BAUER, Chief Judge.

In this appeal, we examine in detail trial procedures in two cases brought under the Age Discrimination and Employment Act, 29 U.S.C. § 621-634 (ADEA). We conclude that the district court erred in its management of these cases and reverse and remand for a new trial pursuant to the provisions of Circuit Rule 36.

A. Background

These cases arise from a reduction in force undertaken by General Electric (G.E.) at its Bloomington, Illinois manufacturing plant between 1982 and 1985. G.E. laid-off, demoted or transferred two hundred salaried-exempt, i.e., management and professional, workers pursuant to the reduction in force (RIF). Wilbur Hany and James Samsel, two employees fired in 1984, sued G.E. on behalf of themselves and others similarly situated in December 1985 ("the Hany suit"). As we discuss more fully below, when private employers are involved, ADEA provides this opt-in class action procedure that preempts the class-action procedure under Federal Rule of Civil Procedure 23(a). 29 U.S.C. § 626(b), (c), 216(b).

Ten other employees subjected to adverse action during the RIF joined the Hany suit as named plaintiffs. See Second Amended Complaint, Pleadings Vol. 1, Doc. 22. The Hany complaint alleged that during 1983 and 1984, G.E. discriminated against older G.E. employees on account of their age. Eleven plaintiffs alleged that they lost their jobs as a result of age discrimination, and the remaining plaintiff, Delbert King, alleged he was demoted for the same reason.

A second group of employees were fired in 1985. Donald Sobin and William Evelsizer filed suit in 1986, Ronald Monkman filed suit in 1987 (the "Sobin suits" or the "Sobin plaintiffs"). These plaintiffs were represented by a single attorney, but elected to file their own, individual lawsuits. G.E. targeted Monkman for demotion or layoff in 1984. Because he was only 39 at that time, Monkman was not protected by the Act. He was not actually terminated until 1985, when he was 40, and thus is entitled to ADEA protection.

Sobin's suit was set for trial for June 8, 1987. On May 10, 1987, Sobin moved for a continuance until after the September trial on the Hany suit. Before the court ruled on the continuance, on June 9, 1987, Sobin moved to consolidate all three Sobin suits with the Hany suit under Federal Rule of Civil Procedure 42(a). In his motion, Sobin contended there was a common question of fact: whether there was a "pattern or practice of age discrimination" at G.E. Despite G.E.'s objections, the district court granted the motion. It ordered that the Hany and Sobin suits be consolidated to litigate the issue of whether G.E. engaged in a "pattern or practice" of age discrimination. See Order dated July 8, 1987, Vol. 2, Doc. 18.

The "pattern or practice" trial was held in January 1988. The Hany plaintiffs' expert testified that during 1983 and 1984, older employees were more likely to be hurt by the RIF than younger employees, and that the difference was statistically significant. When he controlled for the effect of performance and assignment, the expert still found that age was a statistically significant variable affecting the probability of an employee experiencing an adverse action during 1983 and 1984.

Plaintiffs also presented anecdotal evidence of discrimination. There was no statistical evidence supporting the Sobin plaintiffs' allegation that G.E. discriminated in 1985. Six plaintiffs, including Sobin, testified they were demoted or laid-off while G.E. continued to employ younger employees in their respective departments. Witnesses also testified that some G.E. managers made disparaging comments about older workers. The jury found that G.E. engaged in a pattern or practice of discrimination against employees over the age of 40 at the Bloomington plant. No dates appear on the jury verdict form. We assume, therefore, pursuant to the district court's instruction to the second jury, that the first jury found G.E. discriminated from 1982 until the end of the RIF in 1985.

G.E.'s motion for judgment notwithstanding the verdict was denied, and the district court held a scheduling conference to determine the course of the rest of the litigation. The court ruled that the burden of persuasion would shift to G.E. to show that its pattern or practice of discrimination was not responsible for the termination of each individual plaintiff. Because the district court believed trying all eleven individual claims together would be unwieldy, it divided the plaintiffs into groups for trial. The trial for the first group, Hany, Delbert King, Vernon Brickey, and Joseph Callahan, was set for June 1, 1989.

