Full opinion text
MEMORANDUM AND ORDER FOREMAN, Chief Judge: TABLE OF CONTENTS Page I. INTRODUCTION 1562 II. LEGAL FRAMEWORK 1562 A. Jurisdiction 1562 B. Title VII 1562 1. An Individual Disparate Treatment Claim 2. A Disparate Impact Claim 1562 1563 3. A Class Action Pattern & Practice Disparate Treatment Claim 1564 C. A 42 U.S.C. § 1981 Claim 1564 D. Statistical Evidence 1564 III. THE CLASS CLAIMS 1566 A. The Claims and Theories of the Parties 1566 B. An Overview of the Evidence Presented 1567 C. The Plaintiffs’ Evidence 1568 1. Statistical Analysis 2. The Plaintiffs’ Static Work Force Statistics 1568 '1569 3. Anecdotal Evidence 1570 D. The Defendant’s Evidence 1570 1. Statistical Analysis 2. The Defendant’s Anecdotal Evidence 1570 1573 1573 E. The Court’s Analysis of the Plaintiffs’ Case 1. The Accuracy of the Data Upon Which the Plaintiffs Relied 1573 2. The Plaintiffs’ Applicant Flow Analysis 1515 3. The Plaintiffs’ Static Work Force Statistics and Anecdotal Evidence lj^® 4. The Plaintiffs’ Prima Facie Case 1515 F. The Court’s Analysis of the Defendant’s Rebuttal 1576 Page 1. The Distance Factor 2. The Defendant’s Demographic Analysis 1576 1577 G. The Court’s Conclusion Concerning the Class Claims 1579 H. Mister’s Representation of the Class 1581 IV. ROBERT MISTER’S INDIVIDUAL CLAIM 1581 V. ORDER 1582 I. INTRODUCTION The plaintiff, Robert Earl Mister, a black male, filed this action on January 9, 1981, against the Illinois Central Gulf Railroad Company, pursuant to Title VII of the Civil Rights Act of 1964, as amended, 42 U.S.C. § 2000e et seq. (Title VII), 42 U.S.C. § 1981, and 42 U.S.C. § 1985, alleging that the defendant refused to hire him because of his race. In addition to the individual claim, and upon the plaintiffs request, the Court allowed the plaintiff to represent a class of all black persons who had applied for jobs and been wrongfully rejected due to the defendant’s allegedly racially discriminatory practices with regard to hiring within the defendant’s St. Louis, Missouri, Operating Division. On October 11, 1985, the Court granted the defendant’s motion for summary judgment with respect to the plaintiffs’ 42 U.S.C. § 1985 claims. After numerous discovery battles that caused many continuances, the Court, without a jury, heard the liability phase of the case on November 12,13, 14, 15,18, 19, 20, 21, 1985, December 5, 6, 1985 and January 21, 22, 23, 24, 30, 31, 1986. The following memorandum represents this Court’s findings of fact and conclusions of law as contemplated by Fed.R.Civ.P. 52(a). In order to place the discussion of the evidence in the proper perspective, the Court feels that a general overview of the legal standards is appropriate. II. LEGAL FRAMEWORK A. Jurisdiction. This Court finds that it has subject matter jurisdiction of the Title VII claims pursuant to 42 U.S.C. § 2000e-5(f)(3) and that it has subject matter jurisdiction of the 42 U.S.C. § 1981 claims pursuant to 28 U.S.C. § 1343. Venue is proper both under 42 U.S.C. § 2000e-5(f)(3) and 28 U.S.C. § 1391. B. Title VII. Title VII prohibits an employer from failing or refusing to hire an individual because of the individual’s race. 42 U.S.C. § 2000e-2(a)(l). The Supreme Court has identified two theories in Title VII cases— disparate treatment and disparate impact. International Brotherhood of Teamsters v. United States, 431 U.S. 324, 97 S.Ct. 1843, 52 L.Ed.2d 396 (1977). In general, either theory may be applicable to a particular set of facts. Id. at 335 n. 15, 97 S.Ct. at 1854. Regner v. City of Chicago, 789 F.2d 534 (7th Cir.1986). Under the disparate treatment theory, the employer treats some people less favorably than others because of their race, religion, sex, or national origin. Teamsters 431 U.S. at 335 n. 15, 97 S.Ct. at 1854 n. 15. Proof of discriminatory intent is critical. Id. Under the disparate impact theory, employment practices that are facially neutral in their treatment of different groups “fall more harshly on one group than another and cannot be justified by business necessity____” Id. Unlike a disparate treatment case, a plaintiff in a disparate impact case need not show an intent to discriminate. Id. 1. An Individual Disparate Treatment Claim. McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973), is the seminal case concerning the burdens of pleading and proof in an individual disparate treatment case. The Supreme Court therein outlined a three-step formula for such cases. First, the plaintiff must establish a prima facie case. He may do so by showing that a) he belongs to a protected class; b) that he applied for an existing job for which he was qualified; c) that despite his qualifications, he was rejected; and d) after the rejection, the employer continued to seek applicants with the plaintiff’s qualifications. Id. at 802, 93 S.Ct. at 1824. Once the plaintiff establishes a prima facie case, the burden shifts to the defendant to provide a legitimate nondiscriminatory reason for its actions. Id. The purpose of this evidence is to show that the employment decision was not “motivated by discriminatory animus.” Texas Department of Community Affairs v. Burdine, 450 U.S. 248, 257, 101 S.Ct. 1089, 1095, 67 L.Ed.2d 207 (1981). The defendant need not demonstrate nondiscriminatory intent, but must simply produce evidence to rebut the presumption of discrimination raised by the plaintiff’s prima facie case. Id. at 254-55, 101 S.Ct. at 1094. This is not a burden of persuasion, but merely a burden of production. The plaintiff may still persuade the Court that the defendant’s explanation is “unworthy of credence” that “a discriminatory reason more likely motivated” the defendant. Id. at 256, 101 S.Ct. at 1095. The plaintiff at all times retains the ultimate burden of persuasion. The McDonnell Douglas formulation was not “intended to be rigid, mechanized, or ritualistic” in application. Furnco Construction Corp. v. Waters, 438 U.S. 567, 577, 98 S.Ct. 2943, 2949, 57 L.Ed.2d 957 (1978). Nor does it suggest a specific order of proof at trial. Coates v. Johnson & Johnson, 756 F.2d 524, 531 n. 5 (7th Cir. 1985). The ultimate question in a disparate treatment case is whether the defendant’s actions were founded in a discriminatory intent. Teamsters, 431 U.S. at 335 n. 15, 97 S.Ct. at 1854 n. 15. 2. A Disparate Impact Claim. In a disparate impact case, the plaintiff can establish a prima facie case of discrimination by showing that “facially neutral standards in question select applicants for hire in a significantly discriminatory pattern.” Dothard v. Rawlinson, 433 U.S. 321, 329, 97 S.Ct. 2720, 2726, 53 L.Ed.2d 786 (1977). Once the plaintiff establishes a prima facie case of discriminatory impact, the burden of proof shifts to the defendant to demonstrate a business necessity for the practice or that the practice has a “manifest relation to the employment in question” in order to avoid a finding of discrimination. See Griggs v. Duke Power Co., 401 U.S. 424, 431, 91 S.Ct. 849, 853, 28 L.Ed.2d 158 (1971), and Connecticut v. Teal, 457 U.S. 440, 446, 102 S.Ct. 2525, 2530, 73 L.Ed.2d 130 (1982). The Supreme Court has not indicated whether this shift in burden is one of persuasion or simply one of coming forward with the evidence as in a disparate treatment case. See Schlei and Grossman, Employment Discrimination Law, p. 1328 (2d ed. 1983). This Court is inclined to follow those cases holding that the shift is one of persuasion. See Johnson v. Uncle Ben’s Inc., 657 F.2d 