selling on commission, and from managers' guidance as to the types of individuals who had been successful in the past.
EEOC also relies on Sears' Retail Testing Manual, discussed above, which apparently contains the only written descriptions of the desirable characteristics of a commission salesperson. EEOC focuses on the original version of the manual, issued in 1953, which describes a commission salesperson as a man who is "active," "has a lot of drive," possesses "considerable physical vigor," "likes work which requires physical energy," etc. References to males were eliminated in the 1960's. The present version is otherwise substantially similar in content to the original version.
EEOC also presented evidence of Sears' testing practices, EEOC focused its presentation on the Vigorous Dimension of the Thurstone Temperament Schedule. Some of the questions from which the vigor score is determined would more likely be answered affirmatively by males.
C. EEOC's Evidence: Promotions into Commission Sales
EEOC relies almost exclusively on statistical analyses to prove that Sears discriminated against women in promotions from noncommission sales jobs to commission sales jobs. In its analyses, EEOC considered only movements from noncommission sales positions to commission positions. It did not incorporate movements from any other positions into commission sales.
1. Year-end Store Pool Analysis
EEOC's expert, Dr. Siskin, performed two analyses of promotion data. First, he calculated the expected proportion of female promotions for each year using the female proportion of noncommission salespersons in each store at the end of each prior year (year-end store pool) as an estimate of the female proportion of the promotion pool. He compared these figures with the proportion of women actually *1301 hired into commission sales. The resulting disparities and z values associated with those disparities were presented in Table 30 (full time) and Table 33 (part time) of EEOC's Commission Sales Report. The z values exceeded 3 in all but one year in one territory. In performing this analysis, Dr. Siskin made the arbitrary assumptions that all noncommission salespersons are equally interested in and qualified for all commission sales positions in all Sears divisions.
2. Year-end Division Pool Analysis
Dr. Siskin performed a second analysis of the same data, in which he calculated the expected female promotions using the female proportion of noncommission salespersons on a division basis (year-end division pool). This analysis was performed to take into account the possibility that persons in a division in which a promotion occurred had a greater chance of receiving the promotion than persons outside the division.
Tables 31 (full time) and 34 (part time) of EEOC's Commission Sales Report contain the results of this analysis, including z values for the estimated disparities. The z values exceeded 3 in all years and territories except for one territory in one year.
D. Analysis of EEOC's Evidence: Hiring
EEOC has relied almost exclusively on its statistical analyses to meet its burden of persuasion on its commission sales claim. There is so much imprecision in all of the underlying data being used that the mathematical presentation connoting precision is highly misleading. EEOC's statistical evidence is so flawed that it is not sufficient to meet its burden of persuasion.
1. Inadequacies in Data and Variables Analyzed by EEOC
a. Data Analyzed
First, one of the most serious flaws pervading all of EEOC's statistical analyses is in its selection of the applicant pool for commission sales jobs at Sears. Dr. Siskin arbitrarily included in his "sales" applicant pool, on which he based all of his statistical analyses, any individual in the sampled stores who applied for any job at Sears, except those who specifically applied for nonsales jobs only. Thus, he considered that any person who checked the "sales" box on the application, or "sales" and any other box, or even those who checked the "any of the above" box, were applying specifically for a commission sales position at Sears. He also arbitrarily assumed that all of these members of his "sales" pool were applying for all commission sales positions at Sears in all divisions. EEOC presented no material evidence to support either assumption, and Sears' evidence proved that both assumptions were false.
Most written applications in the nonhired sample did not indicate whether the applicant was applying for commission sales. Of those who did indicate a preference for commission sales, over 75% were men. The "sales" pool constructed by Dr. Siskin had a disproportionately high percentage of women. See Written Rebuttal Testimony of Dr. Joan G. Haworth Regarding Commission Sales, at 2-3; Plaintiff's Exhibit 14 at 1-2 (corrected June 7, 1985). By using such an inaccurate, overbroad definition of "sales" applicant, EEOC substantially inflated the female percentage in *1302 the applicant pool used for all of its statistical analyses. This in turn caused an overestimation of the expected proportion of women hires predicted by EEOC's analyses.
In addition, Dr. Siskin admitted that the use of applicant flow data, the type of data he analyzed, is questionable when there are ambiguities concerning what jobs applicants are really applying for. Tr. 5204 (B. Siskin). In this case, there are obvious ambiguities in many applications in the non-hired sample regarding whether the applicants are applying for commission sales or noncommission sales. At Sears, commission sales and noncommission sales are two entirely distinct jobs requiring different qualifications. Even among commission sales jobs at Sears, different qualifications are required for different positions. As Dr. Siskin testified in another case, it is "statistically invalid" to simply aggregate across job categories where the qualifications and, in that case, racial mix of the relevant labor market are so diverse. Sears Siskin Exhibit 3A at 4. In this case, since EEOC's "sales" pool includes people who applied for non-commission sales and commission sales, and people who applied for a wide variety of commission sales jobs requiring different kinds of technical skills, all requiring somewhat different qualifications, EEOC's use of this type of applicant flow data will not accurately predict the expected percentage of female hires.
b. Omission and Inadequate Coding of Important Variables
A second major flaw in EEOC's analysis is its failure to include in its analysis many important factors that significantly affect the hiring process. As described above, Dr. Siskin chose six factors which he determined might be associated with selection for commission sales jobs. These six factors are: (1) job applied for; (2) age; (3) education; (4) job type experience; (5) product line experience; and (6) commission product experience. However, Dr. Siskin is an expert only in applied statistics. He is not an expert in labor economics, has no expertise in the area of retail sales, and has no direct knowledge of Sears. His personal opinion regarding the factors important for commission selling is therefore entitled to little weight.
EEOC presented no credible evidence to show that the six factors chosen by Dr. Siskin are the actual factors most important to selection for commission sales at Sears or to success in commission sales. In fact, all the evidence offered at trial indicates, to the contrary, that a number of important factors were not included in EEOC's analysis, and many of the factors included in the analysis were inadequately coded.
First, with respect to excluded factors, the one single most important factor intentionally excluded by EEOC is the applicant's interest in commission sales and in the product to be sold. Dr. Siskin simply assumed that all applicants in his "sales" pool were equally interested in and qualified for all commission sales jobs at Sears. There was no basis in the evidence for this assumption; the evidence is overwhelmingly to the contrary. As will be discussed below, many applicants for sales jobs at Sears were not in fact interested in selling on commission, and women were considerably less interested in commission sales than men. Moreover, many women were not *1303 interested in selling those product lines which have been called "non-traditional."
Other important factors not controlled for in EEOC's analysis are those characteristics which could be determined only from an interview, not from the written application. These include physical appearance, assertiveness, the ability to communicate, friendliness, and economic motivation. Dr. Siskin admitted that these are factors identified by Sears managers as desirable for commission salespersons. However, no adjustment is made for these factors in his analyses.
Moreover, EEOC's coding of several of the factors it included in its analysis was clearly inadequate to properly adjust for these factors. A prime example of this is EEOC's coding of prior job experience. Dr. Siskin did not distinguish among applicants with widely varying previous sales experience. No distinction was made for those who had worked "outside" sales, those who sold complex or expensive products, or the amount of skill required to sell the product. Among the applicants who indicated only that they had "sales" experience on their applications, women were more likely than men to have been cashiers, and would have been coded with a lower level sales experience had they written "cashier" on the application. Tr. 5965 (B. Siskin), Plaintiff's Exhibit 6 at 6.
EEOC then combined such applicants who indicated only "sales" experience, with applicants who had been sales managers, in the "SalesHigh" category, eliminating any distinction between the two groups. The sales manager group was disproportionately male. Tr. 5750 (B. Siskin); EEOC Commission Sales Report, Appendix 3 at 6. Combining these two groups masked the fact that men in the category were substantially more qualified than women in the category. Tr. 5750 (B. Siskin).
Applicants who had commission selling experience were all coded into the same category. Among the applicants coded as having commission sales experience on the basis of prior sales experience with a product sold on commission at Sears, women were more likely than men to have been cashiers. Therefore, more women than men were coded as having prior commission sales experience who did not in fact have commission sales experience. Tr. 6054-55 (B. Siskin); EEOC Commission Sales Report, Appendix 3 at 9-10. In addition, among men and women coded as having commission sales experience on the basis of their experience selling products sold on commission, men were more likely to have sold the more expensive items and women the less expensive items. Tr. 6050 (B. Siskin). Thus, among those applicants credited with commission sales experience, women had less real commission selling experience, and the experience they had tended to be of a lesser quality.
In addition, EEOC failed to distinguish between varying levels and quality of experience for applicants with non-sales experience. Applicants who owned or managed businesses would be credited with the same level of experience as a person who filled shelves in a stockroom. The coding of this experience did not reflect significant differences in the degree of responsibility or amount of accomplishment of different applicants. Although Dr. Siskin stated on direct examination that refinements within this category would not change anything, he admitted on cross examination that he had misspoken, and that refinements in this area could in fact make a difference. Tr. 4350, 5948-49 (B. Siskin). Male applicants were more likely than female applicants to have had the type of non-sales experience that would increase the chance of being hired into commission sales. See Sears Exhibit *1304 6-9, at 2-3; Written Direct Testimony of Dr. Joan Haworth, ¶ 10(d).
EEOC also failed to code the amount of prior experience an applicant had. An applicant with one month of experience in any area was coded the same as an applicant with 15 years of experience. Male applicants were likely to have more experience than females.
Interruptions in work history were also ignored by EEOC. Men were less likely than women to have interrupted their labor force participation. Among those who did have gaps in their employment, the length of the gaps of women were greater than the length of the gaps of men. These gaps generally lessen the quality and amount of experience of women compared to the men in the same category of prior experience. See Sears Exhibits 6-J-2 and 6-J-3; Written Direct Testimony of Dr. Joan Haworth ¶ 10(b).
Finally, EEOC failed to code other relevant experience recorded on the application. The application forms used after 1973 asked the applicant, "What experience or training have you had other than your work experience, military service or education? (Community activities, hobbies, etc.)." Responses to this question revealed that male applicants were much more likely than female applicants to have a background experience or knowledge in areas relevant to commission sales product lines. For example, males were much more likely to have a background in automotive, building, or mechanical and technical areas (e.g., television repair, industrial arts). This experience generally related to products sold on commission at Sears. Women, on the other hand, were much more likely to have had experience in home furnishings and decorating, or women's apparel and fashions. See Sears Exhibit 6-12. Products related to these experiences are less likely to be sold on commission at Sears. By not coding this experience, EEOC ignored another difference between men and women which could explain some of its estimated disparities.
It is clear from all of these coding decisions that the prior experience characteristics of Dr. Siskin's model do not adequately distinguish between important variances in background experience and knowledge which could significantly affect a hiring decision. Because of the nature of these coding deficiencies, women are credited with greater amounts of experience at higher levels than they actually had. The effect of all these inadequacies is to inflate the percent of women the statistical models estimate should be hired. Since prior experience is a factor which can be highly significant in the decision to hire for commission sales, these inadequacies seriously affect the ability of EEOC's statistical models to make accurate estimates. As Dr. Siskin admitted, if variables are poorly specified or defined, or if important variables are left out, the analysis can lead to the wrong conclusion.
c. Coding Errors
Moreover, errors in EEOC's mechanical coding of information from applications in its hired and nonhired samples also make EEOC's statistical analysis based on this data less reliable. Numerous significant coding errors by EEOC's coding staff were brought to light at trial. Substantial trial time was devoted to this problem. Dr. Siskin attempted to correct these errors by personally recoding commission sales experience and product line experience of full-time commission sales hires. After the recoding by Dr. Siskin, the final percentages of full-time commission sales hires with these experiences were almost identical to the percentages Sears originally presented based on an analysis by Sears' counsel.
However, Dr. Siskin did not recode the "job applied for" variable for full time commission sales hires. EEOC contends that, using its coding, 21.2 percent of the full time sales hires expressed a preference for *1305 commission sales. According to Sears estimates, made by the same individual who did the foregoing Sears coding, 42 percent expressed such a preference. In light of the very high number of errors made by EEOC's coders, and the accuracy of Sears' coding, the court finds that Sears' coding of this variable is far more likely to accurately reflect the prior experience of commission sales hires. This underestimation of the percent of hires who expressed such a preference will also cause an overestimation of the expected proportion of women hires.
These errors in the construction of EEOC's data base and its omission or inadequate coding of important variables, are sufficient together to raise serious doubts about the practical significance of all of EEOC's analyses. EEOC has inflated the female percentage of its "sales" applicant pool by including many applicants who did not apply for commission sales. It has also consistently coded prior experience in such a way that less experienced women are considered to have the same experience as more experienced men. Further, it has made so many general coding errors that its data base does not fairly reflect the characteristics of applicants for commission sales positions at Sears. While some of the statistical inadequacies of the EEOC may have resulted from a misguided attempt to save money, these decisions by EEOC all have the obvious tendency of artificially inflating the percentage of women that EEOC's analysis estimates should have been hired into commission sales by Sears. Therefore, under the principle of "garbage in, garbage out," referred to by Dr. Siskin, results from this analysis are highly suspect.
2. Faulty Basic Assumptions
Even more important than EEOC's errors in constructing its data base and adjusting for factors important to commission sales hires, EEOC's statistical analyses are based on two essential assumptions for which there is no credible support in the record. In designing all of his statistical analyses relating to commission sales, Dr. Siskin made the important assumptions that (1) all male and female sales applicants are equally likely to accept a job offer for all commission sales positions at Sears, and (2) all male and female sales applicants are equally qualified for all commission sales positions at Sears. Plaintiff's Siskin Exhibit 12; EEOC Commission Sales Report at 4. The court finds that both of these assumptions were proven untrue. Without these assumptions, EEOC's statistical analyses are virtually meaningless.[35a]
a. Assumption of Equal Interest
EEOC has offered no credible evidence to support its assumption of equal interest by male and female applicants in all commission sales positions at Sears. Sears, on the other hand, has offered much credible evidence that employees' and applicants' interests, preferences and aspirations are extremely important in determining who applies for and accepts commission sales jobs at Sears. Sears has proven, with many forms of evidence, that men and women tend to have different interests and aspirations regarding work, and that these differences explain in large part the lower percentage of women in commission sales jobs in general at Sears, especially in the particular divisions with the lowest proportion of women selling on commission.
*1306 (1) Store Witnesses
The most credible and convincing evidence offered at trial regarding women's interest in commission sales at Sears was the detailed, uncontradicted testimony of numerous men and women who were Sears store managers, personnel managers and other officials, regarding their efforts to recruit women into commission sales. As discussed above, attracting women to commission sales has been an important priority in Sears' affirmative action programs since the first affirmative action questionnaire was circulated in 1968. Sears managers and other witnesses with extensive store experience over the entire relevant time period testified that far more men than women were interested in commission selling at Sears. Numerous managers described the difficulties they encountered in convincing women to sell on commission. Sears managers continually attempted to persuade women to accept commission selling or other non-traditional jobs. Women who expressed an interest in commission selling were given priority over men when an opening occurred. Managers attempted to persuade even marginally qualified women to accept commission selling positions. They would sometimes guarantee a woman her former position if she would try commission selling for a certain period. Store managers reported that they had interviewed every woman in the store and none were willing to sell on commission. Managers often had to "sell" the job to reluctant women, even though enthusiasm and interest in the positions were qualities management valued highly in commission salespeople. Despite these unusual efforts, managers had only limited success in attracting women to commission sales.
Female applicants who indicated an interest in sales most often were interested in selling soft lines of merchandise, such as clothing, jewelry, and cosmetics, items generally not sold on commission at Sears. Male applicants were more likely to be interested in hard lines, such as hardware, automotive, sporting goods and the more technical goods, which are more likely to be sold on commission at Sears. These interests generally paralleled the interest of customers in these product lines. Men, for example, were usually not interested in fashions, cosmetics, linens, women's or children's clothing, and other household small ticket items. Women usually lacked interest in selling automotives and building supplies, men's clothing, furnaces, fencing and roofing. Women also were not as interested as men in outside sales in general, and did not wish to invest the time and effort necessary to learn to sell in the home improvements divisions. Women often disliked Division 45 (men's clothing) because it sometimes involved taking personal measurements of men.
Custom draperies, however, was one division in which women were willing to sell on *1307 a commission basis, even though it could require some outside selling. More women were willing to sell custom draperies on commission because they enjoyed the fashion and creative aspect of the job, most had past experience in the field, and it was a relatively low pressure commission division. Applicants expressing an interest in the division were almost all women. Very few men were willing to sell draperies.
The percentage of women hired in the various categories generally paralleled their interests and background in the product line involved. This illustrates how much an applicant's interest in the product sold can influence his willingness to accept a particular commission sales position. As is evident from the above discussion, interests of men and women often diverged along patterns of traditional male and female interest.
This lack of interest of women in commission sales was confirmed by the number of women who rejected commission sales positions. Although evidence was presented only for the Eastern territory, this evidence showed that many women turned down commission sales job offers or otherwise expressly indicated that they were not interested in commission selling at Sears. See Sears Exhibit 25.
Women at Sears who were not interested in commission sales expressed a variety of reasons for their lack of interest. Some feared or disliked the perceived "dog-eat-dog" competition. Others were uncomfortable or unfamiliar with the products sold on commission. There was fear of being unable to compete, being unsuccessful, and of losing their jobs. Many expressed a preference for noncommission selling because it was more enjoyable and friendly. They believed that the increased earnings potential of commission sales was not worth the increased pressure, tension, and risk.
These reasons for women not taking commission sales jobs were confirmed in a study performed by Juliet Brudney on behalf of Sears. Ms. Brudney conducted structured interviews of women in nontraditional jobs at Sears, including women in automotive and service technician jobs, and their supervisors. She also interviewed women who were seeking jobs or changing jobs, and found that they perceived commission sales as requiring cut-throat competitiveness that prevents friendships at work. They were also reluctant to sell products with which they were unfamiliar and preferred the security of a steady salary to the risks of making less in commission sales.
*1308 The results of Ms. Brudney's study were supported by the testimony of Sears' expert, Dr. Rosalind Rosenberg, who testified that women generally prefer to sell soft-line products, such as apparel, housewares or accessories sold on a noncommission basis, and are less interested in selling products such as fencing, refrigeration equipment and tires. Women tend to be more interested than men in the social and cooperative aspects of the workplace. Women tend to see themselves as less competitive. They often view noncommission sales as more attractive than commission sales, because they can enter and leave the job more easily, and because there is more social contact and friendship, and less stress in noncommission selling. This testimony is consistent with the uncontradicted testimony of Sears' witnesses regarding the relative lack of interest of women in commission selling at Sears, and with the testimony of Ms. Brudney, and is further evidence that men and women were not equally interested in commission sales at Sears.
(2) Survey Evidence
Sears also introduced a number of surveys taken of Sears employees and applicants which also demonstrate that women were much less interested than men in commission selling at Sears. This evidence also showed that women were particularly less interested than men in selling products in divisions where EEOC found the greatest disparities between the expected and actual proportions of women.
First, Sears presented extensive evidence of differences in the general interests and attitudes of men and women in American society over the past 50 years. This evidence was developed by Dr. Irving Crespi from an exhaustive study of national surveys and polls taken from the mid-1930's through 1983 which related to the changing status of women in American society. Although the evidence presented by Dr. Crespi demonstrated many changes in attitudes over the past 50 years, he made a number of conclusions directly relevant to women's interest in commission sales. Dr. Crespi found that: (1) men were more likely than women to be interested in working at night or on weekends, (2) women were more likely than men to be interested in regular daytime work; (3) men were more likely than women to be interested in sales jobs involving a high degree of competition among salespersons; (4) men were more likely to be interested in jobs where there was a chance of making more money, even though it involved a risk of losing the job if they did not sell enough; and (5) men were more likely than women to be motivated by the pay of a job than by the nature of the job and whether they like it. All of these *1309 conclusions support Sears' contention that women were less likely than men to be interested in commission selling in general.
Sears' evidence also demonstrated that women's attitudes toward work changed measurably between 1970 and 1980. During this time, significant changes occurred in the sexual compositions of the workforce. The number of females in many traditionally male-dominated fields doubled or tripled (e.g., securities and financial services salespeople, lawyers, hardware and building supplies). The female proportion of college students majoring in business subjects rose from 10 percent to nearly one-third in each category. Sears evidence also showed that, by the late 1970's, women were more open to commission sales positions than in the early 1970's, but it was still necessary to "sell" the job to them in many cases. Some reasons for the increased willingness of women to accept commission sales jobs were: (1) commission sales jobs changed from being almost exclusively full time to largely part time, and many more women preferred to work part time; (2) the change in compensation for commission sales from draw versus commission to salary plus commission, which reduced the risk perceived by some women in commission selling; (3) a group of successful commission saleswomen that over time provided role models for other women; and (4) the increased availability of day care, which increased the hours many women were available to work. Thus, female interest in and availability for commission sales increased during the period from 1973 through 1980.
In addition, specific surveys of the interests of Sears employees reveal that far more men than women are interested in commission sales. Sears has taken regular morale surveys of its employees in every retail store approximately once every three years since 1939. Sears has also conducted a number of special surveys to ascertain employee opinions on particular subjects.
The morale surveys demonstrated that most noncommission salespeople were happy with their work, and that more noncommission saleswomen preferred to stay in their present jobs than noncommission salesmen. In 1974-1976, noncommission saleswomen were much less interested than noncommission salesmen in promotion to division management or a higher level. They were far more likely than noncommission salesmen to want to remain in the job they had or one like it. In 1978-1980, only 14% of full time noncommission salesmen and 8.4% of noncommission saleswomen were interested in a different position. Thus, although only a small percentage were interested in other jobs, almost twice as many men as women were interested in new jobs. In addition, almost twice the percentage of female (56.4%) as male (30.3%) full time noncommission salespeople expressed a preference to remain in their present jobs.
The morale of noncommission salespersons was higher than that of all other timecard nonsupervisory employees. Noncommission salespersons were satisfied with their work, and noncommission saleswomen were more satisfied than noncommission salesmen. Most noncommission salespersons, and especially females, enjoyed their work and took pride in their jobs.
Noncommission saleswomen were more likely than noncommission salesmen to believe that their pay levels favorably influenced their attitudes toward their jobs, and were less likely to feel underpaid and to report that their pay at Sears was inadequate to meet their needs. Noncommission saleswomen were also more likely than noncommission salesmen to feel good about their futures at Sears.
*1310 This survey evidence, which has not been challenged by EEOC, demonstrates that noncommission saleswomen were generally happier with their present jobs at Sears, and were much less likely than their male counterparts to be interested in other positions, such as commission sales. These results confirm the reports of Sears' witnesses that women were less interested in commission sales than men.
Sears also presented the results of a special Job Interest Survey taken in 1976. See Sears Exhibit 254-1. This survey was administered to full time noncommission salespersons to determine their interest in big ticket commission selling. The survey was conducted to determine the amount of backpay, if any, Sears might be liable for as a result of EEOC's charges. Twentyfour stores were chosen from the stores in EEOC's investigative sample. The sample was not a random sample, but was chosen to be representative of the entire population in EEOC's investigative sample of stores.
