CHOY, Circuit Judge:
This case concerns the proper use of the “disparate impact” model of proof in employment discrimination cases. Tommie Moore is a black female employee of Hughes Helicopters, Inc. (“Hughes”), a manufacturer of commercial and military helicopters. Moore, suing on behalf of a class of black female Hughes employees, alleges that Hughes has discriminated against black females in violation of Title VII of the Civil Rights Act of 1964, as amended, 42 U.S.C. §§ 2000e,
et seq.
The alleged discrimination occurred in the selection of employees for supervisory and upper-level craft positions from 1975 through 1979. The case went to trial before Judge Robert J. Kelleher in October 1980. At the conclusion of Moore’s case, Hughes moved for an involuntary dismissal pursuant to Fed.R.Civ.P. 41(b). In May 1981, the case was transferred to Judge Manuel L. Real, to whom, by stipulation of the parties, the Rule 41(b) motion was submitted on the transcript and exhibits from the presentation of Moore’s case. On August 27, 1981, the district court granted Hughes’ motion and made an award of costs.
Moore appeals the dismissal of her case, the denial of her earlier motion for summary judgment, the limited scope of the class certification, and the award of court costs to Hughes. We affirm the district court.
I
BACKGROUND
Hughes is an aerospace employer located primarily in Culver City, California. At the time of trial, Hughes employed over 4,000 persons. Among these employees, 1,562, including the present plaintiff class, were within a bargaining unit represented by the Electronics and Space Technicians, Local No. 1553 AFL-CIO (“EAST”). The EAST unit encompasses 192 different job classifications that range in complexity from janitorial positions to highly skilled machinist, mechanic, and electrician positions. The job classifications have each been assigned a labor grade to reflect rates of pay. Labor Grade 1 is the lowest paid job and Labor Grade 20 is the highest.
Moore claims that her class of black female EAST unit employees has been discriminated against in the selection of employees for Labor Grades 15-20, and for the position of “first level supervisor.” The core of Moore’s case is a statistical summary of Hughes’ work force in the disputed job classifications broken down by race and sex. On the dates set forth below, composition of Labor Grades 15-20 by race and sex within the EAST bargaining unit was as follows:
9/24/76
10/2/77
10/1/78
6/3/79
Total 245 (100%) 292 (100%) 323 (100%) 435 (100%)
White Males 209 (85.3%) 246 (84.2%) 260 (80.5%) 339 (77.9%)
White Females 0 (0.0%) 1 (0.3%) 4 (1.2%) 6 (1.4%)
Black Males 8 (3.3%) 14 (4.8%) 23 (7.1%) 35 (8.0%)
Black Females 0 (0.0%) 0 (0.0%) 0 (0.0%) 0 (0.0%)
Males 245 (100%) 291 (99.7%) 318 (98.5%) 427 (98.2%)
Females 0 (0.0%) 1 (0.3%) 5 (1.5%) 8 (1.8%)
Hughes hired and promoted persons into Labor Grades 15-20 based on subjective evaluations of capability and experience.
“First level supervisors” are individuals who had responsibility for directly supervising EAST unit employees. Not all the first level supervisors were themselves members of the EAST unit. During the relevant period, the makeup of the first level supervisory positions was as follows:
Jan. 76
June ’76
June ’77
June ’78
June ’79
Total 56 (100%) 57 (100%) 62 (100%) 76 (100%) 91 (100%)
White Males 43 (76.8%) 41 (71.9%) 44 (70.9%) 56 (73.7%) 64 (70.3%)
White Females 1 (1.8%) 2 (3.5%) 3 (4.8%) 3 (3.9%) 3 (3.3%)
Black Males 5 (8.9%) 8 (14.0%) 8 (12.9%) 8 (10.5%) 10 (10.9%)
Black Females 0 (0.0%) 0 (0.0%) 0 (0.0%) 1 (1.3%) 2 (2.2%)
Males 55 (98.2%) 55 (96.5%) 59 (95.2%) 72 (94.7%) 85 (93.4%)
Females 1 (1.8%) 2 (3.5%) 3 (4.8%) 4 (5.3%) 6 (6.6%)
These employment figures compare to an overall percentage of black female employees in the EAST bargaining unit that began at 3.9% in September 1976 and grew to 5.95% by June 1979. The weighted average of black female representation in the EAST unit throughout the relevant time frame was 4.9%. It is Moore’s contention that the disparity between the percentage of black female employees in upper-level jobs and in the EAST bargaining unit generally suffices to establish a prima facie case of employment discrimination.
