Applying both discrimination models--disparate treatment and disparate impact--we affirm that a black female seeking job referrals from a large but undermanned state employment agency failed to prove a case of discrimination under Title VII. Specific problems with the referral system, established in the record, rebutted any presumption of discrimination, and her statistical analysis critically neglected qualifications and job preferences.

Hazel Hill, a black female, sued the Mississippi State Employment Service (MSES) on her own behalf, alleging that MSES intentionally or recklessly discriminated in its referrals on the basis of race. Her suit stemmed from her dissatisfaction with the treatment which she received when seeking job referrals to prospective positions as a cashier, waitress, or cocktail waitress at MSES's Hattiesburg Office from April 1982 to April 1983, and followed the right to sue letter issued to her by the Equal Employment Opportunity Commission (EEOC) upon the basis of its finding no reasonable cause to pursue the complaint which she had filed against both MSES and the Mississippi Employment Security Commission (MESC).1

MSES's job referral procedures lie at the heart of this case, and we therefore outline them. In almost all respects, MSES complies with the operational guidelines established in the Employment Service Manual (ESMII).2 Operating through numerous local offices throughout Mississippi, MSES refers qualified applicants to area job openings listed with MSES by area employers, but does not guarantee employment. The listing employers indicate the number of applicants sought and the qualifications required, and MSES's clerks process the listings and notify qualified applicants.

To match applicants to jobs, MSES assigns applicants and job orders a nine-digit Dictionary of Occupational Titles Code (Dot Code).3 The code's first three digits identify a particular occupational group, the middle three the worker function ratings, and the final three the alphabetical order of the titles within the six digit code.

Referrals occur in two ways: In one, upon an applicant's request MSES clerks search job orders for vacancies that match the applicant's DOT code and, locating an open order with a corresponding code, refer the applicant if the applicant meets the listed qualifications; In the other, to fill orders, pursuant to ESMII Sec. 1470, MSES clerks search the active files of applicants with a matching occupational title and, if that yields an insufficient number of applicants, search inactive files with matching codes and active files of related occupational qualifications. Ultimately, to match job orders with the most qualified applicant as per ESMII guidelines, MSES clerks primarily rely on the DOT Codes, the type of job, and the applicant's experience and job preference, but as well draw from personal knowledge acquired through prior referral matching, such knowledge including subjective judgments (e.g.: an applicant's attitude, appearance, personality, employment history).

An EEOC officer regularly reviews MSES's referral and other services, assessing compliance with federal and state civil rights laws. Documentation drawn from all codes and applicants, but not from actual interviews, furnishes all supporting information for these reviews. Noting that the reviews to the date of trial highlighted problem areas, the conducting officer rated MSES satisfactory overall.4

Coded for cashier-checker positions (211-492-014) and for waitress positions (311-477-030), Hill alleges discriminatory referral practices and retaliatory practices. At trial, she pointed to several occasions when MSES failed to refer her to open job orders for which she was coded and qualified (qualified, that is, at least to an extent equal to any referred white applicant).5 MSES, for its part, pointed to a record of MSES referrals received by Hill and to Hill's work history, showing ten positions held by Hill prior to the period in question, with the length of each generally fluctuating between a few weeks and a few months.6 All MSES employees with whom Hill had contact specifically denied racial considerations in referral selections, attributing referral problems, such as those reflected in Hill's evidence, to employee error.

Respecting the total number of MSES referrals during the relevant period, Hill and MSES stipulated to white/non-white ratios of applicants in the active file, overall an average of 58.2% white, 41.7% nonwhite.7 As well, both stipulated: MSES referred 473 applicants to orders classified in the 311 DOT Code (255 or 53.9% white and 218 or 46% black) from April 1982 to February 1983, and of the 473 referrals, 233 were referred out of their primary DOT Code (120 or 51.5% white and 113 or 48.4% black.) Additionally, 41 of the 95 job orders processed had more than one referral, 31 of the 41 orders having both black and white applicants, and of the 373 applicants referred to these 31 job orders, 184 were white (49.3%) and 189 black (50.6%).