The results of this special survey were similar to those of the morale surveys. They demonstrate that far fewer women than men were interested in any other job at Sears, including commission sales. Twice as many male as female noncommission salespersons reported that they were interested in another job at Sears (80.6% of the males, 36.6% of the females). See Sears Exhibit 254-3. After stating whether they were interested in any other job at Sears, respondents were asked to list what other jobs they would be interested in at Sears. Almost three times as many males (33.7%) as females (12.3%) stated that they were interested in commission sales. See Sears Exhibit 254-4.
The survey next provided a definition of big ticket commission selling, and asked respondents if they agreed with this definition. Almost all employees agreed with the definition. See Sears Exhibit 254-6. The survey then asked if the respondent would be interested in such a job. After reading the definition, 32% of the men, but only 3.5% of the women, expressed a definite interest in a big ticket commission sales job at Sears. See Sears Exhibit 254-7. The clear majority of females (69.4%) responded that they were definitely not interested in such a job, while only 37.9% of the males responded "definitely no." See Sears Exhibit 254-7. These results demonstrate that at least three times as many males, and possibly as much as eight times as many males, as females were interested in selling big ticket items on commission at Sears.
EEOC has attacked the 1976 job interest survey on several bases, through the testimony of Dr. Stanley Presser. However, the court finds that, although the survey was not taken using ideal methods, none of the problems cited by Dr. Presser are significant enough to affect the essential validity of the results. The survey provides *1311 a sufficiently fair and useful estimate of the interest of noncommission salesmen and saleswomen in big ticket commission sales to corroborate the highly credible testimony of Sears' store witnesses regarding the lack of female interest in commission sales positions.
The results of Sears' National Timecard Nonsupervisory Special Survey ("NTNSS") also confirm this relative lack of interest of Sears' noncommission saleswomen in commission selling at Sears. This survey was taken in 1982, using a sample of 53 Sears "A" and "B" stores. The survey was designed to be given to commission and noncommission salespeople at Sears. Unlike many other surveys taken by Sears, the NTNSS contained many questions requiring highly personal information in response. It asked questions about the attitudes, interests, and the personal beliefs and lifestyles of the employee. To assure respondents of the promised absolute anonymity, the unit of the respondent is not identified on the response sheet.
Question 42 of the survey asked respondents which three positions they wanted to be considered for next at Sears. Commission sales was one of the many alternative responses provided to the question. More than three times as many full time noncommission salesmen as noncommission saleswomen responded that they wanted to be considered for commission sales (20% of the males, 6% of the females), and more than twice as many part time noncommission salesmen as saleswomen (23% of the males, 9% of the females) were interested in commission sales.
EEOC has offered a number of criticisms of this survey. First, it contends that noncommission salespeople could not be adequately identified from responses to the survey. However, although the responses to the question designed to identify noncommission salespersons demonstrated that the question was somewhat ambiguous, Sears was able to adequately identify noncommission salespersons through responses to other questions.
Next, EEOC contends that Question 13C, not Question 42, is the most appropriate predictor of whether a noncommission salesperson is interested in commission sales. Question 13C asks whether the respondent would accept a commission sales job if offered one. However, as Dr. Crespi testified, Question 42 is a better indicator of real interest in the position, since the respondent must affirmatively choose commission sales. A "yes" answer to Question 13C requires less affirmative interest in the position. As discussed above, a strong interest in commission sales is an important factor which Sears managers look for in hiring commission sales applicants. The court therefore finds that Question 42 is the better indicator of the interest in commission sales. In any event, responses to Question 13C also indicate that noncommission salesmen are considerably more interested than noncommission saleswomen in commission selling. See Sears Exhibit 3-B at 13.
As Sears' witness Dr. Roper testified, perhaps the best indication of interest in commission sales can be derived from a combined analysis of the two questions. His combined analysis produced results similar to the responses to Question 42. Of the full time respondents, 20% of the men but only 6% of the women were actively *1312 interested in commission sales. For part time, 22% of the men and 9% of the women were actively interested. A slightly higher percentage of women would prefer another position but would accept commission sales (9% of full time women, 16% of part time women). Equally revealing is that 64% of the full time women were not interested in and would not accept a commission sales job, contrasted with 34% of the men giving the same response.See Written Rebuttal Testimony of Burns Roper, ¶ 4. Thus, even taking into account the responses to Question 13C, the survey shows that noncommission salesmen are clearly far more interested in commission sales than noncommission saleswomen.
In addition to the findings regarding the interest of noncommission salespersons in commission sales determined from direct questions, the NTNSS provided other useful information about the interests of men and women in noncommission sales. Noncommission salesmen, both full time and part time, were more interested in advancement at Sears in general than noncommission saleswomen. See Sears Exhibit 3-B at 9. Noncommission salesmen were more willing to make sacrifices for advancement, including working overtime, working more off-hours and weekends, and being assigned to bigger divisions. See Sears Exhibit 3-B at 10. In addition, among those with prior commission or big ticket sales experience, noncommission salesmen were much more likely than noncommission saleswomen to want to return to commission sales jobs. See Sears Exhibit 3-B at 14. Many other related differences between male and female interests and aspirations were found, all of which are further proof that men and women were not equally interested in commission sales at Sears during the 1973-1980 period.
The court finds that, again, although there are some flaws in the design and administration of the NTNSS, they are not significant enough to undermine the essential validity of the results. The responses described above corroborate the uncontradicted testimony of Sears' store witnesses, the results of the morale surveys, and the 1976 Job Interest Survey discussed above. Together, this evidence provides clear and essentially uncontroverted proof that the interests of men and women in commission sales is far from equal. To the contrary, men have consistently indicated far more interest than women in commission selling, leaving EEOC's foundational assumption of the equal interest of men and women in commission sales without any credible factual support in the record.
(3) Comparing Sears to National Data
In addition to Sears' evidence of actual interest of its employees in commission sales, Sears analyzed national labor force data as another measure of both interest in and availability for commission sales positions. As discussed above, applicant flow data available in this case does not provide a reliable basis for determining whether an applicant was qualified for and interested in a commission sales position at Sears. As an alternative to the applicant flow analyses performed by EEOC, Sears presented *1313 data on the composition of the national labor force as a more reliable and accurate standard by which to measure actual female availability for commission sales jobs at Sears. Sears expert Dr. Joan Haworth analyzed publicly available labor force data to obtain a reasonable benchmark for comparison with Sears. She examined data from the Panel Study of Income Dynamics, collected by the University of Michigan Institute for Social Research, and data from the Income Survey Development Program, based on surveys conducted by the Department of Commerce. She also relied on census data. This data demonstrated that the percentage of women selling on commission in the national work-force was far lower than the percentage of women EEOC estimated were available based on its overinclusive "sales" applicant pool during the years 1976-1980. See Sears Exhibit 6-6 and 6-E-1. This data also revealed that the percentage of women interested in selling a product depended to a great extent on the nature of the product. For example, as late as 1980, females were 83.5 percent of the sales cashiers and 81.8 percent of the persons selling apparel, but only 7.8 percent of the persons selling motor vehicles and boats and 25 percent of the sales workers in hardware and building supplies. See Sears Exhibit 6-7.
Dr. Haworth analyzed the number of male-owned and female-owned businesses by the type of product sold as a measure of interest in the product, since owning a business is the ultimate form of commission sales. She found wide differences in the percent of female-owned business depending on the product involved. For example, women owned 42.3 percent of the apparel and accessory stores, but only 3.9 percent of automobile dealerships and service stations. See Sears Exhibit 6-11. Dr. Haworth also found a wide divergence of male and female interests within product lines. For example, within the apparel and accessories businesses, in 1980, women owned 68.5 percent of women's ready-to-wear stores, but only 13.1 percent of shoe stores, and none of the men's apparel stores. See Sears Exhibit 6-11.
The data also showed a large difference in male and female interest in the various products grouped together by EEOC. For example, for its home furnishings product line, EEOC grouped together furniture, floor coverings and draperies. National data revealed that women owned 60.5 percent of the drapery, curtain and upholstery stores in 1980, but they owned only 14.2 percent of the furniture stores and 1.2 percent of the floor coverings stores.
Although business ownership and general interest in selling a particular product may not be perfectly correlated, this data does give an indication of the differences between male and female interest in selling particular products. This data corroborates other Sears evidence that male and female interest in selling varies significantly by product line.
This difference in male and female interests by product line is important in this case, because most of the disparities produced by EEOC's analysis are concentrated in five product lines: automotive, home building materials, electronics, appliances, *1314 and home improvements. All of these are product lines in which women have been substantially less interested than men.
This national data corroborates Sears' strong evidence of female lack of interest in commission sales and in the particular product lines discussed above. It is further proof that EEOC's assumption of equal male and female interest in commission sales and in all product lines has absolutely no basis in fact.
(4) Other Evidence of Interest
The only evidence introduced by EEOC regarding the interest of women in commission sales is the testimony of several witnesses regarding women's interests and aspirations in the workforce in general. These witnesses described the general history of women in the workforce, and contend essentially that there are no significant differences between the interests and career aspirations of men and women. They assert that women are influenced only by the opportunities presented to them, not by their preferences. They often focused on small segments of women, rather than the majority of women, in giving isolated examples of women who have seized opportunities for greater income in nontraditional jobs when they have arisen.
However, these experts provided little persuasive authority to support their theories. The particular examples of unknown numbers and proportions of women in history to which they refer generally focus on small groups of unusual women and their demonstrated abilities in various historical contexts, not on the majority of women or their interests at the time of this case. The sweeping generalizations these witnesses sought to make are not supported by credible evidence. None of these witnesses had any specific knowledge of Sears, or provided any specific evidence to contradict the strong evidence presented by Sears of the actual differences between the interests of men and women in commission sales positions at Sears.
More convincing testimony in this area was offered by Sears expert Dr. Rosalind Rosenberg. Dr. Rosenberg testified that, although differences between men and *1315 women have diminished in the past two decades, these differences still exist and may account for different proportions of men and women in various jobs. She offered the more reasonable conclusion that differences in the number of men and women in a job could exist without discrimination by an employer.
In conclusion, EEOC's statistical analyses are dependent upon the crucial arbitrary assumption that men and women are equally interested in commission sales jobs at Sears. As is evident from the above discussion, EEOC has provided nothing more than unsupported generalizations by expert witnesses with no knowledge of Sears to support that assumption. Sears has offered a wide variety of credible evidence that, during 1973 to 1980, women in fact were far less interested in commission selling at Sears than men. All the evidence presented by Sears indicates that men are at least two times more interested in commission selling than women. Thus, EEOC's assumption of equal interest is unfounded and fatally undermines its entire statistical analysis. Moreover, a difference in interest of this magnitude clearly would explain the disparities resulting from EEOC's faulty analysis.
b. Assumption of Equal Qualifications
In addition to assuming equal interest, EEOC also assumed that male and female applicants in its "sales" pool are equally qualified for all commission sales positions at Sears. However, as EEOC's Commission Sales Report itself demonstrated, this assumption is also false. The report indicates that, on average, female applicants in the "sales" pool were younger, less educated, less likely to have commission sales experience, and less likely than male applicants to have prior work experience with the products sold on commission at Sears. See EEOC Commission Sales Report at 34, 64; Plaintiff's Exhibit 14; Plaintiff's Exhibit Siskin 36, 37. Although EEOC attempted to adjust for some of these differences in its logit and multivariate analyses, as discussed above, its attempts were not successful. Therefore, its erroneous assumption of equal qualifications of male-female applicants also undermines the validity of its statistical analyses, particularly its unadjusted analysis.
3. Other Deficiencies in EEOC's Statistical Analysis
In addition to all the data problems, the omission and miscoding of important variables, and the faulty assumptions of equal interest and qualifications of men and women in commission sales jobs, there are a number of other important flaws in EEOC's analysis. First, as noted above, Sears is highly decentralized, and each store has a great deal of local autonomy. Each store hires its employees independently of other stores. Each store for the most part has its own individual pool of applicants for jobs at that store. There is generally no swapping of applicants among stores. No territory-wide or nationwide applicant pool has ever existed at Sears from which individual stores could draw the most qualified applicants. EEOC's analysis, however, creates artificial territorial and nationwide applicant pools where none actually existed, and then seeks to evaluate the hiring decisions of individual store managers by comparing those hired with the nationwide and territorial pools it artificially constructed. There is likely to be considerable variation in the quality of applicants *1316 appearing at particular stores, resulting in the hiring of persons who might be less qualified than applicants theoretically available in the artificial national and territorial pools. EEOC's analyses also fail to take into account whether an opening existed at the time any particular applicant applied. Each store had relatively few, if any, commission sales openings in any year. Thus, EEOC did not analyze the hiring situations actually confronted by Sears managers. The weight to be given its statistical analyses must be decreased accordingly.
In addition, the large z statistics relied on by EEOC as proof that their statistical analyses have significance are essentially meaningless. As discussed above, a z statistic measures only the probability that a given difference between two numbers occurred by chance. In the context of this case, the z statistic will have significance only if the numbers compared are the correct proportions of qualified and interested female applicants for commission sales and the correct proportions of female commission sales hires and promotions. Since EEOC did not properly control for interest or qualifications, EEOC never compared these two groups. Without meaningful underlying comparisons in its unadjusted and adjusted analyses, EEOC's z values prove nothing.
EEOC's unadjusted analysis compared the proportion of women hired into commission sales to the proportion of women in EEOC's "sales" pools. As discussed above, the "sales" pool included many persons who were not applying for or interested in commission sales, and contained a disproportionately high percentage of women. This analysis assumed that male and female applicants in the "sales" pool were equally interested in and qualified for commission sales, and made no adjustment for any differences among applicants. Since these assumptions have been proven untrue, EEOC's comparison of the groups provides little useful information to the court. The z values estimating the significance of these comparisons therefore have little meaning.
With respect to EEOC's two adjusted analyses, the z values also have little significance. Both analyzed the same highly overinclusive "sales" pool. Although EEOC attempted to adjust for some factors affecting chance of selection, the adjustments attempted were sorely inadequate, as discussed at length above. No adjustment was attempted for interest in a commission sales position, which might have narrowed the pool down to a more realistic group of applicants interested in commission selling. Therefore, EEOC's "adjusted" analyses cannot provide fair estimates of the proportion of women Sears should have hired into commission sales. Since the comparisons made in the analyses have little value, the z statistics associated with these comparisons also have little value.
In addition, as Dr. Wise testified, EEOC's z values were not properly calculated to reflect the factual circumstances at Sears. As discussed above, Sears is geographically and administratively decentralized. However, by creating and analyzing artificial national and territorial sales pools, EEOC erroneously conducted its analysis as though every applicant at one Sears store is simultaneously an applicant at every Sears store. EEOC also calculated its z statistics as if all hires were drawn from the same pool with a known proportion of women. Instead, they were in fact drawn from many individual store pools with widely varying proportions of women. EEOC estimated the variance between the proportion of women in the aggregated pool and the proportion of women in each *1317 store pool. However, EEOC's statistical expert, Dr. Siskin, did not incorporate these variances into his calculation of the z statistics. Also, EEOC ignored the variances of the factors it did adjust for in its regression analyses.
When these variances were taken into account, the z statistics associated with EEOC's disparities were significantly lower. See Sears Exhibits 5-3, 5-4. For EEOC's nationwide multivariate analysis of full time commission sales hires, z values exceeded 3 in only two years. Since the analyses on which the z values are based do not take into account the wide divergence between male and female interest in commission sales, the court concludes that a fair analysis incorporating applicant interest would not produce statistically significant disparities.
4. EEOC's Applicant Interview Guide Analysis
EEOC has analyzed responses to Applicant Interview Guide questionnaires independently of its regression analyses. These questionnaires allowed applicants to rate their interest, skill and experience in a number of categories. EEOC adjusted its "sales" pool using the interest and experience responses of applicants. In performing this analysis, Dr. Siskin attempted to factor in the extent to which commission sales hires actually had experience according to EEOC's coding. EEOC concluded from its analysis that the expected female hiring rate still exceeded the actual rate in all four product lines.
However, EEOC's analysis does not adjust for any of the factors included in its multivariate and logit analysis. EEOC merely adjusted its "sales" pool, the flaws of which were discussed at length above, by the applicants' own ratings of their skill and experience on a scale from one to five. The court finds that this analysis of the ill-conceived "sales" pool cannot be used as a fair basis for comparison with Sears hiring rates. Sears' analysis of the Applicant Interview Guide questionnaires is discussed below.
5. EEOC's Nonstatistical Evidence
Some of EEOC's nonstatistical evidence also merits discussion. First, EEOC has focused considerable attention on Sears' testing practices. Although EEOC has stated throughout this litigation that it does not claim that Sears' testing practices violate Title VII, it has attempted to show that Sears' psychological tests "disadvantaged" women.
However, many Sears applicants were not tested, and often those who were tested were not tested until after they were hired. The testimony of Sears managers makes clear that, in most cases where employees were tested before being hired, the test had little impact on any decision to hire. This was especially true in considering female applicants for less traditional jobs like commission sales. The credible evidence shows that test scores for women were often completely ignored in hiring women into commission sales.
Moreover, Sears' evidence showed that it used different scoring criteria for males and females, which resulted in a lower satisfactory score for women. In addition, the *1318 actual scores of women on the "vigor" scale of the test shows that the vigor scores of women hired were substantially lower than those of the men hired. Sears Exhibit 6-III, 6-JJJ. There is no credible evidence that a woman's "vigor" score ever prevented her from being hired into commission sales at Sears. The court therefore finds that Sears' testing program did not discriminate against women, and provides no support for EEOC's claim.
EEOC also relies on what it characterizes as Sears' "masculine" description of a commission salesperson in Sears' Retail Testing Manual. However, there is no evidence that this historical description had any influence on hiring decisions, particularly during the years in question. To the contrary, Sears managers often paid no attention to information in the manual. There is no basis in the record for concluding that this description had any negative impact on the selection of women for commission sales.
Finally, EEOC presented two witnesses in its rebuttal case, two women who had applied for commission sales positions at Sears and were not hired. These witnesses were not called specifically to testify that Sears' had discriminated against them. They were called to rebut testimony of a Sears witness regarding coding of applications. Their testimony illustrated the difficulty of determining interest and qualifications of applicants solely from an application form. Neither witness provided any evidence that Sears discriminated against women.
One witness, Alice Howland, applied to a store which had hired only one commission salesperson from outside of Sears for nine years, from 1976 to the present. Consequently, at the time Ms. Howland applied, applications for full time commission sales generally were not even processed. Thus, her testimony provides no proof of discrimination by Sears.
The second witness, Lura Nader, applied for a commission sales position in draperies in response to an advertisement. She was not hired, and testified at trial that, while she didn't think so at the time, she now thought Sears had discriminated against her. However, Sears proved that it hired in that position at that time two other women who had applied on the same date as Ms. Nader and who had better qualifications for the job than she did. Faced with this evidence, Ms. Nader refused to alter her claim of discrimination. Obviously, however, her testimony provides no credible evidence that Sears discriminated against women.
Therefore, none of the nonstatistical evidence offered by EEOC supports its statistical analyses or provides any credible proof of discrimination by Sears. All of the deficiencies in EEOC's evidence discussed above leave its statistical analyses with virtually no persuasive value. Moreover, Sears has presented other persuasive evidence to prove that it did not discriminate against women in commission sales.
E. Sears' Commission Sales Analyses: Hiring
Sears did not perform its own statistical analyses of applicants for commission sales positions. Instead, it offered a variety of other evidence to prove that EEOC's statistical analyses are not entitled to any weight, and to prove that the number of women hired and promoted into commission sales at Sears reflects its affirmative action efforts, not discrimination. Much of this evidence was discussed above in evaluating EEOC's evidence. In addition to that evidence, Sears presented the following facts.
*1319 1. Hiring and Promotion Figures at Sears
During the period from 1973 through 1980, Sears never hired fewer than 20 percent women into full time commission sales and 37 percent women into part time commission sales in any year. Sears hired as many as 40 percent women into full time commission sales, and 52 percent women into part time commission sales. Between 1973 and 1979, the female percentage of full time commission sales hires doubled from 20 to 40 percent. See Sears Exhibits 61, 6T, 6KKK.
Between 1973 and 1975, the female percentage of part time commission sales hires increased from 41.5 percent to 52.3 percent. After 1975, this percentage decreased. This decrease is explained primarily by the fact that the distribution of part time commission sales positions within product lines changed. The number of commission sales positions in product lines in which women were interested decreased, while jobs selling products of greater interest to men increased (e.g., automotive, home building materials).
The proportion of women hired into commission sales varied substantially by product line. For example, the percentage of women hired to sell automotive products was much lower than the percentage of women hired to sell home furnishings. See Sears Exhibit 6-T.
From 1974 through 1980, Sears promoted 11,428 noncommission salespersons to commission sales at 992 Sears stores across the nation. Of those promotions, 53.5 percent went to women. The female proportions of promotions was never lower than 40.7 percent (in 1977), and was as high as 58.3 percent (in 1976). The female proportion of promotions from noncommission to part time commission sales exceeded 50 percent in every year from 1974 through 1980, reaching a high of 62.2 percent in 1975. See Sears Exhibits 6-15-1, 6-LL, 6-PPP.
More than 350 different positions supplied employees for commission sales jobs. Only half of the movement from other positions at Sears into commission sales came from noncommission sales positions. See Sears Exhibit 6-15-2.
The female percentage of commission sales promotions varied substantially by product line. For both full time and part time employees, the proportion of women promoted into commission sales has always exceeded the proportion of women hired to sell both products in which women have been more interested and products which women have not been very interested in selling. See Sears Exhibits 6-T, 6-2. This is the result of the success of Sears' affirmative action efforts in promoting women into commission sales who otherwise would not have applied for the positions.
2. Characteristics of Commission Salespersons
Commission salespeople and noncommission salespeople at Sears have differing distinct characteristics. Commission salespeople are older and have more experience than noncommission salespeople. Differences in age and experience are even more pronounced between big ticket commission salespeople and noncommission salespeople. Both male and female commission sales hires were more likely than nonhires to have been managers, self-employed, or former Sears employees. Most commission sales hires, and to a greater extent big ticket commission sales hires, had this type of related experience.
In addition, among those hired into commission sales, males tended to have more desirable job qualifications than females. Men were more likely to have had previous experience related to commission sales. Among the hires with previous experience, the men had more experience than the women. Even among the hires who had the same number of previous experiences, men had more experience than women. See Sears Exhibits 6-9, 6-J-1. In addition, among the hires with previous related experience, commission salesmen were more educated than commission saleswomen. See Sears Exhibit 6-9C.
*1320 Sears performed a case study of all commission sales hires in all stores in EEOC's non-hired applicant sample. This case study confirmed that male commission sales hires had higher qualifications than female commission sales hires. The study was based on all the information available in the personnel files of applicants. This information is superior to the limited data relied on by EEOC from Sears' payroll tapes.
This Sears case study revealed a number of significant facts. Of full time commission sales hires, 80 percent of the males and 75 percent of the females had at least one of the following three characteristics: (1) a preference for commission sales, or to sell a product sold on commission at Sears; (2) commission sales experience; or (3) training, experience or education relevant to the product line in which they were hired to sell. Of part time commission sales hires, 46 percent of the men and 34 percent of the women had at least one of these three characteristics. Of new full time commission salespersons who transferred from other positions at Sears, 60 percent of the men and 38 percent of the women had at least one of these three characteristics. Of the part time commission sales transfers, 75 percent of the men and 51 percent of the women had at least one of these three characteristics. It must also be noted that a disproportionate number of women hires and transfers were credited with this experience solely because they expressed an interest in or had a background in interior design or draperies, which relates only to a small part of the commission sales product lines.
