This is an age discrimination case1 involving a work force reduction in the Southern Region of Western Electric Company's (Western) Telephone Switching System Installation Operations. Western appeals from a judgment of the district court holding that it committed both individual and "pattern and practice" employment discrimination in its process of selecting installation supervisors for demotion. We reverse.

Western, beginning in late 1974, experienced a substantial downturn in its volume of installation business nationwide. As a result, Western drastically reduced its installation work force at all levels during the years of 1974, 1975 and 1976.

This case concerns only the demotion of installation supervisors in the Southern Region of Western's operations. The Southern Region is divided into five installation areas: the Carolinas; Georgia; Florida; Alabama/Kentucky/Tennessee; and Louisiana/Mississippi. Each installation area is divided into several sub-areas, which consist in turn of several offices or "orbits." The Western employees involved in this action supervised the work of "installers," who are the hourly, unionized Western employees who perform the hands-on work of installing Western's telephone systems.

Installation supervisors are ranked according to five categories of "point" levels: 320, 370, 440, 500 and 590. The point levels correspond with amount of pay and responsibility. The 320's are front line supervisors; 370's, 440's and 500's are responsible for installation activities within orbits; and 590's oversee subordinate supervisors and projects involving many installers within installation areas.

Western, in late 1974, instituted its program of work force reduction among the supervisors because of the downturn of business and consequent discharge of installers.2 Area supervisors were directed to reduce the number of supervisors to maintain an approximate ratio of one supervisor to seven installers.3 Western's written policy guidelines for demotion of supervisors in the Southern Region required that age, sex, race or religion play no role in any decision, and that "ability, performance, potentiality, term of employment, and the needs of the business" be the criteria used in demotion selections. The number of supervisors was reduced by three types of personnel actions: (1) demotion of lower-level supervisors to installers; (2) demotion of higher-level supervisors to lower point levels;4 and (3) certain supervisors who were pension-eligible were offered the opportunity to retire with a separation allowance of a year's pay.5

There were approximately 725 supervisors of all point levels in the Southern Region in mid-1974, before the work force reduction began. Approximately 213 individual supervisors were demoted over the next two and one-half years. Approximately 225 individual supervisors were affected by this total reduction-in-force program.

The present action was brought initially by the Secretary of Labor on May 15, 1978. The Equal Employment Opportunity Commission (EEOC) was substituted as plaintiff in January, 1980. The complaint charged that age considerations tainted Western's selections of installation supervisors for demotion.6 The EEOC's proof was essentially two-fold: first, statistical evidence was offered to show that older supervisors were selected for demotion disproportionately with their percentage representation in the supervisory work force; second, sixteen individual supervisors from the Florida and Carolinas areas in the protected 40 to 65 age group testified that they were selected for demotion because of their age.

Western contended at trial that its actions were not age-discriminatory and were taken solely for legitimate business reasons. Western asserted that the EEOC's statistical analyses were flawed, and that the EEOC's statistics were not legally significant. Western further argued that taking ability into account, any statistical disparity in the demotion of older and younger supervisors was explained. To rebut the accusation of discrimination against the EEOC's sixteen witnesses, Western presented the testimony of the area supervisor responsible for each decision. In each case, the supervisor presented a detailed non-discriminatory explanation for his decision.

Following a bench trial, the district court issued Findings of Fact and Conclusions of Law initially drafted by the EEOC, and entered an Interim Judgment and Order. The court held that Western had willfully violated the ADEA. It essentially found that Western discriminated against the 16 supervisors individually, and that by "pattern and practice," Western had discriminated against supervisors in the Southern Region, ages 40 to 65, as a group.7 The Interim Judgment and Order identified in an appendix 192 supervisors who might have been affected by the violations, and provided for Stage II proceedings to determine the relief due the individuals listed.8 The district court stayed those proceedings pending the disposition of this appeal.

