WIDENER, Circuit Judge:
On April 15, 1975, following a lengthy investigation, the Equal Employment Opportunity Commission (EEOC) filed this action against United Virginia Bank (bank or UVB) alleging that from July 2, 1965 to the date the complaint was filed UVB had discriminated against black applicants in its hiring procedures. The case was tried on its merits and the district court entered judgment for the defendant. The district judge gave only broad conclusory reasons for his decision, and we remanded the case with directions that the district court comply with FRCP 52(a) by entering detailed findings of fact. EEOC v. United Virginia Bank/Seaboard National, 555 F.2d 403 (4th Cir. 1977). We also directed the district court to allow both sides to introduce additional evidence. Id. at 406. On remand, the district court heard additional evidence and again found for the defendant. The EEOC appeals. We affirm for reasons somewhat different in some respects from those given by the district court in its opinion.
[149]*149A part of EEOC’s case was its claim that the bank discriminated against black applicants in maintaining a high school education requirement for employment and that it likewise discriminated in employing credit checks for employees in hiring. EEOC does not appeal from the adverse decision of the district court on these subjects. A part of its initial case also was that the bank had actually discriminated against some fourteen black applicants in'its failure to employ them. On remand this number may have been added to by twenty-two additional applicants. The district court denied relief to all of them, or for their account, and the EEOC acquiesced in the court’s holding in all instances except nine.
Hiring and its attendant initial job assignment is the only issue in this case. Significantly, there is no claim of other racial discrimination in such things as promotions, transfers, pay, etc., which, as often as not, appear in litigation of this nature.
At trial, the EEOC presented the following as evidence of discrimination: (1) The principal part of its case was a statistical comparison of black employees at UVB with black people in the total area work force; (2) a statistical comparison of black and white applicant to hire ratios; (3) specific policies which allegedly discriminate against blacks, e. g., credit checks and a high school education requirement; and (4) individual instances of discrimination, mainly relating to an alleged failure to hire qualified black applicants when openings were available.
The centerpiece and keystone of the EEOC’s ease, both in the district court and on appeal, is that the proper statistical comparison in this case is between the percentage of black employees working in various job classifications at UVB and the percentage of black people in the local labor force. Black people represent 27 percent of the local labor force and 9.5 percent (32/336) of UVB’s 1974 office/clerical workers, 78.6 percent of its service workers (n/i4), 33.3% (%) of its operative employees, and 1.8 percent (2/111) of its managers.
The EEOC also introduced evidence it deemed showed discriminatory distribution of blacks in UVB’s sixty job categories and that 43 of the 60 categories had no black employees.
The fundamental problem with the EEOC’s statistical evidence lies in the fact that UVB’s work force was compared with the work force as a whole. As the district court correctly recognized, this comparison was improper. It is clear that:
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.
Hazelwood School District v. United States, 433 U.S. 299, 308 n. 13, 97 S.Ct. 2736, 2742, 53 L.Ed.2d 768 (1977). Accord, Teamsters v. United States, 431 U.S. 324, 340 n. 20, 97 S.Ct. 1843, 52 L.Ed.2d 396 (1977). Indeed, we spelled out this proposition in our opinion on the first appeal in this case. EEOC v. United Virginia Bank/Seaboard National, 555 F.2d 403, 406 n. 7 (1977). The district court followed our opinion on remand and held that the proper comparison was between UVB’s work force and those individuals with the requisite qualifications for the particular job.
The EEOC, however, rigidly continues to argue that all the black local labor force is qualified for the office and clerical positions at UVB.1 This is simply not true. Tellers must be able to deal with the public, handle and account for money, and operate adding machines, typewriters and other office machines. The district court found that the entire percentage of black people [150]*150in the local labor force would not provide an appropriate statistical group for comparison with UVB black employees. Since this determination was a factual one, it will not be disturbed unless clearly erroneous. FRCP 52(a). We, therefore, need not engage in a detailed re-examination of the job qualifications in this case for the district court’s decision was obviously not clearly erroneous. The above discussion applies a fortiori to the EEOC’s attempt to compare the number of black managers at UVB with the number of black people in the local labor force.