750, 752-53 (5th Cir. 1981), cert. denied, 459 U.S. 967, 103 S.Ct. 293, 74 L.Ed.2d 277 (1982). Even if the defendant meets his burden of persuasion, the plaintiff may still prevail by showing that the defendant was using the practice as a mere pretext for discrimination or that other selection devices without a significant discriminatory effect would also “serve the employer’s legitimate interest____” Albemarle Paper Co. v. Moody, 422 U.S. 405, 425, 95 S.Ct. 2362, 2375, 45 L.Ed.2d 280 (1975). Similar to a disparate treatment case, the allocation of proof set forth above represents a method for analyzing evidence and not a procedure by which evidence is presented. Teamsters, 431 U.S. at 335 n. 15, 97 S.Ct. at 1854 n. 15. A disparate impact theory is most frequently employed in class actions because the employment practice in question, by the very nature of the theory, must affect a large group of persons. Statistics are almost totally determinative in cases under the disparate impact theory. Generally, a disparate impact claim concerns the impact of a facially neutral standards or criteria such as a maximum height requirement, or minimum scores on an objective test. See e.g., Griggs v. Duke Power Co., 401 U.S. 424, 91 S.Ct. 849, 28 L.Ed.2d 158 (1972). Courts are not in agreement as to whether the disparate impact analysis can be applied to subjective employment decisions. See Schlei and Grossman, Employment Discrimination Law, p. 24 (1983-1984 cumulative supplement); Major Issues in the Federal Law of Employment Discrimination, Supplement 1 p. 13. This Court need not get too caught up in this interesting debate because in the recent case of Regner v. City of Chicago, 789 F.2d 534, (7th Cir.1986), the Seventh Circuit has seemingly approved the application of the disparate impact model to a subjective procedure. See also Griffin v. Board of Regents of Regency Universities, 795 F.2d 1281, 1287-1289, nn. 13, 14 (7th Cir.1986). For an excellent discussion of this issue, see Lamber, Discretionary Decisionmaking: The Application of Title VII’s Disparate Impact Theory, 1985 U.Ill.L.R. 869. 3. A Class Action Pattern and Practice Disparate Treatment Claim. Although the disparate treatment theory arises most often in an individual case, class actions may also proceed under this theory. In such a case, the plaintiff generally combines a claim of individual disparate treatment with a claim that the defendant’s treatment of the plaintiff is part of a pattern or practice of discriminatory treatment toward the members of the plaintiff’s class. Plaintiffs who raise a pattern or practice claim must demonstrate that the discrimination has been the regular policy of the defendant, i.e., that “discrimination was the company’s standard operating procedure— the regular rather than unusual practice.” International Brotherhood of Teamsters v. United States, 431 U.S. 324, 336, 360, 97 S.Ct. 1843, 1867, 52 L.Ed.2d 396 (1977). The plaintiff usually establishes a prima facie case by employing statistical evidence to create an inference of classwide discrimination similar to a disparate impact case. Id. at 339, 97 S.Ct. at 1856. Although the plaintiffs almost always offer evidence of individual instances of discriminatory treatment to buttress their case, the plaintiffs are not required to show that each member of the class is a victim of the defendant’s discriminatory policy. Id. at 360, 97 S.Ct. at 1867. Evidence of isolated incidents alone will not establish a prima facie case. However, in some cases, statistical evidence alone may constitute a prima facie case. See Coates v. Johnson & Johnson, 756 F.2d 524, 532 n. 6 (7th Cir.1985). Once the plaintiffs establish a prima facie case, the burden then shifts to the defendant to rebut the inference of discrimination created by the plaintiff’s statistics. The defendant may do so by “demonstrating that the plaintiffs’ proof is either inaccurate or insignificant.” Teamsters, 431 U.S. at 360, 97 S.Ct. at 1867, or by providing a “nondiscriminatory explanation for the apparently discriminatory result.” Id. at 360 n. 46, 97 S.Ct. at 1867 n. 46. At all times, the plaintiffs maintain the burden of persuasion with respect to discrimination. Coates, 756 F.2d at 532. C. A 42 U.S.C. § 1981 Claim. 42 U.S.C. § 1981 provides a redress for racial discrimination in private employment. McDonald v. Santa Fe Trail Transportation Co., 427 U.S. 273, 96 S.Ct. 2574, 49 L.Ed.2d 493 (1976). Traditionally, the courts regarded Title VII and § 1981 cases as involving analogous principles, standards, and theories of discrimination; however, the Supreme Court has eliminated the disparate impact theory in § 1981 cases. General Building Contractors Association v. Pennsylvania, 458 U.S. 375, 102 S.Ct. 3141, 73 L.Ed.2d 835 (1982). In General Building, the Supreme Court held that § 1981 could only be violated by purposeful discrimination. Therefore, the legal standards outlined above for the individual and class disparate treatment claims would apply equally as well to the 42 U.S.C. § 1981 individual and class claims. D. Statistical Evidence. Obvious from the above discussion is that statistics play a dominant role in both disparate impact cases and disparate treatment pattern or practice cases. The two principal statistical comparisons are referred to as “pass/fail” and “population/work force” comparisons. Schlei and Grossman, Employment Discrimination Law, pp. 1332-1358 (2d ed. 1983). In a hiring case such as this, the pass/fail statistics compare the percentage of the protected group (blacks) who meet a given employment requirement or who are actually hired with the corresponding percentage for the majority group (whites). Population/work force statistics in a hiring case compare the availability of the protected group (blacks) in the relevant labor market with the percentage of the protected group in the employer’s work force. The differences resulting from the above comparisons may be measured in a number of ways. The chief barometer employed is a standard deviation. The “standard deviation” is a number that quantifies the degree to which disparities spread out above and below the mean of distribution, thus describing the probability that chance is responsible for any difference between an expected outcome and the observed outcome in a sample consisting of two groups (a binomial distribution). The greater the number of standard deviations, the less likely it is that chance is the cause of any difference between the expected and observed results. The Supreme Court noted in Castaneda v. Partida, 430 U.S. 482, 97 S.Ct. 1272, 51 L.Ed.2d 498 (1977), that under a “two-tail” test of statistical significance, “[a]s a general rule for ... large samples, if the difference between the expected value and the observed number is greater than two or three standard deviations, then the hypothesis that the [disparity] was a random would be suspect to a social scientist.” Id. at 497 n. 17, 97 S.Ct. at 1281 n. 17. The general rule should be applied with caution, however, and its particular applicability may vary from case to case. See D. Baldus & J. Cole, Statistical Proof of Discrimination § 9.03, at 293-97 (1980), 108-12 (Supp.1984); see also American National Bank, 652 F.2d [1176] at 1190-93. Coates v. Johnson & Johnson, 756 F.2d 524, 536 n. 11 (7th Cir.1985). The data used to furnish the components of each comparison can come from a number of sources. The employer’s actual applicant