Western raises two principal contentions on appeal. First, it argues that the district court erred in finding "prima facie cases" of discrimination as to the 16 individual claimants, and that, in any event, Western presented unrebutted business reasons for demoting the 16 individuals. Second, Western argues that the EEOC failed to prove a prima facie case of pattern or practice discrimination. We agree that Western presented unrebutted business reasons for the demotions and with its second contention. Since we reverse for those reasons, it is unnecessary to address the other possible grounds for reversal advanced by Western.

* To establish discrimination in the individual cases, the EEOC must prove by a preponderance of the evidence that "but for" Western's motive to discriminate against an older supervisor, he would not have been demoted. Lovelace v. Sherwin-Williams Co., 681 F.2d 230, 239 (4th Cir.1982); Loeb v. Textron, Inc., 600 F.2d 1003, 1019 (1st Cir.1979). The EEOC may meet its burden "under ordinary principles of proof by any direct or indirect evidence relevant to and sufficiently probative of the issue." Lovelace, 681 F.2d at 239. Alternatively, it may rely on the judicially created proof scheme for Title VII cases, see McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S. Ct. 1817, 36 L. Ed. 2d 668 (1973) and Texas Dept. of Comm. Affairs v. Burdine, 450 U.S. 248, 101 S. Ct. 1089, 67 L. Ed. 2d 207 (1981), which has been adapted for application in ADEA litigation. See, e.g., Fink v. Western Electric Co., 708 F.2d 909 (4th Cir.1983); Lovelace, 681 F.2d 230; Smith v. University of North Carolina, 632 F.2d 316 (4th Cir.1980); Loeb, 600 F.2d 1003.

The McDonnell-Douglas proof scheme involves a three stage process. First, the plaintiff must establish a "prima facie case." By establishing a "prima facie case," the plaintiff creates a rebuttable "presumption that the employer unlawfully discriminated against" him, and the burden of producing evidence on the issue shifts to the defendant. Burdine, 450 U.S. at 254, 101 S.Ct. at 1094. The defendant, at stage two, must rebut "the presumption by producing evidence that the plaintiff was [demoted] ... for a legitimate, nondiscriminatory reason." Id. at 255, 101 S.Ct. at 1094. The employer is not required to prove absence of discriminatory motive, but merely articulate some legitimate reason for its action. Smith, 632 F.2d at 332-33. If the defendant's evidence carries this burden, the presumption in favor of the plaintiff is destroyed. 450 U.S. at 255, 101 S. Ct. at 1094. At the third stage, the plaintiff bears the burden of demonstrating that the proffered reasons were pretextual or untrue. Id. at 256, 101 S.Ct. at 1095. This burden merges with the plaintiff's ultimate burden of persuasion on the motivational issue. Id.; Lovelace, 681 F.2d at 240-41.

As adapted from the McDonnell-Douglas line of cases, the elements of the plaintiff's "prima facie case" in the usual ADEA case are: (1) the plaintiff is in the protected age group; (2) the plaintiff was discharged or demoted; (3) at the time of discharge or demotion, the plaintiff was performing his job at a level that met his employer's legitimate expectations; and (4) following his discharge or demotion, the plaintiff was replaced by someone of comparable qualifications outside the protected class. See Lovelace, 681 F.2d at 239-41; Marshall v. Goodyear Tire & Rubber Co., 554 F.2d 730 (5th Cir.1977). It is readily apparent that the fourth element will not fit well in most reduction-in-force cases. See McCorstin v. United States Steel Corp., 621 F.2d 749 (5th Cir.1980). In such cases, the claimant's job is often phased out and thus he would be unable to point to a replacement. While we do not believe that a claimant in such a case should be precluded, for all practical purposes, from using the McDonnell-Douglas proof scheme by being required to satisfy the fourth element, the claimant should not be permitted to establish a "prima facie case" by satisfying only the first three elements.

