Adams v. Indiana Bell Telephone Co. (2 F. Supp. 2d 1077, 1998)

ORDER ON MOTIONS TO STRIKE AND FOR SUMMARY JUDGMENT

McKINNEY, District Judge.

Currently pending before the Court are two dispositive motions on the claims of the plaintiffs brought under the Age Discrimination in Employment Act (“ADEA”), 29 U.S.C. § 621 et seq., as well as several related motions. The plaintiffs have raised both individual and pattern or practice claims of age discrimination, which they maintain occurred during a corporate-wide resizing process in 1992 and 1993 at Ameritech Services, Inc. (“ASI”). The related motions are three motions to strike the evidence plaintiffs plan to offer through their statistics, labor economics, and industrial psychology experts. If granted, the motions to strike will have the effect of excluding a substantial portion of the evidence on which plaintiffs rely in opposition to the pending summary judgment motions. Thus, the Court will first address the issues raised by these motions.

I. MOTIONS TO STRIKE

A. STANDARDS

According to Rule 12(f) of the Federal Rules of Civil Procedure:

[U]pon motion made by a party within 20 days after the service of the pleading upon the party or upon the court’s own initiative at any time, the court may order stricken from any pleading any insufficient defense or any redundant, immaterial, impertinent, or scandalous matter.

Generally, motions to strike are considered a drastic remedy, and are strongly disfavored. Federal Nat’l Mortgage Ass’n v. Cobb, 738 F.Supp. 1220, 1224 (N.D.Ind.1990); see also New York v. Almy Brothers, Inc., 971 F.Supp. 69, 72 (S.D.N.Y.1997). Accordingly, such motions “are ordinarily not granted unless the language in the pleading at issue has no possible relation to the controversy and is clearly prejudicial.” Cobb, 738 F.Supp. at 1224.; Abdulrahim v. Gene B. Glick Co., 612 F.Supp. 256, 260 n. 1 (N.D.Ind.1985). A trial court, however, clearly has discretion to grant a well-taken motion to strike. Mirshak v. Joyce, 652 F.Supp. 359, 370 (N.D.Ill.1987). In exercising that discretion, the Court must consider the value of the pleading in light of other rules of procedure.

Rule 56(e) requires that affidavits supporting or opposing a motion for summary judgment set forth facts that would be admissible in evidence, and show “affirmatively that the affiant is competent to testify to the matters” presented. Fed.R.Civ.P. 56(e). Courts may permit an affidavit to be supplemented or opposed by depositions, answers to interrogatories, or further affidavits. Id. Although the evidence presented in opposition to or support of summary judgment does not have to be in admissible form, it must be admissible in content. See Winskunas v. Birnbaum, 23 F.3d 1264, 1267-68 (7th Cir.1994) (evidence must be of “evidentiary quality,” meaning that a change in form but not in content would make the evidence admissible at trial). Examples of such evidence include certified documents or sworn testimony, such as depositions or affidavits. Id. at 1267.

To be admissible, testimony must be relevant to the facts the party intends to prove under the legal theory that has been chosen to support the party’s claim. See Charles A. Wright & Kenneth W. Graham, Federal Practice And Procedure: Evidence § 5162. “Relevance is a relationship between the evidence offered and the fact it is supposed to prove.” Id. A piece of evidence may be excluded as irrelevant in one of two ways. First, it may provide satisfactory proof of a fact in dispute, but the fact for which it is offered is not a material fact in the case. Id Second, it may be directed at a material fact, but not have any value as proof of that fact. Id. A material fact is a “fact that is of consequence to the determination of the action.” Fed.R.Evid. 401.

B. RELEVANCE AND MATERIALITY

1. Relevance

Relevant evidence is evidence that has any tendency to make the existence of a fact of consequence to the determination of the matter (i.e. “material fact”) more probable or less probable than it would be without the evidence. Fed.R.Evid. 401. The term “any tendency” is so broad that it has the effect of allowing admission of evidence that has only the slightest bit of probative worth. Charles A. Wright & Kenneth W. Graham, Federal Practice § 5165. In fact, the Seventh Circuit has called the definition of relevance under Rule 401 “expansive.” United States v. Pollard, 790 F.2d 1309, 1312 (7th Cir.1986), overruled on other grnds. by United States v. Sblendorio, 830 F.2d 1382 (7th Cir.1987). Generally, all relevant evidence is admissible, and irrelevant evidence is inadmissible. Fed.R.Evid. 402. The weight of the evidence, however, is for the jury to decide. Judges may not weigh the evidence when determining its relevance. Thus, arguments about the quality or probative value of the evidence advanced to defeat its admissibility are inapplicable to the inquiry.

Once it is found to be relevant, some evidence may still be excluded if its probative value is substantially outweighed by the danger of unfair prejudice, if it brings about confusion of the issues, or if it is misleading for the jury. Fed.R.Evid. 403. It is in this process that the judge may weigh the evidence. Rule 403 presumes a finding that the evidence in question makes some fact in dispute more or less probable. If not, the court may not consider it under Rule 403. By probable it is meant “uncertain, but likely to be true.” Wright & Graham § 5165. The court measures probative value by weighing the evidence in terms of its relationship to the truth or falsity of propositions of fact sought to be proved. Id.

Rule 403 recognizes the discretion of a judge to exclude evidence, but seeks to,limit the exercise of that discretion. Id. § 5212. Even though the court’s role in determining admissibility is limited, it must “weigh” the evidence to determine if its probative value exceeds any procedural problems it creates. See Wright & Graham § 5165. In doing so, the judge must be careful not to usurp the role of the jury under the guise of determining probative value. A judge may exclude evidence only after balancing the “competing considerations” listed in the rule and finding that the procedural harm caused by the evidence substantially exceeds its probative value. Id.

It is this balancing or weighing that furthers the policy behind Rule 403 of limiting judicial discretion in the admission and exclusion of evidence. Id. § 5214. The probative value of the evidence must be substantially outweighed by its prejudicial effect, or by the danger that it might mislead or confuse the jury, before the court would have discretion to exclude it. Id. If the balance goes against the probative worth of the evidence, the court may, but is not required to, exclude the evidence. Id. Thus, application of Rule 403 involves a two-step process: balancing, then exercising discretion.