Sears' study further revealed that, of those remaining employees who did not have one of the three characteristics described above, almost all hires and transfers, and more men than women, were one of the following: (1) a prior Sears employee; (2) a member of a minority group and/or a Vietnam-era veteran; or (c) persons whose files disclosed "patterns of achievement" apart from prior commission sales or product line experience. Of the full time commission sales hires, 95 percent of the males and 90 percent of the females had at least one of the six characteristics, and many had more than one. Of the part time sales hires, 80 percent of the men and 65 percent of the women had at least one of the six characteristics. Thus, the vast majority of commission sales hires, both full time and part time, had at least one, and often more than one, of the six characteristics listed above. Males tended to have more of these characteristics than females.
As EEOC points out, this study would be more valuable if a representative group of nonhired applicants had also been analyzed for comparison with the hires. However, even without this basis for comparison, the study is useful because it demonstrates that Sears' commission sales hires all tend to have at least one characteristic which would make them more likely to be hired than persons without that characteristic, and that the men hired possessed more of these characteristics than the women. Both of these findings are consistent with Sears' evidence that it attempted to hire commission salespersons who demonstrated in some manner that they could perform well, while complying with its own affirmative action plans.
3. Sales Performance Data
Sears also analyzed the sales performance of its commission salespeople. Sales *1321 performance data can provide a good measure of a salesperson's true qualifications for commission selling. It can reflect characteristics related to sales ability which are not easily quantified for a statistical analysis. Sales performance data of commission salespersons at Sears shows that the males have consistently outsold the females in most years, territories and divisions.See Sears Exhibit 6-3, 6-4, 6-5, 6-C. However, relative to their sales, commission saleswomen were paid more than commission salesmen between 1973 and 1980. Among both full time and part time commission salespersons across all territories, commission saleswomen were less likely than commission salesmen to meet or exceed the standard sales per hour set for their division. See Sears Exhibit 6-5. Among full time and part time commission salespeople, the sales performance rates of the males exceeded the rates of the females. This was true among all new hires and promotions, among the top 10 percent of commission salesmen and saleswomen, and among the bottom 10 percent of commission salespeople. Thus, in general, the men Sears hired and promoted into commission sales were more successful salespeople, and therefore, it can be inferred, had more of the true qualifications for commission sales, than the women. These facts are all consistent with affirmative action, and corroborate Sears' other evidence of its affirmative actions program.
Sears expert, Dr. Wise, examined the bottom 10 percent of sales performers and offered another explanation for Sears' hiring patterns. Sears hires relatively few commission salespersons in comparison to the number of job applicants. According to EEOC's analysis, only one of every 250 job applicants is selected for commission sales. Dr. Wise demonstrated that this very small probability of selection can significantly affect the percentage of men and women hired, if men on average are more qualified than women. A difference in average male and female qualifications would not have a great effect on hiring if a large number of applicants were hired. However, when a very small percentage of applicants is hired, assuming that Sears tries to hire the most qualified applicants (those located at the right tail of the distribution curve of job applicants), then far more males than females will be hired. This is true because there are far more males than females at the right tail of the distribution curve due to the difference in average qualifications of males and females. Dr. Wise provided many examples of the effect on the expected proportion of female hires of differences in interest or qualifications that are not incorporated into EEOC's analysis. These examples demonstrate that, even if differences between male and female interests and qualifications are relatively small, these differences can result in grossly inaccurate predictions of the expected proportion of female hires because of the small proportion of applicants hired. See Written Direct and Rebuttal Testimony of Dr. David A. Wise Regarding Commission Sales.
EEOC attacked Dr. Wise's model because it is based on the assumption that Sears can perfectly choose the applicants best qualified to sell on commission at Sears. EEOC witness Dr. Siskin contends that there are likely to be errors in the evaluation of applicants which would make Dr. Wise's model inapplicable to Sears. As Dr. Wise testified, however, any random error that might occur in the evaluation process would inure to the benefit of the lesser qualified group, in this case, the women. Moreover, although Dr. Wise's *1322 model does not perfectly reflect the hiring process at Sears, it provides a useful illustration of the significant effect that EEOC's failure to incorporate differences between males and females can have on the results of its analysis because of the very small number of job applicants hired into commission sales. Dr. Wise's model provides further proof that EEOC's model does not fairly predict the expected proportions of women in commission sales in the absence of discrimination by Sears.
4. Applicant Interview Guides
Sears also analyzed Applicant Interview Guides provided to applicants in certain stores between 1978 and 1980 to measure female interest in areas related to these five product lines in which EEOC found the greatest disparities: automotive, home building materials, electronics, appliances, and home improvements. Applicants were asked to rate their levels of interest, skill and experience in a wide variety of job activities, including some related to commission and noncommission sales. Applicant responses showed great differences between the interests of men and women. Men were more interested in job activities usually associated with commission sales positions at Sears (e.g., selling appliances, hardware, and technical goods). Women were disproportionately interested in activities usually associated with noncommission sales at Sears (e.g., cashier, selling apparel). Similar differences were found in their self-evaluations of skill and experience. See Sears Exhibit 6-B. This evidence corroborates the survey evidence discussed above, which also demonstrated that males were far more interested in commission sales than females.
Dr. Haworth examined certain responses to the Applicant Interview Guides which are related to the five divisions in which EEOC found the greatest disparities. In analyzing these responses, she "normalized" the scores that applicants gave themselves on interest, skill and experience in each of these categories. The scores were normalized to take into account that some applicants might inflate their scores to increase their chances of being hired. Dr. Haworth chose Applicant Interview Guide categories that were roughly analogous to the five product lines. She then adjusted EEOC's expected percentage of female commission sales hires based on the responses to the Guide, to attempt to factor *1323 applicant interest into the analysis. After adjusting for this interest factor, the disparities presented for 1976-1980 in the five product lines were eliminated or greatly reduced.
Sears also analyzed the Applicant Interview Guide data to take into account the fact that very few applicants would be selected, in accordance with Dr. Wise's testimony discussed above. This analysis reduced the expected percentage of females even further, such that the "actual" proportion of female hires as counted by EEOC exceeded the expected percentage of female hires in all five product lines. See Sears Exhibits 5-6; 6-HHH. This was true for both full time and part time commission sales hires.
EEOC has challenged Sears' use of the Applicant Interview Guides on a number of bases. First, it contends that the normalization process unnecessarily deflates the expressed interest of women in these product lines. EEOC asserts that this process is appropriate for comparing interests of a single individual in various products, not for comparing different applicants' interest. It is clear from the testimony that normalization of the scores may in some cases distort an applicant's true interest, and in other cases may appropriately modify an applicant's inflated rating of interest and abilities. The court recognizes that, by normalizing scores, Sears may have undervalued the interest of some female applicants in these product lines. Bearing this in mind, however, the court finds that the results of Sears' analysis do provide probative evidence of relative lack of interest. Although Sears' analysis may understate to some extent female interest, its analysis of the Guides is consistent with other more reliable evidence of female interest discussed above. Collectively, this evidence clearly demonstrates that differences in interest alone accounts for a great proportion of EEOC's disparities. This Sears' analysis also illustrates the important fact that most of the disparities claimed by EEOC arose in product lines in which women tend to have relatively little interest.
EEOC also criticized the Applicant Interview Guide categories chosen by Sears as failing to correspond to the five product groups with which Sears compared them. Although some categories relied on by EEOC in its analysis may be more analogous to the relevant product lines, the categories used by Sears are close enough to the product lines to have some probative value.
EEOC also notes that Sears compared the Applicant Interview Guide responses only to Sears' hiring in the years 1976 through 1980, not for 1973 through 1975. However, the Guides were distributed only in 1978 through 1980. Since the interest of women in nontraditional jobs increased significantly in the 1970's, Sears comparison of the responses only to the years 1976-1980 was reasonable.
Finally, EEOC challenges on two bases Sears' second analysis of the Guides, which takes into account the fact that very few applicants are actually hired. First, EEOC contends that the analysis is flawed because it requires that all hires be from the 99th percentile of persons responding to a particular item. However, as Dr. Wise testified, the model does not assume that persons at the top of the ranking on the Applicant Interview Guides would be among the top one percent of applicants according to actual qualifications. The model only shows that according to the ranking, and including its random error components, the proportions of women would vary dramatically by product line. Second, EEOC asserts that the model assumes that the scores of men and women will be normally *1324 distributed. Dr. Wise admits that the exact results obtained under the model would differ depending on the distributional assumptions made. However, the overall pattern of the results would not change; the model would still show that the percentage of women expected to be hired into the five divisions would be far lower than EEOC's model predicts.
In conclusion, the court finds that Sears' analysis of the Applicant Interview Guides provides some useful information regarding women's lack of interest in commission sales at Sears, and it corroborates Sears' other persuasive evidence to that effect discussed above. However, because of the problems with both parties' analysis of the Guides, the court has not given substantial weight to either side's analyses of the Applicant Interview Guides.
F. Conclusions Regarding Commission Sales Claim: Hiring
Viewing all of this evidence together, considering the credibility of the witnesses and the reasonableness of their testimony, the court finds that EEOC has failed to carry its burden of persuasion on this claim. Its statistical evidence is wholly inadequate to support its allegations. The evidence is replete with flaws which, if any were considered alone, might invalidate the statistical conclusions. Together, all of these flaws leave the analysis with little probative value.
The most eggregious flaw is EEOC's failure to take into account the interests of applicants in commission sales and products sold on commission at Sears. EEOC turned a blind eye to reality in constructing its artificial, overinclusive "sales" pool, and assuming away important differences in interests and in qualifications. The court finds that, because of these errors, EEOC's highly inadequate coding of data available to it, and the other problems with the analyses discussed above, the results of EEOC's statistical analyses are entitled to little weight.
Moreover, evidence presented by Sears provided a more reasonable basis for evaluating Sears, and showed that Sears met all reasonable estimates of the proportion of qualified and interested women. Its evidence demonstrated that, when interest and qualifications are taken into account, EEOC's alleged disparities are virtually eliminated.
Equally important, EEOC has failed to counter Sears' highly convincing evidence of its affirmative action programs. Sears has proven that it has had a long and serious commitment to affirmative action, which applied to the recruitment of women for its commission sales force. Sears' affirmative action evidence established that, not only were women much less interested in commission sales at Sears than men, but also, more importantly, Sears did not have any intent to discriminate against women. Sears' affirmative action efforts to recruit women to commission sales went much farther than is required by law. EEOC has not presented any credible evidence to cast doubt on the commitment of Sears to affirmative action. The court therefore finds that Sears' evidence of its strictly enforced affirmative action programs is strong evidence that it did not in fact intentionally discriminate against women.
Finally, notably absent from EEOC's presentation was the testimony of any witness who could credibly claim that Sears discriminated against women by refusing to hire or promote women into commission sales. It is almost inconceivable that, in a nationwide suit alleging a pattern and practice of intentional discrimination for at least 8 years involving more than 900 stores, EEOC would be unable to produce even one witness who could credibly testify that Sears discriminated against her.*1325 EEOC's total failure to produce any alleged victim of discrimination serves only to confirm the court's conclusion that no reasonable inference of sex discrimination can be drawn from EEOC's statistical evidence regarding hiring into commission sales.
Therefore, based on all of the evidence presented, the court concludes that EEOC has failed to prove that Sears had a pattern or practice of intentionally discriminating against women in hiring into commission sales. On the contrary, the court finds that Sears has proven that it did not have such a pattern or practice. Further, Sears has proven legitimate, non-discriminatory reasons for the alleged statistical disparities between the hiring of men and women into commission sales from 1973 until 1980, and EEOC has not proven them pretextual.
G. Promotion Claim
Much of the evidence discussed above with respect to hiring is relevant to EEOC's promotion claim against Sears, particularly the evidence regarding interest in commission sales, the qualities sought in commission salespeople, the characteristics of commission salespeople, and affirmative action. All of the court's findings above relating to these and all other relevant matters are incorporated here. However, EEOC analyzed promotions to commission sales separately, using a completely different analysis. Its promotion analysis must therefore be discussed separately.
1. EEOC's Promotion Analysis
EEOC analyzed promotions into commission sales by arbitrarily assuming that all noncommission salespersons at Sears were candidates for commission sales positions. EEOC also arbitrarily assumed for its analysis that, during the relevant time period, all noncommission salespersons were equally interested in and qualified for, and had in fact applied for, commission sales jobs at Sears. It's Year-End Store Pool analysis used the female percentage of noncommission salespersons at Sears as the "expected" percentage of female hires into commission sales. It compared these percentages for each year with the female proportion of promotions from noncommission sales into commission sales. EEOC contends that the resulting disparities are proof of Sears' intentional exclusion of females from commission sales.
EEOC made only one effort to "adjust" its analysis, by taking into account the possibility that a noncommission salesperson in a division in which there is an opening might have a greater chance of getting that position than someone outside the division. This analysis did not substantially change the results of the first analysis.
Once again, however, EEOC has deliberately made false assumptions which undermine the validity of its analyses. Its assumption that all noncommission salespersons are in essence applicants for all commission sales jobs at Sears has absolutely no basis in fact, and is contrary to all the credible evidence in the case. As discussed at length above, Sears' evidence of its affirmative action efforts, its survey evidence, and the uncontradicted testimony of *1326 its officers, managers and other store witnesses, prove that relatively few noncommission salespersons were interested in commission sales positions at Sears. Of those who were interested, the men were far more interested than the women in commission selling.
Not only did EEOC erroneously fail to consider the interest of noncommission salespersons in commission sales, it did not even attempt to make any adjustments for differences in qualifications of males and females. EEOC asks this court to assume that all noncommission salespersons would all perform equally well in all commission sales positions, without offering any basis from which the court could reasonably make such an assumption. To the contrary, all evidence regarding qualifications shows that males at Sears were more qualified for commission sales positions than females. The court therefore cannot accept EEOC's baseless assumption of equal qualifications. EEOC's assumptions of equal interest and qualifications render its promotion "analysis" virtually meaningless. Its "adjustment" of its analysis by division pool does nothing to cure these fatal defects.
2. Sears' Evidence Regarding Promotions
Sears expanded EEOC's analysis of promotions by factoring in the indicated interest of noncommission salespersons in commission sales. It relied on responses to Career Aspirations Questionnaires administered to noncommission salespersons at stores in EEOC's non-hired applicant sample. Responses to these questionnaires show that, even in 1982-83, noncommission salesmen were three times as likely as noncommission saleswomen to be interested in moving to commission sales. Applying these figures to EEOC's year-end division pool figures eliminates the overall disparities calculated by EEOC for both full time and part time commission sales promotions.
EEOC criticizes Sears' use of the Career Aspirations Questionnaires, because Sears combined full time and part time responses due to the lack of full time female applicants interested in commission sales. However, there is no indication that the part time responses do not fairly reflect female interest in commission sales. Moreover, very similar results were obtained from the 1976 Job Interest Survey and the National Timecard Nonsupervisory Special Survey ("NTNSS"), both discussed above. The court therefore finds that the Career Aspirations Questionnaires provide a reasonable estimate of the interest of noncommission saleswomen in commission sales. Sears' analysis of these questionnaires demonstrates that interest alone can account for the disparities computed under EEOC's analysis.
Sears also performed a second expansion of EEOC's analysis. First, it adjusted the analysis according to divisions that actually feed into commission sales positions in a given division. EEOC's analysis adjusted only for the division in which a position existed. EEOC assumed that noncommission salespersons in all other divisions had an equal chance of promotion to that division, regardless of how related their experience was to sales in the division with the commission sales opening. EEOC's analysis also adjusted only for those divisions in which there were both commission and noncommission salespersons. Sears' approach reflected the reality that, in many divisions, most persons promoted to commission sales came from a small number of feeder divisions.
Sears performed two adjustments in this regard. First, it adjusted to take into account the possibility that a noncommission salesperson, working in an EEOC selected product line grouping in which a promotion occurred, had a greater chance of receiving the promotion. This analysis substantially reduced EEOC's disparities. See Sears Exhibit 6-5-2 at 1-4. Sears then adjusted to take into account the greater chance of being promoted from the most important feeder divisions, irrespective of whether the product line was related. This analysis also substantially reduced EEOC's disparities. After this analysis, the difference *1327 between EEOC's expected and actual female hiring figures was less than one person per store over the entire period at issue in the case. See Sears Exhibits 6-5-1, 6-5-2. This analysis does not even factor in the aspect of interest in commission selling.
Sears further adjusted this data by controlling for the age and seniority of noncommission salespersons. This adjustment reduced the disparity to only 10% of all promotions into commission sales. See Sears Exhibit 6-5-2 at 5-6. The product lines that contributed most heavily to the remaining disparities were automotive, home building materials and men's apparel. All three are product lines in which women had relatively little interest. Thus, even without controlling for interest, EEOC's asserted disparities have been substantially reduced.
EEOC challenges this second Sears analysis on several grounds, the most important of which relates to the "feeder division" adjustment. EEOC asserts that there is no basis for adjusting for divisions which were shown to feed into commission sales positions in a particular division, unless there is some relationship between the product lines or the experience in the two divisions. EEOC correctly notes that Sears has not shown that the experience in these feeder divisions is related to the type of selling required in the division with the opening. The court agrees that the adjustment for all feeder divisions may be inappropriate. Although there may be legitimate business reasons for the pattern of movement from various divisions into particular commission sales positions, Sears has introduced no evidence to prove this. The court cannot therefore wholly accept Sears' second analysis. However, Sears' adjustment for divisions in the same EEOC product line may reflect related experience. The court therefore finds Sears' second analysis is probative, but recognizes that the analysis may overreduce EEOC's disparities. Despite this, however, the analysis shows that adjusting EEOC's results for just a few relevant factors, excluding interest, substantially reduces the disparities. It is reasonable to conclude that, if the interest factor were added to the analysis, the disparities would be virtually eliminated.
H. Conclusions Regarding Commission Sales Claim: Promotions
All of Sears' evidence merely underscores the complete inadequacy of EEOC's statistical analysis of promotions. EEOC's analysis, based on false assumptions, provides no evidence from which the court could reasonably infer that Sears intentionally discriminated against women in promotions to commission sales.
Moreover, as discussed above with respect to the hiring claims, Sears' affirmative action efforts effectively disprove any claim of intentional discrimination by Sears with respect to promotions. As recounted above, continuous efforts were made to convince noncommission saleswomen to try commission sales. Sears' witnesses' testimony convincingly demonstrated that Sears' entire management, from the highest levels at corporate headquarters to store managers and other personnel, were committed to Sears' affirmative action goals, including the promotion of noncommission saleswomen into commission sales. This testimony stands uncontradicted. The statistical evidence presented by EEOC on its promotion claims does not refute this evidence.
EEOC also failed to present testimony by any Sears employee that she had been unfairly denied a promotion into commission sales at Sears. Once again, despite its allegations of nationwide discrimination in more than 900 stores over 8 years, EEOC was unable to produce one Sears employee to testify that Sears discriminated against her by refusing to promote her to commission sales. This total lack of any testimony to "bring the statistics to life" is but further confirmation of the failures of EEOC's statistical evidence.
Considering all of this evidence, the court concludes, as it did in the hiring claim, that EEOC has failed to prove its claim of a nationwide pattern or practice of intentional discrimination against women in promotions into commission sales at Sears. To the contrary, the court finds that Sears *1328 has proven that it did not have any such pattern or practice. Moreover, Sears has proven legitimate, non-discriminatory reasons for the alleged statistical disparities, and the EEOC has not proven them pretextual.
III. Checklist Compensation
Sears employees are classified as either timecard or checklist. Checklist employees are the company's salaried executive management, professional and administrative employees, most of whom meet the duty requirements for "exempt" status established by the Fair Labor Standards Act. EEOC has charged that Sears had a nationwide pattern and practice of discrimination against women by paying them less than men in 51 specified checklist jobs.
Although EEOC originally alleged pay claims under both the Equal Pay Act and Title VII, EEOC subsequently dropped all of its claims under the Equal Pay Act. Its pay claim will therefore be analyzed under Title VII only. The general legal standards applicable to Title VII cases were discussed above. However, a few issues relating to pay claims under Title VII have arisen which require further discussion.
A. Legal Standards
1. Elements and Burdens of Proof
First, EEOC asserts that the elements and burdens of proof of the Equal Pay Act apply to its claim of a pattern and practice of pay discrimination against Sears under Title VII. Under the Equal Pay Act analysis, a plaintiff need not directly prove intent to discriminate, and the burden of persuasion shifts to the defendant once the plaintiff shows that the defendant pays workers of one sex more than workers of the opposite sex for equal work. The employer then bears the burden of persuading the court that the pay differential is justified under one of the four exceptions specified in the Act. Corning Glass Works v. Brennan, 417 U.S. 188, 196, 94 S. Ct. 2223, 2229, 41 L. Ed. 2d 1 (1974). EEOC thus asserts that it need not prove that Sears had a pattern and practice of intentionally discriminating against women in checklist pay as is required under Title VII analysis, and that it does not bear the ultimate burden of persuasion in the case if it can prove that checklist women received less pay than checklist men in equal jobs. EEOC relies on several cases to support this contention. These cases, in turn, rely on the Bennett Amendment to Title VII, and a Supreme Court case construing this amendment, County of Washington v. Gunther, 452 U.S. 161, 101 S. Ct. 2242, 68 L. Ed. 2d 751 (1981).
The Bennett Amendment, § 703(h) of Title VII, 42 U.S.C. § 2000e-2(h), provides:
It shall not be an unlawful employment practice under this subchapter [Title VII] *1329 for any employer to differentiate upon the basis of sex in determining the amount of the wages or compensation paid or to be paid to employees of such employer if such differentiation is authorized by the provisions of § 206(d) of Title 29 [the Equal Pay Act].
In Gunther, the Supreme Court held that the Bennett Amendment incorporates into Title VII only the four affirmative defenses of the Equal Pay Act, which are that any pay differential resulted from a seniority system, a merit system, a system which measures earnings by quantity or quality of production, or "any other factor other than sex." The Supreme Court took great care to point out that its holding was narrow, and that it was not deciding how sex-based wage discrimination litigation under Title VII should be structured to accommodate the incorporated defenses of the Equal Pay Act, and in particular, the fourth defense. 452 U.S. at 171, 101 S. Ct. at 2249.
At least one of the cases which EEOC cites as support for its contention that the Equal Pay Act analysis applies to Title VII claims of unequal pay for substantially equal work was decided before the Supreme Court's 1981 Gunther opinion. In this case, and other similar pre-Gunther cases, the courts based their decisions to apply the Equal Pay Act analysis on the Bennett Amendment and on the principle that the provisions of Title VII dealing with sex discrimination must be construed in harmony with the Equal Pay Act. DiSalvo v. Chamber of Commerce of Greater Kansas City, 568 F.2d 593, 596 (8th Cir.1978); Orr v. Frank R. MacNeill & Son, Inc., 511 F.2d 166, 170 (5th Cir.1975), cert. denied, 423 U.S. 865, 96 S. Ct. 125, 46 L. Ed. 2d 94 (1975); Ammons v. Zia Co., 448 F.2d 117, 119 (10th Cir.1971); Schultz, 421 F.2d at 266. The post-Gunther cases add, as a supplementary rationale, the argument that the Supreme Court implied in Gunther that the traditional Title VII analysis "does not govern in Title VII actions for sex-based wage discrimination." Miller, 585 F. Supp. at 1512. See also Kouba, 691 F.2d at 875; Schulte, 547 F. Supp. at 339-40.