flow is frequently used in pass/fail statistical comparisons and to determine the relevant labor market in a population/work force analysis. Schlei and Grossman, Employment Discrimination Law, pp. 1332-1358 (2d ed. 1983). General population census data or labor force data may also be utilized to determine the relevant labor market. In addition, work force composition data may be employed on the right side of the population/work force comparison. In ascertaining the relevant labor market, it is essential to define the proper geographic scope and time frame. In defining the geographic scope, the objective is to determine the area from which the employer’s applicants are likely to come. With respect to the time frame, in a Title VII claim the relevant time frame is the period commencing 180 days prior to the filing of the EEOC charge. In the instant case, the relevant time frame is the period commencing on December 8, 1978. The relevant time frame for a 42 U.S.C. § 1981 claim is the period prior to the filing of the lawsuit that corresponds to the analogous state statute of limitations for the state in which the lawsuit is filed. In Beard v. Robinson, 563 F.2d 331, 334 (7th Cir.1977), cert. denied sub nom., Mitchell v. Beard, 438 U.S. 907, 98 S.Ct. 3125, 57 L.Ed.2d 1149 (1978), the Seventh Circuit held that all claims founded on the Civil Rights Acts were governed by the five-year Illinois statute of limitations for actions not covered in other statutes of limitations. Recently, however, in Wilson v. Garcia, 471 U.S. 261, 105 S.Ct. 1938, 85 L.Ed.2d 254 (1985), the Supreme Court held that the most analogous statute of limitations for 42 U.S.C. § 1983 claims is the state’s personal injury statute of limitations. In Illinois, the personal injury statute of limitations is two years. Ill.Rev.Stat. ch. 110, 1113-203. The defendant herein contends, therefore, that pursuant to Wilson, the relevant time frame for the § 1981 claims is two years from the date of the suit. The Court disagrees. In Anton v. Lehpamer, 787 F.2d 1141 (7th Cir.1986), the Seventh Circuit indicated that Wilson would not be applied retroactively. Accordingly, the relevant time frame for the § 1981 claims is the five-year period prior to the filing of this lawsuit, or the period beginning on January 9, 1976. Statistical proof pertaining to a time pri- or to the relevant time period may be admissible as circumstantial evidence of discrimination occurring within the relevant period. Hazelwood School Dist. v. United States, 433 U.S. 299, 309 n. 15, 97 S.Ct. 2736, 2742 n. 15, 53 L.Ed.2d 768 (1977). Statistical evidence covering the period after the filing of the lawsuit is irrelevant. While most courts prefer time-frame statistics, static work force statistics are relevant especially if they demonstrate great disparity. Schlei and Grossman, Employment Discrimination Law, p. 1365 (2d ed. 1983). III. THE CLASS CLAIMS With this legal framework in mind, the Court turns to a discussion of the claims and an analysis of the evidence presented. A. The Claims and Theories of the Parties. At the outset, the Court notes that there existed prior to trial, and even at trial, confusion as to what theory upon which the plaintiffs were relying. The class plaintiffs claimed that this was both a disparate treatment and disparate impact case. The defendant agreed that this was a class disparate treatment case, although it disagreed with the plaintiffs as to what that meant, but argued that disparate impact analysis was inapplicable to this case because the plaintiffs could not point to any objective standard or test used to exclude blacks. The plaintiffs’ response to this latter argument was two-fold. First, the plaintiffs alleged that the defendant utilized a fifty-mile rule to exclude from consideration those applicants who resided outside of a fifty-mile radius from the job location. The plaintiffs argued that this rule, although neutral on its face, had a disparate impact when applied to black applicants. Second, and alternatively, the plaintiffs asserted that if there were no fifty-mile rule, then there existed excess subjectivity in the defendant’s hiring process and such subjectivity had a disparate impact on black applicants. The parties also disagreed as to meaning of a class action disparate treatment case and as to what evidence could be presented to support this theory. The defendant argued that the class claim was a pattern or practice claim consisting of only statistical evidence to demonstrate that the defendant engaged in a pattern or practice of discrimination over the relevant time frame. The plaintiffs seemed to approach the class claims as a number of individual disparate treatment claims, although the plaintiffs recognized that statistics played a major role. The legal analysis outlined above should indicate how this Court viewed the case. First, either the disparate treatment or disparate impact theory may be applicable to a given set of facts. Regner v. City of Chicago, 789 F.2d 534 (7th Cir.1986). Therefore, the plaintiffs could properly approach this case arguing both theories. Second, the Court has construed the class disparate treatment claims as claims alleging a pattern and practice of discrimination on the part of the defendant. The easiest and almost exclusive way for the plaintiffs to demonstrate a pattern and practice of discrimination is by introducing statistics that allow the Court to infer that the employer has a pattern or practice of discrimination. Anecdotal evidence of individual instances of discrimination is not necessary, but can be used to buttress the statistical evidence. International Brotherhood of Teamsters v. United States, 431 U.S. 324, 360 n. 46, 97 S.Ct. 1843, 1867 n. 46, 52 L.Ed.2d 396 (1977). A pattern and practice case is not made by presenting a number of individual disparate treatment claims under the McDonnell Douglas framework. This Court would not have nor could not have certified the class if such were intended. The plaintiffs are still under the impression that individual claims can comprise a class disparate treatment action because they argue, in their proposed findings, that the Court should certify a subclass of those individuals that applied for work on March 27, 1979. Therefore, the Court’s job with respect to the class action disparate treatment claims was to determine if the defendant engaged in a pattern or practice of discriminatory hiring during the relevant time frames. See Coates v. Johnson & Johnson, 756 F.2d 524, 532 (7th Cir.1985). The Court considered the statistical evidence first and foremost. Finally, the Court agrees with the plaintiffs that this case can also be analyzed under the disparate impact theory. See Regner v. City of Chicago, 789 F.2d 534 (7th Cir.1986). As will be discussed more fully later, the plaintiffs have not convinced this Court that there existed a hard and fast fifty-mile rule. But, by the same token, they have convinced the Court that the defendant subjectively applied distance from the workplace in its hiring decisions. It is important to note that a disparate treatment pattern and practice case and a disparate treatment excess subjectivity case are identical in terms of evidence. Schlei and Grossman, Employment Discrimination Law, p. 1288-1289 (2d ed. 1983). [I]n both the disparate treatment pattem-or-practice case and the adverse impact “excessive subjectivity” case, the battle is determined by an evaluation of the