The district court subsequently issued an order resolving several pretrial motions. In response to a motion for clarification of that order, the court reversed its position on the burden of persuasion in the second set of trials. It announced that it would not shift the burden of persuasion to G.E. — the common result of a pattern or practice verdict against an employer. See, e.g., Teamsters v. United States, 431 U.S. 324, 97 S.Ct. 1843, 52 L.Ed.2d 396 (1977). Instead, the district court stated it would continue to place the burden of persuasion on plaintiffs. Each plaintiff would have to establish a prima facie case under the framework set out in McDonnell-Douglas Corp. v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973). Consistent with this change in burdens, in response to plaintiffs' motion, the court ruled that the plaintiffs would "go first" at trial.

The parties hotly contested the proper use of the first jury's verdict. The court determined that the second jury would be instructed of the first jury's finding at the outset. The first jury's finding of fact" could be relied upon by the individual plaintiffs to establish their prima facie cases, and to show that the reasons given by G.E. for terminating the plaintiffs were mere pretexts for discrimination. See Order of May 10, 1985 at 5; Appellant's Supplemental Appendix ("S.A.") at 53. The instruction is set forth in full in the margin.

The second jury returned a verdict in favor of Hany and King, and against Brickey. The jury also found that G.E. willfully violated the Act. The court ordered equitable relief (reinstatement) and King was awarded $34,806 in damages. Hany settled with G.E. after the judgment was entered. G.E. appeals the King verdict and damage award, the willfulness finding, as well as some of the court's procedural and evidentiary rulings. We shall discuss the factual background in more detail where necessary to our analysis.

B. Legal Framework

This court has considered ADEA actions in the context of a reduction in force ("RIF") in several cases. See Oxman v. WLS-TV, 846 F.2d 448 (7th Cir. 1988); Ayala v. Mayfair Molded Prods. Corp., 831 F.2d 1314 (7th Cir. 1987). The ADEA prohibits employer discrimination against any individual between the ages of forty and seventy with respect to compensation, terms, conditions, or privileges of employment, because of the individual's age. 29 U.S.C. § 623(a). Thus, in an age discrimination case, a terminated individual must prove that he would not have been discharged or demoted but for his employer's motive to discriminate against him because of his age. Oxman, 846 F.2d at 452; Ayala, 831 F.2d at 1318; LaMontagne v. American Convenience Prods., Inc., 750 F.2d 1405, 1409 (7th Cir. 1984).

The ADEA incorporates the enforcement provisions of the Fair Labor Standards Act, 29 U.S.C. § 216, by reference. See 29 U.S.C. § 626(d). Before commencing a civil action, a claimant must file a charge alleging unlawful discrimination with the Equal Employment Opportunity Commission ("EEOC"). 29 U.S.C. § 626(d). After filing the charge, the claimant must wait sixty days before filing suit. Id. The EEOC filing and waiting period is intended to give the EEOC an opportunity to investigate the alleged unlawful practice and seek to eliminate it informally through "conciliation, conference and persuasion." Id.

Once the administrative requirements have been satisfied, a claimant may bring a civil suit. The suit may be filed by "any one or more employees for and in behalf of himself or themselves and other employees similarly situated. No employee shall be a party plaintiff to any such action unless he gives his consent in writing to become such a party. . . ." 29 U.S.C. § 216(b). This procedure preempts the class action procedure under Federal Rule of Procedure 23. LaChapelle v. Owens-Illinois, Inc., 513 F.2d 286, 289 (5th Cir. 1975). See, e.g., Arnold v. Postmaster General, 667 F. Supp. 6, 15 (D.D.C. 1987), rev'd on other grounds, 863 F.2d 994 (D.C. Cir. 1988) ("The opt-in requirement is incorporated through section 626(b) of the ADEA and is mutually exclusive with Rule 23."), cert. denied, 493 U.S. 846, 110 S.Ct. 140, 107 L.Ed.2d 99 (1989); Behr v. Drake Hotel, 586 F. Supp. 427 (N.D.Ill. 1984); Allen v. Marshall Field Co., 93 F.R.D. 438 (N.D.Ill. 1982). The Secretary of Labor may also bring suit under 29 U.S.C. § 216(b) and § 217.

A plaintiff in an age discrimination case may prove his claim in two ways. He may try to meet his burden head on by presenting direct or circumstantial evidence that age was the determining factor in his discharge. Or, the plaintiff may use the indirect, burden-shifting method of proof developed in Title VII cases originally outlined in McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973). See Weihaupt v. American Medical Ass'n, 874 F.2d 419, 424 (7th Cir. 1989); Oxman, 846 F.2d at 452. Under either method, the ultimate burden of persuasion remains with the plaintiff at all times. Mathewson v. National Automatic Tool Co., Inc., 807 F.2d 87 (7th Cir. 1986).