However, the Supreme Court in Gunther did not hold or imply that the Equal Pay Act analysis governs Title VII claims of unequal pay for substantially equal work. The passage in Gunther which the cases point to as implying that the Equal Pay Act analysis applies in Title VII sex-based wage discrimination cases is not as equivocal as these courts have presented it to be. As stated above, the Gunther Court *1330 clearly pointed out that it was not deciding how sex-based wage discrimination litigation under Title VII should be structured. 452 U.S. at 171, 101 S. Ct. at 2249. To the contrary, the Gunther Court's discussion of the legislative history of the amendment indicates that it did not intend to imply that the Bennett Amendment incorporates the Equal Pay Act analysis into Title VII. 452 U.S. at 171-76, 101 S. Ct. at 2249-51. This legislative history reveals that Congress did not intend the Bennett Amendment to impose Equal Pay Act burdens and elements of proof to pay claims brought under Title VII. The Bennett Amendment was intended to be a shield, not a sword.
In addition, applying the Equal Pay Act analysis to Title VII claims of unequal pay for substantially equal work creates, rather than prevents, confusion and disharmony between the Equal Pay Act and Title VII. Indeed, confusion has already resulted from such application. For example, the cases which have applied the Equal Pay Act analysis under Title VII do not agree on whether the Equal Pay Act analysis applies in every Title VII case involving *1331 unequal pay for substantially equal work, or whether the Equal Pay Act analysis applies only when the employer asserts the fourth affirmative defense ("factor other than sex"). The courts in these cases are confused with good reason. In Gunther, the Supreme Court recognized that the first three affirmative defenses of the Equal Pay Act are already clearly set out elsewhere in section 703(h) of Title VII. 452 U.S. at 169, 101 S. Ct. at 2248. Therefore, these three defenses are both set out in Title VII and incorporated into Title VII by the Bennett Amendment. Should an employer assert one of these defenses in a Title VII action, it could claim that it is asserting the defense as set out in Title VII and Title VII analysis applies; the plaintiff, on the other hand, could claim that the employer is asserting the defense as incorporated by the Bennett Amendment and the Equal Pay Act analysis therefore applies.
Still more confusion and disharmony would result in a Title VII unequal pay for equal work case where the employer was exempt under the Fair Labor Standards Act. If the Equal Pay Act analysis applies, the employer would lose the benefit of its exemption under the Fair Labor Standards Act and would have to meet the Equal Pay Act's heavier burden of proof. Exemptions under the Fair Labor Standards Act would therefore be rendered superfluous.
Finally, in the recent decision of Patkus v. Sangamon-Cass Consortium, 769 F.2d 1251 (7th Cir.1985), the court stated:
Title VII and the Equal Pay Act are to be construed "in harmony," Grann v. City of Madison, 738 F.2d 786, 788 n. 1 (7th Cir.1984) (citing Orahood v. Board of Trustees, 645 F.2d 651, 654 (8th Cir. 1981)), cert. denied, [___] U.S. [___], 105 S. Ct. 296 [83 L. Ed. 2d 231] (1984), even though Title VII requires a showing of discriminatory intent while the Equal Pay Act creates a type of strict liability in that no intent to discriminate need be shown. Strecker v. Grand Forks County Social Service Board, 640 F.2d 96, 99 (8th Cir.1980). Since the four statutory defenses available under the Equal Pay Act also serve as valid defenses to a claim based on Title VII, County of Washington v. Gunther, 452 U.S. 161, 167, 101 S. Ct. 2242, 2247, 68 L. Ed. 2d 751 (1981); 42 U.S.C. § 2000e-2(h), however, and since we hold that defendants have successfully established a defense to the Equal Pay Act claim, there is no need to address the separate Title VII claim of discrimination in pay.
769 F.2d at 1260 n. 5.
Therefore, following Gunther and other persuasive authority, the legislative history of the Bennett Amendment, and the principle that statutes should be construed harmoniously, this court finds that the Bennett Amendment incorporates into Title VII the four affirmative defenses of the Equal Pay Act, but not the corresponding Equal Pay Act analysis regarding the elements and burdens of proof. Title VII of the Civil Rights Act of 1964 and the Equal Pay Act of the Fair Labor Standards Act have different purposes, coverage, and procedures.*1332 The fact that a plaintiff can bring an identical cause of action under both statutes should not lead the courts to abandon the analysis of Title VII. The court will therefore apply the traditional Title VII analysis in this case, requiring EEOC to bear the ultimate burden of persuading the court that Sears intentionally discriminated against women in checklist pay.
2. "Similar" Work Standard
EEOC also asserts that it need only prove that the persons in the job it analyzed were performing "similar" work. EEOC acknowledges that, under Title VII, a plaintiff establishes a prima facie case of discrimination when it shows that women receive less pay than men for substantially equal work, the standard applied under the Equal Pay Act, 29 U.S.C. § 206(d). Under this standard, EEOC must prove that the persons compared performed work of equal skill, effort and responsibility. Id. This is the "traditional" equal pay analysis, applied under both the Equal Pay Act and Title VII. EEOC has attempted to prove at trial that, during the years 1973-1980, persons with the same job title and job code at Sears performed substantially equal work under this standard.
EEOC also asserts, however, that it can prevail on its pay claim even if it does not prove that the jobs are substantially equal, if it proves that they are "similar." EEOC relies again on Gunther, supra. In Gunther, the employer assessed the worth of male and female prison guards jobs by considering the worth of the jobs and outside markets. It then decided to pay male guards 95 percent of the assessed worth of their jobs, and female guards 75 percent of the assessed worth of their jobs. The jobs performed by male and female guards were not equal. Therefore, no relief could be granted under the Equal Pay Act standard. However, the Court concluded that failure to meet the Equal Pay Act standard does not bar recovery under Title VII.
The Court reasoned that the Bennett Amendment to Title VII, which incorporates the Equal Pay Act defenses into Title VII, does not preclude other types of compensation claims under Title VII not cognizable under the Equal Pay Act. The Court pointed to the broad remedial purposes of Title VII, and found that limiting Title VII to claims cognizable under the Equal Pay Act standard could allow blatant cases of sex-based wage discrimination, which do not fit within the Equal Pay Act analysis, to go unredressed. The Court was faced with such a situation, in which female employees *1333 were not performing work equal to the males, but nonetheless were clear victims of sex discrimination in pay. In these circumstances, the Court concluded that the plaintiffs had established a violation of Title VII without proving that the males and females performed equal work.
The Court specifically declined to define the precise contours of compensation claims under Title VII beyond the facts presented to it. The Court indicated the narrowness of its ruling by stating:
We emphasize at the outset the narrowness of the question before us in this case. Respondents' claim is not based on the controversial concept of "comparable worth," under which plaintiffs might claim increased compensation on the basis of a comparison of the intrinsic worth or difficulty of their job with that of other jobs in the same organization or community. Rather, respondents seek to prove, by direct evidence, that their wages were depressed because of intentional sex discrimination, consisting of setting the wage scale for female guards, but not for male guards, at a level lower than its own survey of outside markets and the worth of the jobs warranted. The narrow question in this case is whether such a claim is precluded by the last sentence of § 703(h) of Title VII, called the "Bennett Amendment."
452 U.S. at 166, 101 S. Ct. at 2246 (footnotes omitted). The Court further clarified, in a footnote, that the sole issue before the Court was whether the respondents' failure to demonstrate equal work precluded them from proceeding under Title VII. 452 U.S. at 166 n. 8, 101 S. Ct. at 2246 n. 8. Throughout its opinion, the Court stressed the uniqueness of the plaintiffs' claim and the strength of the evidence they offered, noting that they had direct evidence of discrimination.
In concluding its opinion, the Court addressed the fears of employers that allowing compensation suits outside of the Equal Pay Act analysis would place "`the pay structure of virtually every employer and the entire economy ... at risk and subject to scrutiny by the federal courts,'" because "`Title VII plaintiffs could draw any type of comparison imaginable concerning job duties and pay between any job predominantly performed by women and any job predominantly performed by men.'" 452 U.S. at 180, 101 S. Ct. at 2253 (quoting Brief for Petitioners 99-101). The Court responded that the employers' fears were unjustified, because claims that called for such comparisons are "manifestly different" from the plaintiffs' claims in Gunther. Id. The plaintiffs' suit did "not require a court to make its own subjective assessment of the value of male and female guard jobs, or to attempt by statistical technique or other method to quantify the effect of sex discrimination on the wage rates." 452 U.S. at 181, 101 S. Ct. at 2253-54.
Thus, the Court indicated that its analysis would not apply to cases unless the evidence of intentional discrimination was direct, or the violation "blatant." The last quotation above strongly suggests that cases requiring subjective assessment of the value of jobs, or relying on indirect evidence of discrimination, such as statistics, would not fall within its analysis. Other courts have limited Gunther to cases where the plaintiff shows a transparently sex-based system for wage determination, or direct evidence of discrimination in pay, and have refused to apply the analysis if it would require a subjective assessment of the differing duties and responsibilities of positions. See, e.g., Plemer v. Parsons-Gilbane, 713 F.2d 1127 (5th Cir.1983). This court concludes that the Gunther Court did not intend to allow comparison of unequal jobs when no direct evidence of blatant intentional discrimination is presented, and when the court is required to make subjective valuations of differing jobs.
In this case, as will be dlscussed below, EEOC has presented no direct proof of any sort of intentional sex discrimination in checklist pay, nor has it identified any discriminatory practice of Sears with respect to checklist pay. Moreover, EEOC's claim would require the court to make a *1334 subjective assessment of the value of male and female jobs, and to use statistics to quantify the effect of sex discrimination on wage rates, which the Gunther Court expressly stated that its holding did not require. The court therefore finds that the Gunther analysis is not applicable to EEOC's claim.
Moreover, even if Gunther is construed not to require direct evidence of sex discrimination, it is clear from the Court's analysis that it did not intend to lessen the standard of proof required. In this case, EEOC has alleged and attempted to prove that males and females in each job code analyzed were performing substantially equal jobs but were paid unequal wages. Unlike in Gunther, it has not attempted to show or offered any proof that there were some predominantly male jobs and other predominantly femalejobs and that the males were compensated on a higher basis than the females. To the contrary, EEOC has attempted to prove that males and females within the same jobs are deliberately paid differently. Its claim therefore falls within the "classic" framework of unequal pay for equal work. As will be discussed below, EEOC has had difficulty establishing that the actual content of the jobs at issue are substantially equal. It has therefore attempted to rely on the Gunther decision to relieve it of its burden of proving the substantial equality of jobs under the traditional analysis, claiming that Gunther permits comparison of merely similar jobs.
However, there is no indication in Gunther or any other case that the Court intended to lessen the standard of proof for a traditional pay discrimination claim. Instead, Gunther only intended to allow redress for claims of discrimination which fall outside the traditional equal pay analysis. The court therefore will not apply the Gunther analysis to allow EEOC to circumvent the elements of proof for its claim that Sears had a pattern and practice of intentionally discriminating in pay against women in certain job codes.
Sears management has always been highly decentralized. Decision-making authority is delegated to the lowest possible levels. After World War II, Sears began a period of rapid expansion, opening new facilities in many communities across the country. During this extended post-war growth period, Sears' decentralization allowed a high degree of flexibility and greatly contributed to its success.
Individual units were given great autonomy in almost every aspect of Sears' operations, including compensation. Sears officials believed that management tools such as job descriptions and organizational charts inhibited individual initiative and growth. Managers emphasized that the person makes the job. Consistent with this philosophy, Sears set compensation based on the individual in the job, his prior work history and salary, and his potential with the company, not solely on the basis of his present job title. Starting salaries were highly dependent on pay levels prior to entering checklist. Employees promoted into checklist would typically receive a raise of 10 percent to 15 percent above their previous pay rate.
*1335 Salary increases were dependent not only upon job performance and potential with the company, but also upon the number of times an employee relocated. Sears managers have always known that they must be willing to relocate to get ahead. As an incentive to relocate, Sears gave employees substantial salary increases for relocating, which were not directly related to the new job the employee would be performing. Thus, each checklist employee's salary reflects, in addition to his present job duties, his pay level before entering checklist, his career path through different checklist positions, the number of times he has relocated and the amount of increases associated with each relocation, his job performance, and Sears' view of his potential with the company. In addition, Sears also has a very strong policy of never decreasing an employee's pay, even if a job is restructured to a lower level or the person has been essentially demoted. All of these policies are consistent with Sears' philosophy of paying the person, not the position. These policies have created substantial differences in the salaries of checklist employees performing similar job functions.
1. Pre-1976 Organization and Compensation
Prior to 1976, as a result of Sears' decentralization, compensation practices and organizational structures varied significantly among different geographic regions and facilities, and by the function of a facility. Job functions and duties were not specifically defined. Persons with the same job title in different units often performed different duties at the discretion of the unit manager. Each manager attempted to use his employees to maximum advantage, varying job duties to meet the needs of the unit or to best utilize the particular talents of an individual employee. Sears had developed a series of job titles and associated job codes, known as "C" codes. However, the job titles were assigned without any precise study of the content of particular jobs. Over the years, "C" codes lost most of their job content meaning. Thus, persons with the same job code could be performing widely varying activities. There was no way to ascertain from general employment information whether persons with the same job were performing the same work.
Compensation of checklist employees was also very loosely controlled. There was no system under which a person with a certain job would be paid a certain salary. Instead, salary decision-making was also highly decentralized. Although checklist salary guidelines existed prior to 1976, checklist salary was generally a product of negotiation between the unit manager and the employee. Store managers made recommendations regarding checklist salaries to the group or zone manager. The store and group manager then attempted to reach an agreement. Although the group managers had final decision-making authority, they usually followed the recommendations of the store manager. Territorial personnel were provided rough guidelines, but generally perfunctorily approved salary decisions.
Compensation for checklist employees in credit operations was set in a different manner. Most credit functions were performed in Credit Centrals, which were separate administrative units. Prior to 1976, the Assistant Credit Manager for each territory was the primary decision maker regarding checklist salaries for Credit Central employees. The Assistant Credit Manager generally consulted the Regional Credit Manager and the territorial personnel department in making salary decisions. Thus, decisions for credit personnel were generally made at the territorial level.
In addition to their salary, almost all checklist employees were eligible for annual bonuses prior to 1977. Thus, a checklist *1336 employee's total yearly compensation consisted of his annual salary plus his bonus. The size of bonuses varied considerably. Bonus guidelines existed, but exceptions were permitted in individual cases. The amount of the bonus was almost entirely within the discretion of the group or zone manager. These managers had broad discretion in determining the total amount of bonus money that would be distributed and the amount of individual bonus. The generosity of group and zone managers in granting bonuses varied widely. The amount of an individual's bonus generally depended on the quality of the employee's performance, the employee's total earnings, and whether the group or zone manager decided to distribute all available bonus funds.
Thus, both salary and bonus decisions were made primarily on a local level, with individual managers exercising broad discretion in this area. As a result, salaries of checklist employees could be widely different, even among employees performing the same job duties. These factors, and the factors which generally affect checklist pay at Sears discussed above, such as job history at Sears, caused substantial differences in compensation among checklist employees of both sexes.
2. New Executive Compensation Program
Many checklist employees at Sears perceived this compensation system as inequitable and unwieldy. By the early 1970's, as Sears' expansion slowed, top management recognized the need for a change in its compensation practices. It sought a new executive compensation system which would be competitive in the marketplace, have internal equity, and reward individual performance. In 1974, Sears hired Hay Associates to help it develop a new system. Two years later in 1976, after a massive company effort, Sears introduced its new Executive Compensation Program.
During these two years, Sears management worked closely with Hay Associates to evaluate its compensation practices. Sears used the Hay Guide Chart-Profile Method of Evaluation ("Hay method") to evaluate jobs. Under this method, each position is measured in terms of know-how, problem solving, and accountability. Evaluators were trained to make systematic judgments about each of these dimensions of a job. The sex of employees holding a particular job was never considered as a factor and was never discussed at any time throughout the entire evaluation and implementation process.
To evaluate all the positions at Sears, Hay consultants and Hay-trained Sears employees wrote position descriptions for each checklist job at Sears. Position descriptions described positions as if they were performed by an average performer, to eliminate the issue of performance during the evaluation process. Committees were set up to evaluate the content of positions for which position descriptions had been written. Committee members were drawn from various levels at Sears. Each committee evaluated positions below the level of its members. The committees included persons familiar with the jobs being evaluated.
The goal of the evaluation committee process was to systematically evaluate positions as they were to be performed under *1337 the new system, not necessarily as they had been performed at or before the time of evaluation. Sears made many changes in job structure and content with a view toward standardizing job content and compensation among units. The evaluation committees found duplication of effort and inefficiencies, which they attempted to eliminate in defining job content.
The committees assigned points to each position for the know-how, problem solving and accountability it entailed. Evaluation points determined the comparative relationship of positions within Sears to provide the basis for structuring salary ranges. A salary range, with a minimum, midpoint, and maximum, was set for each job title and job code in 1976. Sears adopted a salary range of 75 percent to 120 percent of the midpoint salary for each position. These salary ranges were modified annually to maintain Sears' competitive standing in the marketplace.
While the Hay evaluation system helped set the salary ranges, many other factors influenced the actual salary of individual checklist employees. As discussed above, these included performance evaluations, previous job history, number of relocations, seniority, and geographic location. Under the new salary range system, Sears still compensated employees on these bases. However, an effort was made to eventually bring all employees within the given salary range for the job performed. When the new system was introduced, employees whose salaries were below the minimum were immediately given raises to bring them within the salary range. However, because of Sears' policy of not reducing an employee's pay, the process of bringing employees with salaries above the maximum into the salary range took several years to complete. Although salaries were not cut, in adherence to Sears' longstanding policy, salary increases for persons above the maximum were sharply reduced, until the yearly increases in the salary range caught up with the employee's salary to bring him within the maximum salary range. The bonus system was eliminated entirely after 1976.
Salary Increase Guides were introduced to provide time intervals for and percents of salary increases. However, managers still retained discretion as to the actual timing and size of the increase within the intervals provided in the Guide.
The standardization of job activities for a given position sometimes took several years to accomplish. No formal directives changing actual job structures were issued as a result of the Hay evaluation process. Instead, Sears gradually communicated changes to employees over a period of years. Variations in job duties from store to store still exist. In addition, after the initial Hay evaluation process was for the most part complete, Sears continued to make changes in its evaluations of jobs and to create new job categories as needed to better fit the jobs actually performed by employees.
C. EEOC Evidence
As with its commission sales claim, EEOC presented no direct evidence that an employee's sex was considered in making pay decisions. All the direct evidence is to the contrary. EEOC relied on simplistic statistical analyses to attempt to create circumstantial evidence to prove its checklist compensation claim. It performed regression analyses to predict the salaries of male and female employees in each of the 51 checklist jobs at issue, taking into account a number of factors which might influence salary level. EEOC relies on the disparities between its predicted salaries and the actual salaries paid by Sears as proof of Sears' nationwide pattern or practice of discrimination in pay against women.
1. Persons and Jobs Analyzed
EEOC attempted to analyze all Sears employees in the 51 checklist jobs at issue for each year from 1973 through 1980. These 51 job categories were identified by EEOC in terms of job titles and job codes created by Sears when its new executive compensation program was introduced in 1976. *1338 These job categories did not exist prior to 1976.
To prove that the jobs performed by the persons EEOC compared were substantially similar, EEOC relied on Sears' own evaluation of these jobs with Hay Associates in 1974 through 1976. EEOC asserts that the three factors evaluated under the Hay method, problem solving, know-how, and accountability, are equivalent to the Equal Pay Act standards of skill, effort and responsibility. EEOC contends that, since Sears pays employees in each of its job codes within the same salary range, it considers them to be performing substantially equivalent jobs. EEOC therefore relies on Sears' own analysis of its job categories and codes completed in 1976 as proof that the persons in each 1976 job code were performing substantially similar work.
However, prior to 1976, Sears' jobs were identified only by nebulous "C" codes, which were assigned without any analysis of job content. Acknowledging that it could not rely on these "C" code designations to prove substantially similar job duties for employees from 1973-1975, EEOC attempted to work backwards from the 1976 job codes to develop "correspondences" between "C" job codes and the 1976 job codes.
To develop these "correspondences," EEOC compared the 1976 job codes of persons who had not been promoted or transferred in 1976, with the "C" job codes of those persons at year-end in 1975. Since there were often several 1976 job codes associated with one "C" code, EEOC's analyst made judgments from the patterns of movement among jobs to decide which jobs corresponded. He admitted that he made these judgments without any knowledge of the job content of the positions he was analyzing.
Using the "correspondences" thus established, EEOC traced employees backward from their 1976 job codes to compile a group of employees to analyze for 1973-1975. To be counted in a particular position in 1975, an employee had to be in a particular 1976 job code in 1976 and in the corresponding "C" code in 1975. To be counted in 1974, an employee had to have been in the 1976 job code in 1976 and the corresponding "C" job code in both 1974 and 1975. Similarly, to be counted in 1973, an employee had to be in that 1976 job code in 1976, and in the corresponding "C" job code in 1973, 1974 and 1975. Thus, EEOC analyzed only those persons in each year who had not been transferred or promoted from their "corresponding" "C" code job between that year and 1976.
EEOC presented virtually no evidence of the actual job content of employees counted in its analysis for these years. It contends that, based on its determination of "correspondences" with a 1976 job evaluated by Sears, one can assume that persons it identified as having the same job code in 1976 were performing substantially similar jobs in the previous years.
2. Statistical Analyses
Having thus compiled a group of Sears' checklist employees to analyze for each year between 1973 and 1980, EEOC performed several regression analyses to determine whether Sears paid the males and females differently in each job it identified in each year. EEOC's statistical expert, Dr. Siskin, limited his analysis to jobs with *1339 at least 10 men and 10 women in two or more consecutive years.
EEOC first calculated the average salary difference for the males and females they analyzed in each of the 51 jobs for each year, without adjusting for any factors that might affect salary. For some job years, the salary differences were over $5,000, but in most, the differences were between $1,000 and $3,000. Males almost always had the higher average salary. See Table 8, EEOC Report on Checklist Compensation Practices of Sears, Roebuck and Co. (Revised September 18, 1984) [hereinafter referred to as "Checklist Compensation Report"].
EEOC next performed regression analyses to control for some factors which might affect salary. Dr. Siskin developed two models to control for certain factors he believed might affect compensation. In Model 1, he attempted to control for factors "one would expect to affect an employee's compensation." See Checklist Compensation Report at 20. These factors were: (1) length of time in present checklist assignment, overall checklist service or service in the company; (2) territory of the employee; (3) job performance; and (4) Hay points of the job. To measure these factors, he used the following variables: (1) sex; (2) time in present assignment; (3) time in present assignment squared; (4) additional time in checklist (beyond variable 2); (5) additional time in checklist squared; (6) additional time at company (beyond the sum of variables (2) and (4) ); (7) additional time at company squared; (8) territory of employee; (9) job performance, and (10) Hay points.