probative force of plaintiff’s and defendant’s statistical evidence and the evidence with respect to alleged specific instances of discrimination. The proof considerations are no different whether the case be styled a pattern or practice of disparate treatment, on the one hand, or adverse impact resulting from an excessively subjective system of selection for hire or promotion, on the other. Id. at 1289-90 (footnotes omitted). The only important distinction is that in the excess subjectivity case the burden of persuasion would shift to the defendant after the plaintiffs establish a prima facie case. il. An Overview of the Evidence Presented. In support of the class claims, the plaintiffs presented a statistical analysis of the defendant’s hiring decisions over the relevant time frames along with anecdotal evidence. The statistical evidence consisted of an applicant flow analysis prepared by Dr. Leroy Grossman, a labor economist. His rather straight forward approach was to compare the percentage of black applicants for the defendant’s St. Louis, Missouri operating division with the percentage of black hires. The statistical significance of the difference between these two percentages was computed by using a standard deviation analysis. Dr. Grossman’s conclusion was that there existed a statistically significant difference in the percentage of blacks who applied and were hired. The plaintiffs also presented static views of the defendant’s work force for the years 1974 through 1983. Finally, the plaintiffs presented evidence of a number of blacks from the East St. Louis area who applied for jobs at the Carbondale, Illinois, office on March 27, 1979 and who were never hired despite other hirings in the division. The defendant’s evidence consisted of a three-fold attack on the plaintiffs’ statistical evidence. First, the defendant presented evidence to demonstrate that the information Dr. Grossman relied upon to produce his applicant flow analysis was inaccurate. Second, the defendant presented a demographic analysis of the hiring decisions prepared by Dr. David Peterson, a statistician. This method compared the number of actual black hires at each work location during the relevant time frames to the expected number of black hires for each work location based upon the 1980 United States census data. He also utilized a standard deviation analysis to compare the differences. His conclusion was that there was no evidence to suggest that the defendant’s hiring decisions occurred on a basis other than random selection with regard to a hiree’s race. Finally, the defendant presented evidence to demonstrate that distance from the workplace was an important factor in its hiring decisions. In rebuttal, the plaintiffs presented the testimony of Dr. Marc Rosenblum, Chief Economist for the United States Equal Employment Opportunity Commission, who indicated the flaws he saw in Dr. Peterson’s analysis. C. The Plaintiffs’ Evidence. 1. Statistical Analysis. As indicated above, Dr. Grossman employed an applicant flow approach to analyze the defendant’s hiring decisions. He testified that he decided to use this method because, in his opinion, applicant flow information, if available, provides the best method of measuring the employer’s relevant labor market. Unlike a demographic method which relies on an estimate of the employer’s labor market, the applicant flow method utilizes direct evidence of those actually interested in a job. Dr. Grossman also decided to treat the entire St. Louis Missouri Operating Division as one labor market. His reasons for doing so were that historically the railroad industry attracts individuals from greater distances than other industries and because migration is a direct function of employment opportunity. He also added that the fact that the St. Louis Division had only one hiring officer and office supports his decision. Dr. Grossman based his statistical analysis on three alternative sets of data furnished by the defendant. The three sources of data were 1) the defendant’s Monthly Employment Activity Reports (MEARS); 2) the defendant’s answer to supplemental interrogatory 4 which was based on the defendant’s 1591 forms; and 3) the defendant’s answer to interrogatory number 6. The plaintiffs’ statistics were summarized in Plaintiffs’ Exhibit 73. Table No. 1 of Exhibit 73 is an applicant flow analysis using the applicant and hire data from the defendant’s MEARS for the years 1979 and 1980. Dr. Grossman computed a Z-score or standard deviation comparing the percentage of whites hired out of the total number of white applicants to the percentage of blacks hired out of the total number of black applicants. In 1979, the percentage of black applicants hired was 10%, while the percentage of white applicants hired was 31.42%. In standard deviations, the number of blacks hired was 7.58 standard deviation units below the expected number. In 1980, 0% of black applicants were hired while 6.1% of the white applicants were hired. The number of black hires in 1980 fell 2.05 standard deviation units below the expected number. Grossman stated that to his knowledge, two standard deviation units is the customary cut-off point used by statisticians for statistical significance and that the Z-scores for both 1979 and 1980 were significant. Dr. Grossman indicated that the probability of the hiring disparity which occurred in 1979 having been due to chance alone is less than 29 out of 100,000,-000. Table No. 2 of Plaintiff’s Exhibit 73 is an applicant flow analysis comparing applicants in the “laborer” and “no preference” categories with only hires in the laborer category. Dr. Grossman indicated that he made the common sense assumption that applicants who applied for “no preference” would be willing to accept a laborer job. In 1979, according to these figures, 11.0% of black applicants were hired and 39.4% of white applicants were hired. The actual number of blacks hired in 1979 according to these figures was 8.21 standard deviation units below the expected number. The probability that this was due to chance alone is less than 29 out of 100,000,000. Table No. 3 of Plaintiffs’ Exhibit 73 compares the total number of applicants in the “laborer” and “no preference” categories with hires in the laborer, carmen apprentice, and switchman categories for the years 1979 and 1980. For 1979, 12.6% of black applicants were hired and 42.7% of white applicants were hired. The number of black hires under this comparison fell 8.35 standard deviation units below the expected number for 1979. Dr. Grossman stated that the Z-score was not statistically significant for 1980, 0 blacks were hired out of a total of 25 hires. Dr. Grossman pointed out, however, that this was to be expected as the likelihood of a statistically significant Z-score decreases in situations where relatively few hires are made. Table No. 4 of Plaintiffs’ Exhibit 73 is a comparison of applicants in the “laborer,” “no preference,” and “operative” categories with the hires in the laborer, carmen apprentice, and switchman categories. The data in this table, as well as Tables 1-3, was taken from the defendant’s MEARS. For 1979, 10.1% of black applicants were hired while 33.7% of white applicants were hired. The actual hiring of blacks fell 7.87 standard deviation units below the expected number in 1979. Again Dr. Grossman testified that the probability of this having been due to chance alone is less than 29 out of 