II
CLASS CERTIFICATION
A threshold question is whether Moore was improperly denied the right under Fed.R.Civ.P. 23 to bring this action on behalf of a broader class that would include all black and/or all female EAST unit employees. We review the district court’s Rule 23(a)
determination on Moore’s class-certification motion for abuse of discretion or application of impermissible legal criteria.
Yamamoto v. Omiya,
564 F.2d 1319, 1325 (9th Cir.1977).
After briefing and argument, the lower court certified a class of “[a]ll black female employees in the bargaining unit of Electronic [and] Space Technicians, Local 1553 AFL-CIO, who have been employed by Hughes Helicopters at any time on or after December 3, 1975.” The court would not allow Moore to represent white females because Moore had never claimed before the EEOC that she was discriminated against as a female, but only as a black female. The court determined that this raised serious doubts as to Moore’s ability to adequately represent white female employees.
See
Fed.R.Civ.P. 23(a)(4). Moore was not permitted to represent black male employees for two reasons. First, the court determined from a reading of Moore’s deposition that Moore did not believe that black males were discriminated against in the selection of supervisors. This also raised concerns about the adequacy of Moore’s representation. Second, the court determined that Moore could not make out a prima facie case of discrimination against black males since the percentage of black males in the Hughes work force exceeded the percentage of black males in the Los Angeles County labor force.
We affirm the lower court’s class determination on the basis of inadequacy of representation. “[A] Title VII class action, like any other class action, may only be certified if the trial court is satisfied, after a rigorous analysis, that the prerequisites of Rule 23(a) have been satisfied.”
General Telephone Co. v. Falcon,
457 U.S. 147, 161, 102 S.Ct. 2364, 2372-73, 72 L.Ed.2d 740 (1982). Mere membership in a sexual or racial group does not justify a finding that a plaintiff will adequately represent all members of a particular group.
See id.
at 159 n. 15, 102 S.Ct. at 2371 n. 15;
East Texas Motor Freight System, Inc. v. Rodriguez,
431 U.S. 395, 405-06, 97 S.Ct. 1891, 1897-98, 52 L.Ed.2d 453 (1977). Inadequate repre sentation could unfairly lead to a foreclosure of the rights of absent members to seek relief under Title VII in their own behalf.
See Johnson v. Georgia Highway Express, Inc.,
417 F.2d 1122, 1126 (5th Cir. 1969) (Godbold, J., specially concurring),
cited with approval, Falcon, supra,
457 U.S. at 161, 102 S.Ct. at 2372. The lower court’s stated doubts as to Moore’s belief that black males and white females were discriminated against is a reasonable ground for holding that Moore could not adequately represent these groups. We have some qualms about the trial court’s alternate ground for denying Moore the right to represent black males in that it appears to go to the merits of Moore’s case. Although some inquiry into the substance of a case may be necessary to ascertain satisfaction of the commonality and typicality requirements of Rule 23(a), it is improper to advance a decision on the merits to the class certification stage.
Eisen v. Carlisle & Jacquelin,
417 U.S. 156, 177-78, 94 S.Ct. 2140, 2152, 40 L.Ed.2d 732 (1974). Nevertheless, we are satisfied that the district court did not abuse its discretion in limiting its class certification on the basis of inadequate representation.
Ill
DISCRIMINATION CLAIMS
A.
Standard of Review
The findings of fact made by a trial court following the granting of a Rule 41(b) motion for involuntary dismissal will not be set aside unless clearly erroneous.