When balancing evidence, the court should be mindful of the nature of the evidence. For example, evidence that.proves an ultimate fact without any inferential step, except for a credibility determination, is deemed at this stage to be more probative than evidence that would require an inference to be drawn before the ultimate fact may be supported. Id. § 5214. The weighing process under Rule 403 requires the judge to assume the credibility of a witness, because to do ■ otherwise would usurp the jury’s function. Id. Thus, the probative worth of eyewitness testimony as to an ultimate fact is presumed to be 100%. Id. Nevertheless, very little evidence is of such caliber. More often a witness or document provides circumstantial evidence, or evidence that requires an inferential step before it can prove an ultimate fact. Some evidence may require several inferential steps, which lessens its probative value over evidence requiring none or one. Evidence that requires an inference before it can prove a material fact may be countervailed by offerings of the other party that make the inference less reasonable.

Probative value is not a precise mathematical term. Id. Rather it is a reflection of the economic concepts of supply and demand. Id. The need for the evidence is assessed against the possible supply of alternative means of proving the fact. See id.

2. Materiality

Evidence may be entirely relevant to an issue in dispute, but the issue itself may not be material to the action. Such evidence would be deemed irrelevant, but on the basis of the lack of materiality of the ultimate fact it is intended to prove. Materiality is determined with reference to the substantive law being applied, and it depends on the legal theories the plaintiffs seek to employ to support their claims.

Here, the plaintiffs have brought their claims under the Age Discrimination in Employment Act (“ADEA”). 29 U.S.C. §§ 621 et seq. The ADEA provides in part:

the purpose of this chapter [is] to promote employment of older persons based on their ability rather than age; to prohibit arbitrary age discrimination in employment;

It shall be unlawful for an employer—(1) to discharge any individual or otherwise discriminate against any individual ... because of such individual’s age; (2) to limit, segregate, or classify employees 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 age;

It shall not be unlawful for an employer ... (1) to take any action otherwise prohibited ... [in] this section where age is a bona fide occupational qualification ... or where the differentiation is based on reasonable factors other than age ....

29 U.S.C. §§ 621(b), 623(a)(1), 623(a)(2), 623(f)(1).

Employees may enforce the ADEA against their employer either in their individual capacities or as nominal plaintiffs in an opt-in class action. See 29 U.S.C. § 626(b) (incorporating 29 U.S.C. § 216(b) by reference); Grayson v. K Mart Corp., 79 F.3d 1086, 1096 (11th Cir.), cert. den., — U.S. —, 117 S.Ct. 435, 136 L.Ed.2d 332 (1996). The employees here have done both. To state an individual claim for age discrimination under the ADEA, plaintiffs must offer direct evidence of discrimination, or use the indirect, burden-shifting, method of proof described in McDonnell Douglas v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973). Using the direct method, a plaintiff may prove discrimination by providing evidence that the fact-finder may interpret as the employer’s acknowledgment of discriminatory intent. Chiaramonte v. Fashion Bed Group, Inc., 129 F.3d 391, 396 (7th Cir.1997). For it to succeed, such evidence must clearly demonstrate the motivation of the person who made the contested employment decision. Id. (citing Cheek v. Peabody Coal Co., 97 F.3d 200, 203 (7th Cir.1996)). To succeed on the individual age discrimination claim, a plaintiff must prove that age was the determining factor in that decision. Id.

That task is more difficult in the context of a corporate-wide workforce resizing and restructuring, as happened here. See Hartley v. Wisconsin Bell, Inc., 124 F.3d 887, 889 (7th Cir.1997); Testerman v. EDS Technical Products Corp., 98 F.3d 297, 303-04 (7th Cir.1996) (in a ease involving a reduction in force (“RIF”), a claimant must show that age “tipped the balance” in favor of his or her discharge). When an employer must reduce its workforce, “it is foreordained that some employees will lose their jobs; and ... the defendant inevitably will be able to point to some imperfection as a reason for the discharge.” Testerman, 98 F.3d at 304. However, if the employer fires an older employee who would otherwise have survived the RIF but for his age, the ADEA is violated. Id. In this context, the court must deal with “small gradations, with an employer’s subjective comparison of one employee to another, and it is incumbent upon [the court] to remember that what is at issue is not the wisdom of the employer’s decision, but the genuineness of the employer’s motives.” Id.

In Chiaramonte, the plaintiff attempted to prove motivation of the decision-maker, president John Elting, by offering evidence of statements made by the company’s chief executive officer, John Singer. Chiaramonte, 129 F.3d at 396. Singer allegedly told Chiar-amonte that “age had to be a factor” in the decision to terminate him. Id. at 397. However, Chiaramonte had no evidence that Elt-ing and Singer had discussed the reasons for Chiaramonte’s termination, nor could he provide evidence that Elting made statements that could be interpreted as age-related animus. Id. Elting’s deposition testimony indicated that he was the sole decision-maker, although he does not unambiguously deny having discussed the decision with Singer. Singer, however, unequivocally denied discussing the decision with Elting. Id. “Statements by a non-decision-maker that amount to mere speculation as to the thoughts of the decision-maker are irrelevant to [the] inquiry.” Id.

Chiaramonte also offered as evidence of discrimination comments made by Debbie Lunn, a lower-level supervisor, that the company planned to get rid of “all you old people.” Id. Lunn, who often socialized with Elting, denied making the statement. Id. Even if she had, it would have no probative value the court said, because Lunn had no control over the decision to discharge Chiara-monte. Id. “[A]ctions and comments by employees not involved in a discharge decision cannot provide a basis for charging other employees with discrimination.” Id.

If an employee cannot provide direct evidence of discrimination, he or she may adopt the traditional burden-shifting method of proof. However, the McDonnell Douglas method is not an “efficient formula” for large-scale workplace restructuring cases. Hartley, 124 F.3d at 889. Ordinarily, this approach is based on the legitimate assumption that the plaintiff was doing fine in his or her job, then lost it, and was replaced by someone younger, or white, or male, or not disabled. Hartley, 124 F.3d at 889. Applying that rationale to a RIF case is problematic because it “ignores the circumstances surrounding [the] termination and it presumes [the plaintiff] was singled out for termination.” Id. When an employer must reduce its workforce, employees who were doing fine in their jobs may lose them. Id.; see also Sheehan v. Daily Racing Form, Inc., 104 F.3d 940, 943 (7th Cir.), cert. den., — U.S. —, 117 S.Ct. 2480, 138 L.Ed.2d 989 (1997). Nevertheless, adapting this method to the RIF context, an employee may establish a prima facie case by showing:

1. Membership in a protected age group;

2. Satisfactory job performance;

3. An adverse employment action, such as a discharge; and

4. That substantially younger, similarly-situated employees were treated more favorably.

Chiaramonte, 129 F.3d at 398 (citing O’Connor v. Consolidated Coin Caterers Corp., 517 U.S. 308, 116 S.Ct. 1307, 1310, 134 L.Ed.2d 433 (1996)).