In order to establish a prima facie case in a reduction in force case under the McDonnell Douglas framework, the plaintiff must establish the following: (1) she was within the protected age group; (2) she was performing according to her employer's legitimate expectations, (3) she was terminated or demoted, and (4) others not in the protected class were treated more favorably. Oxman, 846 F.2d at 455. We developed this variation on the traditional prima facie case because in RIF cases there often is no replacement of an older employee by a younger one. Therefore, "[t]his formulation merely requires an employer that releases a protected employee while simultaneously hiring (or not "bumping") younger employees to fill positions for which the older employee was qualified, to explain its actions, without forcing the protected employee to uncover that elusive `smoking gun.'" Id. at 455-56, overruling Matthews v. Allis-Chalmers, 769 F.2d 1215 (7th Cir. 1985). In Ayala v. Mayfair Molded Products Corp., 831 F.2d 1314, 1318 (7th Cir. 1987), we enunciated a slightly different prima facie standard which required that an employee show she was qualified to assume another position at the time of the RIF, and yet was not retained. Because this standard drew upon our ruling in Matthews, we believe that it is no longer controlling, and that Oxman represents the state of the law in this Circuit. See also Burlew v. Eaton Corp., 869 F.2d 1063, 1066 (7th Cir. 1989) (adopting Oxman standard); Howard Eglit and Martin Malin, 3 Age Discrimination, § 17.61, at n. 12 (1991) (noting disparity between Oxman and Ayala).

Once a prima facie case is established, the burden of production shifts to the defendant to provide some legitimate, non-discriminatory reason for the demotion or discharge. Once the defendant articulates a lawful reason for the discharge, the presumption of discrimination dissolves, and the burden of production shifts back to the plaintiff to show the employer's reasons are pretextual. Oxman, 846 F.2d at 453 (citing LaMontagne v. American Convenience Prods. Inc., 750 F.2d 1405, 1409 (7th Cir. 1984)). This burden shifting approach is designed to help the trial judge manage the presentation of evidence at trial, and to evaluate motions for summary judgment. See Texas Dep't of Community Affairs v. Burdine, 450 U.S. 248, 252-53, 101 S.Ct. 1089, 1093, 67 L.Ed.2d 207 (1980); Lovelace v. Sherwin-Williams, 681 F.2d 230, 240-41 (4th Cir. 1982).

Normally, at this stage of the proceedings, "after a full trial on the merits, we need not concern ourselves about the method of proof [the plaintiff] used. The only question for us is whether there was sufficient evidence for a reasonable jury to find that age was a determining factor in [the employer's] decision to fire [the plaintiff]." Brown v. M M/Mars, 883 F.2d 505, 507 (7th Cir. 1989). See also Lockhart v. Westinghouse Credit Corp., 879 F.2d 43, 49 (3d Cir. 1989). In this case, however, because the trial court seemed to allow the methods of proof to cloud its handling of the ultimate factual questions, we believe it is necessary to consider the shifting burdens framework applied below.

Title VII provides another enforcement mechanism that played a pivotal role in the resolution of the case below. This is the class-based "pattern or practice" trial. The term "pattern or practice" derives from § 707(a) of the Civil Rights Act of 1964, 42 U.S.C. § 200e-6(a). The section states:

When the EEOC has reasonable cause to believe that any person or group of persons is engaged in a pattern or practice of resistance to the full enjoyment of any rights secured by this subchapter, and that the pattern or practice is of such a nature and is intended to deny the full exercise of rights herein described, the EEOC may bring a civil action. . . .

Id. The Supreme Court has stated that "[t]he pattern or practice language in § 707(a) of Title VII was not intended as a term of art, and the words reflect only their usual meaning." Teamsters, 431 U.S. at 323, n. 16, 97 S.Ct. at 1855, n. 16. In making this finding, the Court relied in part upon legislative history.