The EEOC failed to present any evidence as to the percentage of persons in the labor force qualified to hold the various positions at UVB. The bank did introduce a report entitled “Manpower Information for Affirmative Action Programs” prepared by the Virginia Employment Commission. The report, based on U. S. Labor Statistics in part, shows the percentage of black and other minorities in various occupations in 1975 for the Norfolk, Virginia Beach, Portsmouth Standard Metropolitan Statistical Area (SMSA). The district court took the data in this report and converted it into percentages which could be compared to the major job categories at UVB, and summarized these percentages in a table.2 The district court may have mistakenly thought, however, that the Virginia Employment Commission report was for 1970 rather than 1975. If so, it compared the SMSA percentages with UVB’s 1970 work force when it should have compared them to the 1974 work force (the last full year before this action was initiated).
We do not intimate that the SMSA table necessarily demonstrates the percentage of black people in the labor force qualified to perform clerical work. However, the burden was on the EEOC to prove discrimination and to produce evidence to support its position. The SMSA report constitutes the only evidence which even remotely speaks to qualifications. Without these figures, the EEOC’s case is virtually without any statistical evidence to support it.
Further problems emerge when an attempt is made to compare the SMSA percentages with the UVB employees. When the EEOC prepared its figures for black employees and applicants, it made no effort to exclude employees hired prior to the effective date of Title VII (July 2, 1965). Thus, the figures EEOC presents are weighted against UVB to the extent that white employees hired prior to the time Title VII was in effect .were included in the employment figures. The Supreme Court has clearly stated that an employer who, after the effective date of the Act, “made all its employment decisions in a wholly nondiscriminatory way would not violate Title VII even if it had formerly maintained an all-white work force by purposefully excluding Negroes.” Hazelwood School District v. United States, 433 U.S. 299, 309, 97 S.Ct. 2736, 2742, 2743, 53 L.Ed.2d 768 (1976). Hazelwood involved a public employer, but it is clear that private employers are equally included within its rationale. Thus, in order to be valid, a statistical analysis such as the one presented here must exclude those persons hired prior to July 2, 1965 (the effective date of Title VII for private employers). To include pre-Act hires in the statistical analysis in this case would improperly weight the evidence and would tend to show present discrimination by an employer if it had discriminated prior to the effective date of the Act but had not discriminated after the Act took effect. It is therefore clear that the June 9, 1975 employment list as a whole was improperly used against UVB because the EEOC made no attempt to factor out [151]*151the pre-Act hires. Our own examination of the employment records (the June 1975 list) shows at least 92 1975 employees who were hired prior to July 2, 1965 and are almost certainly bound to have been employed in June 1974. How many other pre-Act employees quit, retired, or died during the year 1974 — 75 is not argued, but a brief reference to a disputed termination list indicates that the few pre-Act hires terminated during the period of June 1974-June 1975 were white, which of course would cause the CastanedaHazelwood analysis of standard deviations to be weighted further toward the defendant than it already is shown to be by the analysis which follows.
Furthermore, even if the pre-Act hires are not excluded, a comparison of the number of black employees of UVB with the number of black employees in similar jobs in the SMSA demonstrates that the difference between the two figures is probably not statistically significant. The following table, first prepared by the defendant from the district court’s table, which we have recalculated, shows the following comparisons:
Table I
Category
Official/ Managerial 4.8 111 5.33 2 2.25 1.48
Office/ Clerical 14.0 336 47.04 32 6.36 2.36
Operative 45.4 3 1.36 1 .87 .41****
Service 49.6 14 6.94 11 1.87 (-)2.14
[152]*152In Castaneda v. Partida, 430 U.S. 482, 496, 97 S.Ct. 1272, 51 L.Ed.2d 498 (1977), the Supreme Court indicated that figures falling outside two or three standard deviations from the expected values constitute a statistically significant difference. The number of operative workers involved here probably was too small for a statistically significant comparison, and in all events UVB was well over the mark in that category. We are left then with the official/managerial, office/clerical, and service categories which show that the observed numbers of black employees were 1.48, 2.36 and ( — )2.14 standard deviations from the expected value on a particular day in June 1975. This difference is borderline at best in view of the Supreme Court ruling that a variation of two or three standard deviations would normally be considered significant, and in any event the numbers in the case at bar are far different than those in Castaneda where the observed value was more than 29 standard deviations from the expected value over an 11-year period.