By establishing these elements of a prima facie case of age discrimination, the employee shifts the burden of production to the defendant, who must come forward with a legitimate, non-discriminatory explanation for the decision. Provided that the defendant succeeds in meeting this. obligation, the presumption of discrimination created by the prima facie case is rebutted. The burden then shifts back to the plaintiff to prove that the employer’s stated reasons are pretextual, either by showing that they are unworthy of credence or that a discriminatory reason more likely motivated the decision. Id. The ultimate burden of persuasion on the issue of discrimination, however, always lies with the plaintiff. Id.

A crucial difference in a RIF case, is that the defendant can meet its burden of production by explaining the reasons behind the downsizing and the process used to select employees for discharge. Thus, the burden quickly returns to the plaintiff to show that those reasons were a pretext for discrimination, or that the process was deliberately designed to weed out older workers. Unlike an age discrimination ease when no reduction in force occurred, the question is not whether the employee was or was not performing satisfactorily, but whether the employer “honestly believed that [the plaintiff] was the weakest member of the department.” Sirvidas v. Commonwealth Edison Co., 60 F.3d 375, 378 (7th Cir.1995). Plaintiffs lose at the pretext stage if the court finds that the defendant “honestly believed in the nondiscriminatory reasons it offered, even if those reasons are foolish or trivial or even baseless.” Hartley, 124 F.3d at 890.

Another difference in the RIF context is seen when plaintiffs sue on behalf of those similarly situated based on an alleged pattern or practice of age discrimination. In that context, the prima facie case is established by showing that unlawful discrimination has been the company’s standard operating procedure, the regular rather than the unusual practice. EEOC v. Chicago Miniature Lamp Wks., 947 F.2d 292, 297 (7th Cir.1991); Coates v. Johnson & Johnson, 756 F.2d 524, 532 (7th Cir.1985). It usually consists of “statistical evidence demonstrating substantial disparities in the application of employment actions as to ... the protected group, buttressed by evidence of general policies or specific instances of discrimination.” Coates, 756 F.2d at 532 (citing International Bhd. of Teamsters v. United States, 431 U.S. 324, 336, 360, 97 S.Ct. 1843, 52 L.Ed.2d 396 (1977)). Evidence of discrimination against particular individuals may form part of the proof for a pattern or practice claim. However, to elevate individual claims to the level of a pattern or practice case requires evidence that the discriminatory treatment is typical of the employer’s employment practices, or that those practices are motivated by a policy of discrimination against the protected class that pervades the entire business. King v. General Elec. Co., 960 F.2d 617, 623 (7th Cir.1992). Another way to satisfy the evidentiary requirement would be to show that a policy of discrimination that is reflected in the employer’s other employment practices, is the same as reflected in the practice at issue. Id. The employee must prove by “significant evidence” the regular, purposeful, less-favorable treatment of members of a protected group. Id.

Once a pattern or practice prima facie case is established, the burden shifts to the defendant to show that the plaintiff’s statistical proof is inaccurate or insignificant, or that a nondiseriminatory explanation exists for the disparity. Coates, 756 F.2d at 532-33. The strength of the evidence needed to accomplish this depends on the strength of the plaintiffs proof. Id. Interestingly, a defendant has two possible courses of action open for rebutting the prima facie case. First, it may offer proof that negates the pattern or practice, such as evidence of the inaccuracy or invalidity of the plaintiffs statistics. Id. In fact, the most effective way to rebut a statistically-based prima facie case is to offer more accurate statistics. Id. at 544. By offering statistics that include an omitted factor and showing that the statistics no longer indicate discrimination, a defendant shifts the burden back to the plaintiff. Likewise, a defendant may offer a non-discriminatory explanation for the apparently discriminatory overall result. Id. at 537. Second, the defendant may proffer a legitimate nondiscriminatory reason for the employment decision made with respect to each class member. Coates, 756 F.2d at 533. Offering a non-discriminatory reason only for the discharge of each named plaintiff will weaken, but not destroy, the class action for a pattern or practice of discrimination. Id.

In light of these burdens and using the direct method to prove either type of claim, the material facts a plaintiff must prove include those that tend to show the employer’s intent or motivation. These could be discriminatory statements or comments made by the decision-maker, or attributed to the decision-maker, or those made by others to which the decision-maker has agreed. Likewise, circumstantial evidence, such as a statistical imbalance in the employer’s workforce, can be material to the inquiry under the direct method of proof. See Perfetti v. First Nat’l Bank of Chicago, 950 F.2d 449, 450 (7th Cir.1991), cert. den., 505 U.S. 1205, 112 S.Ct. 2995, 120 L.Ed.2d 871 (1992) (describing four methods of proving employment discrimination).

However, under the indirect method for individual claims, the material facts are those tending to prove each of the elements of the prima facie case as they relate to an individual plaintiff. Also material is the defendant’s explanation for the decision against each plaintiff and evidence about the plaintiff’s allegations of pretext. For example, evidence that similarly-situated, but substantially younger, workers were retained for the same types of jobs the plaintiffs had been performing would be material to an indirect method of proof. Likewise, direct evidence of a contradiction between the employer’s expressed justification for a decision and documentation created at the time of the decision would be material to the pretext inquiry. Perfetti, 950 F.2d at 451. Circumstantial evidence of pretext, such as showing that a proffered justification was not a genuine job requirement, or that the criteria or rules were inconsistently applied to others, would also be material. Id.

