Opinion
Civil Action No. 03-cv-01167-WDM-CBS.
September 2, 2005
ORDER
This matter is before me on the defendant's Motion in Limine to Strike Testimony of Plaintiff's Proposed Expert Robert A. Bardwell and Request for a Daubert Hearing. The parties have submitted extensive briefs and exhibits both before and after a scheduled Daubert hearing at which only Dr. Bardwell testified, followed by counsels' arguments. The target of defendant's Motion to Strike, a Preliminary Report on the Impact of Race and National Origin on Workforce Utilization and Compensation at Allright Parking 1992-2000, prepared by Dr. Bardwell on April 16, 2004, and marked as Exhibit A (April 16, 2004 Bardwell Report), was supplemented by Dr. Bardwell's testimony. Defendant's challenge is based upon the opinion of its own expert, Dr. George F. Rhodes, who submitted a Report on Statistical and Economic Issues Associated with Race and National Origin at Central Parking, dated June 9, 2004, and marked as Exhibit B (June 9, 2004 Rhodes Report). Plaintiff responded with an affidavit of Dr. Bardwell, apparently dated December 1, 2004, (December 1, 2004 Bardwell Affidavit). Dr. Rhodes then filed a supplemental report, dated December 20, 2004, and marked as Exhibit A to defendant's reply in support of its motion (December 20, 2004 Rhodes Report).
Plaintiff responds, arguing in favor of the testimony of an individual identified as Dr. Andrew Bardwell. Although plaintiff persists in so identifying this expert ( see June 13, 2005 Motion for Leave to Reply), Dr. Bardwell's affidavit and his testimony before me identifies the proper name as Robert A. Bardwell.
A ruling based upon some sworn, live testimony some by affidavit or affirmation, some by deposition and some not sworn at all, is not the best procedure to resolve Daubert issues. However, Fed.R.Evid. 104(a) allows me to decide issues of witness qualification and admissibility of evidence without strict compliance to the rules of evidence. How I perform the gate-keeping function under Daubert is left to my discretion. See Kumho Tire Co., Ltd. v. Carmichael, 526 U.S. 137, 152 (1999); Gobell v. Denver and Rio Grande Western Railroad Co., 215 F.3d 1083, 1087 (10th Cir. 2000). "The most common method for fulfilling this function is a Daubert hearing, although such a process is not specifically mandated." Gobell, at 1087. I may satisfy this role by ruling on a motion in limine as presented here, on a trial objection, or on a post-trial motion, so long as I have sufficient evidence to meet "the task of insuring that an expert's testimony both rests on a reliable foundation and is relevant to the task at hand." Daubert v. Merrell Dow Pharmaceutical, Inc., 509 U.S. 579, 597 (1993). I conclude that the parties' presentation was adequate to enable me to make intelligent findings in my "gate-keeping" function on this matter.
Discussion
In general, defendant moves to strike the April 16, 2004 Bardwell Report as unreliable. Plaintiff responds that the report meets the Daubert standards, that any shortcomings result from defendant's inadequate record keeping or production, and that any deficiency goes to the weight of the evidence and not its admissibility.Rule 702 of the Federal Rules of Evidence, as amended after Daubert, provides the basic framework for analyzing these issues. Neither side challenges the qualification of the other's expert. The remaining issues, therefore, are whether "(1) the testimony is based upon sufficient facts or data, (2) the testimony is the product of reliable principles and methods, and (3) the witness has applied the principles and methods reliably to the facts of the case." Fed.R.Evid. Rule 702.
When presented with this pretrial challenge of expert testimony I am to make a "preliminary assessment of whether the reasoning or methodology underlying the testimony is scientifically valid and of whether that reasoning or methodology properly can be applied to the facts in issue." Daubert, 509 U.S. at 592-93. This gate-keeping function "is to ensure the reliability and relevancy of expert testimony. It is to make certain that an expert, whether basing testimony upon professional studies or personal experience, employs in the courtroom the same level of intellectual rigor that characterizes the practice of an expert in the relevant field." Kumho Tire Co., 516 U.S. at 152. Hence, the key determinations are whether the expert opinion "both rests on a reliable foundation and is relevant to the task at hand." Daubert, 509 U.S. at 597. In determining reliability I may consider the following factors: whether the theory or technique can be (and has been) tested, whether it has been subjected to peer review and publication, the known or potential rate of error, and whether it is generally accepted within the relevant expertise. Id. at 593-94.
In performing my gate-keeper function I must determine that the methodology is scientifically valid and that it was properly applied. See Dodge v. Cotter Corp., 328 F.3d 1212, 1221-22 (10th Cir. 2003). In this case, the dispute focuses on the second issue, namely the application of the scientific methodology as opposed to the methodology itself.
Defendant and its expert agree that regression analysis would be appropriate circumstantial evidence of discrimination if properly performed. December 24, 2004 Rhodes Report, pp. 11-12.