It was never intended that a "prima facie case" of employment discrimination could be established only by mechanically applying the precise requirements set forth in McDonnell-Douglas. The Supreme Court in McDonnell-Douglas recognized that "the facts necessarily will vary" among cases, and said that "the specification above of the prima facie proof required ... is not necessarily applicable in every respect to differing factual situations." 411 U.S. at 802 n. 13, 93 S. Ct. at 1824 n. 13. The fourth element required to establish a "prima facie case" of discrimination was intended to demonstrate unequal treatment by the relatively objective evidence of replacement by a member of a non-protected class. See McCorstin, 621 F.2d at 753. Although a plaintiff in a reduction-in-force case normally is unable to show that he was replaced by such a person, he can nevertheless establish a "prima facie case" by producing some other evidence that the employer did not treat age neutrally. See Fink, supra, 708 F.2d at 916-19; Williams v. General Motors Corp., 656 F.2d 120, 129-30 (5th Cir.1981), cert. denied, 455 U.S. 943, 102 S. Ct. 1439, 71 L. Ed. 2d 655 (1982).

The district court in this case permitted the EEOC to establish a prima facie case based on a showing of only the first three elements, namely that "each claimant (1) was a member of the protected age group; (2) was qualified for the job and was performing up to his employer's expectations; and (3) was demoted and/or subsequently retired, despite his qualifications and performance." This was an error of law. Before requiring Western to produce evidence of legitimate business reasons for the demotion selections, the district court should have required the EEOC either to show that, upon the claimants' demotions, persons outside the protected age class were retained in the same positions, or to produce some other evidence that Western did not treat age neutrally in its demotion selections. It is true that the district court in a minority of the individual cases found a claimant demoted and a younger supervisor retained. That is not a sufficient evidentiary basis, however, upon which it can be inferred that this was true in the other individual cases. In any event, Western presented legitimate business reasons why it demoted each of the 16 individuals instead of others in their areas. The district court subsequently found the reasons advanced by Western for demoting the 16 individuals to be pretextual. Assuming then that the EEOC had established a prima facie case requiring Western's response, the evidence yet leaves us with the definite and firm conviction that Western demoted all 16 individuals for legitimate business reasons which were not pretextual.9 Those reasons were grounded on qualification and geographical considerations. Whether we view that evidence as rebutting the prima facie case of these individuals or as establishing legitimate business reasons, the trial court's holdings as to the individuals must be reversed.

The EEOC attempted to establish prima facie cases for the 16 individual claimants primarily by their testimony. In rebuttal, Western presented evidence of the guidelines area supervisors were directed to follow10 and the testimony of the area supervisors responsible for each particular demotion. The EEOC has withdrawn its claim on appeal as to one supervisor, Otis Foster.11 Of the remaining fifteen, seven at the time of their demotion were ages 40 through 46, and eight were age 50 and over.

The evidence concerning the seven supervisors who were demoted at ages 40 through 46 demonstrates the weakness of these cases. Robert Shampy, for example, was demoted to installer at age 42. He was demoted, however, along with J.L. Allred, age 40, the only other supervisor at Laurinburg, because work in Laurinburg had decreased to the point that no supervisor was needed.12 Arch Pugh, at age 41, was also demoted to installer. Pugh admitted that his demotion was the result of his limited expertise, and that he had never requested advanced training. In some cases, such as that of Thomas Harrington (age 41), older supervisors had been retained instead, and in other cases, such as those of Charles Harris (age 41) and Eugene Bain (age 44), the younger supervisor retained had been a supervisor longer. Indeed, all demotees in this age group were very close in age to most of the younger supervisors who were not demoted.

An illustrative example in the 50 and over group is Neal Levine, who was demoted at age 54 from a level 370 to a level 320 supervisor. The area supervisor testified that Levine was demoted because he "was not a good supervisor" and let his subordinates "walk all over him," and cited specific examples. His further uncontroverted testimony was that supervisors who were not demoted lacked these deficiencies and were otherwise good supervisors. Samuel Gantt, another example, was demoted at age 51 from a level of 440 to a level of 370. While his job responsibilities did not change, his point level was reduced because activity diminished. Indeed, Gantt testified that he had no reason to believe that age was a factor in his demotion.