The framework for the indirect method of proof in the pattern or practice type of claim is comparable to that in McDonnell Douglas, except the content of the specific stages will be different. Coates, 756 F.2d at 533. For example, the focus in the class action is on a pattern of discriminatory decision-making. Id. This means that specific allegations of individual discrimination may be material, but not unless the number of instances is significant. Id. On the other hand, strong statistical evidence of a pattern of discrimination may be rebutted with more accurate statistics. Once a prima facie showing is made, each class member will benefit from a presumption of discrimination for their individual claims, which must be rebutted separately by the defendant. See King, 960 F.2d at 623; Coates, 756 F.2d at 532. However, the defendant may rebut a class pattern or practice prima facie case with evidence showing a non-discriminatory reason for the overall disparity, or by challenging the statistics.

C. RELIABILITY AND USEFULNESS

In addition to being relevant and material, scientific or other specialized evidence must also be reliable before it can be admitted into evidence. In Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993), the Supreme Court established a standard for determining the admissibility of scientific evidence and expert testimony. Bradley v. Brown, 42 F.3d 434, 437 (7th Cir.1994); O’Conner v. Commonwealth Edison Co., 13 F.3d 1090, 1106 (7th Cir.), cert. den., 512 U.S. 1222, 114 S.Ct. 2711, 129 L.Ed.2d 838 (1994). To assist district courts with assessing reliability, the Court set up a framework of analysis that fits into the preliminary analysis of evidence required by Rule 104.

1. Daubert Standard

Under the federal rules of evidence, if scientific, technical, or other specialized knowledge will assist the trier of fact to understand the evidence and determine the facts in issue, a witness qualified as an expert by knowledge, skill, experience, training, or education, may testify thereto in the form of opinion or otherwise. Fed.R.Evid. 702. Under the Daubert. framework, the court must assess whether the expert is qualified and will testify to reliable scientific knowledge that is helpful to the trier of fact. The witness is qualified as an expert by “knowledge, skill, experience, training or education,” and the expert’s testimony will be helpful if, among other things, it is relevant. The latter inquiry corresponds to the relevance determination made with respect to non-technieal or scientific evidence under Rule 401. If an expert’s testimony is not based on rehable scientific knowledge, or if it is based on such knowledge but fails to relate to any material facts, then it is not us.eful.. If the expert’s testimony is not helpful, it is not relevant.

The reliability of scientific knowledge is assessed in relation to this non-exclusive list of factors:

1. whether a theory or technique can be or has been tested;

2. whether a theory or technique has been subjected to peer review and publication;

3. the known and potential rate of error; and

4. the “general acceptance” of the theory or technique.

After this preliminary assessment of the scientific validity of the evidence to be offered, the court determines whether the reasoning or methodology properly can be applied to the facts in issue (that is, whether it fits the facts). Daubert, 509 U.S. at 593-94, 113 S.Ct. 2786; Bradley, 42 F.3d at 437.

Even before Daubert the Seventh Circuit had discussed the proper use of expert witness testimony: See Mid-State Fertilizer v. Exchange Nat’l Bk., 877 F.2d 1333 (7th Cir.1989). That discussion guides the court’s assessment of whether the testimony will be useful to the factfinder. Applying Federal Rule of Evidence 705, the court may allow an expert to present “naked” opinions. Id. at 1339. Admissibility, however, does not mean utility. Id. With fact witnesses, to overcome a summary judgment motion a proffered affidavit must “set forth facts” that would be admissible. Fed.R.Civ.P. 56(e). In the case of an expert witness (who is a non-fact witness), the affidavits or reports should set forth “a process of reasoning beginning from a firm foundation.” Mid-State, 877 F.2d at 1339. Expertise is a rational process, and a “rational process implies expressed reasons for judgment.” Id. (citing FPC v. Hope Natural Gas, 320 U.S. 591, 627, 64 S.Ct. 281, 88 L.Ed. 333. (1944)). Moreover, an opinion is only as good as the sources behind it. “An expert who supplies nothing but a bottom line supplies nothing of value to the judicial process.” Id. This is especially true when the opinion makes no sense in light of the circumstances. Judges must look behind an expert’s ultimate conclusion and analyze the. adequacy of its foundation. Id. at 1339. If the factual context makes the claim implausible, then the plaintiff must provide, more persuasive evidence than would otherwise be necessary. Id.

Only after discerning the reliability and usefulness of the proffered evidence, will the court balance its probative value against the danger of a prejudicial effect, confusion of the issues, or other factors found in Rule 403. Even though an expert may be qualified to testify about the results of some scientific study, and the study might prove helpful to the jury, the court may exclude the testimony if it finds that the evidence is of so little probative value that the prejudicial effect will substantially outweigh it. “Expert evidence can be both powerful and quite misleading because of the difficulty in evaluating it. Because of this risk, the judge in weighing possible prejudice against probative force under Rule 403 ... exercises more control over experts than over lay witnesses.” Daubert, 509 U.S. at 595, 113 S.Ct. 2786 (quoting Weinstein, Rule 702 of the Federal Rules of Evidence is Sound: It Should Not Be Amended, 138 F.R.D. 631, 632 (1991)). Finally, if the court is satisfied that the evidence is relevant, reliable, and admissible, it will then consider the evidence as a whole to determine if it suffices to raise a genuine issue of material fact for trial.

The court’s role in determining admissibility of scientific or other technical evidence is that of a gatekeeper. General Elec. Co. v. Joiner, — U.S. —, —, 118 S.Ct. 512, 517, 139 L.Ed.2d 508 (1997); Daubert, 509 U.S. at 589, n. 7, 113 S.Ct. 2786. It performs that role by examining the methodology and reasoning behind the expert’s opinion, in the context of his or her conclusions, which are not “entirely distinct” from methodology. Joiner, 118 S.Ct. at 523.

Trained experts commonly extrapolate from existing data. But nothing in either Daubert or the Federal Rules of Evidence requires a district court to admit opinion evidence which is connected to existing data only by the [bare assertion] of the expert. A court may conclude that there is simply too great an analytical gap between the data and the opinion proffered.

Id. (citing Turpin v. Merrell Dow Pharm., Inc., 959 F.2d 1349, 1360 (6th Cir.1992)). The district court in Joiner had done precisely that, and the Supreme Court ruled that it was error for the Eleventh Circuit Court of Appeals to reverse that finding. Id. 118 S.Ct. at 519.