"[A] pattern or practice would be present only where the denial of rights consists of something more than an isolated, sporadic incident, but is repeated, routine, or of a generalized nature. There would be a pattern or practice if, for example, a number of companies or persons in the same industry or line of business discriminated, if a chain of hotels or restaurants practiced racial discrimination throughout all or a significant part of its system, or if a company repeatedly and regularly engaged in acts prohibited by the statute. . . . The point is that single, insignificant, isolated acts of discrimination by a single business would not justify a finding of a pattern or practice. . . ."

110 Cong.Rec. 14270 (1964) (Statement of Senator Humphrey) (quoted in Teamsters, 431 U.S. at 336-37, n. 16, 97 S.Ct. at 1855, n. 16). The Court has noted the "manifest" difference between individual claims of discrimination and a class action alleging a general pattern or practice. Cooper v. Federal Reserve Bank of Richmond, 467 U.S. 867, 876, 104 S.Ct. 2794, 2799, 81 L.Ed.2d 718 (1983). Proving isolated or sporadic discriminatory acts by the employer is insufficient to establish a prima facie case of a pattern or practice of discrimination. See Teamsters, 431 U.S. at 336, 97 S.Ct. at 1855.

The Cooper Court explained this distinction, drawing upon its holding in General Telephone Co. of Southwest v. Falcon, 457 U.S. 147, 102 S.Ct. 2364, 72 L.Ed.2d 740 (1982). In Falcon, the Court held that an individual's claim that he was not promoted because of racial animus did not necessarily make him an adequate representative of a class composed of persons who had allegedly been refused employment for discriminatory reasons. Id. The Court noted the "wide gap" between individual claims and class claims of discrimination. Falcon, 457 U.S. at 157, 102 S.Ct. at 2370.

[T]o bridge that gap [the claimant] must prove much more than the validity of his own claim. Even though evidence that he was passed over for promotion when several less deserving whites were advanced may support the conclusion that respondent was denied the promotion because of his national origin, such evidence would not necessarily justify the additional inferences (1) that this discriminatory treatment is typical of petitioner's promotion practices, (2) that petitioner's promotion practices are motivated by a policy of ethnic discrimination that pervades petitioner's . . . division, or (3) that this policy of ethnic discrimination is reflected in petitioner's other employment practices, such as being in the same way manifested in the promotion practices.

Cooper, 467 U.S. at 876-77, 104 S.Ct. at 2800 (quoting Falcon, 457 U.S. at 157-58, 102 S.Ct. at 2370-71).

The Cooper Court further observed that "a class plaintiff's attempt to prove the existence of a companywide policy [(i.e., a pattern or practice)], or even a consistent practice within a given department, may fail even though discrimination against one or two individuals has been proved." 467 U.S. at 878, 104 S.Ct. at 2800. This observation reinforces the distinction between individual and class claims. It is quite proper, the Court noted, to recognize that a "court must be wary of a claim that the true color of a forest is better revealed by reptiles hidden in the weeds than by the foliage of countless freestanding trees." Id. at 879-80, 104 S.Ct. at 2801 (quoting NAACP v. Claiborne Hardware Co., 458 U.S. 886, 934, 102 S.Ct. 3409, 3436, 73 L.Ed.2d 1215 (1982).

In order to prove that the employer has engaged in a pattern or practice of discrimination, then, the plaintiff must show that there is regular, purposeful, less-favorable treatment of a protected group. See, e.g., Teamsters, 431 U.S. at 335, 97 S.Ct. at 1854 (government's theory was "that the company . . . regularly and purposefully treated Negroes and Spanish-surnamed Americans less favorably than white persons."). We have explained that

plaintiffs in a pattern-or-practice government or class discrimination claim have the initial burden of showing that the unlawful discrimination was the employer's regular policy. A plaintiff must `establish a prima facie case that such a [discriminatory] policy existed.' Teamsters, 431 U.S. at 360 [ 97 S.Ct. at 1867]. In other words, a plaintiff would have to prove `by a preponderance of the evidence that [the] discrimination was the company's standard operating procedure — the regular rather than the unusual practice.' Id. at 336, 97 S.Ct. at 1855.