In each individual case, the area supervisor responsible for the demotion testified concerning legitimate business reasons for each demotion. They testified that while most of the supervisors demoted were good supervisors, poor business conditions required a reduction in the number of supervisors, and the supervisors selected for demotion were less qualified, based on their performance and expertise, than those supervisors retained. We find no reasons legally sufficient to discredit this testimony. It is true that the trial court held that the qualification assessments were subjective. That appellation, however, does not convert an otherwise legitimate reason into an illegal one. See Burdine, 450 U.S. at 256-57, 101 S.Ct. at 1095-96; Ward v. Westland Plastics, Inc., 651 F.2d 1266, 1270 (9th Cir.1980).

The evidence in its entirety demonstrates that the district court erred in each of its findings wherein it found employment discrimination on the basis of age against each of the 16 claimants.

The ultimate factual issues in the EEOC's group discrimination action are whether there was a "pattern or practice" of disparate treatment against the group and, if so, whether the differences were based on age considerations. Teamsters v. United States, 431 U.S. 324, 97 S. Ct. 1843, 52 L. Ed. 2d 396 (1977).13 The EEOC, as plaintiff, bears the initial burden of making out a prima facie case of discrimination. Id. at 336, 97 S.Ct. at 1855. In alleging systemwide pattern or practice discrimination, the EEOC was required to prove more than mere occurrence of isolated or sporadic discriminatory acts. Id. Its task was to establish by a preponderance of the evidence that age discrimination was Western's "standard operating procedure--the regular rather than the unusual practice." Id.

The EEOC presented its case to the district court alleging pattern or practice discrimination on the basis of age in the demotion of installation supervisors ages 40 to 65 in the Southern Region. The EEOC's evidence clearly does not establish a prima facie case of age discrimination against that group. The EEOC's only statistical expert admitted at trial that there was no statistically significant difference between the number of supervisors in that age group who were actually demoted and the number which would have been expected to be demoted, in each installation area and in the Southern Region as a whole. The district court, therefore, clearly erred in finding a prima facie case of "pattern and practice" discrimination against that group, and the EEOC concedes this on appeal.

The EEOC contends on appeal, however, that the evidence established a prima facie case of age discrimination against a different group--the 50 to 65 age group14 for four of the installation areas, excluding Georgia.15 We disagree. The EEOC did not plead, or even argue, to the district court the sub-group case it now presents to us on appeal. The EEOC does not contend that it lacked sufficient discovery or opportunity to properly plead and try its case, and Western's defense focused on the pattern or practice claim for the 40 to 65 age group which the EEOC presented at trial. The district court's findings of fact and conclusions of law are premised on that group claim. Indeed, most of those findings and conclusions were adopted almost verbatim from the proposed findings of fact and conclusions of law drafted by the EEOC.16 The EEOC now concedes that those findings and conclusions of "pattern and practice" discrimination against the 40 to 65 age group were clearly erroneous.

This creates at least a suspicion that the EEOC strained the litigated facts to produce the findings it drafted and which were accepted by the district court. It is difficult for an appellate court under these circumstances to generate enthusiasm for accepting yet a new theory on appeal based on a proposed factual finding not stressed at trial. The propriety of dividing a group into smaller units for statistical analysis depends on the circumstances of each case. See E.E.O.C. v. United Virginia Bank, 615 F.2d 147, 153 (4th Cir.1980); E.E.O.C. v. Datapoint Corp., 570 F.2d 1264, 1269 (5th Cir.1978). The facts in the case sub judice vary considerably from those in United Virginia Bank and Datapoint, but here as in those cases we do not feel the creation of a new sub-class is warranted.