Without accounting for employer qualifications, which may have rendered some applicants unacceptable for referral, MSES filled relevant job orders from April 1982 to February 1983 as follows: processing six cocktail waitress orders, shorting four, and referring in total 21 whites and no blacks; processing 39 waitress orders, shorting 14 (11 of which referred no blacks), and referring in total 113 whites and 46 blacks (but referring no blacks to 24 of the 39 orders); and processing 146 cashier job orders and referring in total 998 whites and 457 blacks (but in 57 of the 146, referring only white applicants).8

Himmelstein, Hill's statistics expert, analyzed the actual referral pool derived from the stipulations above and the unreferred pool as manifested in data generated by a computer program developed to reflect referrals in the 211 and 311 codes on the basis of race. MSES controlled and supplied all information used to develop the program and Hill's computer expert ran the program on MSES's computers.9 The program searched MSES's relevant files to show an applicant's DOT Code on a given date, and the data generated reflects black and white unreferred applicants with either a 211 or 311 code during the relevant time period. Looking for a pattern of referral based on race and not attributable to chance, Himmelstein factored the data, generated by running the program with the three digit numbers of the occupational codes, into a formula to determine the number of standard deviations. For cocktail waitress, waitress, and cashier positions, respectively, he calculated standard deviations 4.83, 8.78, and 3.41, all of which are considered statistically significant. Applying the formula to data generated from the entire nine-digit code, Himmelstein calculated a standard deviation of 2.58 for the 211 Job Orders and 1.92 for the 311 Job Orders. Few applicants or referrals matched the nine-digit code.

MSES's statistical expert, Piette, considered the unreferred pool generated by the program inaccurate, whether generated from the nine or the three digit codes. Piette identified several additional problems in the program: inability to distinguish part time from full time jobs and an applicant's preference; inability to assign value to important factors such as education or experience; and inability to consider difficulties or inabilities in contacting particular applicants to apprise them of referrals.10 Piette, however, admitted that changes in the program would only reduce the standard deviations calculated by Himmelstein.

Examining the racial composition of the referred applicant pool, Piette concluded MSES's referrals and referral procedures to be neutral. Based upon the print-outs generated by Hill's program, Piette found 65.9% of all applicants (referred or unreferred) to be non-black compared to 34.1% black in the 211 code, and 50.1% non-black compared to 49.9 black in the 311 code. Only 13.5% of the total 211 coded applicants (69.2% non-black, 30.8% black) and 31.5% of the total 311 coded applicants (54.5% white, 45.5% black) received MSES referrals. Analyzed by racial composition, referrals nearly equalled the racial composition of the applicant pool. The proportions of out-of-code referrals by race among total referrals for the 211 Code were 70.4% black and 64.3% white; for the 311 Code, 78.9% black and 73.9% white. For out-of-code referrals to 211 positions, 32.8% were black and 67.2% white; for 311, 47.1% black and 52.9% white.

To locate our review today in the proper legal context, we briefly condense the Supreme Court's methodology for analyzing Title VII actions. The Court probes Title VII actions under two well-established models for ferreting out unlawful discrimination: disparate treatment and disparate impact analysis. The first concentrates on discriminatory motive, the second on discriminatory effect. On the issue of discrimination, in either case, the plaintiff bears the ultimate burden of persuasion; but the onus of producing sufficient evidence for us to assess whether the plaintiff met this burden shifts. Wards Cove Packing Co., Inc. v. Antonio, 490 U.S. 642, 109 S. Ct. 2115, 2120, 104 L. Ed. 2d 733 (1989). Hill asks that her case be analyzed under both models, and we do so.

Pinpointing both overt and covert discrimination, disparate treatment analysis reveals differential treatment of similarly situated persons on the basis of impermissible factors such as race or sex. Dislodging the status quo of prior discrimination, disparate impact analysis discloses the disproportionate impact of a facially neutral but non-job-related policy upon a protected group such as blacks or women. Griggs v. Duke Power Co., 401 U.S. 424, 429-30, 91 S. Ct. 849, 852-53, 28 L. Ed. 2d 158 (1971). Either model is applicable to employment practices that utilize discretionary and subjective criteria. Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 108 S. Ct. 2777, 101 L. Ed. 2d 827 (1988).