Further, Daubert does not say that a “failure to adhere to the customary methods for conducting a particular kind of scientific inquiry is irrelevant to the admissibility of the scientist’s testimony. On the contrary, ... it is relevant.” Braun v. Lorillard Inc., 84 F.3d 230, 235 (7th Cir.), cert. den., — U.S. —, 117 S.Ct. 480, 136. L.Ed.2d 375 (1996). If an expert departs from the generally accepted methods in her field, the court may appropriately insist that the departure be grounded in “demonstrable and scrupulous adherence to the scientist’s creed of meticulous and objective inquiry.” Id. Braun was a case in which the plaintiffs attorney hired an expert who specialized in detecting asbestos in buildings, after the experts he had hired who specialized in detecting it in humans had failed to achieve the desired results. . Id. The court stated that this is the kind of abuse Daubert sought to prevent. Id.

In sum, as the Seventh Circuit recently noted, courts must determine if the proffered evidence is “genuinely scientific, as distinct from being unscientific speculation offered by a genuine scientist.” Rosen v. Ciba-Geigy Corp., 78 F.3d 316, 318 (7th Cir.), cert. den., — U.S. —, 117 S.Ct. 73, 136 L.Ed.2d 33 (1996). This does not mean that judges must determine the essence of “real” science. Instead, “the goal is to make sure that when scientists testify in court they adhere to the same standards of intellectual rigor that are demanded in their professional work.” Id. at 318.

2. Use of Statistics

Statistical evidence is a type of “other specialized knowledge” that may be useful in a discrimination case, but only when the quality of the data can be verified. See Federal Judicial Center, Reference Manual on Scientific Evidence 341 (West Pub.1994). That is because “statistics are so manipulable that some skepticism may be justified” when they are offered in the evidentiary context. Coates, 756 F.2d at 529. Verification involves determining if the measurement process was reliable and valid, and whether the measurements were recorded accurately. Manual on Scientific Evidence 341. Reliable statistical evidence is that which was obtained from a measure or model that produces consistent results that can be reproduced repeatedly. Id. The measuring instrument or model should be capable of being validated, that is, there must be some independent and highly accurate way of measuring the variable being analyzed. Id. at 342. For example, the validity of an employment test is measured by comparing the results to actual job performance. Finally, the data must have been recorded properly, with no gaps or mistakes present. Id. If gaps or mistakes exist, they should appear randomly so as not to distort the results. Id.

When all of these conditions are met, the measuring device has been verified, but it still must be applied to the correct data. “Statistical analysis appears to be a science and, when properly and fairly used, it can be of great assistance to the factfinder.” Coates, 756 F.2d at 539. Thus, it is important to determine how the data to be measured has been selected and whether there is any evidence of selection bias. For example, “[a] census measures some characteristic of every unit in a population of individuals or objects.” Manual on Scientific Evidence 343 (emphasis in original). With this measure, it is the population that is selected. As a type of observational study, a census yields useful results only when the population is relevant to the inquiry for which the statistical analysis is being performed. See id. at 343. Moreover, the analysis of census results may be affected by the omission of some units from the population. Id. If this occurs, the expert should ask whether the missing data is likely to differ in some systematic way from the data actually measured. Id. If so, a supplemental study may be necessary.

Parties often attempt to prove causation by using statistical associations identified in an observational study. The associations among variables are expressed as percentages, proportions, ratios, correlation coefficients or slopes of regression lines. Moore’s Federal Practice Reference Manual on Scientific Evidence 362 n.96 (Fed.Jud.Ctr. 1995). Even though an association is identified, however, there is a second issue: is the association causal? Id. Thus, identifying an association is only the first step.

To explore a possible cause and effect relationship further study is required. The investigator should then observe the effect on one variable, the dependent variable, of manipulation of another, an independent variable. Id. at 348. Independent variables represent the possible “causes” of the “effect” being studied. If the study does not contemplate the results of manipulating various independent variables, it does little to prove causation. That is because the real cause of the observed results may be variables other than the one with which the dependent variable is associated. When an association is identified among variables, a controlled experiment will help identify which independent variable caused the effect. In fact, this is the best way to eliminate other variables as possible causes of observed associations. Id. at 347. If a controlled experiment is not conducted, then the investigator should at least measure the group in which the effect occurred against a comparable group in which the effect is not expected or observed. See id. at 349. The key to success, however, is that the second group actually be comparable.

Another issue that arises with observational studies is that the investigator has no control over who or what is observed. Id. at 350. Thus, a controlled experiment is precluded, and the investigator should attempt to control for “potentially confounding variables,” such as socioeconomic status, education, or worsening economic conditions. Id. at 351. A confounding variable is one that is correlated with both the dependent and independent variables, making it difficult to discern which variable has caused the effect in the dependent variable. Id. at 348. Without collecting data on the confounding variable, or controlling for other factors that may influence the outcome, the investigator is unable to draw any conclusions about causation from the associations among variables. Id. at 348, 352. When an association is found using different studies, or the same study and different groups, the results of the observational study provide more useful evidence. Id. at 351. This is because the other studies control for the possibility that the association is due to some defect in the study itself or a peculiarity among group members. Id.

Alternatively, an observational study could be useful if the association among variables holds true even when the investigator takes into account, by appropriate statistical techniques such as comparing smaller, homogeneous groups, the effects of plausible confounding variables. Id. Finally, observational studies may provide useful information when there is also a plausible explanation for the effect of the independent variable. In that case, the causal link does not depend only on the association between the independent and dependent variables. Id. at 352. If the explanation for linking the dependent variable’s response to other variables is less plausible, it is possible to conclude that the independent variable being tested is the cause of the response. Id.

When none of these methods of validating an observational study have been used, experts will tend to disagree about the study’s interpretation. Id. Ultimately, deciding whether associations are causal is not a matter of statistics, but of good scientific judgment. Id. If the data comes from an observational study, the court should inquire about how the units that were studied were formed, whether they are comparable, what are, and what are the factors correlated with, the independent variables, what adjustments were made to account for those factors, and were those adjustments sensible. Id. Only with satisfactory responses to these inquiries can a court be sure of the reliability of statistical evidence from observational studies in proving causation.