Furthermore, the statistical significance which might be attached to these figures falls out when it is remembered that this table includes at least 92 pre-Act hires. Eight of these (5 black and 3 white) were in the service department. Thirty-five (33 white and 2 blacks) were in the office/clerical category, and 49, all white, were in the official/managerial category. None were operatives. When these pre-Act hires are removed from Table I, the following are the results:
Table II
Category
Official/ Managerial
Office/ Clerical
Operative
Service
% Black Employees in SMSA_
4.8
14.0
45.4
49.6
Total No. Employed by UVB in 1974
Expected No. Black Employees
2.55
42.14
1.36
2.98
Observed No. Black Employees
Standard Deviation _
1.56
6.01
.87
1.25
No. Standard Deviations _
.35
2.02
.41
(-12.41
When the employees who were hired pri- or to the effective date of the statute are excluded from the table, we see that what doubtful statistical significance the statistical evidence had is reduced still further. The numbers of standard deviations in the two largest categories, comprising 354 out of 363 employees, were reduced in the official/managerial category from 1.48 to .35, and in the office/clerical category from 2.36 to 2.02. The standard deviations for the operative category remained the same, while in the service category the standard deviations increased from -2.14 to -2.41. But in the service category, which is yet less than 3 standard deviations, the number of employees in the statistical sample was reduced to 6, probably too small a number for statistical comparison. Mayor v. Educational Equality League, 415 U.S. 605, 620—621, 94 S.Ct. 1323, 39 L.Ed.2d 630 (1974). Even if the number of 6 employees, however, should be considered as a large enough statistical sample, the EEOC offered no evidence at all that any of these 6 employees who, being employees, must be presumed to be readily available as witnesses, had applied for any position other than in the service category. Even more importantly, neither did it offer any evidence at all that any of the 6 had the qualifications to hold positions in another category, especially in the office/clerical or official/managerial categories. The operative category seems not to have been seriously considered by anyone.
[153]*153Taking all of these things into consideration, we are of opinion the statistical evidence we have just discussed does not suffice to prove a prima facie case of discrimination.
Having shown that the district court incorrectly considered certain parts of the statistical evidence, its ruling that a prima facie case was established, if correct, must rely on the balance of the statistical information it considered of consequence. It found that there were no black employees in 43 of 60 specific job assignments and that those were largely the positions most highly paid; that black employees were in only 2 of 109 managerial positions; that in 1975 no black employees were employed in 5 specific job assignments in a total of 84 positions, and all 14 custodial positions were held by black employees. The district court considered this evidence as “the failure of UYB to place any blacks in the majority of positions and job categories.” The question we have is whether or not this will support its ruling that a prima facie case was established, having shown that a part of the statistics which it relied on were improperly considered.
Even if the entire local labor force were the appropriate group for statistical comparison, there are still serious problems with these statistics. It is grossly misleading to say that 43 of the 60 job positions were occupied only by white employees. The bank points out, without contradiction, that 22 of these categories contained only one employee, and 41 of the categories contained less than 5 employees. Absent other evidence, it is ordinarily improper to draw statistically significant comparisons from such small samples. Teamsters v. United States, 431 U.S. 324, 340 n. 20, 97 S.Ct. 1843, 52 L.Ed.2d 396 (1977); Roman v. ESB, 550 F.2d 1343, 1352 (4th Cir. 1976). While some statistical significance might be attached to such figures in some cases, in the case before us we feel that they lack statistical significance because, for practical purposes, they stand alone. With the exception of the individual employees mentioned elsewhere in the opinion, out of what must have been thousands of black applicants, the EEOC offered no evidence to show that any black applicant had the qualifications to hold one of these positions. Neither did it offer any evidence to show the seniority of the people in the positions, it being uncontradicted that the bank prefers to promote from within. It is significant to note here there is no claim of discrimination in promotions or transfers. Neither is there any claim of unequal pay because of race for. people performing the same work. There not being statistical proof of discrimination within the larger group taken as a whole, we think it is not proper to break the group down into sub-components with so few members as one and to find a prima facie case within a sub-group when such finding is not permissible for the larger group of which the sub-group is a part. Accordingly, we are of opinion the EEOC did not prove a prima facie case of discrimination by statistical evidence.
