The opinion of the court was delivered by: Charles P. Kocoras, District Judge
Defendant EMC Corporation ("EMC") has filed a motion to strike the declaration, opinions and testimony of Plaintiffs' expert Dr. Jerry Goldman on the basis that minimum indicia of reliability are lacking as to his proffered submissions. There are a variety of grounds cited in support of EMC's motion, some of which are described in this opinion. The principal criticism is that Dr. Goldman never uses the same statistical model in analyzing a series of years. It is alleged that he "cherry-picked" among a variety of methodologies and models to find the one that meets a pre-determined result for each year and that, had he analyzed the data consistently for all of the years studied, he could not have reported the results Plaintiffs wanted.
The Plaintiffs' response to the criticisms, in substance, is that Dr. Goldman's analyses required adjustment from 2001 to 2004 because of the following: changes to compensation structure, fluctuations in workforce data, and characteristics of the workforce. Plaintiffs also argue that the details presented by their expert go to the weight to be accorded his views and not to their admissibility, even while recognizing that general reliability considerations address admissibility of testimony in the first instance.
The Plaintiffs seek to certify a class of current and former female EMC sales employees subjected to EMC's alleged discriminatory pay and management selection policies and practices. One of the two practices challenged by the Plaintiffs was the assignment of sales quotas, accounts and territories. The other challenged practice was the grooming and selecting of managers. The core of Plaintiffs' allegations is that EMC managers manipulated sales quotas to prevent women from achieving their goals. A significant aspect of this charge against EMC and its managers is whether women and men attain their sales quotas at differing rates.
At his deposition, Dr. Goldman agreed that the assignment of sales quotas was the assignment of the opportunity to make a certain amount of variable compensation through commissions and bonuses. Despite that acknowledgment, Dr. Goldman admitted that he never looked at the quota assignments of men and women to see whether the assignment of quota to men and women was even-handed. And he never looked to see whether the quota to attainment ratio for men and women was significantly different, even though the data was available to do so for each of the four years under consideration in this case. Nor did he bother to study the percentage of women versus the percentage of men who attained their quotas using readily available data. Dr. Goldman ultimately conceded that none of the studies that he did for 2001 and 2002 got at the practice that he was trying to study; none of them studied the compensation effect of the practice as accurately as he wished. Hearing transcript at 172.
In addition to not using available data to study the compensation practices of EMC condemned by the Plaintiffs, Dr. Goldman chose to arbitrarily eliminate relevant portions of data that could have been used to arrive at more complete and accurate assessments of EMC's practices as regards women. For example, in 2001 Dr. Goldman established an upper sales quota limit of $10.5 million or less for sales employees with established quotas. The effect of setting this arbitrary limit was to eliminate 42 of the highest female earners at EMC, or approximately 40% of the available pool of women. This lopping off of a significant group of people for analytical purposes was a regular and recurring feature of Dr. Goldman's analyses in a number of other studies which were offered as support for his conclusions. It is not surprising that when the entire group of men and women who were similarly situated are added in, the statistical conclusion that men are favored over women cannot stand based on Dr. Goldman's work.
Dr. Goldman conceded that a number of his regression studies had similar dollar quota limitations as that described above, resulting in the use of a little over half of the available data and, necessarily, producing a skewed result. The reasons offered by Dr. Goldman for this limitation is of importance, for it provides insight into his assumptions and approaches to his studies.
At both his deposition and during his hearing testimony, Dr. Goldman affirmed that he did not rely on anything Dr. Joan Madden (another expert for the plaintiffs) did for his own report, including definitions of people similarly situated. At another point in his testimony, he testified that he had several conversations with Dr. Madden and knew about her findings with respect to promotion. The implication of this assertion was that Dr. Goldman's groupings and limitations were based on Dr. Madden's findings, a resource previously and expressly disavowed. There is simply no way to reconcile these conflicting statements by Dr. Goldman.
At another point in his testimony, Dr. Goldman suggested that the allegations in the Plaintiffs' Complaint were used in deciding on the cut-offs and the groupings in the study. The allegations in any plaintiff's complaint are simply that, assertions of fact which, if true, may amount to unlawful discrimination and private wrongs committed against a person or group of people. When a putative statistical expert assumes the truth of the complaint's allegations and structures the protocols of his testing procedures accordingly, then any pretense of the use of mathematics and statistics in any objective, independent or disciplined way cannot be sustained. Instead of fairly assessing the fullness of data readily available in order to see if the analyses statistically support Plaintiffs' allegations, almost half of the top portions of the sales quota data is ignored. Instead, we are asked to accept the accuracy of the bottom half of the results as representative of the whole.
Other decisions made by Dr. Goldman are equally questionable. In his regression studies, Dr. Goldman took out the EMC seniority variable and excluded it from his study. His explanation was twofold:
First of all, that its value could have been confounded on account of bias with the result of discrimination; but then, secondly, by deposition of the company's representatives that the various skills and prior experience that a person had were incorporated within their job code and as well particularly within their salary to boot. Hearing transcript at 299.
Not dissimilar to Dr. Goldman's arbitrary sales quota used in most of his analyses, if one assumes discrimination and structures studies with that as an operative assumption, the statistical results produced will be skewed ab initio. Additionally, with seniority data readily available, to ignore it because some other categories may reflect some components of seniority experience is to purposely avoid the pure effect of EMC seniority experience on compensation and related issues.
Another choice made by Dr. Goldman was to use median studies and not consider average or mean studies. Dr. Goldman said that he worried about the possibility that there were a few outliers for highly compensated women who would skew the results. Rather than test the possibility, however, he just ignored the information available from studies of averages. Results vary materially based on whether the average or median method of measure is employed, and it would have been far more instructive to consider both.
In addition to those particularized above, EMC has leveled a number of other criticisms of Dr. Goldman's work. Many of their criticisms are valid. For example, Dr. Goldman included in his database employees for the years 2001 through 2004 if they were full-year employees who were there for five years or less. In other words, the database included people irrespective of whether they were in sales jobs or not; people were studied who were not members of the putative class. Also, anyone employed more than five years did not count. Nor did Dr. Goldman differentiate between managers and sales ...