Plaintiff essentially argues, and some authority exists for the proposition, that a challenge to the application of the scientific methodology, as opposed to the methodology itself, goes to the weight of the evidence and not its admissibility. See United States v. Gibson, 383 F.3d 689, 697 (8th Cir. 2004). Plaintiff urges that defendant may challenge plaintiff's expert and his use of variables before the jury and let it decide what weight, if any, should be given to Dr. Bardwell's statistical correlation. A review of the Tenth Circuit authority does not disclose a clear bifurcation that would, in effect, admit the expert testimony even if a recognized methodology is questionably applied.
Even the Eighth Circuit recognizes that exclusion may be warranted if the methodology "were so altered [by a deficient application] as to skew the methodology itself." United States v. Beasley, 102 F.3d 1440, 1448 (8th Cir. 1996).
Other circuits have adopted this view that if the expert's application of an otherwise reliable methodology is somehow flawed, the flaws impact the weight and credibility of the evidence, not its admissibility. See United States v. Shea, 211 F.3d 658, 668 (1st Cir. 2000).
The Tenth Circuit has not so sharply bifurcated the issues between methodology and its application although it has concluded that improper application may render expert opinion evidence irrelevant. See United States v. Davis, 40 F.3d 1069, 1074 n. 6 (10th Cir. 1994) (if protocol not followed "the scientific evidence may not be relevant under Daubert's second prong because improperly applied science cannot assist the trier of fact").
Defendant's expert concludes that the Bardwell opinion on pay differentials, based upon a regression analysis, fails to meet any of these Daubert standards and is hence unreliable and without sufficient foundation. December 20, 2004 Rhodes Report, pp. 2-5. Dr. Rhodes criticizes Dr. Bardwell's estimates as based solely upon comparison of pay with race or national origin without including the variables of skill, education, and experience that Dr. Rhodes views as essential. Without considering those variables, he opines no valid inference may be drawn with regard to race or any other variables. He points to several different examples where pay differentials may well be based upon education, experience, and other factors. Dr. Rhodes concludes that, without the variables being included, plaintiff's regression analysis does not meet any of the four Daubert considerations: (1) regression equations without these factors are not testable; (2) because they are not testable it is impossible to know or estimate the rate of error; (3) the analyses done are not based on peer review and publication; and (4) analyses without the variable factors are not generally accepted in the scientific community. Id. pp. 3-4.
Plaintiff's response is based upon the December 1, 2004 Bardwell Affidavit. Dr. Bardwell essentially concedes that having the additional variables in the regression analyses is preferred but blames the defendant for their absence, observing that defendant did not provide sufficient documentation to allow their inclusion. On the basis of a self-constructed "burden-shifting paradigm," he suggests plaintiff need only present prima facie evidence, which the defendant can rebut "by providing additional business related factors that explain the apparent disparity. Defendant's failure to produce data cannot preclude a demonstration of disparity." Id. at ¶ 9 (original emphasis). Dr. Bardwell justifies this opinion — which is essentially legal — by citing to Bazemore v. Friday, 478 U.S. 385, 404 (1986).
Dr. Bardwell characterized defendant's production as "the most egregiously flawed production of incomplete and potentially altered employee data that I have encountered in litigation." Dec. 1, 2004 Bardwell Affidavit, ¶ 4.
Defendant disagrees, arguing there is no legal basis for Dr. Bardwell's "burden-shifting process" and regardless, urging me to rely on Dr. Rhodes's opinions that the omission of essential variables renders the regression analysis unreliable and that no inference should be drawn therefrom. December 20, 2004 Rhodes Report, p. 4. As a consequence, defendant asserts, the evidence is inadmissible under Daubert and Fed.R.Evid. 702 principles.
Plaintiff (and his expert) rely heavily on Bazemore. A comparison of the factual background of that case with this renders Bazemore tenuous support for plaintiff's position. Although pre- Daubert, the case remains authority for the basic proposition that regression analyses are acceptable even though not all measurable variables were included. "Normally, failure to include variables will affect the analysis' probativeness, not its admissibility." 478 U.S. at 400. However, this statement was footnoted with the observation that "there may, of course, be some regression so incomplete as to be inadmissible as irrelevant; but such clearly was not the case here." Id. at note 10. Indeed, as compared to Dr. Bardwell's analysis, involving only starting salaries and race or national origin, the multi-year regression analyses in Bazemore included race, education, tenure, job title and, in one instance, job performance. Id. at 398. The district and appellate courts had held that the analysis failed to "include all measurable variables thought to have an effect on salary level," including differences in counties being analyzed. Id. at 399 (emphasis in original). It was this absolutist view which was rejected by the Court. It is a much different circumstance here where none of those factors were included in plaintiff's analysis. Further, as the Court observed in Bazemore, there was other evidence besides the regression analysis to support the allegations of discrimination. Here, plaintiff provides no real independent evidence of discrimination to support his statistical evidence other than reference to differences in management appointments.