In Model 2, Dr. Siskin added factors which he considered as potential factors, but about which he was "uncertain whether or not they should affect compensation." These factors are: (1) education, (2) hire as college trainee, and (3) location in an urban or suburban facility. Model 2 evaluated these factors in addition to the factors incorporated into Model 1. To attempt to measure these additional factors, Dr. Siskin added to Model 1 the following variables: (1) employee hired as college trainee; (2) facility in urban area; (3) facility location unknown, (4) facility location in a suburban area; (5) grammar school completed; (6) high school completed; (7) two years of college completed; (8) four years of college completed; (9) advanced education completed; (10) other education; and (11) education unknown.
Regression analyses using both Model 1 and Model 2 were performed on two separate levels. First, both models were run for all employees in a particular job as identified by EEOC for each year. Second, Dr. Siskin grouped what he considered to be related positions into "job families," and ran both models for these job families.
After reviewing the results of the analyses under both models, he concluded that the variables added to the Model 2 "did not have a consistent effect." He therefore based his report and conclusions only on his analysis of Model 1. When analyzed at the job level, in nineteen of the fifty-one jobs at issue, there were no differences in compensation which were statistically significant at the three standard deviation levels. Salary differences for the remaining thirty-two jobs were significant above the three standard deviation levels in most years. The job family analyses produced similar results.
EEOC presented no nonstatistical evidence to support its claim. No individuals testified of situations in which women were paid less than men for substantially equal work.
*1340 D. Deficiencies in EEOC's Evidence
After considering all the evidence presented with respect to checklist compensation, the court finds that the EEOC has not proven a nationwide pattern or practice of intentional pay discrimination by Sears on the basis of sex. EEOC failed to prove that the employees analyzed performed substantially equal work. In addition, EEOC excluded a substantial number of checklist employees from the group of employees analyzed for 1973-1975, relied on inaccurate data, failed to consider important variables which affect salary, and ignored many legitimate employment and pay policies of Sears which affect checklist compensation.
1. 1973-1975 Analysis
EEOC's analysis for 1973 through 1975 is inadequate. EEOC offered no credible evidence that the persons it compared were actually performing the same or substantially the same job. It sought to rely on Sears' 1976 evaluation of its own jobs by tracing employees in the 1976 job codes back into 1975, 1974 and 1973. However, the 1976 job codes did not exist in these years. A relatively small number of "C" codes had been divided into a large number of 1976 codes. As discussed above, Sears evaluated positions as they were to be performed under the new system, not necessarily as they had been performed in the past. The job reevaluation process resulted in a number of changes in job content. In addition, especially prior to 1976, there could be substantial variation from store to store in the job duties performed by persons with the same job title. Sears' decentralization permitted individual store managers to completely change the job duties of employees in their stores based on the abilities of the individual employee and the manager's needs.
Moreover, the job "correspondences" upon which EEOC relied to select employees for its analysis were not based on any analysis of job content. EEOC's analyst merely examined patterns of movement from 1975 "C" job codes into the 1976 job codes, and used his individual judgment to determine which jobs corresponded, without any knowledge of the actual job content of any of the jobs he was analyzing. EEOC's analyst testified that, by identifying corresponding or "equivalent" jobs, he did not mean "to imply (that) equivalent means the same jobs or that the jobs are equal or that they're the same job content." Tr. 7,741 (Cupingood). Thus, equality of job content cannot be inferred from EEOC's proffered "correspondences" between pre-1976 "C" code jobs and 1976 job codes and titles. No evidence of actual job content was presented.
In light of all of these facts, even assuming that Sears' 1976 job evaluation process establishes that persons with the same job title and code were performing substantially equal work after 1976, there is no basis for relying on the 1976 evaluations as proof of job content before 1976. EEOC has provided no other credible evidence of the job content of the persons it compared in these years. It has therefore *1341 failed to meet its burden of proving that the persons compared were performing substantially equal work.
In addition, the analysis for 1973-1975 is inadequate because, by its method of selecting employees to be analyzed, EEOC excluded substantial numbers of employees from its analysis. It excluded: (1) employees who were in one of EEOC's "corresponding" jobs who were promoted or transferred into a different, noncorresponding job before 1976; (2) employees in a corresponding job if they were promoted to another corresponding job; and (3) employees who terminated their employment with Sears before 1976, even if they were active employees sometime from 1973 through 1975. Since so many employees were excluded from its analysis, even if EEOC had given sufficient evidence of job equality, its analysis for 1973-1975 would be entitled to little weight.
For all of these reasons, the court concludes that EEOC's analysis of employees for 1973-1975 is insufficient to create a reasonable inference of a pattern or practice of sex discrimination by Sears. It has therefore failed to meet its burden of proof for these years. In addition to the reasons discussed above, however, EEOC's proof for the years 1973 through 1975 shares the deficiencies in statistical analyses discussed below with respect to all years.
2. 1976-1980: Equality of Jobs
As discussed above, for the years 1976 through 1980, EEOC relies on Sears' own evaluation of its checklist jobs to prove that checklist employees with the same job codes were performing substantially equal work. To meet its burden of proof on this claim, EEOC must prove the equality of jobs in terms of the Equal Pay Act standards of equal skill, effort, and responsibility. "[T]he EEOC must establish, based upon `actual job performance and contentnot job titles, classifications, or descriptions,' that the work performed ... is substantially equal." EEOC v. Mercy Hospital and Medical Center, 709 F.2d 1195, 1197 (7th Cir.1983) (quoting Gunther v. County of Washington, 623 F.2d 1303, 1309 (9th Cir.1979), affirmed, 452 U.S. 161, 101 S. Ct. 2242, 68 L. Ed. 2d 751 (1981)). Thus, EEOC must introduce proof of the jobs as they are actually performed. It may not rely on the fact that employees bear the same job title, job code or written position description to prove equality.
EEOC asserts that the Hay evaluation process establishes equality of job performance for each of the jobs at issue in this case. As discussed above, using the Hay Guide-Chart Profile Method, Sears evaluated its checklist jobs in terms of "know-how," "problem-solving," and "accountability." In a pamphlet entitled "An Introduction to Sears Executive Compensation Program," Sears Exhibit 221, Sears gave simplified definitions of these terms as follows. Know-how is the knowledge, experience and skill needed to get the job done. Problem solving is the use of the know-how in getting the job done. Accountability is the degree to which a position has impact on results. Although these factors appear to be subsumed to a large extent within the Equal Pay Act categories, they are not the equivalent of equal skill, effort and responsibility.
More importantly, Sears' evaluations measured only idealized position descriptions.*1342 As the courts in Gunther and Mercy Hospital held, reliance on job titles and position descriptions is insufficient. Although the Hay evaluation of the position description developed by the committees would be helpful in determining whether jobs are substantially equal, they are not a substitute for proof of actual performance. EEOC has produced no evidence of how jobs were actually performed at Sears. Without proof of actual job performance and content, the Hay evaluations of position descriptions are inadequate to prove equal skill, effort and responsibility.
Moreover, Sears made changes in positions as it developed descriptions and evaluated jobs. It also attempted to standardize job duties as much as possible. However, changes in job duties resulting from this process were not immediately communicated to employees. Instead, Sears gradually introduced these changes, sometimes over a period of years, through personnel directives and other means. Thus, the actual job content of many positions could still vary for some time after the new executive compensation program was implemented. Differences in actual job performance and content from store to store still exist today.
EEOC bears the burden of proving that the jobs it evaluated were substantially equal. Not only has the EEOC failed to introduce any evidence of the actual performance of jobs, but also Sears has shown through credible testimony that there were differences in job content of employees with the same job code. EEOC was obligated to demonstrate that these differences were not significant, and that the persons compared performed "substantially" equal duties. It failed to introduce any such evidence. The court therefore concludes that EEOC has not met its burden of proving the substantial equality of the jobs analyzed.
Moreover, even if the Hay evaluations were deemed adequate to prove the substantial equality of the jobs in general, this proof would be insufficient in a number of instances. Since EEOC relies wholly on the accuracy of Sears' evaluation of its jobs to prove job content, it is also limited by the inaccuracies of that process. In a number of positions, Sears found that it had improperly evaluated jobs. A sizeable number of credit collection employees were misclassified under the new program in 1976. Persons performing different duties were sometimes classified in the same job code, and persons performing the same duties were sometimes classified into different job codes. Since EEOC relies solely on Sears' analysis of these jobs, it has completely failed to prove the substantial equality of jobs performed by employees in these job codes.
3. Overall Statistical Analyses
In addition to its failure to properly analyze employees in 1973-1975, and its failure to prove equality of jobs, EEOC's statistical analyses are so flawed that they lack any persuasive value. Its data base has serious errors, it omitted important variables from its analyses, and it used a model *1343 that fails to consider Sears' method of actually compensating individual employees.
a. Data Base
The data base analyzed by EEOC does not accurately reflect the compensation or other factors EEOC attempted to analyze. EEOC relied exclusively on the computer extract tapes provided by Sears, which contained inaccurate data and omitted important data. First, the data on bonuses is inaccurate. The only year for which bonus data is accurately recorded is 1974. In 1973 and 1975, bonuses recorded on the tapes included lump sum payments for delayed bonus reapportionment, in addition to the earned bonus. Also, actual bonus amounts were not always recorded. Without accurate bonus information, EEOC could not properly analyze the total compensations of checklist employees in 1973 through 1976.
Overwrites earned as a timecard Division Manager or a checklist Catalog Sales Office Manager, and changes in monthly salary resulting from reapportionment of the overwrite, often were not identified on the tapes. In addition, transfer and relocation histories, and the dates of assignment to present position and entrance of checklist data were often inaccurate.
EEOC possessed paper records which it could have compared with the computer information, but it chose not to check the data for accuracy. Sears checked the computer data against the paper records. Dr. Haworth used the paper records to correct the date of present assignment and the date of checklist entry. Correcting the data for these two fields only, using EEOC's analysis, Sears reduced by over thirteen percent the number of job/years in which there were statistically significant disparities in salary. See Sears Exhibit 6-TTT. Thus, the overestimate of EEOC's disparities between males and females caused by data inadequacies was quite substantial.
In addition, job performance data was missing for many employees. Virtually no job performance data was available before 1977, and data for 1977-1980 is very limited. Whenever there was no data in the job performance field on the extract tapes, for the years 1977-1978, EEOC arbitrarily assigned a job performance score of 42, the midpoint of the "adequate" rating category. For 1979-1980, EEOC assigned a score of 57, the midpoint of the "very good" rating category. However, these scores were never used by Sears managers for any purpose. Imposing such arbitrary performance ratings for employees with missing job performance data severely limits the usefulness of this variable in the analysis.
These inadequacies in the data analyzed by EEOC make the results of its analyses highly suspect. As Dr. Siskin admitted, errors in a number of these variables could artificially create a pattern of negative sex coefficients adverse to women. The weight given to EEOC's statistical evidence must therefore be reduced accordingly.
b. Omitted Variables
More important than the inadequacies discussed above, EEOC's statistical analyses lack persuasive value because EEOC omitted important variables which influence checklist compensation at Sears. It is important to bear in mind that, with the type of statistical model used by both EEOC and Sears, all important factors *1344 which affect compensation must be included in the model to obtain reliable results. The analyses produce a sex coefficient, which is intended to measure the effect of sex on compensation. However, the sex coefficient reflects not only the effect of sex, but also the residual effect of any factor which affects salary that is not included in the model. Thus, if important variables are omitted, the effect of sex on compensation estimated by the model will be artificially inflated. The court agrees with both Dr. Wise and Dr. Haworth that, in this situation, it is better to err on the side of including variables when their effect on salary is uncertain, than to exclude them and obtain a biased estimate of the sex effect.
Among the measurable variables EEOC failed to analyze are: prior non-Sears experience, the level of prior timecard responsibility at Sears, the number of relocations, accurate performance measures, organizational levels, unit size, merchandise groups, and leaves of absence. As discussed above, two of the most important variables affecting checklist starting salary are prior non-Sears experience and prior timecard experience with Sears. Data on prior non-Sears experience was generally not available. However, from the limited information available, men had more pre-Sears experience than women. Omission of this variable would therefore bias the results and produce artificially high unexplained salary disparities. EEOC also failed to factor in prior timecard experience. As discussed above, Sears set compensation to a large extent on the basis of the employee's last salary at Sears. Often, employees entering checklist received a ten to fifteen percent salary increase above their last timecard pay level. Men often entered checklist from a higher timecard level than women. This variable therefore should also have been factored into EEOC's analysis.
Another factor which significantly affected checklist salary at Sears was the number of relocations. As discussed above, substantial salary increases were given to employees who relocated, and such increases were unrelated to the job responsibilities the relocated employees were to assume. Employees willing to relocate not only received greater salary increases, but were promoted from one position to another more quickly. Therefore, an employee's salary could be significantly higher than that of another employee with fewer relocations who was performing the same job. Sears' evidence showed that males were more willing to relocate than females. However, EEOC completely failed to control for the number of relocations in either of its models.
EEOC included a variable to control for the territory in which an employee worked. However, because of Sears' decentralization, salary decisions were usually made at the unit or group level, with only nominal approval by territorial officials. By failing to control for the individual unit, group or zone of an employee, EEOC omitted another important factor affecting salary.
In addition, compensation for various buyer positions was related to the merchandise group in which an employee worked. Employees working with more profitable products were often compensated at a higher level than others. A buyer's merchandise group could therefore be a significant factor affecting his salary. EEOC also failed to control for this factor.
Other measurable factors that could affect salary which EEOC did not include in its model are: veteran status, marital status and size of family, leaves of absence and college major. Sears expanded *1345 EEOC's model to include all of these factors, except pre-Sears experience for which there was no data. The number of statistically significant disparities was reduced by 50 percent in 1976 and 1980, and 75 percent in 1979. Over the entire 1976 to 1980 period, the number of statistically significant disparities dropped by an average of more than 62 percent. See Sears Exhibit 6-TTT.
Thus, EEOC's statistically significant disparities were substantially reduced by attempting to adjust for only those measurable variables discussed above. However, even the variables included in the analysis could not be accurately measured. For example, only the number of relocations could be controlled for, not the size of the salary increase associated with the relocation. Also, as previously indicated, no data was available for pre-Sears experience. Further, as noted above, EEOC's job performance variable was wholly inadequate to measure job performance. EEOC arbitrarily substituted scores never relied on by Sears for employees whose scores were missing. Since no accurate job performance scores were available for many employees, Sears could not include an accurate job performance measure for these employees. The disparities remaining after Sears expanded EEOC's model could be explained by the inability to accurately adjust for these measurable factors alone.
In addition to the measurable variables, however, a number of unmeasurable variables which could significantly affect checklist compensation at Sears were not analyzed. These include commitment, loyalty, dedication, and motivation. As Dr. Siskin admitted, these factors may have a greater impact on compensation than other variables that are more easily measured. Although it would be virtually impossible to measure these variables for inclusion in a regression model, they do legitimately affect compensation. Their omission from the model must therefore be taken into account when evaluating the results of the analyses.
c. Other Flaws in EEOC Models
In addition to problems with the data and omitted variables, the model itself used by EEOC is flawed. The most important flaw in EEOC's entire analysis is that it aggregated its analysis for each job on a nationwide basis. Salary decisions were almost always made by individual store managers in conjunction with group or zone managers, and salary levels varied considerably from group to group. To make a valid comparison of male and female employees at Sears, EEOC should have analyzed employees at the same unit level where possible, or otherwise at the same group or zone level. Although the number of checklist employees performing equal work may have been small, EEOC is not limited to proof by statistical analyses. It could have proven that, within units, groups or zones, in which salaries were determined on the same basis, men and women performing substantially equal work were given unequal pay. It chose not to attempt to prove any discrimination at the unit, group or zone level. By failing to do so, it failed to compare men and women who were similarly situated. Because it did not compare employees who were similarly situated, its measures of statistical significance based on nationwide aggregations have little meaning.
EEOC could have analyzed each territory separately to control for differences between territories and variances within territories. As Dr. Siskin acknowledged, the territorial variable included in his models did not adjust for variances within each territory. When he ran the analysis separately by individual territory for one position, the disparities and associated t-values were substantially reduced. See Plaintiff's Exhibit Siskin (Pay) 43. Performing the entire analysis by individual territory could *1346 have substantially reduced salary disparities and associated t-values for all jobs.
EEOC's model is also flawed because it analyzes an employee's salary in each year as though salary is set anew every year. In fact, an employee's salary for any year is highly dependent upon his salary in prior years. EEOC's analysis fails to take this into account. As will be discussed below, Sears designed a model to take this fact into account, and estimated salary disparities were substantially reduced.
In addition, there is no basis for EEOC's analysis of job families. Jobs were grouped together into so-called families without any knowledge of whether factors important to compensation for the jobs were similar. Sears never made any such grouping of its jobs for any business purpose. EEOC's analysis of these artificial job families therefore provides little useful information to the court.
As the foregoing discussions indicate, EEOC made no attempt to reflect the decision-making process at Sears. In fact, its expert stated that there was no reason to do so. As a result, its model has little persuasive value. EEOC's omission of important factors, its use of an inadequate data base with significant errors, and its use of a flawed statistical model, all render its statistical evidence insufficient to support a reasonable inference of intentional discrimination by Sears.
This conclusion is strongly reinforced by the utter lack of testimony by any of the supposed thousands of victims of this alleged nationwide pattern or practice of discrimination. It is hard to imagine that, if Sears had been systematically paying women less than men performing the same work on a nationwide basis for 8 years, not one person would be available to so testify. Without any witnesses to provide some basis for making the leaps of faith required to accept EEOC's statistical analyses, no reasonable inference can be made from its statistical evidence that Sears had a nationwide pattern or practice of discrimination against women in pay during the period 1976 through 1980.
In addition, as discussed above, EEOC's analysis of employees in 1973 through 1975 is unacceptable for other reasons as well, and EEOC has failed to sufficiently prove equality, or even similarity, of job content. For all of these reasons, the court concludes that EEOC has failed to carry its burden of proof with respect to its checklist compensation claim. The court's conclusions with respect to EEOC's evidence is also supported by Sears' evidence on this subject, discussed below.
E. Sears' Evidence
Sears presented both direct and circumstantial evidence to prove that it did not discriminate against women with regard to checklist compensation. Sears presented the testimony of numerous Sears employees concerning checklist compensation practices at Sears. It also performed both multiple regression analyses and a cohort analyses of its employees.
1. Regression Analysis
Sears performed its own regression analyses using computerized data supplemented with more accurate and complete data from employees' Personal Record Cards. To reflect that salary is a function of starting salary plus salary increases, Sears used a model which analyzed both the salary level, which is the entering salary component, and a rate of change component. Sears' statistical expert, Dr. Joan Haworth, used an exponential model, which takes into account the diminishing influence of certain variables and the rate of salary change.
To reflect compensation practices at Sears, its model attempted to measure the following variables: seniority, education, job performance, relocations, organizational level, unit size, merchandise lines, prior timecard salary, and prior timecard experience. The court finds that all of these factors can affect checklist compensation at Sears. Sears' model also controlled for marital status, number of children, veteran *1347 status, and leaves of absence. These factors will be discussed below.
Adjusting for all of these factors, Sears calculated the "adjusted percent difference" between male and female salaries in terms of both the level and rate components of salary. T-values for these terms were also calculated. The overall regression analysis showed a salary difference favoring men which ranged from 3 percent to 6 percent. Of the 195 job/year combinations analyzed for the years 1976 through 1980, only 3 had disparities statistically significant at the three standard deviation level.See Sears Exhibits 6-CC, 6-EE, 6-222. Further analysis of the two positions with statistically significant disparities, through which employees not similarly situated were eliminated, reduced the disparities below the statistically significant level.
Sears also performed its regression analysis for the years 1973 through 1975, using data available for persons in EEOC's "artificial workforce." Under Sears' regression analysis, three jobs had adjusted salary disparities with t-values in excess of three. Two of these jobs (Satellite Appliance Store Managers I and II (R362, R361)) could not be properly analyzed because accurate overwrite data was not available from the computer or paper records. For the third job (CSO Manager IWithout Service), Sears' evidence showed that lack of overwrite data in the computer records caused the salary disparity.
Sears refined its analysis for 1973 through 1975 by excluding statistical "outliers," because their characteristics skewed the results. An "outlier" was defined in this case as an employee whose predicted salary is more than two standard errors from his actual salary. Exclusion of such outliers is an accepted statistical practice. When outliers were excluded from the "artificial workforce" analyzed by EEOC, all statistically significant disparities except for one job/year were eliminated. See Sears Exhibit 6-XXX.
Although Sears performed its regression analyses for the years 1973 through 1975, Sears' expert testified, and the court agrees, that meaningful results cannot be obtained from these regression analyses because bonus data was inaccurately recorded on the computer tapes in those years. Thus, there is no accurate way to determine the amount of total compensation earned by employees in those years.
In addition, for all years analyzed, but particularly for 1973 through 1975, important data regarding timecard salary and experience and non-Sears experience was missing. As discussed above, an employee's entering checklist salary was highly dependent upon his prior timecard experience and salary. Basing entering salary on prior timecard experience and pay level was a legitimate way of compensating employees on the basis of the quality of their previous experience. Since prior timecard data was often missing, this important variable cannot be adequately factored into any statistical analyses. A large part of the *1348 estimated salary disparities could be attributable to this missing data.
Sears demonstrated that this is the case by analyzing the 1980 salaries of checklist employees who had entered checklist jobs between 1976 and 1980. The computer data for these employees in 13 jobs was more complete. Sears analyses showed that differences between men's and women's salaries in these jobs were significantly greater when the quality and level of prior timecard experience was not included in the analysis. See Sears Exhibit 6-VVV. Thus, it is very likely that a significant part of the salary disparities estimated by both Sears' and EEOC's analyses are the result of missing data, not discrimination by Sears.
The inability of the model to fully control for other factors included in the model also results in an overestimation of salary disparities. For example, the checklist seniority variable is inadequate because it does not account for the quality of work performed; it accounts only for the amount of time spent in previous checklist positions. Sears illustrated the importance of this by comparing a male and a female employee in the same position, Assistant Buyer IV (Buy Level 5+) (MO28), in 1979. The male employee had six prior checklist positions as well as significant timecard positions. The female had only one prior checklist position and clerical timecard positions. The actual monthly salary difference between the two employees was slightly more than $700. Sears' model predicted a gap of only $300. This disparity most likely occurred, at least in part, because the model did not control for the quality of prior experience.
The same is true of other variables. For example, with respect to relocations, although the number of relocations can usually be determined, there is no way to ascertain the size of the salary increase associated with a relocation to accurately adjust for this factor. In addition, although Sears performed regressions on a territorial basis, these analyses could not fully reflect Sears' decision-making, which was performed at lower organizational levels. The number of comparable employees at the unit, group or zone level was too small to perform a meaningful regression analysis at these levels. This is yet another example of the inability of regression models to properly analyze the facts of this case.
In addition to these problems relating to measurable variables, the statistical models are unable to adjust for important unmeasurable variables, such as motivation, commitment and loyalty. Yet, the testimony clearly establishes that these may be the most important variables that affect salary. However, despite all of these limitations of regression analyses, Sears' model showed relatively small disparities between the salaries of males and females. See Sears Exhibit 6-YYY.