Even if we considered it appropriate to address that revised claim, the EEOC has not carried its burden of proving by a preponderance of the evidence that Western engaged in pattern or practice discrimination against the 50 to 65 sub-group because of their age. The principal statistical evidence relied on by the parties consisted of standard deviation analysis tables.17 The EEOC introduced at trial four standard deviation tables, which are reproduced in an appendix to this opinion.18 These tables analyze by area and the Region the disparity between the expected and actual number of supervisors demoted ages 50 and over for the period July 1, 1974, to July 1, 1976.19 The expected number of demotions were calculated by multiplying the percentage of supervisors ages 50 and over in a particular area or the Region by the total number of supervisors demoted in that area or the Region. The standard deviations in tables 99A and 100A were calculated using the binomial formula,20 and the standard deviations in tables 99B and 100B were calculated using the hypergeometric formula.21 In general, the tables show that the number of standard deviations for the whole Southern Region for the over 50 age group is greater than three, and that the number of standard deviations for each of the five areas varies between -.5 and 5.2,22 with the analyses using the hypergeometric formula noticeably favoring the plaintiff.

The Supreme Court has provided some guidance to lower courts in determining the legal significance of statistical evidence. See Hazelwood School District v. United States, 433 U.S. 299, 97 S. Ct. 2736, 53 L. Ed. 2d 768 (1977); Castaneda v. Partida, 430 U.S. 482, 97 S. Ct. 1272, 51 L. Ed. 2d 498 (1977). As to methodology, the Supreme Court has considered percentage comparisons and standard deviation analyses of those comparisons. In interpreting Hazelwood and Castaneda, we have held that courts in this circuit must apply a standard deviation analysis before drawing conclusions from straight percentage comparisons. Moultrie, 690 F.2d at 1082. The cases from the Supreme Court and this court discussing the probative value of statistics relative to the number of standard deviations were based on statistical evidence utilizing the binomial distribution formula.23 The Supreme Court has said that "[a]s a general rule ..., if the difference between the expected value and the observed number is greater than two or three standard deviations, then the hypothesis that the [selection process] was random would be suspect...." Castaneda, 430 U.S. at 497 n. 17, 97 S. Ct. at 1281 n. 17; see Hazelwood, 433 U.S. at 308 n. 14, 97 S. Ct. at 2742 n. 14. We have further said that courts "should be extremely cautious in drawing any conclusions from standard deviations in the range of one to three." E.E.O.C. v. American Nat. Bank, 652 F.2d 1176, 1192 (4th Cir.1981).

The Supreme Court has noted that the usefulness or weight of statistical evidence "depends on all of the surrounding facts and circumstances." Teamsters, 431 U.S. at 340, 97 S.Ct. at 1856. For example, the Supreme Court in Hazelwood said: "When special qualifications are required to fill particular jobs, comparisons to the general population (rather than to the smaller group of individuals who possess the necessary qualifications) may have little probative value." 433 U.S. at 308 n. 13, 97 S. Ct. at 2742 n. 13. Like any other type of circumstantial evidence, statistical evidence "must not be accepted uncritically,"24 and because many statistical models used in discrimination cases by professional statisticians are sophisticated and complex, courts must give "close scrutiny [to the] empirical proof" on which the models are erected.25 Courts should guard against the use of data which may have been "segmented and particularized and fashioned to obtain a desired result."26

The EEOC and the district court in its findings of fact and conclusions of law apparently relied primarily on the standard deviation studies based on the hypergeometric distribution formula. The hypergeometric formula obviously is different from the binomial formula which the Supreme Court and our cases have relied on to date, and as can be seen from the EEOC's exhibits, the hypergeometric tables clearly favor the EEOC. It is not feasible, therefore, without corresponding conversion, to discuss compliance of such an analysis with the binomial distribution numerical guidelines. That process, however, is not necessary to the decision of this case. Even relying on the more favorable hypergeometric studies without conversion, they are insufficient, in view of all the evidence, to establish a prima facie case of pattern or practice age discrimination against the sub-group of supervisors alleged by the EEOC on appeal.