It is simply not realistic to say that every member of the labor force has the qualifications to be a bank teller for example. It is even more unrealistic to say that every member of the labor force has the qualifications to be a bank manager or official. Despite the examination of thousands upon thousands of employee records covering a period of nine years, the EEOC did not present evidence of one person who was qualified to be a bank official or manager, who had applied for the job and who was denied employment. While the district court found adversely to the claims of the individual cases which were presented to it, even if those findings were clearly erroneous, out of what is bound to be several thousand applications for employment by black members of the labor force, the EEOC has come up with at most only 36 examples of claimed discrimination, and it pursues only 9 of those on appeal. These numbers are simply not enough to prove a prima facie case. They show no hiring practice or procedure which is discriminatory-
[154]*154Another set of statistics argued by the EEOC deals with the number of applicants compared with the number of hires.3 The evidence shows that in 1973-74 4058 persons applied for jobs at UVB. 2882 of the applicants were white and 1176 were black. 358 whites were hired and 52 blacks were hired. Thus, 12.4 percent of the white applicants were hired and 4.4 percent of the black applicants. The EEOC argues that this is prima facie evidence of discrimination since whites had a nearly three times better chance of being hired than blacks. Once again, however, the EEOC, not carrying the burden of proof which is on the plaintiff, has not produced any evidence as to the relative numbers of blacks and whites, although the applications were apparently at hand, who were qualified for the jobs at UVB. The EEOC again asks us to find that all black people in the local labor force are qualified for the jobs available at UVB. To repeat, for the reasons stated in our discussion of the general hiring statistics, we reject this argument. We have repeatedly rejected such arguments and we adhere to those opinions. EEOC v. Radiator Specialty Co., 610 F.2d 178 (4th Cir. 1979); Hill v. Western Electric Co., 596 F.2d 99 (4th Cir. 1979); Patterson v. American Tobacco Company, 535 F.2d 257 (4th Cir. 1976); Roman v. ESB, Inc., 550 F.2d 1343 (en banc) (4th Cir. 1976).
We do not mean to say that our statistical analysis is conclusive. We can only base our calculations on evidence in the obviously insufficient record. However, the burden of proof was on the EEOC, and it failed to present any evidence of the post-Act hires or on the percentage of black people in the labor force qualified to work at UVB in its various job categories. We have taken the time to show that even on the sparse evidence in the record the EEOC has not proved its case, and we note that failure of proof is the primary basis for our decision.4
[155]*155EEOC also complains that the bank’s standards for hiring were subjective in some aspects, particularly in granting employment interviews and in the remarks made by a receptionist prior to such interviews. Out of the thousands of applications received by the bank during the nine years in question, EEOC'introduced evidence that 12 of them contained comments made by a receptionist, not by the person conducting the employment interviews, which it claims indicate racial discrimination in its hiring policy. More accurately, any intended derogation from some of the comments we think might be characterized as racial slurs. The district court found that the comments were not racially discriminatory, and we cannot say its findings are clearly erroneous.
Nevertheless, the comments deserve some examination. We are of opinion five of the twelve comments could have had nothing to do with race. They are: “NFI Neat. Had a tendency to be aggressive and pushy in getting a job;” “Neat and pleasant — not a lot of personality;” “Quiet and vague. Didn’t impress either in dress or manner;” “15” (meaning black); “Note teller material. May be okay for other openings.” “Note” in the last comment, may have meant “not.” In either event, nothing about race can be drawn from the comment.