The limitations of regression analyses were also illustrated by a Sears case study. Sears analyzed two male employees in one of the job/years in which Sears found a statistically significant disparity after performing its regression analysis. The study was based on information in their personal record cards. The case study showed that these employees' salaries were substantially different, by as much as $600 per month, and that this gap could not be explained by Sears' multiple regression model. Regression models can only predict earnings based upon what the measurable variables indicate about employees' earnings differences. If these employees had been of different sexes, much of this salary disparity would have been attributed to sex discrimination. This case study illustrates another important limitation of regression analyses. Although these analyses can provide useful information in some cases, when applied to analyze highly complex matters, such as the factors which affect management salary decisions, their results can be of limited value.
EEOC has criticized Sears' regression analysis on two major grounds. First, EEOC questions the inclusion of a number of the variables in Sears' model. These variables are marital status, number of children, leave, veteran status, business *1349 major, merchandise categories, timecard salary and experience, and unit size.
With respect to marital status and number of children, the theory for including these variables is that they tend to have different effects on males than females. Earnings of males tends to increase if they are married and as the number of children increases, purportedly due to the increased need to support their families. The effect on women's earnings tends to be the opposite, particularly as the number of children increases, since demands in the home are usually greater for women. Sears included these variables in the model because, if they do affect salary and are omitted from the analysis, their impact on salary will be reflected in the residual sex coefficient and therefore attributed to sex discrimination. However, the explanatory power of these variables is uncertain. When a variable has an obvious correlation with sex, as these variables do, but no clear relationship with salary, their inclusion in a model can drive down the model's estimated effect of sex on salary and the associated t-values. Although these variables seem to have some relationship with salary and therefore there is some basis for including them in the analysis, the court has taken into account that these variables can also artificially reduce estimated salary disparities and the associated t-values.
The same is true of the business major variable. This variable seems to have little explanatory power in the model. Since more male than female college students major in business, inclusion of this factor may also bias the results of the analysis.
EEOC also challenges inclusion of the veteran status variable. Although there are many conflicting opinions of the value of military service in civilian employment, at least in some cases, military service is related to higher starting salaries, and can be related to higher job performance. If nothing else, the variable measures a type of pre-Sears experience which, like any pre-Sears experience, can affect checklist starting salary. The court therefore finds that inclusion of this variable in the model is justified. Similarly, the leave variable affects at least the quantity of experience, and may in some cases affect the quality of prior experience. It may therefore legitimately be included in the model.
The court also finds that EEOC's criticisms of the timecard salary and experience, merchandise category and unit size variables are without merit. Sears' evidence shows that these factors can have an impact, sometimes significant, on checklist salary. They are therefore legitimately included in the model.
As discussed above, with this type of statistical analysis, it is probably better to err on the side of including variables when their effect on salary is unclear. However, the court must also consider the possible bias that inclusion of these variables can cause, and has discounted Sears' results to take this into account.
EEOC's second major criticism of Sears' regression analysis relates to Sears' measures of statistical significance. EEOC contends that the t-values Sears calculated for the salary rate and level variables of its analysis are not valid measures of statistical significance. EEOC's expert, Dr. Siskin, testified that the t statistic can only measure the statistical significance of a single dependent variable, such as total salary. Dr. Haworth's model measured the impact of various factors on two dependent variables, salary level and the rate of change. According to Dr. Siskin, when two dependent variables are analyzed, calculations of statistical significance must be based on both of the dependent variables together. He asserts that a "joint F" statistic, not the t statistic, will give a meaningful *1350 measure of the statistical significance of both the level and rate variables together.
Sears' expert, Dr. Joan Haworth, testified that a joint F statistic should not be used when data are incomplete as in this case. However, Dr. Haworth responded to Dr. Siskin's criticism by applying a joint F test to 13 jobs for the year 1980 for which there was relatively complete data. None of the 1980 salary differences for these jobs was significant at the 1 percent level (2.58 standard deviations), and only three were significant at the 5 percent level (1.96 standard deviations). Sears' Haworth Rebuttal (Pay) Exhibit 2. Thus, when complete data was available, salary disparities were not significant even applying the joint F test.
In addition, Dr. Wise testified that, although the F statistic may be able to measure the joint statistical significance of both variables, given the imprecision of all the statistical models in this case, the best way to determine the significance of the results of the analyses is to look directly at the adjusted salary disparities estimated by the model. According to Dr. Wise, these disparities should be judged in light of the factors left out of the model and the other failings of the model.
This approach seems the most reasonable. Although the t statistic cannot provide the most useful measure of the statistical significance of the results of Sears' regression analysis, the adjusted disparities produced by the model provide valuable information regarding the effect of various factors on checklist compensation at Sears. See Sears Exhibits 6-BBBB, 6-AAAA, and 6-EEEE. Considering the inability of the model to accurately reflect important, legitimate, non-sex based factors which affect checklist compensation at Sears, the possibility of bias from the inclusion of variables discussed above, and all the evidence in this case, the court concludes that the remaining disparities are attributable to deficiencies in the statistical analysis, not sex discrimination by Sears.
The foregoing discussion and Dr. Wise's testimony illustrate the severe limits of regression analyses in evaluating complex decision-making processes of this nature. Neither EEOC's nor Sears' regression models could fairly analyze male and female checklist compensation at Sears. Recognizing the limitations of its own regression analysis, Sears also performed a cohort analysis to more accurately compare male and female checklist compensation. The results of the cohort analysis support the court's conclusions with respect to Sears' regression analysis.
2. Cohort Analysis
In Sears' cohort analysis, it attempted to compare the careers of similarly situated male and female checklist employees. All employees entering checklist in a given year were analyzed together in terms of their starting salaries and their rates of salary increase over the succeeding years. An analysis of yearly cohorts provides a comparison of the compensation patterns of those who entered checklist positions under similar market conditions and pursued differing career paths at Sears.
Women from the 1973 cohort of all checklist employees received higher percentage increases in salary than men for every year from 1974 through 1980. See Sears Exhibit 6-17. Similar results were obtained for the 1974 through 1979 cohorts. When checklist employees in the 51 jobs only at issue were analyzed from 1973 to 1980, women received higher percentage increases in each cohort in almost every year.
In addition, Sears measured the rates of increase for all checklist employees in general, not just those in the cohorts discussed above. Year-to-year percentage increases *1351 of women equaled or exceeded males in every year. See Sears Exhibits 6-19, 6-20, and 6-21. Sears' studies also showed that, for the jobs at issue, among employees in the same job for two consecutive years, women's actual dollar increases were equal to or greater than men's increases from 1976 through 1980. A similar pattern is shown for employees in EEOC's "artificial workforce" for 1973 through 1975. See Sears Exhibit 6-22.
Sears adjusted its cohort analysis to determine whether the results favoring women were related to differences in individual characteristics. Even after adjusting for job performance, position in the salary range, and time in current job, the increases favored women. In fact, in the 22 comparisons made, the only statistically significant differences are in the 13 that favored women. See Sears Exhibit 6-HHHH. The same pattern is found after analysis of all cohorts of entering employees who remained in the same job for two consecutive years.
Sears also tested its results to determine whether women were given greater salary increases because they had lower starting salaries. It adjusted for the quartile of the salary range an employee was in. Its analysis showed that women had higher percentage increases even adjusting for the employees' position in the salary range. Thus, the greater increases for women did not depend on whether the woman was at the lower end of her salary range. See Sears Exhibit 6-HHHH.
Thus, Sears' analyses show a clear pattern of consistently higher salary increases for women than men. In addition, Sears' evidence showed that in all years from 1973 through 1980, except 1976 (the year in which the new compensation program was introduced and promotions cannot be accurately identified), Sears promoted females into all checklist jobs at a higher rate than it promoted males. See Sears Exhibits 6-28, 6-HH. Checklist women were thus given larger salary increases and were promoted more often than checklist men.
Entering salaries of checklist cohorts for all years were also analyzed. When Sears compared the small number of men and women entering checklist with similar timecard job level and seniority, there is no pattern of significant disparity. See Sears Exhibit 6-AA. However, there are relatively few employees with comparable experience.
Sears also analyzed all checklist employees entering jobs at issue in each year combined. It adjusted for the initial position, territory, timecard experience, and individual attributes controlled for in Sears' regression analysis discussed above. Salary disparities ranged from 3.57 percent in 1976 to 1.08 percent, all in favor of males. See Sears Exhibit 6-YYY. The limitations of Sears' regression model discussed above are applicable here, and could easily explain these small disparities.
Sears also analyzed the salary increases given to men and women upon promotion to checklist in 1976 through 1980. This analysis showed that there was no pattern of disparities adverse to women. See Sears Exhibits 6-23, 6-24, 6-25. Although there is some disparity between male and female increases in 1974 and 1975 when all checklist jobs are aggregated, the differences vary significantly and strongly favor men in some jobs and women in others. See Sears Exhibits 6-26 and 6-27. No clear pattern of disparity against women can be discerned in these years either.
Thus, Sears' evidence shows no pattern or practice of discrimination in the increases given to checklist employees when promoted from timecard positions. This is further evidence that there were no significant disparities between men and women in starting checklist salary. As discussed above, Sears' analyses of salary increases and rates of promotion show that women were treated more favorably than men. Together, these analyses are strong evidence that Sears did not have a pattern or practice of discrimination against checklist women with respect to compensation. This evidence demonstrates that, in many cases, Sears actually treated checklist women more favorably than checklist men. This is consistent with its affirmative action program.
*1352 EEOC attempts to dismiss this evidence with the assertion that these cohort analyses do not address the issue in this case whether Sears paid females less than males in the same job. See Plaintiff's Final Argument at 91. However, these cohort analyses provide the most complete evidence in this case regarding Sears' treatment of similarly situated checklist men and women. Although these analyses do not include checklist employees hired before 1973, they nevertheless provide very useful information regarding Sears' treatment of checklist employees in relatively similar circumstances during the years at issue in this case. These analyses support the court's conclusion that the pay disparities that remain after Sears' and EEOC's regression analyses are not the result of sex discrimination by Sears, but rather are the result of the inability of the regression analyses to accurately reflect Sears' complex, nondiscriminatory decision-making processes.
3. Sears' Witnesses
More important than any statistical evidence in this case is the testimony of Sears' witnesses. Sears employees from many areas of the company testified regarding Sears' checklist compensation practices and treatment of women. Each of these witnesses had substantial personal knowledge of the checklist system at Sears. The court found them to be highly credible witnesses and was persuaded by their testimony. They testified about the organizational structure of Sears, its decentralization, and the resulting differences in compensation and job duties across individual units. They also testified that it was Sears' long-standing practice to compensate checklist women on an equal basis with men. The witnesses testified that they had carried out this policy, they had seen others carry it out, and they knew of no instances where women were treated less favorably than men. They testified very convincingly that sex was not a factor in establishing the Executive Compensation Program with Hay Associates in 1976, or in determining at any time the pay of checklist employees. Sex was not discussed and did not enter into pay considerations in any way.
The witnesses' uncontradicted testimony helps "bring to life" Sears' statistical evidence, and lends strong support to the conclusion drawn from Sears' statistical evidence, that Sears compensated checklist men and women in a non-discriminatory way. The fact that Sears' affirmative action programs were in effect throughout this period, making it unlikely that Sears would be intentionally discriminating against women in pay at the same time, corroborates their testimony.
In stark contrast to Sears' presentation, EEOC presented no credible witnesses with personal knowledge of Sears who could contradict the testimony of Sears' witnesses, or give any life to EEOC's inadequate statistical data and analyses. EEOC did not prove even one individual instance of pay discrimination by Sears, and it presented no credible evidence of a nationwide pattern or practice of pay discrimination. As EEOC's statistical expert admitted, statistical analysis alone cannot prove causation, or, in this case, that the salary disparities it estimated were caused by Sears' intentional discrimination. Without more, EEOC's statistical evidence is not sufficient to permit a reasonable inference of pay discrimination by Sears. Sears, on the other hand, has provided far more convincing statistical evidence and highly credible witnesses with personal knowledge of Sears to support its denial of pay discrimination.
F. Conclusion: Checklist Compensation
Based on all the evidence presented with respect to checklist compensation, *1353 the court concludes that EEOC has failed to prove that Sears had a pattern or practice of discrimination against women in compensation in the 51 checklist jobs at issue. On the contrary, the court finds that Sears has proven that it did not have such a pattern or practice of pay discrimination against women. Further, Sears has proven legitimate non-discriminatory reasons for the alleged statistical disparities between the compensation of checklist men and women, and EEOC has not proven them pretextual.
Accordingly, based on the above findings of fact and conclusions of law, it is hereby adjudged and ordered that judgment is entered against plaintiff and in favor of defendant on all claims at issue in the trial of this case, and plaintiff's claim for relief is hereby denied.NOTES
 In its First Request for Information, the EEOC asked Sears for responses to the following questions:
Rule 7. State whether at any time since February 1, 1964, the Company has restricted the working hours of females in any job on the basis of any state or local law or regulation.
Rule 8. If yes, state by designated facility for each job so restricted:
a. The name and citation of the state or local law or regulation upon which the restriction was based;
b. The dates during which the restriction was in effect.
The EEOC summarized Sears' responses by territory in an Attachment to its Memorandum in Support of its Motion for Partial Summary Judgment.
 In her affidavit, Jane L. Dolkart states that she is the custodian of the investigative file of Brown v. Sears, Roebuck and Co., EEOC Charge No. XXX-XX-XXXX, and that the Exhibits attached to her affidavit are from that investigative file. The following Exhibits are attached to Jane L. Dolkart's affidavit: (1) the actual responses to the EEOC's First Request for Information (see supra n. 1); (2) a copy of Sears' "Store Doctor's Guide" dated July 1, 1969; and (3) various excerpts from Sears' Personnel Manual relating to the alleged discriminatory policies. As established in the ensuing discussion, only three claims remain for the court to rule on in this motion for partial summary judgment. Those claims, and the sections of the Sears' Personnel Manual which Jane L. Dolkart attached to her affidavit and which relate to those claims, are as follows:
1. Sears violated Title VII by maintaining a policy of failing to offer the same protection from lay-off to employees on pregnancy leave as that offered to employees absent with other temporary disabilities. Sears' Personnel Manual, Section 2303(e) (dated January 1, 1975) and Sections 7151-7159 (dated December 1, 1970).
2. Sears violated Title VII by maintaining a policy of involuntary transfer of pregnant employees "when appearance is a factor." Sears' Personnel Manual, Section 2254 (dated September, 1974).
3. Sears violated Title VII by maintaining a policy of providing a day's paid absence to male employees when their wives gave birth, but not to female employees who gave birth. Sears' Personnel Manual, Section 7257(a) (2) (dated June, 1966).
 Sears contends in "Attachment B" to its Response that its responses to the EEOC's First Request for Information cannot be construed as anything more than citations to the applicable state or local laws or regulations and the dates the laws or regulations went into effect. The affidavits of Ronald W. Borkowski, William G. Carney, John J. Conners, James W. Hoffman, and Warren W. Sheridan all support Sears' contention that it did not limit the number of hours female employees could work pursuant to state or local laws or regulations.
 Sears attached to its Response the affidavits of Jean Shoffner and Kenneth M. Cook. In her affidavit, Jean Shoffner, a nurse at Sears since 1969, states that Sears did not exclude women from hiring on the basis of their pregnancy in the late 1960's, and that she personally knows of one incident in which the merchandise manager at the Greensboro, North Carolina facility hired a woman who was approximately four months pregnant to perform clerical duties. Kenneth M. Cook, Manager of Personnel Research for Sears, states in his affidavit that in 1973, within seven months of their hire date, 27 women went on pregnancy leave of absence and 496 women voluntarily terminated their employment due to pregnancy with no intention to return.
 In her affidavit, which is attached to Sears' Response, Vivian J. Butler, Manager of Timecard Personnel for Sears, contends that employees receiving disability benefits were not guaranteed greater protection from reductions in force than employees taking either illness or pregnancy leave.
 Vivian J. Butler also states in her affidavit that she knows of no instance in which a pregnant employee was transferred involuntarily from the position she occupied prior to her pregnancy.
 The EEOC conceded in its Reply that the Sears' affidavits, which stated that none of the facilities in question in fact restricted the hours of women in a manner different from that applied to men, did raise a genuine issue of material fact, and that summary judgment was therefore inappropriate on that issue. As noted in the ensuing discussion, the EEOC withdrew this claim after this motion was fully briefed. The court therefore denies the EEOC's motion for partial summary judgment as to this claim.
 The EEOC's Reply Brief, filed on March 12, 1982, was the last pleading filed with regard to this motion for partial summary judgment.
 Transcript of Proceedings, October 7, 1982, pp. 159-161.
 In the October 15, 1982 Statement of Issues, the EEOC identified the date of the beginning of the relevant period of time as August 30, 1971, which is exactly two years prior to the filing of the EEOC's charge on August 30, 1973. The EEOC argues in its Memorandum in Support of its Motion for Partial Summary Judgment that the period of Sears' potential liability should begin May 31, 1969, four years prior to the filing of a charge by the Women's Equality Action League. Memorandum in Support of Plaintiff's Motion for Partial Summary Judgment, p. 4, n. 2. In the Order dated January 27, 1984, this court ruled that the EEOC would not be permitted to expand its case beyond the claims listed in the April 7, 1982 Statement of Issues, but the court also stated that it would allow the EEOC to identify the time parameters relevant to its claims in the October 15, 1982 Statement. The court finds that the EEOC clearly set out the date of the beginning of the period of Sears' potential liability in the October 15, 1982 Statement of Issues, and the EEOC will not now be permitted to expand the time parameter beyond that set out in the October 15, 1982 Statement. The August 30, 1971 date therefore stands as the date of the beginning of the relevant period of time.
The October 15, 1982 Statement of Issues also sets out the dates upon which Sears deleted from its Personnel Manual the policy statements that the EEOC now challenges. According to the October 15, 1982 Statement, the policy of providing a day's paid absence to male employees when their wives gave birth, but not to female employees who gave birth, was deleted in March, 1975; the policy of failing to offer the same protection from lay-off to employees on pregnancy leave as that offered to employees absent with other temporary disabilities, in 1978; and the policy of involuntarily transferring pregnant employees "when appearance is a factor," in 1978.
 In Coates, the Seventh Circuit looked to Teamsters as providing an example of what a plaintiff needs to show in order to establish a prima facie case in a pattern or practice action. According to the Coates court, "[t]he plaintiffs' prima facie case will thus usually consist of statistical evidence demonstrating substantial disparities in the application of employment actions as to minorities and the unprotected group, buttressed by evidence of general policies or specific instances of discrimination." Coates, 756 F.2d at 532 (footnote omitted).
 Because the court finds that the EEOC failed to establish a prima facie case that the policies existed in fact, and therefore denies the EEOC's motion for partial summary judgment, the court does not now address the question of whether these three policies violate Title VII.
 Before trial, EEOC also abandoned any claim with respect to full time commission sales hires in Sears' Eastern, Midwestern, and Southern Territories for the years 1979 and 1980.
 In judging the credibility of each witness and the weight to be given the testimony of each, the court has taken into account for each witness the intelligence, the ability and opportunity to observe, the age, the memory, the manner while testifying, any interest, bias, or prejudice the witness may have, and the reasonableness of the testimony considered in the light of all of the evidence in the case.
Even though substantial portions of the extensive evidence in this case (including virtually all of EEOC's evidence) involved statistics, these findings of credibility and weight to be given testimony were of crucial importance in deciding the outcome of the case. As will be discussed below, the massive statistical evidence presented must be evaluated in light of the facts and circumstances at Sears. Therefore, the testimony of nonstatistical witnesses was as important as the testimony of the statistical experts. The credibility of statistical experts and the weight to be given their testimony were also of great importance. The court made specific credibility and weight findings on various issues testified to by each of the experts. The court has only included certain specific credibility and weight findings to the extent that they could practically be incorporated into this lengthy opinion.
 The court sought to make as complete a record as possible in this case. However, due to the vast amount of evidence presented, the court could not discuss all the evidence presented in a wieldy fashion. Although the court considered all the evidence presented in reaching its decision, only the most important bases for the court's decision are discussed in this opinion. Similarly, the court was able to make only limited citation references to the specific evidence relied upon in this opinion. The court's limited reference to exhibits follows the wording usually used to mark the trial exhibits "Plaintiff" and "Sears."
 Section 703(a) (1) provides:
(a) It shall be an unlawful employment practice for an employer
(1) to fail or refuse to hire or to discharge any individual, or otherwise to discriminate against any individual with respect to his compensation, terms, conditions, or privileges of employment, because of such individual's race, color, religion, sex, or national origin;....
 As a matter of convenience only, the court has occasionally used the masculine pronoun to refer to both men and women collectively.
 Section 703(a) (2) provides:
(a) It shall be an unlawful employment practice for an employer
. . . . .
(2) to limit, segregate, or classify his employees or applicants for employment in any way which would deprive or tend to deprive any individual of employment opportunities or otherwise adversely affect his status as an employee, because of such individual's race, color, religion, sex, or national origin.
 In addition, the court finds that the EEOC waived its right to pursue a disparate impact theory. As noted above, the EEOC gave little hint that it would proceed under this theory until after trial commenced. It made only an equivocal reference to the theory in a footnote of its trial brief submitted before trial. Not until closing argument did the EEOC, however tenuously, state its position on the subject. However, in the final pretrial order, EEOC specifically stipulated that it alleged that Sears discriminated against women under Section 703(a) (1) of Title VII, 42 U.S.C. § 703(a) (1). Final Pre-Trial Order, Tab III.A.1.a., Stipulation A-8. As discussed above, the disparate impact theory evolved under § 703(a) (2), with the Griggs court specifically quoting only that provision of Title VII. By relying only on § 703(a) (1), EEOC in effect told the court and Sears that it was not pursuing a disparate impact claim. As set forth in the written procedures issued by the court to both parties, the final pretrial order controls the subsequent action of the case. A party cannot, by the terms of the order, alter or expand its theories of the case after trial has begun without the consent of the parties and the court, or by order of court to prevent manifest injustice. (Final Pretrial Order and Pretrial Procedure Order); Fed.R.Civ.P. 16(e). EEOC did not attempt to amend the pretrial order. Further, the court finds that to allow EEOC to alter the theory of its case after trial would be manifestly unjust to Sears. The court therefore holds EEOC to its original stipulation of its claim, which states that Sears has violated only § 703(a) (1), the provision relating to disparate treatment, and will not permit EEOC to now attempt to rely on the disparate impact theory based on § 703(a) (2).
 Many courts have acknowledged this. E.g., Bazemore v. Friday, 751 F.2d 662, 672 (4th Cir. 1984) ("[A]ll measurable variables thought to have an effect on salary" should be included in a multiple regression model.), cert. granted, ___ U.S. ___, 106 S. Ct. 379, 88 L. Ed. 2d 333 (1985); Eastland v. Tennessee Valley Authority, 704 F.2d 613, 623 (11th Cir.1983) (quoting Fisher, Multiple Regression in Legal Proceedings, 80 Colum.L. Rev. 702, 713 (1980) ("[A] properly done study begins with a decent theoretical idea of what variables are likely to be important.")), cert. denied sub nom., James v. Tennessee Valley Authority, 465 U.S. 1066, 104 S. Ct. 1415, 79 L. Ed. 2d 741 (1984); Valentino v. United States Postal Service, 674 F.2d 56, 71 (D.C.Cir.1982) ("[A] regression model [that] ignores information central to understanding the causal relationships at issue [can]not [be accepted as proof] adequate to raise an inference that sex discrimination accounts for salary difference."); Wilkins v. University of Houston, 654 F.2d 388, 402 (5th Cir. 1981), vacated, remanded on other grounds, 459 U.S. 809, 103 S. Ct. 34, 74 L. Ed. 2d 47 (1982), aff'd on remand, 695 F.2d 134 (5th Cir.1983).