100,000,000. Table No. 6 of Plaintiffs’ Exhibit 73 is an applicant flow analysis comparing the number of applicants in all categories with the number of hires in all categories using as its data base defendant’s answer to interrogatory number 6. Dr. Grossman indicated through this table that in each year 1977-1980 there is a statistically significant discrepancy between the number of blacks actually hired and the number expected to be hired. More importantly, Dr. Grossman pointed out that the combined Z-score for 1977, 1978, and 1980, excluding the year 1979, was 3.21. Dr. Grossman testified that this year after year pattern is highly significant and simply could not have occurred by chance. Table No. 7 of Plaintiffs’ Exhibit 73 is an applicant flow analysis for all hires and all applicants using the application data from defendant’s answer to interrogatory number 6 and the hire data from defendant’s answer to supplemental interrogatory number 4, the 1591 forms. Dr. Grossman stated that the 1591 hiring data used in this table was more likely to be accurate since the 1591 hires were in some cases greater than the hires listed in the MEARS. The expected hiring of blacks in the years 1977, 1978, 1979, and 1980, according to this table fell 3.45, 4.08, 13.40, and 2.93 standard deviation units short of that expected, respectively. In Dr. Grossman’s opinion, not only is the Z-score for each year statistically significant, but the year by year pattern shows overwhelmingly that this pattern of hiring could not have been due to chance. Furthermore, setting aside the year 1979 with the highest Z-score of 13.40, the combined Z-score for 1977, 1978, and 1980 was 5.54. Dr. Grossman testified that the probability of this having been due to chance alone is again less than 29 out of 100,000,-000. Table No. 8 of Plaintiff's Exhibit 73 is Dr. Grossman’s response to the defendant’s contention that the MEARS applicant data for the year 1979 is inaccurate. A comparison of the MEARS applicant data and the actual count of applications for 1979 indicated that the actual number of applications was 1,577 while the MEARS report listed 1,196 applications for the year. Dr. Grossman concluded that the higher number was more likely to be correct because there were actual applications to support that number. In Table No. 8, Dr. Grossman assumed that the additional 381 applications which were in the possession of the defendant and not apparently listed on the MEARS were all from white applicants. Dr. Grossman indicated that this assumption would increase the availability of white applicants and correspondingly decrease the expected number of black hires — an assumption favorable to the defendant. Comparing the resulting applicant data with the 1591 hiring data under this assumption, Dr. Grossman arrived at a Z-score of 8.67 standard deviation units for the year 1979. 2. The Plaintiffs’ Static Work Force Statistics. In addition to the applicant flow analysis, the plaintiffs also presented a static view of the defendant’s work force for the years 1974 through 1983. Plaintiffs’ Exhibit 80 summarized the defendant’s EEO-1 forms for those years. This exhibit showed that blacks were excluded entirely from some job categories and that the highest percentage of black employees was found in the lowest paying job categories. 3. Anecdotal Evidence. A good portion of the plaintiffs’ case consisted of evidence concerning an incident on March 27, 1979 in Carbondale, Illinois, where a number of blacks from the East St. Louis, Illinois area applied for jobs. The plaintiffs presented the testimony of a number of the class members who applied for a job at Carbondale on the day in question. These individuals testified that they were referred to the Carbondale office by Capitol Employment Agency, a private employment agency. Each individual testified that a large group of predominately black individuals (between 120 and 179) was present on March 27, 1979 in Carbondale seeking a job. The testimony of each individual indicated that he was present on March 27, 1979 in Carbondale, that he filled out an application, that he was ready, willing, able to work, that had asked, he would have been willing to relocate, and that he was never contacted by the railroad. The plaintiffs also presented the testimony of a few white individuals who were hired by the defendant during the relevant time frame and who resided at a distance of greater than fifty miles from their work location. D. The Defendant’s Evidence. 1. Statistical Analysis. Prior to performing his extensive demographic analysis, Dr. Peterson indicated that he analyzed the accuracy of the defendant’s data and attempted to verify the defendant’s position that distance from the workplace was a factor in its hiring decisions. Dr. Peterson testified that he collected a large amount of data and documents from the defendant in order to analyze defendant’s labor markets and hiring practices. He indicated that he interviewed various company officials and he was advised that defendant had a long standing policy of applying distance of an applicant from the initial job location as a factor in its hiring decision. Dr. Peterson collected all of the MEARS available. He collected all of the applications available. He also utilized the defendant’s Form 1591, the defendant’s federal withholding form, which he concluded was the most accurate evidence of each of the defendant’s hiring decisions in its St. Louis division between 1976 and 1980. The 1591 form also indicated the place of residence of each applicant hired and his initial job location. Dr. Peterson testified that he checked the accuracy of his data bases against documents entitled “Year End Tapes,” which are computer tapes maintained by the defendant at its Chicago headquarters and which contain numerous bits of data concerning each of the defendant’s employees at the end of each year being considered by the court. Dr. Peterson also testified that he confirmed the accuracy of his data base by cross-checking that data against payroll log books that were maintained by the defendant at its Chicago headquarters. In order to confirm or refute defendant’s contention that its labor markets were local and directly related to its job locations, and to confirm or refute defendant’s contention that it used distance of an applicant’s residence from the job location as a factor in its hiring decisions, Dr. Peterson indicated that he analyzed each of defendant’s hiring decisions between 1976 and 1980 in its St. Louis division. Dr. Peterson and his staff, utilizing the data in the defendant’s 1591 withholding forms, actually measured the distance that each person hired on the division lived from the individual’s initial job location at the time of hire. This measurement data was then programmed into a computer at Dr. Peterson’s Personnel Research offices in Durham, North Carolina. This computer printout was contained in Defendant’s Exhibit 4 and is entitled “LOCDISTF.” Based upon the information contained in this printout, Dr. Peterson generated various graphs and maps pertaining to defendant’s labor markets and geographic hiring which have also been admitted into evidence. Dr. Peterson concluded that the defendant consistently hired 80% or more of its employees into each of its numerous job locations in the St. Louis division from a distance of less than 50 miles. Dr. Peterson also analyzed the accuracy and completeness of various types of