Four of the comments had to do with hair style, either Afro, bush, or mod. In view of the fact that bank employees must deal with the public when on company assignment, we cannot see that comments concerning unusual hair styles indicate racial discrimination. “Good grooming regulations reflect a company’s policy in our highly competitive business environment. Reasonable requirements in furtherance of that policy are an aspect of managerial responsibility.” Fagan v. National Cash Register Company, 157 U.S.App.D.C. 15, 25, 481 F.2d 1115, 1125 (D.C.Cir. 1973). See also Woods v. Safeway Stores, Inc., 420 F.Supp. 35 (E.D.Va.1976), aff’d 579 F.2d 43 (4th Cir. 1978), which held not discriminatory a regulation forbidding beards when the affected employee was medically required to wear a beard to correct facial conditions afflicting more black than white people.
The remaining three comments describe one applicant as “Light, very pleasant girl — interested in clerical work;” another as “. . . 15 with red hair, but very pleasant;” and the last as “Okay guy. Nob like Sam Weeks or Floyd Shorter, but okay.” (Weeks and Shorter were black employees at UVB.)
While not everyone would consider that the comments with respect to these last three people describing skin color as light, and hair color as red, and comparing a black applicant with black employees, as being racial slurs or indicating discrimination, nevertheless, it is arguable that such was the intent of the comments. Even assuming that the finding of the district court as to these three was clearly erroneous, we do not believe that such comments on three applications out of thousands indicate, or are any substantial evidence of, discriminatory hiring procedures by the bank. Discriminatory as the comments may be construed, taken in their worst light, they are just too few in number. Mayor v. Educational Equality League, 415 U.S. p. 621, 94 S.Ct. 1323.
The district court held that the plaintiffs’ prima facie case, which it had held to have been established by statistical evidence, was rebutted because there was no showing of discrimination against any specific person, no showing of any specific discriminatory employment practice, and also because the defendant had and pursued with commendable vigor its affirmative action program.
Our opinion has previously mentioned the specific practices which undoubtedly were those referred to by the district court in its opinion. We have also discussed the statistical evidence. The affirmative action program of the bank also deserves mention.
[156]*156The facts concerning the affirmative action program of the bank are uncontradicted, and we briefly recite some of them here. At least as early as 1967, the bank actively solicited minority students for part-time positions. Of the students employed by the bank participating in the Norfolk City School Cooperative Education Program, a majority have been black, and the bank has made specific requests to the school system for referral of black students. There has been a steady increase in the number of black employees at the bank. Once having hired a black employee, the bank has exercised diligence in his training and engaged in efforts to train black employees in addition to the efforts exercised in training white employees. The bank undertakes active recruiting at predominately black colleges in the area, and black bank officers participate in an effort to interest black students in banking in general, and in UVB in particular as a potential employer. The bank maintains an affirmative action file in which minority applications are maintained, and, when applicants are being considered to fill a vacancy, refers to the affirmative action file in addition to referring to the other applications on file.
The bank, supported by the finding of the district court, maintains that its extensive efforts in an affirmative action program should be sufficient to rebut any prima facie statistical case made out by the EEOC. See Patterson v. American Tobacco Co., 8 EPA 6003 (E.D.Va.1974), remanded in part on other grounds, 535 F.2d 257 (4th Cir. 1976); Park v. Firestone Tire and Rubber Co., 8 EPD 6062 (N.D.Ohio 1974). We need not decide this question because, as stated, we are of opinion the plaintiff did not prove a prima facie statistical case, or, even if a weak statistical case might be said to have been proven by a view of a part of the statistics, the plaintiff did not carry its burden of persuasion. New York City Transit Authority v. Beazer, 440 U.S. 568, 587, n. 31, 99 S.Ct. 1355, n. 31, 59 L.Ed.2d 587 (1979).
Finally, the EEOC raised claims of individual discrimination other than those relating to the comments we have just discussed above. We have reviewed the district court’s opinion and the record with respect to these claims and are of opinion that its findings as to these were not clearly erroneous.
Accordingly, we are of opinion the EEOC failed to prove its case of discrimination against UVB with respect to hiring.
The judgment of the district court is accordingly
AFFIRMED.
mistakenly labeled 1970 in district court’s opinion.