Sime v. Trustees of the California State University and Colleges,
526 F.2d 1112, 1113 (9th Cir. 1975). The court below found that Moore had failed to establish a prima facie case of employment discrimination and that, even assuming a prima facie case, Hughes had successfully overcome any presumption of discrimination. Our past cases are divided as to whether our review of a lower court’s ultimate finding regarding establishment of a prima facie case is
de novo
or under a “clearly erroneous” standard.
See generally Gay v. Waiters’ and Dairy Lunchmen’s Union, Local No. 30,
694 F.2d 531, 539-46 (9th Cir.1982). In this case, as in
Gay,
we reach the same conclusion under either standard and thus need not decide which standard is the correct one.
See id.
at 545-46. Nevertheless, since strong arguments may be made that
de novo
review is
both analytically superior to “clearly erroneous” review, and better in harmony with past Title VII cases,
see id.
at 539-46, we will consider the district court’s determination that Moore failed to establish a prima facie case under the
de novo
standard of review.
B.
Application of Disparate Impact Analysis
Moore has insisted that her action be litigated solely under the disparate impact model of employment discrimination. In an “impact” case, an employee must show that a facially neutral employment practice has a “significantly discriminatory” impact upon a group protected by Title VII.
Connecticut
v.
Teal,
457 U.S. 440, 446, 102 S.Ct. 2525, 2531, 73 L.Ed.2d 130 (1982). Discriminatory intent need not be proven.
International Brotherhood of Teamsters v. United States,
431 U.S. 324, 335 n. 15, 97 S.Ct. 1843, 1854 n. 15, 52 L.Ed.2d 396 (1977);
Bonilla v. Oakland Scavenger Co.,
697 F.2d 1297, 1303 (9th Cir. 1982),
petition for cert. filed,
51 U.S.L.W. 3775 (U.S. April 15, 1983) (No. 82-1699). Once an employee has demonstrated that an employment practice has a disparate impact upon protected persons, the employer must demonstrate that the particular employment practice has “a manifest relationship to the employment in question,”
Griggs v. Duke Power Co.,
401 U.S. 424, 432, 91 S.Ct. 849, 854, 28 L.Ed.2d 158 (1971), or that the employment practice is justified by “business necessity,”
Gay, supra,
694 F.2d at 537 (citing
Contreras v. City of Los Angeles,
656 F.2d 1267, 1275-80 (9th Cir.1981),
cert. denied,
455 U.S. 1021, 102 S.Ct. 1719, 72 L.Ed.2d 140 (1982)). The employer may also rebut the employee’s prima facie case by showing the inaccuracy of the employee’s statistics.
Contreras v. City of Los Angeles,
656 F.2d 1267, 1273 (9th Cir.1981),
cert. denied,
455 U.S. 1021, 102 S.Ct. 1719, 72 L.Ed.2d 140 (1982). If the employer demonstrates a business justification the employee may still prevail by demonstrating that the employment practice is a “mere pretext for discrimination,”
Teal, supra,
457 U.S. at 447, 102 S.Ct. at 2531, or that the employer’s purpose could be served by selection devices with less discriminatory impact,
Dothard v. Rawlinson,
433 U.S. 321, 329, 97 S.Ct. 2720, 2726, 53 L.Ed.2d 786 (1977).
Normally, when a Title VII plaintiff alleges widespread, systemic employment discrimination, such as in this case, courts have analyzed the claims under the disparate treatment mode of analysis.
See, e.g., Teamsters, supra.
This is particularly true in the case of subjective hiring systems that select employees in a manner disproportionately adverse to persons protected by Title VII. Subjective hiring systems “provide a convenient pretext for discriminatory practices,”
Nanty v. Barrows Co.,
660 F.2d 1327, 1334 (9th Cir.1981), and are thus well suited to the disparate treatment focus on intentional discrimination.
Since disparate impact analysis is traditionally concerned with “employment practices that are facially neutral,”
Teamsters, supra,
431 U.S. at 335 n. 15, 97 S.Ct. at 1854 n. 15, there is some question as to whether it may be applied at all to subjective employment decisionmaking. The law in this court is unsettled.