employment records that he obtained from the defendant. Dr. Peterson stated that the purpose of this analysis was to determine the most proper form of study to be utilized. Dr. Peterson testified at length at trial that an accurate and adequate applicant flow analysis could not be performed in this case due to the incompleteness and inaccuracy of the applicant data available. He testified that his analysis showed that the applicant data contained in the MEARS was inaccurate. He indicated that there were large chronological gaps in the actual applications retained by the company and that there was no way to create an accurate applicant count. In addition, he pointed out that none of the actual applications received into evidence contained a racial designation for the applicant. Dr. Peterson concluded that a proper applicant flow analysis could not be conducted based upon the application data contained in the MEARS because that document did not contain any information concerning the residence of applicants and did not contain any information pertaining to the locations of jobs available at the time any application was pending. In Dr. Peterson’s opinion, the only available method to analyze the defendant’s hiring practice in its St. Louis division between 1976 and 1980 was a demographic analysis comparing actual black hires to expected black hires based upon the 1980 United States census data. Dr. Peterson testified that this 1980 U.S. Census data was the most accurate data available for use in a demographic study. Dr. Peterson performed three different types of demographic analyses of defendant’s hiring practice in its St. Louis division. He performed each of these three analyses on three different periods of time: 1976 through 1980; December 8, 1978 through January 9, 1981; and finally, 1979 and 1980. The first demographic analysis performed by Dr. Peterson was described as a county by county analysis. Dr. Peterson identified each of the counties within the defendant’s St. Louis operating division. He testified that he obtained U.S. Census data for each county which identifies the availability of black persons for hire in various job categories. To determine the availability of blacks in each county within which the defendant hired, Dr. Peterson in this first analysis utilized what he has described as a diminished civilian labor force. The diminished civilian labor force availability numbers that he used excluded officers, managers, professionals, technicians, and sales workers. Dr. Peterson indicated that he utilized this diminished availability number because the evidence had shown that defendant made only eight hires into the categories excluded from the availability pool between 1976 and 1980. In Dr. Peterson’s first demographic analysis, he compared the actual number of blacks hired residing in each county to the availability of blacks in the diminished civilian labor force for that individual’s county of residence. Dr. Peterson then performed, on a county by county basis, statistical calculations to show whether or not the defendant’s hiring decisions in each county were occurring by random chance. He expressed his results in terms of standard deviations. Defendant’s Exhibit 4 contains computer printouts which show the results of these calculations. Dr. Peterson stated that his calculations of standard deviations county by county were well within the most stringent two standard deviation level recognized by courts for the determination of statistical significance in discrimination cases. Dr. Peterson statistically aggregated the results from the county by county analysis over all the years studied. Dr. Peterson concluded that defendant hired a total of 1,230 employees, including rehires, (and excluding 4 hires from Michigan, Indiana and Kansas), from 1976 to 1980. One hundred nineteen of those hired were black. Dr. Peterson’s statistical calculations based upon the census data showed that defendant’s overall expected black hires during that period should have been 99.386 individuals. In terms of standard deviations, defendant’s black hires compared to its actual hires in this first analysis shows a positive standard deviation of 2.13797. This indicates that the defendant had hired more blacks than would be expected if the defendant’s selections were by random chance. (Defendant’s Exhibit 4 § DCLFALLF pp. 21-22). In Dr. Peterson’s second demographic analysis, he compared the actual number of blacks hired with the number of blacks expected to be hired, not only county by county, but also EEO category by EEO category. The census data available for each county in the defendant’s St. Louis operating division contains availability data broken down by EEO category such as clerical, laborer, operative, and crafts. County by county and EEO category by EEO category Dr. Peterson compared the number of black hires in the Division within each category to the census data pertaining to the availability of blacks in each county and in each EEO category within the county. Dr. Peterson performed this analysis by both including and excluding rehired individuals. Dr. Peterson expressed his results in terms of standard deviations broken down EEO category by EEO category for each county. In reviewing Defendant’s Exhibit 4 which contains the computer printout, Dr. Peterson concluded that category by category within each county the defendants were well within the two standard deviation level recognized by courts in considering discrimination issues. Again, in this second analysis, Dr. Peterson statistically aggregated all of these calculations into one overall test of the defendant’s hiring practices between 1976 and 1980, county by county, EEO category by EEO category. Between 1976 and 1980, defendant hired a total of 1,227 employees (excluding 4 hires from Michigan, Indiana, and Kansas and 3 hires for which the initial job could not be determined), 119 of whom were black. Comparing this number broken down by EEO category with the availability data contained in the census for each county based upon EEO category, Dr. Peterson concluded that 118.089 blacks would have been the expected number of blacks hired if defendant’s hiring choices were being made by random selection. Thus, defendant’s actual black hires exceeded the expected black hires utilizing this type of analysis. Dr. Peterson expressed this in terms of a positive standard deviation of .0939966. (See Defendant’s Exhibit 4, § EEOAVLF pp. 33-34). Excluding rehires in the second application of this analysis, Dr. Peterson concluded that defendant hired a total of 1,148 employees between 1976 and 1980 in its St. Louis division. There were 111 people hired who were black. There were 112.34 expected black hires in the St. Louis Division based upon 1980 census data broken down by EEO category. This is expressed in terms of a negative standard deviation of 0.17038 which, Dr. Peterson concluded, was well within the most stringent 2.00 standard deviation test applied by courts for statistical significance. Dr. Peterson’s third demographic analysis was organized by job locations. This study shows the counties from which employees were hired for each job location. This study also shows the availability of blacks based upon census data for each job location EEO category by EEO category. Dr. Peterson statistically weighted each county’s participation in the black availability pool for each job location by mathematically qualifying