D. ANALYSIS

Plaintiffs offering scientific or technical evidence have the burden of establishing the qualifications of the experts, the reliability of their testimony, its relevance, and its usefulness to the jury. Daubert, 509 U.S. at 587 n. 10, 113 S.Ct. 2786 (citing Bourjaily v. United States, 483 U.S. 171, 175-76, 107 S.Ct. 2775, 97 L.Ed.2d 144 (1987)). The defendants have challenged the admissibility of a report and testimony from Dr. Richard Wertheimer (“Wertheimer”), a labor economist that plaintiffs plan to offer as an expert witness, under the Daubert standards. According to the defendants, Wertheimer’s statistical analyses are flawed to the point of inadmissibility, in part because he combined all managers at ASI, regardless of job function, experience, education, or skills, when measuring the number of older workers that were selected for termination relative to the number of younger ones selected. By doing so, defendants argue, he treated non-fungible managers as interchangeable, possibly skewing the results. In addition, defendants assert that Wertheimer failed to control for any variable other than age in his statistical study, he compared the ASI workforce to a dissimilar “U.S. Workforce,” and he varied the methods used to achieve the results his clients sought.

However, plaintiffs have strongly disagreed with defendants about whether Wer-theimer improperly aggregated statistics about non-fungible managers at ASI. According to the plaintiffs, this type of challenge goes to the probative value of the statistical evidence, not its admissibility. They argue that because part of their ADEA claims depend on proof that a pattern or practice of age discrimination was present, Wertheimer’s pooling of all the CRESP selection process results provides probative evidence on that issue.

In addition, defendants claim that Wer-theimer’s subsequent report, filed within thirty days of their experts’ reports, was “unauthorized” because it violated the terms of the case management plan and Fed. R.Civ.P. 26(a)(2)(C). According to the defendants, this report contains new opinions based on evidence that was in the possession of the plaintiffs at the time of the original report. The new report includes analyses of selection rates based on salary grades and vice-presidential groups in support of plaintiffs’ contention that the gross disparities in selection rates between those over and those under forty occur even when you look at subgroups. See Pis.’ Ex. 57, Wertheimer’s Rebuttal Report. Wertheimer’s rebuttal report was filed on May 2,1996, more than one year after his initial report. Defendants object to consideration of this report because it came after the final date listed in the case management plan for expert witness discovery. However, the rebuttal report was filed within days of the last expert discovery deposition, and more than thirty days before defendants’ deadline for fifing dispositive motions. It appears to be a response to challenges raised by defendants’ experts regarding Wertheimer’s initial report. The Court can discern no prejudice to defendants from the fifing of this rebuttal report.

According to the defendants, the second report is also unreliable in that it contains most of the same flaws as the original. Chief among them is the failure, again, to analyze the data in terms of the actual facts relating to the process used to select employees for termination. Secondarily, Wertheimer formed sub-groups for this report that bear no relationship to the CRESP process. A hint of this fact is revealed in a footnote in the rebuttal report relating to the slot machine hypothetical, in which Wertheimer admits that “the situation is more complex than what has been presented because the ASI managerial workforce is not evenly split between younger and older employees.” Pis.’ Ex. 57, Wertheimer Rebuttal at 2, n.4. In fact, prior to the resizing, workers at least age forty or more comprised 58.1% of ASI’s workforce, while those under age forty constituted 41.9%. Defs.’ App. 104, McCabe Report. The difference in age composition of the workforce and among the various CRESP groups was not accounted for in Wertheimer’s analyses of the effects of the resizing. In addition, Wertheimer included all of the employees who had volunteered to take early retirement among those described as “selected.” As noted below, the Court finds this departure from the facts to be unsupported by sound scientific principles.

Plaintiffs have conceded the “limited value” of the profile of the U.S. managerial workforce. Pis.’ Mem. In Oppos. To Defs.’ Mot. To Strike Testimony of Expert Witness at 21. As defendants had noted, that- comparison group was not sufficiently controlled for factors that would make it actually comparable to the management workforce at ASI. In addition, the plaintiffs admit that their expert did not consider all of the possible variables affecting who was selected for termination. Id. However, they argue that they need only show a pattern of discrimination, and the burden shifts to the defendants to show that the “statistical disparity is due to a non-discriminatory variable or variables that plaintiffs failed to consider.” Pis.’ Mem. in Oppos. to Mot. to Strike at 22. In response, defendants argue that plaintiffs are attempting to shift their burden at the admissibility stage to them. This Court agrees.

In essence, plaintiffs are describing use of the McDonnell Douglas burden-shifting method of proof, which is not employed when determining the reliability of an expert’s report. Instead, reliability is founded on whether the expert is qualified and will testify to reliable scientific knowledge that is helpful to the trier of fact. Moreover, by framing the issue in terms of whether their expert must consider “all of the possible variables,” the plaintiffs overlook the possibility that their expert may have failed to consider one of the major factors causing the disparity. See Bazemore v. Friday, 478 U.S. 385, 106 S.Ct. 3000, 92 L.Ed.2d 315 (1986) (per curiam) (reversing lower court finding that plaintiff’s statistical analyses were erroneous because they omitted information about differences caused by geographic location). In Bazemore, the Court said that an analysis that included less than “all measurable variables” may still prove a plaintiffs case. Id. at 400, 106 S.Ct. 3000. However, to be considered acceptable as evidence of discrimination the analysis must account “for the major factors” affecting the results. Id.

Citing Bazemore, the plaintiffs argue that any-failure to include variables in a statistical analysis goes to its weight, not its admissibility. Id. Nevertheless, the Supreme Court noted that an analysis can be so incomplete that it is inadmissible as irrelevant. Id. n. 10. To determine if the analysis can carry the plaintiffs ultimate burden of proof a court must consider the factual context of the case, in light of all the evidence presented. Id. Despite the plaintiffs’ contention, the defendants did not argue that plaintiffs’ expert must account for all of the possible variables before his statistics would be meaningful. Instead, defendants asserted that major factors had been overlooked, such as, the non-fungibility of the managers, or the unique characteristics of the CRESP process that affected selection. If such omissions render the analysis incomplete, then Wertheimer’s testimony would be inadmissible. With that understanding, the Court now turns to an evaluation of the proffered statistics in the context of the facts of this case.