 Sears' retail stores are classified as A, B, C, D or E stores, generally according to the assortment of merchandise and the size of the store. "A" stores are generally the largest, with the full range of Sears merchandise. "E" stores are the smallest, operating as "satellite stores" to a nearby "A" or "B" store, and containing primarily the major home appliances. From 1973-1980, approximately 95% of all Sears' commission sales hires were in "A" or "B" stores.
 Applications used before 1974 asked the applicant to write the type of work applied for in a blank space on the application. The application form used from 1974 through 1980 allowed the applicant to indicate the type of work applied for by placing a check mark next to any of the five categories listed: sales, office, mechanical, warehouse, or other. For the "other" category, the applicant could specify the type of work sought in a space provided. Both applications asked whether the applicant sought part time or full time work, permanent or temporary employment, and the hours available for work.
 Ray Graham had recommended in 1968 that Sears submit itself to the requirements of Executive Order No. 11,246, 3 C.F.R. 336 (1965). He believed it would help Sears implement its program and would provide a good example for other American businesses. However, during the next decade, no other national retail merchant followed Sears' lead. After declaring itself a government contractor, Sears was repeatedly found to be in compliance with federal affirmative action requirements during the relevant period of 1973-1980.
The General Services Administration ("GSA") was selected by OFCCP to review Sears' compliance. GSA organized workshops to help companies develop affirmative action plans. It sometimes asked Sears to take leadership roles in these workshops. On occasion, GSA even requested that Sears assist in the training of the government's equal opportunity personnel. See Written Direct Testimony of Ray J. Graham at 15-16. GSA thus considered Sears a model company for others to follow with respect to affirmative action. Government correspondence affirmed this.
The court finds the testimony of Ray J. Graham to be highly credible and entitled to great weight.
 Under the 1970 plan, units were required to formulate goals for hiring and developing women and minority candidates into supervisory positions. Managers were required to develop programs and report the results.
 In 1971, Sears convened a two-and-one-half day meeting of the officials of top manufacturers and suppliers. The purpose of the meeting was to explain to the company representatives the law, Sears' commitment, Sears' program, and Sears' expectations regarding the affirmative action programs of those who sold to Sears. EEOC Chairman, Williams H. Brown, and two other EEOC commissioners, EEOC's General Counsel, and several of its key staff members attended. EEOC officials questioned Sears officials at the meeting to demonstrate the issues employers face in various aspects of affirmative action programs.
 At this meeting, Sears officials and managers were told that their future success with the company would depend on their success in carrying out the company's affirmative action goals.
 These witnesses included, among others, Ray J. Graham, Charles F. Bacon, Sandra Hagerty, Pamela Balistee, Ed Michaels, Effie Tate, Thomas Dowd, Raymond Valdez, Daniel Mihalovich, Gary Rutherford, Thomas Biczak, Channing Phillips, and Linda Rollins. The court finds their testimony is entitled to great weight.
 EEOC eventually called two witnesses in its rebuttal case, but the purpose of their testimony was to contradict Sears' witnesses' testimony regarding the coding of certain applications. As will be discussed below, neither provided any credible evidence that Sears discriminated against them.
 The method of selecting the stores is described in EEOC's Report on Selection of Non-Hired Applicants Sample, prepared by EEOC witnesses Dr. Eugene Ericksen and Leonard Cupingood.
 EEOC's Report on Descriptions of Sampling Procedures, Sears Employees Sample, prepared by Dr. Eugene Ericksen, describes the method of selecting the sample persons hired into commission sales.
 See Table 1 (full time) and Table 16 (part time), Report on Sears, Roebuck & Co's Commission Sales Hiring and Promotion Practices, by Dr. Bernard Siskin (revised September 10, 1984) hereinafter referred to as Commission Sales Report.
 Problems with this extremely broad definition of "sales applicants" will be discussed at length below.
 Each characteristic was grouped into various categories as follows. The Job Applied For characteristic was grouped into eight mutually exclusive categories. Persons were assigned to the categories on the basis of the types of work they checked and/or wrote on their application. One of the categories was "commission sales." This category included persons who wrote in the words "commission sales" or "big ticket," or who indicated an interest in sales and also indicated divisions or products sold on commission at Sears, and did not indicate any other specific interest. The other categories were: commission sales plus non-sales, noncommission sales, noncommission sales plus non-sales, sales, sales plus non-sales, and preference not indicated.
The Age characteristic was divided into seven age group categories.
The Education characteristic was grouped into four categories: (1) less than 12 years; (2) 12 years; (3) 13-15 years; and (4) 16 or more years.
The Job Type Experience characteristic was divided into six categories: (1) commission sales (persons whose applications specifically indicated commission sales experience); (2) sales-high (prior experience not identified as commission sales or as lower level sales positions such as cashiers); (3) sales-low (generally cashier-type experience); (4) unspecified (job experience illegible or of indeterminate type); (5) non-sales; and (6) no experience. No attempt was made to code for amount of job experience, to differentiate between those with very little experience and those with many years of experience.
The Product Line Experience characteristic was designed to take into account experience with a product, whether or not that experience was in sales. Thirteen product line groupings were used, consisting of twelve product lines and an "all others" category.
The Commission Product Sales Experience characteristic was developed to identify sales experience that was likely to have been commission sales but was not specifically indicated as such on the application. A person was generally given credit for commission product experience if the application indicated experience with a product sold on commission at Sears.
See EEOC's Commission Sales Report, Appendix 3.
 The text of EEOC's Commission Sales Report (at p. 32A-33A) discussing Table 10 indicates that the table reflects the relative chance of selection after adjusting with the logit analysis. However, the title of Table 10 indicates that it contains the relative chance of selection after adjusting with the multivariate analysis. The court assumes that the text, and not the title of the table, is correct. In any event, this assumption will not be material to the outcome of the case.
 Four of the six characteristics as controlled for by Dr. Siskin were regrouped. Job Applied For was regrouped from eight categories into six categories; Age was regrouped from seven categories into two categories: (i) 16-24, and (ii) 25 and over; Education was grouped from four categories into two categories: (i) 12 years or less, and (ii) 13 years or more; and Job Type Experience was regrouped from six categories into two categories: (i) commission sales and sales high, and (ii) all others.
 Some of the categories on the Applicant Interview Guide on which applicants are asked to rate themselves are: (1) Face to face selling, persuading customers; (2) Sell major appliances: stoves, refrigerators, freezers, etc.; (3) Decorating, interior design; (4) Parts Dept.: stock, catalog, locate. Auto, mechanical; (5) Sell home improvement jobs: Kitchen, fence, roof, etc.; (6) Sell wearing apparel; (7) Deluxing: furniture refinishing, merchandise setup, preparation; (8) Deliver merchandise to customer home; (9) Sell hardware, paint, electrical: Technical goods; and (10) Construction: Carpenter, sheetrock; Plumbing, etc. See Sears Exhibit 6-13.
 As noted above, Dr. Siskin's pool of "sales" applicants consists of every person in the non-hired applicant sample except those who specifically applied for non-sales jobs only. Dr. Siskin assumed that all applicants in this pool are equally qualified for and interested in all commission sales jobs at Sears in all divisions.
 For this analysis, Dr. Siskin used his product line analysis data based on his coding of commission sales hires at Sears. He determined the percent of hires with experience and without experience by product line for the four product lines he chose to examine. As the applicant pool to compare with the hires with experience, he used respondents to the Applicant Interview Guide who rated themselves 3, 4 or 5 on experience in a related product line. As a pool to compare with those hired into commission sales without experience, he used respondents who rated themselves 1 or 2 on experience in the relevant category, and who also rated themselves 4 or 5 on interest in that category. He combined these two figures to reach his expected proportion of women hires based on responses to the Applicant Interview Guide. See Tr. 6356-60 (B. Siskin).
 Dr. Siskin matched the following Applicant Interview Guide categories with the four Sears product line groups:
AppliancesSell major appliances: stoves, refrigerators, freezers, etc.
AutomotiveParts Dept.: stock, catalog, locate. Auto, mechanical
MaterialsSell home improvement jobs: Kitchen, fence, roof, etc.
Home ImprovementsSell hardware, paint, electrical: Technical goods
 The results of Dr. Siskin's Applicant Interview analysis are as follows:
Multivariate Cross- Applicant classification Interview Analysis Guide Appliances 42.4 48.7 Automotive 28.9 51.0 Home Building 32.5 47.4 Materials Home Improvements 43.1 41.5 (Pl. Ex. Siskin 72).
 Questions asked include: "Do you have a low pitched voice?" Do you swear often?" "Have you ever done any hunting?" "Have you played on a football team?" However, as will be discussed below, men and women were evaluated using different scales.
 EEOC also introduced evidence at trial based on a judgment sample designed in the investigatory stages of this case. However, EEOC did not rely on this evidence in its closing arguments. The court therefore has not specifically addressed it, although the court has considered it along with all of the evidence in the case. The court notes, however, that the judgment sample was designed in a suspect manner, and the evidence EEOC introduced based on that sample is entitled to little weight.
 Dr. Siskin was unable to identify any policy or practice at Sears that was discriminatory against women. Instead, he contended that there was "something in the process" that caused the disparity between the "expected" and "actual" figures that he produced.
 As Sears' expert Dr. J. Haworth noted, by excluding "non-sales" applicants, EEOC may have sought to create the impression that it was restricting its pool of applicants. However, its definition of "sales" applicants was already overinclusive, and because the non-sales applicants were disproportionately male, excluding non-sales applicants had the effect of increasing the female proportion of the pool. EEOC attempted to justify its exclusion of non-sales applicants on the basis that only 3.3 percent of the non-sales applicants were hired into commission sales. EEOC included, however, 16-19 year old sales applicants, most of whom were female, even though they had an even smaller chance of being hired. Thus, it is evident that EEOC constructed its "sales" applicant pool with a view toward maximizing the female percentage of the pool.
 Applicant flow data is data about persons who actually applied for the job being analyzed.
 The court recognizes that these factors are not easily quantified for purposes of a statistical analysis, and that data relating to these factors was generally not available to EEOC from the application forms it chose to rely upon. However, the court finds that these factors, and those discussed above, are characteristics in commission sales applicants sought by managers and are important qualifications for commission sales hires. Therefore, Dr. Siskin's analyses are entitled to less weight to the extent they do not incorporate these factors.
 This principle has been graphically described as "garbage in, garbage out." Tr. 5622. (B. Siskin).
[35a] In essence, EEOC's statistical case is based on the erroneous underlying presupposition that all relevant Sears employee applicants and hires, as well as the jobs involved, are fungible. It is assumed that, not only are all the jobs alike, but also all men and women are exactly alike except for a few specific simplified characteristics that have been somewhat quantified in a limited way. Since the attempt to differentiate between men and women applicants and hires is inadequate, and no attempt is made to distinguish between jobs, this results in misleading statistics that do not accurately reflect the considerations and decisions that actually go into the individual hiring process for each Sears employee for each job in each store and unit throughout the vast Sears organization. See, e.g., Written Direct Testimony of Dr. David A. Wise. EEOC's pay analysis suffers from the same erroneous presupposition and analysis.
 Sears' witnesses with extensive experience testified that the ratio of men interested in commission sales to women was at least 8 or 10 to 1.
The court finds the testimony of the Sears officials and employees in this case to be very credible and persuasive. Their testimony in the case was supported by a substantial number of other employees whose uncontradicted written testimony was also persuasive.
 Outside sales were required in Division 32 (fencing), Division 37 (floor coverings), Division 42 (plumbing and heating), Division 64 (building materials), and Division 65 (kitchens, dishwashers). Outside sales often required night calls on customers.
 These divisions included the following. (1) Division 32fencing and garden tractors. Prices could be as high as $5,000. Selling required technical knowledge of product and repair and maintenance of tractors. Selling fencing often required outside work. (2) Division 42plumbing, heating, air conditioners and fences. Selling these products required substantial technical knowledge of the product, could require contact with dirty machinery and salespeople could often be on call around the clock. (3) Division 64building materials. Selling in this division required a general background in construction to sell "inside" items such as insulation, paneling, cement mixers, and drop stairs. Outside sales required dealing with roofs, gutters, crawl spaces, and the general construction of a house. (4) Division 65kitchens, dishwashers. The most technically complex of the Installed Home Improvements Division, it required knowledge of carpentry, electricity, gas, and architecture.
 By Dr. Siskin's own count, there were 680 refusals of offers of commission sales or indications of a lack of interest in commission sales. Tr. 17577; Sears Siskin Rebuttal Ex. 1. Although Dr. Siskin and EEOC contend that there were far fewer refusals of actual job offers, it is obvious from Sears Exhibit 25 that there were a larger number of documented commission sales job refusals by women, and a very large number of specific statements of lack of interest by women in commission sales during the entire relevant time period, from 1973 until the 1980's.
EEOC also argues that evidence of female job refusals is not helpful unless the number of male refusals is also known. However, even without data on male refusals, this evidence corroborates Sears' witnesses' testimony that sincere and extensive efforts were made to recruit women into commission sales, and that many women were not interested.
 Juliet Brudney is a consultant and writer on employment issues, particularly issues concerning women.
 In a structured interview, standard questions are asked of each interviewee. No control group was used. Although the study was in no way comprehensive of all women in the job studies, Ms. Brudney was a credible witness and her study confirmed testimony given by Sears managers.
With respect to the automotive and service technician jobs, Ms. Brudney found that women had little or no training or experience in these fields before Sears hired them. Sears trained them on the job and in special training courses at company expense. They were assisted by male employees at Sears, and did not encounter the male resistance often found with other employers. This willingness of Sears to train women in this manner is further proof of their sincere and costly attempts to recruit women into nontraditional jobs.
 Dr. Rosenberg is an associate professor of history at Barnard College, Columbia University, where she specializes in American Women's history. The court found her to be a well-informed witness who offered reasonable, well-supported opinions. She was a highly credible witness.
 Neither Ms. Brudney nor Dr. Rosenberg contend that all women have these tendencies or preferences, and the court has not drawn any such inference from their testimony. They have merely attempted to describe the overall tendencies of many women. EEOC presented witnesses with contrary views, whose testimony is discussed below. Suffice it to say at this point that few sweeping generalities can be accurately made about women (or men) overall in the workplace or in society. The court continually throughout trial exhorted witnesses to quantify their generalizations by estimating some percentage of women with the interests or views being discussed. Few witnesses were able to do so. The court therefore gave little weight to much of the testimony generalizing about women in the workforce from the actions of a few. However, the testimony of Ms. Brudney, although not based on a scientific study, reflected actual views of women at Sears, and was corroborated by the testimony of credible Sears' witnesses discussed above. The testimony of Dr. Rosenberg was also consistent with the experience of Sears' managers and other personnel, and with other evidence discussed below, and has also accordingly been given some weight by the court.
 Although EEOC witness Dr. Stanley Presser attempted to discredit Dr. Crespi's testimony on the basis that it is impossible to measure a change of attitude unless identical questions are asked of the same population over time, Sears' witness Burns Roper convincingly testified that identical questions are not necessary to draw conclusions about change in attitudes over time. Based on this testimony, the court finds Dr. Presser's criticism without merit.
 The results of these surveys of timecard employees are given to store managers to aid them in understanding employee perceptions and attitudes. Individual store policies are often changed as a result of the surveys.
 The survey was hurriedly developed and administered under severe time constraints for use in conciliation meetings with the EEOC.
 The results of the 1976 Job Interest Survey are contained in Sears Exhibit 254-3.
 Dr. Presser is the Director of the Detroit Area Study for the Sociology Department of the University of Michigan. He testified for EEOC as an expert on survey techniques.
 First, Dr. Presser testified that the results of the survey cannot be relied on because the sample of stores surveyed was a judgment sample, requiring use of human judgment in selecting stores, rather than a random sample. The court finds, however, that the stores sampled in the survey would not unfairly affect the results of the survey. Although certain characteristics of the selected sample, such as age and education of employees, may be different than the overall population, Dr. Smith has adequately adjusted the results to compensate for this and found that the number of males and females definitely interested in commission sales did not change. Dr. Presser also criticized the description of big ticket selling in the survey as providing only a measure of interest in particular big ticket jobs, not of all big ticket jobs in general. However, the definition in the survey is sufficiently general that it would not mislead respondents. This is confirmed by the overwhelming agreement of respondents with the proffered definition of commission sales. In any event, the responses to questions 1 and 2 of the survey, which Dr. Presser admits are not affected by the definition of commission sales, shows that non-commission salesmen were far more likely to be interested in commission sales than noncommission saleswomen.
 Question 13a asked: "Are you or have you ever been in non-commission sales at Sears?" The possible responses were: (1) currently, (2) in the past but not currently, and (3) never. This question apparently was ambiguous, because far more persons responded affirmatively than were in noncommission sales at the sampled stores. Through responses to a number of other questions, however, Sears was generally able to eliminate those persons "in non-commission sales" who were not actually salespersons at the time of the survey. See Written Direct Testimony of Dr. Irving Crespi, Appendix 3.
 Even Dr. Siskin admitted that "[T]here is a big difference between (saying an applicant would accept a job) and saying that's tremendous interest, because, you know, there is interest and there is interest." Tr. 4248-49 (B. Siskin).
 For part time, 49 percent of the women and 21 percent of the men were not interested in and would not accept a commission sales job.
 EEOC has also criticized the survey on other bases. It contends that Sears should have pretested its method of identifying non-commission salespersons. However, as Burns Roper pointed out, pretesting in this situation would not likely have revealed the problems encountered when the survey was given to the entire population sampled. See Written Direct Testimony of Burns Roper, ¶ 5.
 See, e.g., Crespi Written Testimony, ¶¶ 30, 33, 34, 36, 37.
 Although EEOC witnesses testified that women responding to the surveys discussed above may have been influenced by a perceived lack of opportunity for women in commission sales, they offer no concrete evidence to support or quantify this claim, and no persuasive reason to believe that women would be so reluctant to express their interest in commission sales.
 Although the surveys and testimony discussed above relate primarily to the interests of women already employed by Sears, the court finds that this evidence also provides a good indication of the interest of women applying for sales positions at Sears.
 Dr. Joan Haworth was Sears' principal expert witness. She testified as an expert in both labor economics and statistics. She had a very thorough knowledge and understanding of her fields of interest as well as Sears itself. The court found her to be a highly credible witness whose testimony was entitled to great weight.
 Dr. Haworth analyzed this data on all commission sales workers in the work force, which includes all types of commission sales, from the insurance industry to industrial sales. Dr. Siskin criticized Sears' use of national data on commission sales in general as an inappropriate benchmark, asserting that data in retail sales would provide a more appropriate basis for comparison. However, retail industry data includes all noncommission and commission salesworkers, and therefore would not provide a reasonable benchmark for commission sales. The data on commission salesworkers provides a more reasonable basis for comparison, particularly because the product lines into which 95 percent of Sears' full time commission salespersons were hired included items such as hardware, building supplies, paint and appliances. Persons selling these products are represented in the wholesale and industrial trades, as well as the retail trade.
 EEOC asserts that national data does not provide a reliable benchmark for commission sales because it reflects past discrimination. However, EEOC presented no evidence of past discrimination in the commission sales field in general or in any of the areas evaluated by Dr. Haworth. Although the court has considered the possibility of past discrimination in evaluating this evidence, since there is no proof of such discrimination, much less any quantification of it, the court finds that this data does provide a useful basis for comparison with Sears.
 These witnesses were Dr. Alice Kessler-Harris, Dr. Janice Fanning Madden, and Eileen Appelbaum.
The court also notes that EEOC analyzed Sears Applicant Interview Guide data. This analysis was performed separately from EEOC's principal multicell and logit analyses. Dr. Siskin adjusted the data to reflect the relative experience of Sears' commission salesforce. This evidence relates only to a few product line groupings, and, because of Dr. Siskin's adjustment of the data, EEOC's analysis provides no specific data on applicant interest in commission sales that is useful to the court in determining applicant interest in commission sales. EEOC's analysis of the Guide data is discussed below.
 See, e.g., Written Testimony of Dr. Janice Fanning Madden.
 For example, Dr. Kessler-Harris testified about the experience of women during both World Wars, who took jobs such as welders, shipfitters, and crane operators, as well as similar experiences of other women in unusual circumstances throughout history. It is not an issue in this case that some women are both capable and interested in holding commission sales jobs in traditional male product areas, such as automotive, plumbing, furnaces and fencing. This is obviously true. The real question is what percentage of women versus men, who applied for those jobs at Sears stores during 1973 to 1980, were capable and interested.
 For example, Dr. Kessler-Harris insisted that numerical differences between men and women within jobs in the workforce can only be explained by sex discrimination by employers. Written Testimony of Alice Kessler-Harris; Tr. 16,545-46. She offered no evidence to support this bald assertion. Also, Dr. Madden insisted that women's preferences play no part in their decisions regarding jobs. Written Direct Testimony of Dr. Janice Fanning Madden, at 1-2. She later admitted that interest may have some effect on female job decision-making.
 Sears has also introduced evidence of applicant interest in commission sales through its responses to EEOC requests for information, known as "Recruitment 2." These Sears responses show a very low percentage of female applicants for commission sales. However, the court has not relied on this evidence because of its questionable reliability. Only 43 of the 112 responding A and B stores actually separated commission sales applicants from other applicants. Uniform instructions for compiling the responses apparently were not given. In addition, at one unit, all male applicants were treated as commission sales applicants, and all females were treated as non-commission sales applicants. Although Sears did not include this store's responses in its analysis, it calls into question the reliability of all responses. The court has therefore not relied on this evidence as proof of lack of interest in commission sales.
 Probably recognizing this, EEOC has not relied on its unadjusted analysis in its final argument or its proposed findings of fact and conclusions of law.
 Dr. David A. Wise is the John F. Stambaugh Professor of Political Economy at the John F. Kennedy School of Government, Harvard University. He testified as an expert statistician on behalf of Sears. The court found him to be a very credible witness and gave great weight to his testimony.
 EEOC expert Dr. Eugene Ericksen estimated the variances for the nationwide estimates of the female proportion of the nonhired applicant pool.
 In 1973, the corrected z value was 4.12, and in 1974, the corrected z value was 3.86. See Sears Exhibit 5-4.
 During its pre-suit investigation, EEOC made no findings of discrimination with respect to Sears' testing practices. In fact, internal EEOC memoranda show that Sears' tests for certain selling positions were "approved" during EEOC's pre-suit investigation. Also, when the court asked EEOC counsel in a pre-trial conference whether Sears' test procedures were an issue in the case, the court was assured that they were not.
 The testimony of EEOC expert witness Dr. Lopez on Sears' testing practices was not persuasive. Dr. Lopez had little knowledge of Sears' practices and had no reasonable basis for his criticisms. The court finds that his testimony in this case had little credibility.
 Sears did not perform its own analysis of applicant flow data because of the claimed impossibility of identifying commission sales applicants from the applications and other available data. As discussed above, Sears is not required to provide its own regression analysis to refute EEOC's analyses. It may prevail by other evidence and by showing that EEOC's analyses are so faulty that they do not provide a reasonable basis for inferring that Sears intentionally discriminated against women.