each county’s participation based upon the numbers of employees the ICG actually drew from each county. The results of Dr. Peterson’s third demographic analysis were not significantly different from the results of the first two analyses discussed previously. In terms of standard deviations, Dr. Peterson concluded that the defendant’s performance was well within the norms adopted by courts for a determination of statistical significance in discrimination cases. In addition to this analysis of the five-year time frame, Dr. Peterson also performed the same type of analyses on more focused periods of time: December 8, 1978 through January 9, 1981, and 1979 and 1980. Dr. Peterson indicated that the purpose for refocusing the analysis to the December 8, 1978 through January 9, 1981 time period was that it conforms with the time parameters of the class claims under Title VII. He stated that the analysis of 1979 and 1980 was performed because of defendant’s contention that under § 1981 the time frame of plaintiffs’ claims was limited to those two years. The type of analyses performed by the defendant for each of those segments of time was exactly the same analyses performed for the broader period of 1976 through 1980 and the results were similar. Ironically, shortly after Dr. Peterson became involved in the case, he performed an applicant flow analysis substantially similar to that performed by Dr. Grossman on information provided by the defendant. The results of this analysis were similar to the results Dr. Grossman obtained. Dr. Peterson stated that he performed such an analysis to ascertain the defendant’s potential exposure in this case. 2. The Defendant’s Anecdotal Evidence. Division superintendent, James Johnson, testified about the defendant’s policy of hiring locally. He also testified at length about the defendant’s reasons for having such a policy. Mr. Johnson stated that it was his experience over many years oh the railroad that employees in general will bump back to job locations closer in proximity to their place of residence. These bumping rights are guaranteed to employees under their union contracts. Mr. Johnson testified at length about the additional expense and administrative difficulties that bumping practices create for the defendant. It was Mr. Johnson’s testimony that it was more beneficial for the company to hire individuals close to their initial job location than it was to hire individuals from great distances who would usually end up bumping back to job locations nearer to their residence. Mr. Johnson pointed out that, under certain circumstances, the defendant is obligated pursuant to collective bargaining agreements to pay moving expenses and other expenses for certain employees who are moved from jobs close to their residence to jobs farther away. Mr. Johnson also testified that the only opportunity that the defendant had to obtain a local work force was at the time of initial hire. Superintendent Johnson testified that absenteeism and tardiness are high among persons who live great distances from their job location. He testified that, as a matter of policy, it is better for the company to attempt to hire individuals who reside close to their job location in order to correct for these problems. Superintendent Johnson also testified that in emergencies, defendant is required to call laborers and other employees in on short notice and that it is difficult to do so if the employees live great distances from the job site. In order to correct for this situation at the time of initial hire, the defendant for many years has had a policy of hiring individuals close to job locations who would be more readily available for emergency service. E. The Court’s Analysis of The Plaintiffs’ Case. 1. The Accuracy of the Data Upon Which the Plaintiffs Relied. Prior to analyzing the statistical evidence, the Court feels obliged to comment at length on the accuracy of data upon which the plaintiffs’ expert relied. One of the defendant’s major attacks on Dr. Grossman’s applicant flow analysis was that the applicant information was inaccurate both in terms of numbers of applicants and their race designation. This attack generated numerous arguments and several motions both in the few months before trial and during trial. The plaintiffs argued that the defendant should be bound by its interrogatory answer and the data contained in its MEARS. In September of 1985, the plaintiffs asked the Court to allow them to obtain an additional expert to perform a demographic analysis because the applicant data upon which Dr. Gross-man relied to perform his applicant flow analysis was inaccurate and because Dr. Grossman was not qualified to perform a demographic analysis. The defendant venomously opposed a second expert arguing that the plaintiffs were aware or should have been aware of the discrepancies in the MEARS and the answers to the interrogatories in March of 1984. Viewing the dispute as merely involving a difference of approach to the case, the Court denied the plaintiffs’ request, and set the case for trial on its November, 1985 docket. After listening to all the arguments, and the testimony concerning the preparation of the MEARS, the Court finds that any inaccuracies in the MEARS did not substantially affect the plaintiffs’ applicant flow analysis. Dr. Grossman obtained the applicant information from the MEARS, the only available source of applicant data other than the applications themselves. The MEARS were the monthly employment activity reports and were used to monitor the defendant’s affirmative action plan. They were the only source of applicant information supplied to the defendant’s equal employment opportunity department (EEO). Mary Annette Lane, the defendant’s hiring person for the St. Louis Missouri division, testified that she prepared the MEARS monthly in Carbondale, Illinois and sent them to Ellen Grossberg, the defendant’s EEO director. The MEARS contained a tally of both the number of applicants and the number of hires by sex, race, and EEOC category. (See for example Plaintiffs’ Exhibit 26). Lane testified that the applicant numbers were arrived at by merely counting the number of applicants she received for a given month. Lane indicated that after receiving the application from the applicant, she or her assistant Ellen Mertz place either a “1” or “2” on the application. The number “1” indicated that the applicant was a white male, and the number “2” indicated that the applicant was a black male. Lane added that if no number appeared on the application she would assume that the applicant was a white male. She admitted that this assumption would make the MEARS inaccurate. She also speculated that a possible explanation for the higher number of applications than that reflected on the MEARS for certain months could have resulted from the fact that if an applicant were hired, his application was placed in his employment file and taken from the pile that she would count to arrive at the applicant side of the MEARS. She also alluded to the fact that some applications could have been lost. However, Lane testified that during the years in question she believed that the MEARS were accurate. As indicated earlier, Dr. Peterson performed an analysis to check the accuracy of the applicant count data and the hiring data contained in the MEARS. Dr. Peterson made a comparison of the MEARS applicant count or applicant flow data with the actual applications in