EEOC v. Sears, Roebuck Co., 839 F.2d 302, 354-55 (7th Cir. 1988). Once plaintiffs establish a broad pattern or policy of discrimination, courts have found it is reasonable to infer that the employer discriminated against particular individuals, unless the employer can prove otherwise. Thus, once the pattern is established, the burden shifts to the defendant to show there was no discrimination. The Court explained the logic of shifting burdens in Title VII cases in Teamsters, using Franks v. Bowman, 424 U.S. 747, 96 S.Ct. 1251, 47 L.Ed.2d 444 (1976) as an example:

[In Franks, the class] alleged a broad-based policy of employment discrimination, upon proof of that allegation there were reasonable grounds to infer that individual hiring decisions were made in pursuit of the discriminatory policy and to require the employer to come forth with evidence dispelling that inference.[45]

[45] The holding in Franks that proof of a discriminatory pattern and practice creates a rebuttable presumption in favor of individual relief is consistent with the manner in which presumptions are created generally. Presumptions shifting the burden of proof are often created to reflect judicial evaluations of probabilities and to conform with a party's superior access to the proof.

Teamsters, 431 U.S. at 359, 97 S.Ct. at 1866. Accord Price Waterhouse v. Hopkins, 490 U.S. 228, 246 n. 11, 109 S.Ct. 1775, 1788 n. 11, 104 L.Ed.2d 268 (1988).

In order to support the "judicial evaluation" (that it is more probable than not that an employer that routinely discriminates, discriminated against a particular individual), there must be significant evidence of the alleged routine. In Teamsters, for example, the government first presented statistical evidence showing grave disparities between the percentages of blacks in the population at large, those employed by the defendant, and those employed in particular positions. 431 U.S. at 337-38, 97 S.Ct. at 1855-56. In fact, no blacks were employed in some positions until after the suit was filed. Id. at 337-38, nn. 17-18, 97 S.Ct. at 1855-56, nn. 17-18. The government then supported this evidence with individual testimony recounting over forty specific instances of discrimination. Id. The Court held that this evidence was sufficient to carry the government's burden of proof. Id. at 337, 97 S.Ct. at 1855.

Other courts considering what evidence is necessary to show that an employer routinely and purposely discriminated have also required substantial proof of the practice. Thus, in Chisholm v. United States Postal Service, 665 F.2d 482, 495 (4th Cir. 1981), twenty class members testified about individual discrimination to support the statistical evidence. Without significant individual testimony to support statistical evidence, courts have refused to find a pattern or practice of discrimination. One court has warned that "the definition of a pattern or practice is not capable of a mathematical formulation. . . ." Ste. Marie v. Eastern R. Ass'n, 650 F.2d 395, 406 (2d Cir. 1981). In Ste. Marie, the court found seven individual discriminatory acts coupled with problematic statistical evidence were insufficient to support a finding of a pattern or practice of discrimination. Id. See also Cooper v. Federal Reserve Bank of Richmond, 467 U.S. 867, 104 S.Ct. 2794, 81 L.Ed.2d 718 (1983), (citing with approval the opinion of the court below, EEOC v. Federal Reserve Bank of Richmond, 698 F.2d 633, 643-44 (4th Cir. 1983) contrasting 20 or 40 class members testifying of class discrimination (sufficient) to three or even seven incidents of discrimination over a period of years (insufficient)); Goff v. Continental Oil Co., 678 F.2d 593, 597 (5th Cir. 1982) (even if all three witnesses' accounts of racial discrimination were true, this evidence would not have been enough to prove a pattern or practice of company-wide discrimination).

Evaluating the sufficiency of the evidence supporting the claim is not the only difficult task faced by a reviewing court in an ADEA case. One problem that arises when courts borrow procedures designed to enforce Title VII and apply them to ADEA cases is that the fit can be imprecise. Title VII cases are tried to a judge whereas ADEA cases are tried to a jury. Cf. Loeb v. Textron, Inc., 600 F.2d 1003 (1st Cir. 1979) (discussing adjustment of Title VII's McDonnell Douglas framework to ADEA). Altering definitions and procedures can be a complicated and confusing process. Title VII permits the Attorney General to bring civil actions to stop widespread practices that prevent protected groups from fully exercising their rights. 42 U.S.C. § 2000e-6(a). Individuals harmed by the challenged practice may join the government's suit.