Statistical evidence may prove discrimination when it shows a pattern of conduct against a protected group that, if unrebutted, would allow the inference of intentional discrimination against individual members. Barnes v. GenCorp, Inc., 896 F.2d 1457, 1466 (6th Cir.), cert. den., 498 U.S. 878, 111 S.Ct. 211, 112 L.Ed.2d 171 (1990). The statistics, however, must show a “significant disparity” and must eliminate the “most common nondiscriminatory explanations for the disparity.” Id. They must also bear a logical connection to the facts and circumstances being analyzed. See O’Connor, 116 S.Ct. at 1310; Coser v. Moore, 739 F.2d 746, 751 (2nd Cir.1984). The Barnes court indicated that the most obvious explanations for a statistical disparity in termination decisions are random chance or lower proficiency. Barnes, 896 F.2d at 1466. However, the facts and circumstances of any given case will influence what constitutes an obvious explanation, or the major factors.

The statistics offered by the plaintiffs have only eliminated random chance as a possible explanation for the disparities. See Pls.’ Ex. 56, Wertheimer Report at 8, ¶¶ 14,16, 20-22, 28; Defs.’ App. 108, Rothman Report at 4; see also EEOC v. Sears Roebuck & Co. 889 F.2d 302, 323 n. 20 (7th Cir.1988) (standard deviation and other measures of statistical significance merely attempt to eliminate chance as reason for results, do not prove causation). That method would be valid if it is reasonable to assume that the job skills the defendants needed to retain were equally distributed across age groups. See Barnes, 896 F.2d at 1466. If it is not reasonable so to assume, a defendant can show that the statistical method is faulty, or that chance does explain the discrepancy, and the statistical evidence may lose its effectiveness altogether. Id. at 1469. If it is reasonable, the statistical evidence that shows a greater number of terminations among certain age groups than would be expected due to chance, may state a prima facie case for age discrimination. Id.

In Barnes, the court allowed the plaintiffs to use such statistical evidence to satisfy their prima facie case of age discrimination, but found those same statistics insufficient to prove that the defendants’ proffered legitimate explanation was a pretext. Id. at 1468-69 (statistics cannot determine whether the more likely cause of the disparity was defendants’ bias or a legitimate selection criterion). The court noted that when it appears that job skills are not evenly distributed across age groups, such statistics become much less probative. Id. at 1468. In the context of determining the reliability of proffered statistical evidence, the proponent has the burden of showing the reasonableness and validity of the expert’s assumptions. See Daubert v. Merrell Dow Pharm., Inc., 43 F.3d 1311, 1316 (9th Cir.), cert. den., 516 U.S. 869, 116 S.Ct. 189, 133 L.Ed.2d 126 (1995) (party presenting expert must show that expert’s findings are based on sound scientific method). Wertheimer’s analyses assumed that the job skills to be retained were equally distributed across age groups. Plaintiffs argue that ASI treated its management workforce as fungible, so it was proper for their expert to do so as well.

In a class action pattern or practice claim of age discrimination, a plaintiff must show that the “regular policy of the employer” was to discriminate on the basis of the protected trait. Coates, 756 F.2d at 532. The plaintiffs prima facie case often consists of statistical evidence showing a significant disparity between the employment decisions made regarding members of the protected group, and those made as to younger workers. Id. The statistics, however, will usually be “buttressed by evidence of general policies or specific instances of discrimination.” Id. The focus is on whether there has been an overall pattern of discriminatory decision-making. Id.

That pattern may be defeated, despite any amount of statistical evidence, if the defendant articulates sufficient nondiscriminatory reasons for each separate discharge. Id. If, however, the defendant only explains the discharges of the named plaintiffs or class representatives, the pattern or practice claim does not necessarily fail. Id. at 533. When strong statistical evidence establishes the prima facie case, a defendant’s successful rebuttal of each particular decision to discharge weakens, but does not defeat, the class claim. Id. On the other hand, with respect to individual claims only, a defendant’s showing that a discharge was based on a neutral assessment of the particular plaintiffs skills in relation to those of other employees, will defeat the claim. Id. The statistics proffered by the plaintiffs here are offered in support of both the pattern or practice claim and each individual plaintiff’s claims.

Turning to the report in controversy, the Court notes that in the conclusion of his report, Wertheimer states that in his professional opinion all of the evidence is consistent with the “1992-1993 Workforce Resizing Program treating employees at least age 40 differently from employees under age 40.” Pls.’ Ex. 56, Wertheimer Report at 25. He supports this conclusion with his finding that ASI management employees at least fifty or more years of age are “significantly underrepresented” in relation to a “comparison group” of United States workers. The plaintiffs have already conceded the “limited value” of this comparison, and the Court sees no reason to further examine it. See Hazelwood, 433 U.S. at 308, 97 S.Ct. 2736 (dealing with issue of proper comparative groupings for purposes of determining from overall statistics whether an employer is discriminating). That is because, even if it were supported, Wertheimer’s conclusion does not allow a finding of intentional discrimination by the defendants. Instead, it could only support a finding that the resizing program may have had a disparate impact on older workers. Such a finding says little about the defendants’ intent.

Even in a pattern or practice discrimination case, a plaintiff must still prove discriminatory intent. Chicago Miniature, 947 F.2d at 298. Intent means more than knowledge that a certain action will cause discriminatory results. Id. at 297. It means a subjective desire or wish for that result. Id. Here, the plaintiffs attempt to prove the defendants’ wrongful intent with statistical evidence of a disparate impact from which they want the factfinder to infer an intent to discriminate. Cf. Id. (prima facie case can be established by statistical evidence ... buttressed by general policies or specific instances).

For this to occur, several inferential steps must be taken, and each one of them must be reasonable in light of the evidence. See Salus v. G.T.E. Direct. Serv. Corp., 104 F.3d 131, 137 (7th Cir.1997) (reasonableness of inference questioned, but not considered clear error). The first step would be a finding that the resizing program in fact had a disparate impact on older workers, which would require the finders of fact to choose which expert’s testimony and reports to accept as true. The Court’s task here is to determine whether it is proper to put them to that choice. Next, the factfinder would have to conclude that the impact occurred because the selection process was tainted, either by the actions of the individual decisionmakers, or by some systemic defect, such as a criterion designed to target the older workers. Yet, the factfinder would also have to attribute the taint in the selection process to a conscious and deliberate choice by the defendants. See Chicago Miniature, 947 F.2d at 297 (intent means a subjective desire or wish). If the statistical evidence does not relate in some meaningful way to these steps, it will not be useful to the jury.