the possession of the defendant. Dr. Peterson testified that this comparison showed that for some months there were more applications than the number of applicants reported on the MEARS. Dr. Peterson also testified that for other months the defendant had more actual applications than the number of applicants reported on the MEARS. Dr. Peterson’s analysis of plaintiffs’ data base is contained in Defendant’s Exhibits 39 through 44. Based upon this evidence, Dr. Peterson concluded that the monthly count of hires and applicants contained in the MEARS was completely inaccurate. There is no question in the Court’s mind that the information contained in the MEARS was incomplete and to a certain extent inaccurate. The more important issue, however, is whether these flaws affected Dr. Grossman’s analysis. The Court finds that they did not. Dr. Grossman’s analysis was not based on the actual numbers of black and white applicants, but on the relative percentage of black applicants to white applicants. Unless it can be shown that either the numerator alone or the denominator alone of this percentage is inaccurate or that both are distorted other than on a random basis, the percentages are not affected. Here, only one flaw could have affected the numerator or denominator other than on a random basis and that was Lane’s assumption that no number on the application indicated that the applicant was a white male. However, the bias created by this assumption could only favor the defendant because it would increase the number of white applicants resulting in a lower percentage of blacks that one would expect should be hired. There is nothing to indicate that the losing of applications, the noninclusion of hired applicants, or other errors occurred other than on a random basis. Of course, the Court is mindful of the fact that because the data was incomplete, the plaintiffs’ analysis is based only on a sample of the actual population of applicants. However, because the sample size was so close to the actual population, the Court is convinced that the sample was an accurate representation of the entire applicant pool. Although not necessary because of the above finding, the Court is persuaded by the plaintiffs’ argument that the defendant should be estopped from attacking the inaccuracy of the data it supplied the plaintiffs. In the typical Title VII hiring case, where statistics play a major role, it is the defendant who argues for the use of an applicant flow analysis and the plaintiff who attacks the accuracy of the defendant’s data. Here the table is reversed. There is a certain amount of irony in an employer defending in part a major discrimination case on the theory that the information it keeps for its own affirmative action plan and which it supplies to the plaintiffs via an interrogatory answer is inaccurate and should be discarded. Further, its interesting to note, that neither Lane, the person in charge of hiring, nor Grossberg, the director of the defendant’s EEO department, questioned the accuracy of numbers which, taken at face value, showed major differences between the number of blacks who applied and who were hired. It was not until after Dr. Peterson performed an applicant flow analysis and then later tested the data upon which it was based that the accuracy of the data was questioned. Once the defendant learned of the inaccuracies, it did not inform the plaintiffs, nor did it amend its interrogatory answer. The defendant points out that the plaintiffs could have very easily verified the accuracy of the MEARS and the interrogatory answer because they had the applications and they knew of the existence of the defendant’s 1591 forms. However, the Federal Rules of Civil Procedure do not require a party to verify an opposing party’s answer to an interrogatory; they do require that a party update his prior response if he obtains information upon the basis of which he knows that the prior response was incorrect. Fed.R.Civ.P. 26(e)(2). The Court realizes that the defendant cannot be held liable for keeping bad records nor should it be penalized for doing so. However, where the plaintiffs are totally at the mercy of the defendant to provide them with data, and the defendant does not immediately notify the plaintiffs of problems with the data, the defendant should not be able to gain an advantage by arguing that the Court should disregard the plaintiffs’ analysis because it was generated from inaccurate and incomplete data. 2. The Plaintiffs’ Applicant Flow Analysis. While the Court has found that the irregularities of the date upon which Dr. Gross-man based his applicant flow analysis did not affect his analysis, the defendant has also argued that Dr. Grossman made a number of faulty assumptions. First, the defendant indicates that in preparing his applicant flow analysis, Dr. Grossman assumed that an application filed at any time during the year was active throughout the year, an assumption contrary to the testimony of Lane who testified that an application remains pending for only six months. Second, and in a related vein, Dr. Gross-man matched applications filed in the later months of a year with hires that occurred earlier in the year. The Court finds that neither assumption substantially affected the validity of Dr. Grossman’s analysis. Both assumptions deal with timing problems inherent in any applicant flow analysis. However, where the analysis encompasses a number of years, as here, these timing problems are virtually eliminated. For example, while in year one, applications filed late in the year are matched against hires made earlier in the year, in year two, applications filed early in the year are matched against hires made early in the year even though those hired probably filed their applications late in year one. This “rollover” effect nullifies any serious timing problem so long as a number of years are considered. Here, Dr. Grossman’s analysis included the years 1977 through 1980. A good deal of the defendant’s argument is undermined by the fact that Dr. Peterson’s applicant flow analysis made the same assumptions. The Court concludes therefore, that Dr. Grossman’s analysis was not affected by flaws in the data base, nor by the two above assumptions. 3. The Plaintiffs’ Static Work Force Statistics and Anecdotal Evidence. In addition to the above applicant flow analysis, the plaintiffs presented two additional pieces of evidence to buttress the results indicated by that analysis. First, the plaintiffs presented a great deal of testimony concerning the March 27, 1979 incident, wherein a number of the class members applied for jobs at the Carbon-dale, Illinois office. Second, the plaintiffs offered into evidence various exhibits that showed the composition of the defendant’s St. Louis Division work force by race at various times between 1976 and 1980. At the outset, the Court emphasizes again that it has viewed this case essentially as a statistical analysis case. Neither of the above two types of evidence alone, or together would establish a prima facie case for the plaintiffs. The Court finds from the static work force statistics that the defendant employed, for the period between 1976 and 1980, a great deal more whites than blacks and that the highest concentration of blacks appeared in the lower paying unskilled jobs. The Court also finds, however, based on the evidence presented by the defendant that this disparity is a result of hiring decisions mad