The EEOC concedes that it failed to establish its pattern or practice case for all supervisors ages 40 to 65 in the Southern Region. Viewing the plaintiff's binomial tables for the ages 50 to 65 supervisors, the number of standard deviations exceeds three for only two of the areas, and the corresponding hypergeometric tables show that the number of standard deviations exceeds three for only three areas. The EEOC's statistics for the Georgia area show, if anything, that older supervisors were favored.

The data base upon which the EEOC conducted its statistical analyses was accepted as correct by the trial court over Western's protestations. The trial court discredited Western's evidence as to how it rated supervisors for performance and expertise, thus rejecting Western's contentions that supervisors of unequal qualifications were included by the EEOC in the same statistical base. In addressing the individual claims of discrimination in Part I, we hold that the trial court was clearly erroneous in finding that qualifications played no role in Western's process of selecting supervisors for demotion. It is clear then that Western considered relative qualifications and geographic location of supervisors in making demotion selections. Thus, the probative force of the EEOC's statistical evidence is even weaker than if the analyses had taken into account relevant qualifications and location. See Hazelwood, 433 U.S. at 308 n. 13, 97 S. Ct. at 2741 n. 13; Federal Reserve Bank of Richmond, 698 F.2d at 659-60; United Virginia Bank, 615 F.2d 149.

The statistical evidence produced by the EEOC admittedly has some probative force, but at most it is inconclusive circumstantial evidence. The EEOC, moreover, is unable to bolster its marginal statistical case with individual cases of discrimination because it failed to establish discrimination as to any of the 16 witnesses.27 To the contrary, the evidence in its entirety creates strong inferences that Western attempted to achieve its overall reduction-in-force objectives in a fair and legitimate way. We are thus of the firm and definite opinion that the evidence did not establish a pattern or practice of discrimination as to supervisors in the Southern Region ages 40 to 65 or as to any sub-group, and that the district court was clearly erroneous in its findings to that effect.28

The judgment of the district therefore is reversed.

Reversed.

     ANALYSIS OF SOUTHERN REGION DEMOTIONS AND RECLASSIFICATIONS
                    AND ALL REQUESTED RETIREMENTS
       BY AGE - COMPARISON OF GROUP 50 YEARS OF AGE AND OLDER
                  WITH GROUP UNDER 50 YEARS OF AGE
               FOR PERIOD JULY 1, 1974 TO JULY 1, 1976
                                2          3          4         5
                     1        Total        %        Total       %
                   Total     50 And     50 And      Under     Under
     Area          Supv.     Over **     Over        50         50
---------------------------------------------------------------------
Carolinas           136         21       .1544       115      .8456
Georgia             124         22       .1774       102      .8226
La. Mississippi      96         21       .2188        75      .7812
Florida             204         38       .1863       166      .8137
Ala., Kentucky,     165         37       .2242       128      .7758
 Tennesee
Total Region        725        139       .1917       586      .8083
*1019_                                (TABLE CONTINUED)
                                8
     6            7         EXPECTED          9
   TOTAL     DEMOTIONS,    DEMOTIONS,     Difference                    11
DEMOTIONS,      REQ.          REQ.         Between         10        Number of
   REQ.      RETIREMENTS   RETIREMENTS      Actual      Standard     Standard
RETIREMENTS  & Reclass.    &  Reclass.     & Expected    Deviation   Deviations
& Reclass.    50 & Over   50 & Over (a)      (b)           (c)          (d)
-------------------------------------------------------------------------------
     58          16           8.9552        7.0448       2.7518     2.5601
     28           4           4.9672       - .9672       2.0214      .4785 *
     14           9           3.0632        5.9368       1.5469     3.8379
     85          22          15.8355        6.1645       3.5896     1.7173
     40          21           8.9680       12.0320       2.6377     4.5615
    225          72          43.1325       28.8675       5.9046     4.8890