Wertheimer found that the “termination process selected a significantly higher fraction of employees at least age 40 than of employees under age 40 at each stage of the process.” Pis.’ Ex. 56, Wertheimer Report at 25. Specifically, he noted that the termination rates “generally rose with age,” which means his study found an association between the dependent variable of termination rates, and the independent variable of age. Again, for this association to support a finding of discriminatory intent, a factfinder would have to determine that the termination process was purposefully designed to achieve that result.

The termination rate is the number of terminated employees belonging to an age category divided by the number of workers in that age category before the terminations. Id. at 7, ¶¶ 11, 14. Essentially that figure represents the percentage of employees in a given age category that were terminated. Once Wertheimer calculated these percentages for all employees at least age 40 or more and for all employees less than age 40, he contrasted the percentages and found them “highly statistically significant,” because the probability of these results occurring by chance was .009%, using a two-tailed test. Id. That equates with a 2.61 standard deviation, or less than a one in a million chance of it occurring by random chance. Id. According to Wertheimer, this difference is highly statistically significant.

However, Wertheimer’s analysis fails to account for any other independent variable that might explain the association between age and termination rates, including job skills, education, experience, or self-selection. For an ADEA claim, in a work force resizing or restructuring context that includes an early retirement incentive, it is imperative that the analysis take these “unusual circumstances” into account. See Hartley, 124 F.3d at 889. Because Wertheimer’s failure to do so is a departure from generally accepted methods in the field of statistical analysis, it must be shown to be grounded in “demonstrable and scrupulous adherence to the scientist’s creed of meticulous and objective inquiry.” Braun, 84 F.3d at 235. In other words, he must have a scientifically-based reason for departing from the customary method of statistical analysis.

Instead, Wertheimer has explained his procedures by stating that he used information provided by the plaintiffs’ counsel about the characteristics of the workforce he studied. He admitted to a lack of familiarity with the actual process used for the involuntary terminations. See Def.App. 118, Wer-theimer Dep. at 160-62, 218, 238. In fact, when questioned about the source of information on which he based his judgment, Wertheimer stated that he relied on “information that was conveyed to [him] by [plaintiffs’ counsel] ... [who] gave [Wertheimer] ... an impressionistic view of [the process]. But it certainly wasn’t a detailed or specific view of this process.” Id. at 161-62. He also admitted that he did not know what criteria were used in Stage I of the process. Id. at 165. Moreover, Wertheimer disregarded the defendants’ classification of certain employees as voluntary terminees, without an explanation for having done so. Pis.’ Ex. 56, Wertheimer’s Report at 9, ¶ 19. The Court does not find this to be an adequate explanation for Wertheimer’s departure from generally accepted statistical procedures. See Robinson v. PPG Indus., Inc., 23 F.3d 1159, 1163 (7th Cir.1994) (“Truly voluntary retirements do not give rise to an inference of age discrimination.”). Thus, to the extent his testimony relates to a comparison of overall selection and termination rates for those employees aged forty and above, and those under forty, it should not be admitted.

The same problem weakens Wer-theimer’s analysis of selection rates. Wer-theimer concluded that the rate at which employees were selected for possible termination (selection rate) was higher for employees at least age forty than the rate for employees less than forty. Pis.’ Ex.- 56, Wertheimer’s Report at 25, ¶ 28(d). As noted, this analysis included those employees who may have been selected, but who took a voluntary early retirement package. Id. at 9, ¶ 19. He found the selection rates to be as follows: aged forty or more, 16.5%, under age. forty, 10.9%. The disparity between the two rates was equivalent to 6.68 standard deviations, which Wertheimer found to be “highly statistically significant.” Although he has not adequately explained his reason for doing so, Wertheimer considered the employees who took early retirement to be “involuntarily terminated,” rather than self-selectors. Pis.’ Ex. 56, Wertheimer Report, at 9, ¶ 19. The Court has found this to be a departure from generally accepted methods of scientific inquiry, and will not admit evidence about selection rates that ignores the factor of self-selection.

Wertheimer’s report provides an indication of this factor’s effect on the results. When the termination rates (instead of selection rates) of employees aged forty or more are compared with those for under age forty, the disparity diminishes significantly. The termination rate for employees who were at least forty was 11.7%, while the rate for those less than forty was 9.7%. Pis.’ Ex. 56, Report at 25, ¶ 28(b). That reflects a difference of 2%, or 2.61 standard deviations, which Wertheimer also found to be “highly statistically significant.” Id.; Report, Table D-5. According to defendants’ expert, because the workforce resizing program used two simultaneous methods to reduce the number of managers, a voluntary and an involuntary one, the statistical analysis should present the results of each method separately. Defs.’ App. 104, “Analysis of Employment Data for Ameritech Services, Inc.” by George P. McCabe, Ph.D. (“McCabe Report”) at 12. The voluntary method included an early retirement package with incentives, and the involuntary method used the CRESP termination process. Id.

The percent of total terminated employees who were aged forty or more was 62.6%, while the percent of terminees under age forty was 37.4%. Id. at 8, ¶ 15. When compared to the proportion of employees prior to the resizing program that were aged forty or more, 58.9%, and those under forty, 41.9%, the two sets of numbers do not appear very different to the untrained eye. Nevertheless, they are different, and the plaintiffs say they demonstrate that a higher number of employees under age forty were left in the workforce after the resizing than were there before the resizing. Conversely, a lower number of employees aged forty or more were left in the workforce after the resizing.

The plaintiffs basically offer those statistics, and various permutations and computations related to those figures, as support for a finding that the defendants had a regular and ongoing policy of treating employees aged forty or more less favorably than it treated those under age forty. In essence, by showing a disparate impact the plaintiffs seek to elicit a finding of deliberate and intentional discrimination because of age. The problem with this strategy is that the statistics do not reveal the cause of the disparity, and when only disparate treatment is available to prove discrimination, the analysis of causation is crucial.

Given Wertheimer’s admission that he did not know much about the process used by ASI to resize its workforce, or about the criteria used for selecting terminees, and that he assumed that all of the 6,695 managers involved in the CRESP process were interchangeable, the Court is not comfortable relying on the results of his statistical analy-ses. His departures from generally accepted scientific metho