Compare Heagney v. University of Washington,
642 F.2d 1157, 1163 (9th Cir.1981) (subjective salary decisions not susceptible to impact analysis),
with Wang v. Hoffman,
694 F.2d 1146, 1148 (9th Cir.1982) (“In order to prevail on [disparate impact] theory, [plaintiff] need only demonstrate the lack of objective criteria and a disparity in job promotions.”).
Be
cause it is not necessary for us to attempt to resolve this conflict, we assume, without deciding, that Hughes’ employment decisions are susceptible to analysis under the disparate impact method of proof.
Disparate impact analysis benefits Title VII plaintiffs by enabling them to shift a substantial burden of persuasion to the employer upon establishing a prima facie case.
See Gay, supra,
694 F.2d at 537 n. 4;
Contreras, supra,
656 F.2d at 1271. However, this benefit must be weighed against the requirements of a disparate impact prima facie case, which are in some respects more exacting than those of a disparate treatment case. A disparate impact plaintiff “must not merely prove circumstances raising an inference of discriminatory impact; he must prove the discriminatory impact at issue.”
Johnson v. Uncle Ben's, Inc.,
657 F.2d 750, 753 (5th Cir.1981),
cert. denied,
- U.S. -, 103 S.Ct. 293, 74 L.Ed.2d 277 (1982). The parties in this case clash sharply on what constitutes the required proof of impact. They disagree in particular over the question of which statistical labor pool is measured for purposes of determining impact. A related question is which party bears the burden of demonstrating whether a proffered labor pool is or is not overinclusive.
1. Appropriate Labor Pool
Bearing in mind that the focus of a disparate impact inquiry is those employment practices that have a discriminatory effect,
see Griggs v. Duke Power Co.,
401 U.S. 424, 431, 91 S.Ct. 849, 853, 28 L.Ed.2d 158 (1971), “[identification of the appropriate candidate pool and its racial [or ethnic or sexual] makeup is usually the starting point for impact analysis.” Bartholet,
Application of Title VII to Jobs in High Places,
95 Harv.L.Rev. 945, 970 (1982). The best evidence of discriminatory impact is proof that an employment practice selects members of a protected class in a proportion smaller than their percentage in the pool of actual applicants, or, in promotion and benefit cases, in a proportion smaller than in the actual pool of eligible employees.
See Rowe v. Cleveland Pneumatic Co., Numerical Control, Inc.,
690 F.2d 88, 93 (6th Cir.1982);
Hester v. Southern Railway Co.,
497 F.2d 1374, 1379 (5th Cir.1974).
Disparate impact should always be measured against the actual pool of applicants or eligible employees unless there is a characteristic of the challenged selection device that makes use of the actual pool of applicants or eligible employees inappropriate. Commonly, such a characteristic would occur in discriminatory hiring cases where the employment practice in question is in the nature of an “entrance requirement.” In these cases, persons who lack the challenged requirement will self-select themselves out of the pool of applicants. Examples of such entrance requirements include height and weight specifications,
see Dothard v. Rawlinson,
433 U.S. 321, 97 S.Ct. 2720, 53 L.Ed.2d 786 (1977), and high school graduation,
see Griggs, supra.
When an employer requires such qualifications, the makeup of the pool of actual applicants does not fairly reflect the pool of individuals affected by the challenged requirement.
See Dothard, supra,
433 U.S. at 330, 97 S.Ct. at 2727. In these cases, disparate impact may be established through reference to a reasonable proxy for the pool of individuals actually affected by the alleged discrimination. The choice is usually between general population statistics and the statistics of a relevant labor market.
General population statistics are useful as a proxy for the pool of potential applicants, if ever, only when the challenged employer practice screens applicants for entry level jobs requiring little or no specialized skills.
If special skills are re
quired for a job, the proxy pool must be that of the local labor force possessing the requisite skills.