(a) (Value in Column 3) X (Value in Column 6)

(b) (Value in Column 7) - (Value in Column 8)

(c) Standard Deviation Determined in Accordance with Formula

Adopted by the Court in EEOC v. United Virginia Bank,

615 F.2d 147, 151 (4th Cir. 1980) - Square Root of

(Col. 6 X Col. 3 X Col. 5)

(d) (Value in Col. 9) k (Value in Col. 10)

Note FNBNFN * Favors 50 Years of Age and Older Group

Note FNBNFN ** Based on Ages as of July 1, 1975 Except for Those Demoted Age is Based

on Date of Demotion.

     ANALYSIS OF SOUTHERN REGION DEMOTIONS AND RECLASSIFICATIONS
                    AND ALL REQUESTED RETIREMENTS
       BY AGE - COMPARISON OF GROUP 50 YEARS OF AGE AND OLDER
                  WITH GROUP UNDER 50 YEARS OF AGE
               FOR PERIOD JULY 1, 1974 TO JULY 1, 1976
                                2          3          4         5
                     1        Total        %        Total       %
                   Total     50 And     50 And      Under     Under
     Area          Supv.     Over **     Over        50         50
---------------------------------------------------------------------
Carolinas           136         21       .1544       115      .8456
Georgia             124         22       .1774       102      .8226
La. Mississippi      96         21       .2188        75      .7812
Florida             204         38       .1863       166      .8137
Ala., Kentucky,     165         37       .2242       128      .7758
 Tennesee
Total Region        725        139       .1917       586      .8083
                               (TABLE CONTINUED)
                                8
     6            7         EXPECTED          9
   TOTAL     DEMOTIONS,    DEMOTIONS,     Difference                    11
DEMOTIONS,      REQ.          REQ.         Between         10        Number of
   REQ.      RETIREMENTS   RETIREMENTS      Actual      Standard     Standard
RETIREMENTS  & Reclass.    &  Reclass.     & Expected    Deviation   Deviations
& Reclass.    50 & Over   50 & Over (a)      (b)           (c)          (d)
-------------------------------------------------------------------------------
     58          16           8.9552        7.0448       2.0917     3.3680
     28           4           4.9672       - .9672       1.7858      .5416 *
     14           9           3.0632        5.9368       1.4372     4.1308
     85          22          15.8355        6.1645       2.7483     2.2430
     40          21           8.9680       12.0320       2.3028     5.2249
    225          72          43.1325       28.8675       4.9069     5.8830

(a) (Value in Column 3) X (Value in Column 6)

(b) (Value in Column 7) - (Value in Column 8)

(c) Standard Deviation Determined From Hypergeometric

Distribution - A refinement for large samples of

the Formula Adopted by the Court in EEOC v. United

Virginia Bank, 615 F.2d 147, 151 (4th Cir. 1980)

(d) (Value in Col. 9) k (Value in Col. 10)

Note FNBNFN * Favors 50 Years of Age and Older Group

Note FNBNFN ** Based on Ages as of July 1, 1975 Except for Those Demoted. For Those

Demoted Age is Based on Date of Demotion.