 The data on hired applicants analyzed by EEOC was derived from Sears' payroll tapes. These tapes were used only for payroll purposes, and data not used directly for paying employees was more likely to be unavailable or inaccurate. Not only was more information available to Sears through the personnel files, but it was likely to be more accurate. EEOC had access to the personnel files but chose to rely on the payroll tapes.
 Of the full time hires in this category, 83 percent of the men and 58 percent of the women had one of these second three characterictics; of the part time hires in this category, 63 percent of the males and 39 percent of the males had one of these three characteristics. See Written Direct Testimony of Channing D. Phillips.
 Sales performance is evaluated on the basis of sales per hour, adjusted by the average sales per hour in the salesperson's division, to allow for differences in price and sales volume among divisions.
 For example, if approximately 30 percent of the women applicants had been more qualified and interested than the average man, this would explain the difference between EEOC's expected hiring rate and the actual hiring rate. See Sears Exhibit 5-5.
 Dr. Wise also points out that, based on the sales performance data for hires, it is not reasonable to assume that evaluation error played a dominant role in actual hiring and promotion decisions. He notes that one would expect that the error would be considerably less when ranking current employees considered for promotions than when ranking outside applicants. However, the relationship between sales performance of men and women promoted into commission sales is very similar to the pattern observed among hires. See Sears Exhibit 6-3-2. This similarity would not be expected if there were a high evaluation error. In addition, Dr. Wise demonstrated that Dr. Siskin's estimations of Sears' inability to choose the best applicant were grossly overstated. See Written Rebuttal Testimony of Dr. David A. Wise, ¶¶ 13-21.
 EEOC expert Dr. Siskin also criticized the sales performance data on the basis that it reflects the greater seniority of males. Dr. Siskin correctly notes that seniority and sales performance are correlated. However, males outperformed females not only overall, but also among hires and promotions in their first year of hire. See Sears Exhibit 6-3-2. EEOC attempts to discount this evidence by arguing that Sears Exhibit 6-3-2 measures the first calendar year of performance, including those hired at the beginning of the year and those hired at the end of the year. EEOC contends that since the number of females hired into commission sales was generally increasing over time, there probably were more females hired at the end of the year than earlier. Thus females would have less seniority, thereby explaining their poorer performance. However, EEOC offered no factual evidence to support this assertion that more females were hired at the end of the year. The court was not persuaded by Dr. Siskin's unsupported speculation that this factor could have a significant effect on performance rates.
 Applicant Interview Guides were given to applicants to stores in the Denver, Colorado area from approximately 1978 through 1980, and in Waco, Texas during 1980.
 Normalization is a commonly used statistical technique. To normalize the scores, each person's average response was calculated, and then the response in each category was compared with that average. The procedure penalizes persons who have more and varied interests and experience. However, since the experience, skill and interest scores on all questions were higher for men, more men than women would be penalized by this procedure.
 For example, to measure interest in the Home Building Material product line, the category analyzed by EEOC (Sell home improvement jobs: Kitchen, Fence, Roof, etc.") is probably more appropriate than the category chosen by Sears ("Construction: Carpenter, Sheetwork, Plumbing, etc."). However, the construction category is related to the type of products sold in the Home Improvements Division, and the court has taken into account the likelihood that very few women would express interest, skill or experience in this category, and decreased the weight given to this evidence accordingly.
 Contrary to Dr. Siskin's assertions, the model does not assume that responses to the Applicant Interview Guide Categories will be normally distributed. It assumes that the underlying interest, skills and experience will be normally distributed.
 EEOC has argued that victims of discrimination often do not know that they have been discriminated against, and that therefore the court should not view its failure to produce persons who were discriminated against in a negative light. The number of Title VII suits filed by individuals alone seems to fairly refute EEOC's contention. More specifically, however, this dispute between EEOC and Sears has lasted for over ten years. With EEOC's vast investigation, and the enormous amount of information provided by Sears to it, EEOC had ample time and information available to identify at least some members of the alleged huge class of victims it purports to represent.
"When the statistical evidence does not adequately account for `the diverse and specialized qualifications necessary for [the positions in question],' strong evidence of individual instances of discrimination becomes vital to the plaintiff's case." Valentino, 674 F.2d at 69 (citations omitted). See generally B. Schlei & P. Grossman, Employment Discrimination Law 1391-94 (2d ed. 1983).
Moreover, particularly in the area of promotions, the EEOC should have presented evidence of actual applications for commission sales positions that were discriminatorily denied, or evidence that, had a woman employee who claims discrimination known of an opening in commission sales, she actually would have applied for that position. See Box v. A & P Tea Co., 772 F.2d 1372, 1377 (7th Cir.1985).
 The EEOC has added to and subtracted from its list of jobs at issue on numerous occasions. A final list was arrived at only through court order, with a final "final" list being produced on April 16, 1984.
 The Equal Pay Act of 1963, 29 U.S.C. § 206(d), provides that an employer cannot discriminate by paying workers of one sex less than workers of the opposite sex for jobs requiring equal skill, effort and responsibility, and which are performed under similar working conditions, except where such payment is made pursuant to (1) a seniority system; (2) a merit system; (3) a system which measures earnings by quantity or quality of production; or (4) a differential based on any factor other than sex.
 The EEOC cites the following cases as support for its contention that when a Title VII claim is also actionable under the Equal Pay Act, the Equal Pay Act analysis applies: Kouba v. Allstate Insurance Co., 691 F.2d 873 (9th Cir. 1982); Odomes v. Nucare, Inc., 653 F.2d 246 (6th Cir.1981); Strecker v. Grand Forks County Social Service Board, 640 F.2d 96 (8th Cir.1980), aff'd en banc, 34 FEP cases 1008 (8th Cir.1981); Schultz v. Wheaton Glass Co., 421 F.2d 259 (3d Cir.1970), cert. denied, 398 U.S. 905, 90 S. Ct. 1696, 26 L. Ed. 2d 64 (1970); Chang v. University of Rhode Island, 606 F. Supp. 1161 (D.R.I.1985); Miller v. Kansas Power & Light Co., 585 F. Supp. 1509 (D.Kan.1984); Schulte v. Wilson Industries, Inc., 547 F. Supp. 324 (S.D.Tex.1982); Marcoux v. State of Maine, 35 FEP Cases 553 (D.Me. 1984). Although the court in each of these cases applied the Equal Pay Act analysis to a Title VII claim of unequal pay for substantially equal work, in none of the cases did the court present a reasoned analysis in support of this application.
 Schultz, supra.
 These cases refer to the section of the Gunther opinion in which Justice Brennan asserts that the Court's interpretation of the Bennett Amendment will not render the amendment superfluous. 452 U.S. at 169-71, 101 S. Ct. at 2248-49. Justice Brennan pointed out that, although the first three affirmative defenses incorporated into Title VII by virtue of the Bennett Amendment mirror provisions already found in section 703(h) of Title VII, the Bennett Amendment guarantees that courts and administrative agencies will not adopt inconsistent interpretations of the similar provisions. As for the fourth affirmative defense ("any other factor other than sex"), Justice Brennan stated:
[I]ncorporation of the fourth affirmative defense could have significant consequences for Title VII litigation. Title VII's prohibition of discriminatory employment practices was intended to be broadly inclusive, proscribing "not only overt discrimination but also practices that are fair in form, but discriminatory in operation." Griggs v. Duke Power Co., 401 U.S. 424, 431, 91 S. Ct. 849, 853, 28 L. Ed. 2d 158 (1971). The structure of Title VII litigation, including presumptions, burdens of proof, and defenses, has been designed to reflect this approach. The fourth affirmative defense of the Equal Pay Act, however, was designed differently, to confine the application of the Act to wage differentials attributable to sex discrimination. H.R.Rep. No. 309, 88th Cong., 1st Sess., 3 (1963), U.S.Code Cong. & Admin.News 1963, p. 687. Equal Pay Act litigation, therefore, has been structured to permit employers to defend against charges of discrimination where their pay differentials are based on a bona fide use of "other factors other than sex." ... Although we do not decide in this case how sex-based wage discrimination litigation under Title VII should be structured to accommodate the fourth affirmative defense of the Equal Pay Act, ... we consider it clear that the Bennett Amendment, under this interpretation, is not rendered superfluous.
452 U.S. at 170-71, 101 S. Ct. at 2248-49 (footnote omitted). This passage does not, as the cases have asserted, imply that the Equal Pay Act analysis applies in Title VII sex-based wage discrimination cases. If this passage implies anything at all, it implies only that incorporation of the fourth affirmative defense may alter the structure of Title VII disparate impact cases, because the "any other factor other than sex" defense is different from and broader than the "business necessity" defense established in Griggs. See B. Schlei & P. Grossman, supra, at 479.
 In its discussion of the legislative history of the Bennett Amendment, the Gunther court pointed out that the House of Representatives did not add discrimination based on sex to the proscriptions in Title VII until two days before the House voted on Title VII. At that time, the House did not discuss any possible inconsistencies between Title VII, as amended, and the Equal Pay Act.
The Senate did not refer the Civil Rights bill to any committee before it considered the House version; therefore, neither House of Congress conducted a formal analysis on the implications of the overlapping jurisdiction of the two statutes. In the Senate floor debates, however, several Senators did express concern that insufficient attention had been paid to the overlapping jurisdiction of the statutes.
Senator Bennett proposed his amendment to address this general concern over possible inconsistencies, not to address any specific potential conflict between the statutes. In the short discussion on the floor of the Senate, Senator Bennett emphasized the technical nature of the amendment and the need to prevent disruption of the effective administration of the Equal Pay Act. He explained that the amendment assured that the provisions of the Equal Pay Act would not be nullified in the event of a conflict with Title VII. The Supreme Court interpreted his explanation as referring only to the affirmative defenses of the Equal Pay Act. 452 U.S. at 174, 101 S. Ct. at 2250. In other words, Senator Bennett's concern in proposing the amendment was for the benefit of employers that the four affirmative defenses be available to employers in Title VII, as well as Equal Pay Act, cases.
Senator Dirksen was likewise concerned that the defenses be available to employers universally in sex-based wage discrimination cases. In the Senate discussion, he commented that all the Bennett Amendment does is recognize the exemptions to coverage in the Fair Labor Standards Act, of which the Equal Pay Act is a part. These exemptions make the Fair Labor Standards Act applicable to a class of employers narrower than that class to which Title VII applies. According to the Supreme Court, Senator Dirksen's remarks are consistent with its view that the Bennett Amendment assures that the Equal Pay Act defenses are available to employers even where only Title VII applies, because, for instance, the employer is exempt from the Fair Labor Standards Act. 452 U.S. at 175, 101 S. Ct. at 2251.
After the brief discussion, the Bennett Amendment passed in the Senate without recorded vote. Back in the House, there was no debate on the Bennett Amendment when the Senate version of the Civil Rights bill returned for final approval.
Given the summary treatment of the Bennett Amendment in both Houses of Congress, it is clear that Congress did not intend the Bennett Amendment to have the radical effect of carving out of Title VII jurisprudence the one and only exception where Title VII analysis would not apply in a Title VII case. Also, the expressed concerns of Senators Bennett and Dirksen, as interpreted by the Supreme Court, do not support the interpretation that the Bennett Amendment incorporates into Title VII the Equal Pay Act analysis. As stated above, Senators Bennett and Dirksen intended the amendment to make the Equal Pay Act defenses universally available to employers in sex-based wage discrimination cases. The interpretation that the Bennett Amendment incorporates the Equal Pay Act analysis would penalize, rather than protect, employers in Title VII cases by requiring them to meet a greater burden of proof.
 The court in Chang, supra, followed the view that the Equal Pay Act analysis applies in every Title VII case involving unequal pay for substantially equal work.
 In Kouba, supra, the court stated:
In County of Washington v. Gunther, 452 U.S. 161, 170-71, 101 S. Ct. 2242, 2248-49, 68 L. Ed. 2d 751 (1981), the Supreme Court recognized that very different principles govern the standard structure of Title VII litigation, including burdens of proof, and the structure of Title VII litigation implicating the "factor other than sex" exception to an equal-pay claim (though the Court reserved judgment on specifically how to structure an equal-pay claim under Title VII). Accordingly, we have held that even under Title VII, the employer bears the burden of showing that the wage differential resulted from a factor other than sex.
691 F.2d at 875 (emphasis added and citations omitted).
 The broad remedial purpose of Title VII is to "prohibit all practices in whatever form which create inequality in employment opportunity due to discrimination on the basis of race, religion, sex, or national origin." Franks v. Bowman Transportation Co., 424 U.S. 747, 763, 96 S. Ct. 1251, 1263, 47 L. Ed. 2d 444 (1976) (citations omitted). The Equal Pay Act, on the other hand, has a more narrow purpose. As the Supreme Court stated in Corning Glass Works, supra:
Congress' purpose in enacting the Equal Pay Act was to remedy what was perceived to be a serious and endemic problem of employment discrimination in private industrythe fact that the wage structure of "many segments of American industry has been based on an ancient but outmoded belief that a man, because of his role in society, should be paid more than a woman even though his duties are the same."
417 U.S. at 195, 94 S. Ct. at 2228 (citation omitted).
 Title VII exempts certain persons and entities from its coverage in its definitional section, section 701 of Title VII, 42 U.S.C. § 2000e, and in section 702, 42 U.S.C. § 2000e-1. These exemptions are not as far-reaching as those found in section 13 of the Fair Labor Standards Act, 29 U.S.C. § 213.
 In Gunther, the Supreme Court pointed out that the Equal Pay Act, unlike Title VII, does not require plaintiffs to file administrative complaints and await administrative conciliation efforts. Also, the Equal Pay Act has a more generous statute of limitations for backpay relief. 452 U.S. at 175 n. 14, 101 S. Ct. 2250 n. 14.
 As will be discussed below, EEOC also asserts that it need not prove that persons in the jobs it analyzed are performing equal work, that it is sufficient under Gunther to prove that the jobs were only "similar." Although the Seventh Circuit Court of Appeals has not yet decided whether the Bennett Amendment incorporates the Equal Pay Act analysis into "traditional" Title VII pay claims of unequal pay for equal jobs, it has decided that the Equal Pay Act analysis does not apply to claims of unequal pay for "similar" work. Boyd v. Madison County Mutual Insurance Co., 653 F.2d 1173 (7th Cir.1981), cert. denied, 454 U.S. 1146, 102 S. Ct. 1008, 71 L. Ed. 2d 299 (1982). Therefore, even if the "similar" jobs standard were applied to this case, the Title VII elements and burdens of proof would be applicable.
 It is also important to note that, even if Gunther were applied to this case to allow EEOC to compare only similar jobs, EEOC has presented absolutely no evidence to the court regarding the job content of the "similar" jobs from which it could make the required subjective comparison of male and female jobs and pay. See discussion below.
 In a footnote addressing arguments raised in the dissent, the Court confirmed its intention to provide a remedy for victims of discrimination whose claims do not fit within the Equal Pay Act analysis. The court noted, "The dissent attempts to minimize the significance of the Title VII remedy in these cases on the ground that the Equal Pay Act already provides an action for sex-biased wage discrimination by women who hold jobs not currently held by men. Post, at 2264. But the dissent's position would still leave remediless all victims of discrimination who hold jobs never held by men." 452 U.S. at 179 n. 19, 101 S. Ct. 2253 n. 19 (emphasis in original). Thus, the Court's obvious concern was with plaintiffs who would not state a claim cognizable under the Equal Pay Act standard.
 Retail stores in metropolitan areas are organized into administrative units called retail groups, which are headed by group managers. Retail stores not located in metropolitan areas are organized into administrative units called retail zones, headed by zone managers.
 The only checklist employees not eligible for bonuses were Catalog Sales Office ("CSO") Managers and Eastern Territory Satellite Appliance Store Managers. These employees received a monthly "overwrite," a payment based on a fraction of one percent of their unit's sales.
 Sears' salary levels were considered to be high in relation to other retailers and industries. Sears clearly had the right under Title VII "to change and revise the job-evaluation and pay system." Patkus, 769 F.2d at 1261.
 Hay Associates is a well-known and well-regarded consulting company specializing in job evaluation and compensation, organizational structure and development, and management appraisal and development.
 The committees looked at position descriptions of persons presently in the job being evaluated. In some cases, only one position description was referred to. In others, several were needed to understand the basic duties of a position.
 Sears has labeled this group of employees analyzed by EEOC for the years 1973 through 1976 EEOC's "artificial workforce."
 The only evidence relating to job content presented by EEOC was the testimony of Jon M. Laking. Mr. Laking is a former general partner of Hay Associates, who worked on the Sears job evaluation project from 1974 until 1975, when he was removed from the project at Sears' request. He testified about "relationships" between certain "C" code jobs, and recodings of jobs after 1976. His testimony briefly discussed only a few pre-1976 jobs, and did not discuss specific job content or variations in jobs from store to store. See note 109 infra.
 For the years 1976-1980, EEOC analyzed all employees in the 51 job codes at issue.
 Dr. Siskin stated that, for statistical reasons, he believed there should be a minimum number of men and women in a position for analysis. He stated that these conditions would ensure that, at least for certain levels of analysis, there would be a sufficient number of observations for meaningful results, and that those results could be considered for a period of at least two consecutive years.
 EEOC attempted to prove that Sears followed a policy of paying "heads of households" higher salaries. However, although Sears may have had such a policy at one time in the past, it is clear from the credible testimony of Sears' witnesses that no such policy existed after 1953.
 EEOC's analyst, Leonard Cupingood, admitted that he would probably change some of his judgments in light of information now available to him.
 Mr. Cupingood had originally labeled his "corresponding" jobs as "equivalent" jobs. He explained to the court his reason for changing the label to "corresponding" jobs as quoted above.
 As noted above, the only evidence of job content presented by EEOC was the testimony of Jon M. Laking, formerly of Hay Associates. See Report on Hay Compensation Systems and Sears, Roebuck & Co.'s Executive Compensation Program and Checklist Compensation Practices, by Jon M. Laking, Plaintiff's Exhibit 9, Appendix 1. Mr. Laking testified as to "relationships" between certain "C" code jobs. However, neither EEOC's analyst, Leonard Cupingood, nor its statistical expert, Dr. Siskin, relied on his testimony in making their judgments about "corresponding jobs." Moreover, Mr. Laking's testimony does not affect the court's conclusion that there were substantial variations in job content from store to store.
In addition, the court found that much of the testimony of Mr. Laking was not credible. He was removed from the Sears project after little more than a year because his performance was unsatisfactory. He was subsequently asked to resign as a general partner at Hay. Jobs evaluated under his direction had to be completely reevaluated after he left. There were also a number of inaccuracies in his testimony. His memory on cross-examination was also highly selective, and he had an obvious basis for prejudice against Sears. Based on all of these factors, and the court's observation of Mr. Laking at trial, little weight was accorded to his testimony. On the other hand, the court found the testimony of Charles Bacon and Ian Sym-Smith, among others, to be highly credible and entitled to great weight.
 In addition, records of employees in the Pacific Coast Territory were not computerized until 1975. Since computerized data was not available for these employees, EEOC did not include them in their analyses.
 Sears' evaluation committees relied on varying numbers of position descriptions to determine what were the important job duties. Although there could be substantial variation from unit to unit, the committee formulated what was in many cases an ideal position description of the duties persons in a particular job should be performing. The committee then evaluated the position in terms of the three Hay factors.
 The affected job codes are A281 and A282 (Collection Manager II and ICredit Central), A286 (Division Manager), and A303 (Collection ManagerCredit Central). EEOC is not excused from proving actual job content merely because it has chosen to challenge 51 different jobs on a nationwide basis. The elements of proof do not change with the size of the case.
In addition, even if somehow the Gunther standard of "similar" jobs were applied to this case, EEOC would still be required to prove the actual job content of positions as performed. It could not rely solely on the Hay evaluations of positions. Two other courts have rejected reliance on the Hay evaluation system to prove substantial equality of jobs. See Marshall v. J.C. Penney Co., Inc., 464 F. Supp. 1166, 1191 (N.D. Ohio 1979); Wheeler v. Armco Steel Corp., 471 F. Supp. 1050, 1052 (S.D.Tex.1979).
 The computer tapes from which the extracts were made were used for payroll purposes, not to store information which would be important in making compensation decisions.
 A job/year is counted for each job in which there is a statistically significant disparity in a year. Thus, if there were statistically significant disparities for one job code in three years, three job years would be counted.
 EEOC based this decision to increase the score it assigned for 1979-1980 on a change in the written policy expressed in Sears' Executive Compensation ManualBasic Company Policy, under which Sears would assign a rating of 57 to employees too new to be rated. See EEOC's Checklist Compensation Report, Appendix 2.
 EEOC objects to the inclusion of these variables in the model. It asserts that they do not have sufficient explanatory value to be included in the model, and have been added merely to drive down the t-values of the disparities. However, as Dr. Wise testified, it is difficult to ascertain statistically the precise explanatory effect that these variables may have. Sufficient evidence has been presented to indicate that each of these variables did affect compensation at Sears to some extent. The court therefore finds their addition to the model useful. However, Sears Exhibit 6-TTT indicates that there are substantial reductions in salary differences and associated t-values even if these variables are omitted from the model. Therefore, their inclusion in the analysis is not essential to the court's conclusions. These variables will be discussed more fully below.
 The adjusted percent difference is the mean percent difference between men's and women's salaries, evaluated at the mean of employees' checklist seniority.
 The 3 job/years were: Personnel Manager IIIRetail Store (R606) for 1976 and 1978, and Assistant Buyer III (Buy Level 4) (MO29) for 1979.
 Overwrite data for the year 1974 was also missing for this third job (CSO Manager IWithout Service). However, Sears supplemented the 1976 computer records with paper records for 1976, which included overwrite data, and found that the statistical disparities associated with the computer records resulted from the lack of overwrite data. Although overwrite data is not available from computer or paper records for 1974, it is reasonable to infer that this data would also explain the 1974 disparities.
 In addition, all statistically significant disparities that occurred for these years without excluding outliers were in the entering salary level. Little data was available regarding the prior timecard experience or non-Sears experience of these employees. However, the data available indicate that males had more pre-Sears experience than the females. These factors would have a significant impact on income, and could easily explain the disparities in starting salary.
 Sears' "step-wise" breakdown of its regression analysis shows the differences in statistically significant t-values and adjusted percent differences when the marital status and number of children variables are omitted from the model. Sears Exhibit 6-BBBB. Although the effect of these variables on the results varied, they generally did not significantly reduce the t values or adjusted percent differences. Thus, their inclusion in the model does not significantly affect its results.
 When all checklist employees were analyzed, men received greater increases in only one year for the 1974 cohort, and one year for the 1976 cohort. See Sears Exhibit 6-17.
 Men received higher increases in only three instances. In two instances where men's percentage increase was higher, it was higher by less than 1 percent. In the third instance, it was higher by less than 2 percent. Sears Exhibit 6-18.
 Data on employees entering checklist before 1973 is even more limited than the data for those entering checklist in 1973 and later.
 Because of the great length of the trial and the large volume of evidence Sears proposed to present, the court in effect limited the number of witnesses Sears could present. EEOC did not offer contradicting testimony, so the court sought to avoid cumulative and repetitious testimony.