ADEA has no parallel provision, but courts nevertheless have adopted the pattern-or-practice terminology and the shifting burden of persuasion to ADEA actions. EEOC v. Western Electric, 713 F.2d 1011 (4th Cir. 1983); EEOC v. Sandia, 639 F.2d 600 (10th Cir. 1980). Moreover, in some cases plaintiffs have attempted to prove that the employer engaged in a pattern of discrimination in order to bolster their individual claims. See, e.g., Graefenhain v. Pabst Brewing Co., 827 F.2d 13 (7th Cir. 1987). In Graefenhain, the two plaintiffs presented evidence intended

to show that Pabst's reduction in force explanation was merely a pretext for terminating older, higher salaried employees. In other words, [plaintiffs] attempted to show that their termination was part of a plan to cut costs by replacing older, higher salaried employees with inexperienced, younger employees who could be paid less.

Id. at 17. We must reiterate, however, that isolated, sporadic discrimination is not sufficient. Because of the seriousness of the charge, it is essential that plaintiffs adequately support allegations that an employer routinely engages in wide-spread discriminatory practices.

Finally, as we review the course of the proceedings below, we bear in mind the appropriate standard of review. In reviewing a jury's verdict, we determine whether there was evidence "sufficient for a reasonable jury to conclude that age was a determining factor" in the employer's decision to fire the plaintiff. Holzman v. Jaymar-Ruby, Inc., 916 F.2d 1298 (7th Cir. 1990). All evidence is viewed in the light most favorable to the successful party, and all evidentiary conflicts and reasonable inferences are drawn in that party's favor. Id. See also Brown v. M M/Mars, 883 F.2d 505, 507 (7th Cir. 1989).

One of the difficulties of adapting the Title VII pattern-or-practice framework to ADEA cases becomes evident when we compare standards of review. Because a judge tries Title VII cases, we review those verdicts under a clearly erroneous standard. Sears, 839 F.2d at 309. The district court's factual findings in Title VII cases are subject to the same standard of review. Bazemore v. Friday, 478 U.S. 385, 106 S.Ct. 3000, 92 L.Ed.2d 315 (1986). A district court's factual findings are accorded special deference when they are based upon statistical evidence. Sears, 839 F.2d at 310 (citing Soria v. Ozinga Brothers, Inc., 704 F.2d 990, 995 n. 6 (7th Cir. 1983)). We need not resolve the question of whether a jury's evaluation of statistical evidence is entitled to the same special deference. In the instant appeal, we are convinced that there was simply not enough evidence to support the pattern-or-practice verdict.

C. Analysis

Neither the McDonnell Douglas proof structure system, nor the pattern or practice terminology is talismanic designations. The McDonnell Douglas formula, though appropriate and workable, was not intended to be "rigid, mechanical, or ritualistic," nor the exclusive method for proving a claim of discrimination. Loeb v. Textron, Inc., 600 F.2d 1003, 1017 (1st Cir. 1979). The trial court seemed to be mesmerized first by the pattern or practice terminology in Sobin's motion to consolidate, and then by the Oxman incarnation of McDonnell Douglas. We believe that the district court erred in its management of the two cases below. Obviously trying ADEA cases, with the somewhat convoluted burden-shifting procedures, can be a complicated task, and any missteps are understandable.

In this case, the Hany plaintiffs brought an action on behalf of themselves and all others similarly situated. They alleged simply that beginning in January 1983, G.E. "began discriminating against older G.E. employees on account of their age." Complaint, Appellant's S.A. at 10. The Hany complaint went on to allege that G.E. gave employees false performance ratings, trumped up disciplinary charges against older employees, and terminated "many of its older employees while retaining younger employees whose performance and background were inferior to the older employees who were laid off or terminated." Id. at 11. The Hany complaint also alleged that G.E. applied its "lay off policies and practices" discriminatorily to older employees. Id. The complaint asserted that the violations were willful. Basically, then, the Hany plaintiffs asserted that G.E. pursued a policy of age discrimination in its reduction in force.

The Sobin plaintiffs, suing individually, complained that G.E. willfully violated ADEA by discharging them because of age. These discharges occurred in 1985. Sobin, Monkman, and Evelsizer did not allege any other facts or violations as to any other employees, to each other, or to older employees as a group. Despite this lack, their attorney (the three were represented by the same attorney) moved to consolidate the trials based upon an asserted common question of fact.