     ANALYSIS OF SOUTHERN REGION DEMOTIONS AND RECLASSIFICATIONS
       BY AGE - COMPARISON OF GROUP 50 YEARS OF AGE AND OLDER
                  WITH GROUP UNDER 50 YEARS OF AGE
               FOR PERIOD JULY 1, 1974 TO JULY 1, 1976
                                2          3          4         5
                     1        Total        %        Total       %
                   Total     50 And     50 And      Under     Under
     Area          Supv.     Over **     Over        50         50
---------------------------------------------------------------------
Carolinas           136         21       .1544       115      .8456
Georgia             124         22       .1774       102      .8226
La. Mississippi      96         21       .2188        75      .7812
Florida             204         38       .1863       166      .8137
Ala., Kentucky,     165         37       .2242       128      .7758
 Tennesee
Total Region        725        139       .1917       586      .8083
                                             9
                              8         Difference                      11
    6           7         EXPECTED        Between         10        Number of
  TOTAL     DEMOTIONS     DEMOTIONS       Actual       Standard      Standard
DEMOTIONS   & Reclass.   &  Reclass.     & Expected     Deviation    Deviations
& Reclass.  50 & Over   50 & Over (a)       (b)           (c)          (d)
-------------------------------------------------------------------------------
    57          15          8.8008         6.1992       2.7280     2.2724
    28           4          4.9672        - .9672       2.0214      .4785 *
    11           7          2.4068         4.5932       1.3712     3.3498
    81          21         15.0903         5.9097       3.5041     1.6865
    36          17          8.0712         8.9288       2.5023     3.5682
   213          64         40.8321        23.1679       5.7450     4.0327

(a) (Value in Column 3) X (Value in Column 6)

(b) (Value in Column 7) - (Value in Column 8)

(c) Standard Deviation Determined in Accordance with Formula

Adopted by the Court in EEOC v. United Virginia Bank,

615 F.2d 147, 151 (4th Cir. 1980) - Square Root of

(Col. 6 X Col. 3 X Col. 5)

(d) (Value in Col. 9) k (Value in Col. 10)

Note FNBNFN * Favors 50 Years of Age and Older Group

Note FNBNFN ** Based on Ages as of July 1, 1975 Except for Those Demoted. For Those

Demoted Age is Based on Date of Demotion.

     ANALYSIS OF SOUTHERN REGION DEMOTIONS AND RECLASSIFICATIONS
       BY AGE - COMPARISON OF GROUP 50 YEARS OF AGE AND OLDER
                  WITH GROUP UNDER 50 YEARS OF AGE
               FOR PERIOD JULY 1, 1974 TO JULY 1, 1976
                                2          3          4         5
                     1        Total        %        Total       %
                   Total     50 And     50 And      Under     Under
     Area          Supv.     Over **     Over        50         50
---------------------------------------------------------------------
Carolinas           136         21       .1544       115      .8456
Georgia             124         22       .1774       102      .8226
La. Mississippi      96         21       .2188        75      .7812
Florida             204         38       .1863       166      .8137
Ala., Kentucky,     165         37       .2242       128      .7758
 Tennesee
Total Region        725        139       .1917       586      .8083
                                             9
                              8         Difference                      11
*1019_     6           7         EXPECTED        Between         10        Number of
  TOTAL     DEMOTIONS     DEMOTIONS       Actual       Standard      Standard
DEMOTIONS   & Reclass.   &  Reclass.     & Expected     Deviation    Deviations
& Reclass.  50 & Over   50 & Over (a)       (b)           (c)          (d)
-------------------------------------------------------------------------------
    57          15          8.8008         6.1992       2.0868     2.9707
    28           4          4.9672        - .9672       1.7858      .5416 *
    11           7          2.4068         4.5932       1.2970     3.5414
    81          21         15.0903         5.9097       2.7276     2.1666
    36          17          8.0712         8.9288       2.2193     4.0233
   213          64         40.8321        23.1679       4.8312     4.7955

(a) (Value in Column 3) X (Value in Column 6)

(b) (Value in Column 7) - (Value in Column 8)

(c) Standard Deviation Determined From Hypergeometric

Distribution - A refinement for large samples of

the Formula Adopted by the Court in EEOC v. United

Virginia Bank, 615 F.2d 147, 151 (4th Cir. 1980)

(d) (Value in Col. 9) k (Value in Col. 10)

Note FNBNFN * Favors 50 Years of Age and Older Group

Note FNBNFN ** Based on Ages as of July 1, 1975 Except for Those Demoted. For These

Demoted Age is Based on Date of Demotion.