See Valentino v. United States Postal Service,
674 F.2d 56, 71 (D.C.Cir.1982);
Kinsey v. First Regional Securities, Inc.,
557 F.2d 830, 839 (D.C.Cir.1977);
Rich v. Martin Marietta Corp.,
467 F.Supp. 587, 609-10 (D.Colo.1979);
cf. Dothard, supra,
433 U.S. at 338, 97 S.Ct. at 2731 (Rehnquist, J., concurring);
Hazelwood School District v. United States,
433 U.S. 299, 308 n. 13, 97 S.Ct. 2736, 2741 n. 13, 53 L.Ed.2d 768 (1977) (similar analysis of proper labor pools in disparate treatment case);
United States v. Ironworkers Local 86,
443 F.2d 544, 551 (9th Cir.) (necessity of accurate statistical comparison for inference of discrimination), cer
t. denied,
404 U.S. 984, 92 S.Ct. 447, 30 L.Ed.2d 367 (1971). This rule is a natural deduction from the commonsense observation that if a person is not otherwise qualified for a job, he is not deterred from applying by any particular employment requirement unrelated to his ability to perform the job.
2. Burden of Proof in Establishing the Appropriate Labor Pool
Both sides have drawn upon the Fourth Circuit case of
EEOC v. Radiator Specialty Co.,
610 F.2d 178 (4th Cir.1979), to support their arguments concerning the allocation of the burden of proof regarding whether general statistics or a qualified labor pool must be used to determine disparate impact.
In
Radiator Specialty,
the court divided the possible burden allocations into three categories. First, there will be cases where it is manifest as a matter of law that no special skills or qualifications are required for a job. In these cases, the plaintiff may establish his prima facie case using general population statistics.
Id.
at 185. At other times, the need for special qualifications will be manifest as a matter of law. In these cases, the plaintiff will be required to show a disparate impact on the qualified labor market to establish his prima facie case.
Id.
In still other cases, it will not be immediately obvious that a job requires any special qualification. In these cases, it will be the defendant’s burden to establish that generalized statistics do not adequately reflect the pool of presumptively qualified individuals.
Id.
We agree with the allocation of the burden of proof as described in
Radiator Specialty.
We should note that when the
allocation is applied to promotion cases, it will be necessary for the plaintiff to show that the pool of eligible employees is qualified for promotion to a position for which it is manifest that special skills are required.
C.
Discrimination in Labor Grades 15-20
Turning now to the merits of Moore’s case, it is apparent that Moore is relying on little more than an inference of discrimination from the bare absence of black female employees in Labor Grades 15-20 within Hughes’ EAST bargaining unit. She points out that black females are concentrated in the lower grades, and implies that the absence of a uniform percentage of black females throughout all labor grades is indicative of a discriminatory employment practice. We rejected this very inference in
Pack v. Energy Research & Development Administration,
566 F.2d 1111 (9th Cir.1977) (per curiam). In
Pack,
the plaintiff sought to establish a prima facie case of sex discrimination by a statistical showing that 98% of the professional employees above a certain level were male, while 95% of the professional employees below that line were female. We agreed with the lower court’s finding that no prima facie case was established, stating, “No evidence whatsoever was introduced to demonstrate that the lower-grade professional women were qualified to occupy the higher positions or that there elsewhere existed a pool of qualified women applicants.”
Id.
at 1113. The same may be said of Moore’s showing regarding the distribution of black females at Hughes.
Nor may a prima facie case be established by assuming that any generalized pool of black females within the EAST unit is the appropriate pool for measuring the disparate impact of Hughes’ promotions into the upper labor grades. The district court found as a matter of law that the upper-level positions had a manifest requirement for persons possessing special skills. We agree. These grades were composed of jobs whose mere identification signaled the need for qualified employees.
Because of the skills required for these upper-level jobs, and because Hughes does not provide training to enable individuals to acquire the requisite skills, 323 of the 470 openings in Labor Grades 15-20 that occurred during the period covered by this suit were filled by outside hires. Persons hired into the upper labor grades
averaged
10 years of experience in the trade in which they were hired.