A district court's decision to consolidate cases is subject to review only for an abuse of discretion. See United States v. Knauer, 149 F.2d 519 (7th Cir.), aff'd, 328 U.S. 654, 66 S.Ct. 1304, 90 L.Ed. 1500 (1946). We are not sure the Hany and Sobin cases should have been consolidated. In fact, we believe that the district court may have abused its discretion in the matter. The Sobin plaintiffs never alleged any company-wide discriminatory policy or practice — only that each individually was fired because of his age. The Hany plaintiffs loosely pleaded a pattern or practice of discrimination; their complaint alleged that G.E. generally discriminated against older workers. Their allegations were confined to the two-year period when they were fired or demoted. The difficulties posed by the consolidation are compounded because the several complaints involved different time periods. The Hany plaintiffs were allegedly injured in 1983 and 1984, the Sobin plaintiffs in 1985. We conclude, because of the different allegations and time frames, the district court abused its discretion by consolidating these actions.

Even if the consolidation itself was not an abuse of discretion requiring reversal, we do not believe plaintiffs (particularly the Sobin plaintiffs) presented sufficient evidence to support the jury's finding that a company-wide pattern of discrimination existed at G.E. between 1982 and 1985. Certainly the Sobin plaintiffs, who were terminated in 1985, failed to present a sufficient evidentiary basis for the jury's finding that G.E. engaged in a pattern of discrimination in 1985. There was no statistical evidence supporting this claim. Although statistical evidence is not strictly required, in this case the Hany plaintiffs' evidence compares figures for 1983 and 1984 to figures from 1985 to show the significance of the age disparities in 1983 and 1984.

Further, almost all the anecdotal evidence cited by appellees concerns alleged individual discrimination occurring in 1983 and 1984. See Appellees' Br. at 6-9. Even if we were convinced that the evidence presented was sufficient to show a discriminatory practice in 1983 and 1984, this evidence provides scant support for the allegation that G.E. discriminated in 1985.

As we noted in EEOC v. Sears, Roebuck Co., 839 F.2d 302, 329 (7th Cir. 1988), statistical evidence cannot serve as a basis for proving discrimination beyond the time period analyzed. In Sears, we refused to allow plaintiffs to use data from a two-year hiring period to show discrimination over a seven-year period. Id. Although the extrapolation is not so great in this case — two years of data for five years of practice — we still believe the data an inadequate measure of G.E.'s practices over the five-year period.

Moreover, the statistical evidence itself was not overwhelming even for the years analyzed. "Although statistics are useful, they are not magical. The finder of fact must always consider the extent to which the statistical evidence reasonably supports an inference of discrimination." Howard Eglit and Martin Malin, 3 Age Discrimination § 17.65 at 17-208 (1989).

The usefulness of statistical evidence, of course, depends on the surrounding facts in any particular case. Teamsters, 431 U.S. at 340, 97 S.Ct. at 1857. The Hany plaintiffs' expert applied a chi-square and a logistic regression analysis to the figures for G.E.'s reduction in force. He also compared the relative percentages of workers affected by the RIF, and found it was more likely for older workers to be affected than younger ones. Some courts insist that experts apply a standard deviation analysis to population data. See, e.g., Moultrie v. Martin, 690 F.2d 1078 (4th Cir. 1982) (interpreting Casteneda v. Partida, 430 U.S. 482, 496-97 n. 17, 97 S.Ct. 1272, 1281 n. 17, 51 L.Ed.2d 498 (1976) and Hazelwood School District v. United States, 433 U.S. 299, 308-09 n. 19, 97 S.Ct. 2736, 2744 n. 19, 53 L.Ed.2d 768 (1977) to require it.). We do not believe that it is necessary to go so far, so long as the statistical analysis properly controls for factors that might lead to differences in raw percentages.

Straight percentage comparisons which show the relative representation of a group in two populations alone may not be statistically significant. See Moultrie, 690 F.2d at 1082. Some method of measuring the disparities, such as the standard deviation analysis is therefore necessary. The regression analysis used by the Hany expert purports to control for these factors. Indeed, we noted in EEOC v. Sears, Roebuck Co., 839 F.2d at 324, that "[m]ultiple regression analyses, designed to determine the effect of several independent variables on a dependent variable, which in this case is hiring, are an accepted and common method of proving disparate treatment claims."