Statistical evidence may well create an inference of discrimination but it must cross a threshold of reliability and not "be so flawed as to render it insufficient to raise a jury question." Martinez v. Wyoming Department of Family Services, 218 F.3d 1133, 1138 (10th Cir. 2000). Other circuits have agreed. See In re Paoli R.R. Yard PCB Litig., 35 F.3d 717, 745 (3d Cir. 1994) (any step that renders analysis unreliable means expert opinion inadmissible); Amorgianos v. Nat'l Passenger Corp., 303 F.3d 256, 267 (2nd Cir. 2002) ("To warrant admissibility . . . it is critical that an expert's analysis be reliable at every step.")
With this background, the issue distills down to whether a recognized methodology using so few variables was still properly applied so as to serve as evidence from which discrimination may be inferred. Given Tenth Circuit authority on the use of statistical evidence in discrimination cases, I conclude it cannot.
Use of statistical data has long been recognized in this Circuit to "create an inference that an employer discriminated against individual members of the class." Fallis v. Kerr-McGee Corp., 944 F.2d 743, 746 (10th Cir. 1991). However, to use such statistical evidence to create that inference "the statistics must show a significant disparity and eliminate non-discriminatory explanations for the disparity. . . . In other words, a plaintiff's statistical evidence must focus on eliminating non-discriminatory explanations for the disparate treatment by showing disparate treatment between comparable individuals." Id. (Emphasis in original).
Significance of the disparity is not an issue here.
The Tenth Circuit has consistently followed this principle to deny admissibility of simple statistical comparisons where the evidence did not focus on eliminating non-discriminatory explanations for any disparity. See Cone v. Longmont United Hospital Ass'n, 14 F.3d 526, 532-33 (10th Cir. 1994) (comparison of ages alone insufficient to show age discrimination); Rea v. Martin Marietta Corp, 29 F.3d 1450, 1456 (10th Cir. 1994) (statistics without adjustment for various performance evaluations and rankings inadmissible); Doan v. Seagate Tech., Inc., 82 F.3d 974, 979 (10th Cir. 1996) (evidence is flawed when failed to eliminate non-discriminatory explanations for numerical disparities).
These authorities rejecting statistical evidence that does not address non-discriminatory variables demonstrate the failure may result in summary judgment (Cone v. Longmont United Hospital Association) or even judgment as a matter of law (Doan v. Seagate Technology, Inc.) without imposing any burden upon the defendant to present evidence in response to the claimant's insufficiently supported expert opinion. Contrary to plaintiff's argument, no "burden-shifting paradigm" is suggested in these cases. Instead, they represent the basic process of a claimant being required to present evidence to establish a case sufficient to reach a jury without imposing upon the defendant some independent obligation to come forward with evidence the plaintiff has failed to provide. Accordingly, I conclude that the admissibility of plaintiff's expert evidence is not determined by the nature of the records maintained by defendant.
Focusing on the issue of reliability of the stated opinion of Dr. Bardwell in context of this Tenth Circuit authority requiring the elimination of non-discriminatory reasons for numerical disparities, I find the plaintiff's evidence to be lacking. The non-discriminatory variables should have been included in the plaintiff's analysis and were not. As defendant's expert concludes, their absence renders any inference from the regression analysis unreliable. The failure of the defendant to produce evidence necessary for a proper regression analysis, for whatever reason, does not make unreliable evidence admissible either outright or subject to defendant disproving the unreliable inference with more evidence. I accept Dr. Rhodes's opinion that the limitation of the variables to race or national origin and starting salary fails each of the four Daubert tests. Given the limitations, Dr. Bardwell's conclusions are not testable and the rate of error cannot be determined. The methodology as applied here would not be accepted by the scientific community and would not be based upon peer review.
Dr. Rhodes concludes that in significant experience he had never "come across any information that would lead me to believe that the equations in Dr. Bardwell's report would be accepted by the scientific community." Dec. 20, 2004 Rhodes Report, p. 4.
Concerning the effects of the report dealing with management positions, it is my understanding that plaintiff does not seek a management position and hence this analysis would otherwise be irrelevant except to the extent it might be used to support the initial claim of differential pay. To the extent it is so used, I again find and conclude that the methodology used is not reliable because it fails to take into account the actual application process implemented by defendant in hiring management staff. Rather than using the actual applicant pool, Dr. Bardwell conducted his analysis with an assumed applicant pool consisting of existing employees. In fact, the actual applicant pool may include some fraction of existing employees but also many applicants from the general population. December 20, 2004 Rhodes Report, pp. 6-8. There is also no consideration of variables of education, experience and so forth. As a consequence, the analysis cannot reliably be used to infer racial bias and cannot be used to bolster plaintiff's claim.
In general, Dr. Bardwell's opinions violate Fed.R.Evid. 702 because they are not based upon sufficient facts or data and he did not reliably apply recognized principles or methods to the facts presented. Accordingly, it is ordered:
1. Defendant's motion to strike (Document No. 52) is granted and Dr. Bardwell's preliminary report (Exhibit A) and his testimony concerning the same are not admissible.
2. Plaintiff's motion for leave (Docket No. 91) is granted.