Applying the proper method of proof, Moore cannot demonstrate disparate impact in either the promotion or hiring of individuals to fill the skilled positions in Labor Grades 15-20. Moore has not even attempted to show disparate impact in promotion to the upper-level jobs through comparison with the pool of actual promotion applicants. This is understandable. Not a single member of the plaintiff class applied for promotion or transfer into a position in Labor Grades 15-20 during the period covered by this lawsuit. Moreover, Moore has failed to establish that there were any black female employees qualified for the positions in Labor Grades 15-20. Indeed, the district court found specifically that there were no class members in the “Occupational Family Group Jobs” immediately below the upper level positions from which promotions might have been made. Accordingly, even assuming that a proxy pool of EAST unit employees should be created to test for disparate impact in promotion, Moore has failed to show that any such pool would include class members. There is no prima facie case of disparate impact in the promotion of employees into Labor Grades 15-20.
Nor is there any evidence that the plaintiff class was subject to discrimination in the 323 hires into Labor Grades 15-20. Again, not a single class member applied for any of these positions. There was no evidence of the presence of class members
among the pool of eligible individuals. Even when Hughes’ hiring is compared with the external Los Angeles labor market, there is little to suggest that its hiring practices had a disparate impact on black females.
In short, Moore has failed to produce any evidence whatsoever of disparate impact in the selection of employees for Labor Grades 15-20.
D.
Discrimination in First Level Supervisory Positions
During the relevant time frame, there were 76 openings for first level supervisors. Two of these openings (2.6% of the total) were filled by black females. Since the weighted average of black female representation in the EAST bargaining unit throughout the relevant time was 4.9%, it is Moore’s contention that there should have been 3.7 (4.9% of 76) black female first level supervisors. We agree with the district court that the identification of a job position as “supervisor” implies at least a modicum of job-related experience. It was thus Moore’s responsibility to identify a pool of employees eligible for supervisory positions who possessed the requisite experience. Moore presented no evidence of black females with appropriate skills who were available for promotion, transfer or hire into first line supervisory positions.
Close analysis of the evidence reveals that Moore would in any case be unable to create a prima facie case of disparate impact in these positions. Of the 76 supervisory openings, 11 were filled by persons from outside the EAST unit. One of these 11 (9.1%) was a black female. Of the 65 promotions, 17 went to salaried, non-EAST unit employees who primarily supervised other non-EAST, salaried employees. EAST unit employees were eligible for promotion to these supervisory positions over salaried employees only in a theoretical sense. Of the remaining promotions, all but eight involved promotions from jobs in which there were
no
black females eligible
for promotion; of the remaining eight, one (12.5%) was given to a black female. This is not evidence of an employment practice having “a significantly discriminatory impact.”
Connecticut
v.
Teal,
457 U.S. 440, 447, 102 S.Ct. 2525, 2531, 73 L.Ed.2d 130 (1982).
Even if we were to accept Moore’s contention that the percentage of black females selected as first level supervisors should equal the overall percentage of black females in the EAST unit, there is still insufficient evidence of significant discriminatory impact. Where the expected mean distribution of black females in a group of 76 persons could be achieved by an increase of less than two black females, we are unwilling to agree that a prima facie case of disparate impact has been proven.
See Knutson v. Boeing Co.,
655 F.2d 999, 1001 (9th Cir.1981) (where addition of one female among 19 or 20 openings would have resulted in percentage of females greater than in pool of potential applicants, no prima facie case of disparate impact).
IV
COSTS
Moore was assessed $2,624.25 in court costs as provided by Fed.R.Civ.P. 54(d). Moore argues that costs should be assessed against a Title VII plaintiff only under the same conditions that would permit an assessment of attorney’s fees against the plaintiff. This court has already rejected Moore’s suggestion.
National Organization for Women v. Bank of California,
680 F.2d 1291, 1294 (9th Cir.1982). While we agree that a court may consider the limited resources of a Title VII plaintiff when assessing costs,
id.,
we will only disturb an award of costs upon a showing of abuse of discretion.
Id.
Here, the only costs allowed were Hughes’ $20 statutory docket fee and the deposition costs of nine persons identified by Moore as witnesses to be called at trial.. An abuse of discretion has not been shown. We affirm the lower court’s assessment of costs.
V
CONCLUSION
Because we hold that the district court properly granted Hughes’ Rule 41(b) motion, we need not address Moore’s contention that denial of her motion for summary judgment was improper. We affirm the judgment of the district court in all other respects.
AFFIRMED.