Regardless of our view of the statistical evidence provided by the Hany plaintiffs, they have failed to provide the amount of anecdotal evidence claimants in other cases have supplied to support a finding of a pattern or practice of discrimination. As we noted above, more than forty examples of individual discrimination were provided in Teamsters, 431 U.S. at 337, 97 S.Ct. at 1855. In Chisholm, 665 F.2d at 495, twenty employees testified about individual discrimination. Further, the Hany plaintiffs benefitted from the presentation of the Sobin plaintiff's anecdotal evidence. See, e.g., Testimony of Sobin, Trial Transcript, Vol. 6 at 1031. We are not convinced that without the Sobin evidence, the jury would have returned a verdict finding a pattern of discrimination. We do not hold that in every case where improperly joined parties testify about alleged discrimination that the verdict as to all plaintiffs must be reversed because of prejudice to the defendant. Cf. Lockhart v. Westinghouse Credit Corp., 879 F.2d 43, 51 (3d Cir. 1989) (discussing opt-in plaintiffs who were not "similarly situated"). Still, due to the difficulties with the consolidation, the meager anecdotal evidence adduced to support the statistical data, and because the jury verdict was not confined to either of the relevant time-frames, we believe that it cannot stand. Although not unmindful of the significant resources dedicated to the trial below, that reason alone cannot persuade us to preserve the verdict.

The district court had an opportunity to cut its losses when G.E. moved for judgment notwithstanding the verdict. "A judgment n.o.v. is properly granted where the evidence presented, combined with all reasonable inferences that may be drawn from it, is insufficient . . . when viewed in the light most favorable to the non-moving party." Graefenhain v. Pabst Brewing Co., 827 F.2d 13, 15 (7th Cir. 1987). "We do . . . weigh the evidence to the extent of determining whether the evidence is substantial; a mere scintilla of evidence will not suffice." Id. As we have discussed, we believe there was not substantial evidence to support the jury's finding that G.E. engaged in a pattern or practice of discrimination from 1982 until 1985.

We also are forced to note that the district court compounded the problems in this litigation when it failed to accord the jury's verdict the proper weight. Once the government or a class of individual plaintiffs has demonstrated that a pattern or practice of discrimination exists, they establish a presumption that individual class members have been discriminated against. Franks v. Bowman, 424 U.S. at 772, 96 S.Ct. at 1268. In other words, the proof of the pattern or practice supports an inference that any particular employment decision, during the period in which the discriminatory policy was in force, was made in pursuit of that policy. Teamsters, 431 U.S. at 362, 97 S.Ct. at 1868. Thus, the burden shifts to the employer to show that this particular employee was not fired or demoted pursuant to the discriminatory policy.

Although the district court initially applied the proper shift in the burden, it reversed its position on the effect of the first jury's verdict between the trials. In response to a motion for clarification, it determined that the plaintiffs would have to start afresh in the individual trials, and establish a classic McDonnell Douglas prima facie case under Oxman. It mistakenly held that "[p]laintiffs must establish a prima facie case and setout [sic] the elements." Order of May 10, 1989, at 4, Appellant's App. at 62.

Because the first jury verdict cannot stand, we do not believe the second jury's verdict can be salvaged either. Because the first verdict was not adequately supported, the plaintiffs were not entitled to the instruction based upon the first jury's finding. Arguably, the verdict in G.E.'s favor against Brickey could stand despite the problems with the litigation because he lost despite receiving the undeserved instruction. Nevertheless, both trials were fraught with difficulties, and we are not comfortable with this thumbnail determination of the relative equities. Therefore, we reverse both verdicts and remand the cases for new trials.

Consolidation of cases to determine common factual questions does not merge the suits or change the rights of the parties. See Miller v. United States Postal Serv., 729 F.2d 1033, 1036 (5th Cir. 1984); Midwest Community Council, Inc. v. Chicago Park Dist., 98 F.R.D. 491 (N.D.Ill. 1983). We believe, therefore, when the cases are retried, the suits of the Hany and Sobin plaintiffs should be separated. The Hany plaintiffs are free to bring a common action to attempt to show a pattern or practice in 1983 and 1984; if they succeed in establishing such a pattern, they would be entitled to a shift in the burden of proof as explained in Teamsters and its progeny. We do not believe the Sobin plaintiffs, because they did not allege any pattern or policy of discrimination in their complaint, can pursue a similar strategy. But we leave that determination to the district court. Again, we are aware of the burden this litigation has placed on the district court and the parties. We regret the need to start the process again. Nevertheless, for the foregoing reasons, the decisions of the district court are

REVERSED AND REMANDED.