Only the disparate treatment model requires proof of discriminatory animus, and it requires such proof whether a given action alleges individual or class-wide (pattern and practice) discrimination. Although, for individual allegations, a single proven instance of unlawful disparate treatment suffices, showing class-wide disparate treatment requires more than isolated or exceptional instances of differential treatment. Hill does not assert class-wide disparate treatment. In either case, the necessary animus may be shown directly or inferred from circumstantial evidence.

Statistics play an important evidentiary role: circumstantial evidence of disparate treatment often includes (but need not) statistical evidence, and a finding of disparate impact requires statistically significant disparities. Statistics are not irrefutable: as the Supreme Court cautioned, "they come in infinite variety." Teamsters v. United States, 431 U.S. 324, 340, 97 S. Ct. 1843, 1856, 52 L. Ed. 2d 396 (1977). Indeed, we ourselves have noted that soundly compiling and assessing statistics is "a task both complex and arduous." Wilkins v. University of Houston, 654 F.2d 388, 410 (5th Cir.1981). Particularly in a case such as Hill's, alleging discriminatory referrals, we must carefully balance the danger of prejudice to a defendant posed by drawing inferences of discrimination from a handful of referral decisions and the unfairness to the Title VII claimant of imposing unattainable standards for proof of discriminatory intent. Pegues v. Mississippi State Employment Services, 699 F.2d 760, 769 (5th Cir.1983). Hence, we must be acutely aware that spotting mistaken understandings or mischaracterizations of numerical data either to suggest or to camouflage discrimination calls for a finely tuned quantitative and qualitative eye.

With the proper methodology in mind, we turn to Hill's evidence. The case requires from us two standards of review. Whether MSES unlawfully discriminated against Hill constitutes the ultimate fact at issue in this Title VII action, and consequently we address this issue de novo. Even so, unless clearly erroneous, the trial court's findings of subsidiary fact bind us. Wilkins v. University of Houston, 654 F.2d 388, 390 (5th Cir.1981).

The district court properly divided its factual inquiry into three stages: (1) Hill's prima facie case; (2) MSES's rebuttal; and (3) Hill's showing of pretext. We follow suit. First, we find no flaw in the prima facie test set forth by us under similar facts and applied here by the District Court. See Pegues at 774. Second, with reservations noted below we accept MSES's proffered rebuttal. Third, we find that Hill failed to show pretext.

Following our lead, to establish her prima facie case the court required Hill to prove by a preponderance of the evidence (1) her membership in a protected group, (2) her application and coding for an occupation for which MSES was making referrals, (3) her failure to secure a referral, and (4) MSES's later referral of non-members of the protected group. We agree that the specific incidents of non-referral (supra, n. 5) establish Hill's prima facie case and engender a presumption of discrimination.

MSES countered, attributing its practices to inefficiencies inherent in a large, heavily-utilized, bureaucratic, and undermanned system with limited resources. Specifically, MSES cited employee error or oversight, inefficient communication between staff members, inability to contact applicants, and applicants' preferences. These, MSES contends and the district court found, satisfy MSES's burden of producing a legitimate non-discriminatory reason for its practices. Although we note our reservations below, we must agree.

Accepting general inefficiency as a legitimate business reason troubles us, for it seems a likely candidate for use as a shield for almost any employer--including a discriminating one--to raise.11 As the Supreme Court emphasizes, the desire to home in on the critical and telling evidence schools our tripartite analysis. Texas Department of Community Affairs v. Burdine, 450 U.S. 248, 256-57, 101 S. Ct. 1089, 1095-96, 67 L. Ed. 2d 207 (1981). An employer's mere affirmations of good faith in individual selections add no factual assistance and cannot adequately rebut a prima facie showing of racial discrimination. International Bhd. of Teamsters v. United States, 431 U.S. 324, 342, n. 24, 97 S. Ct. 1843, 1858, n. 24, 52 L. Ed. 2d 396 (1977); Pegues v. Mississippi State Employment Service, 699 F.2d 760, 766 (5th Cir.1983).

Likewise, an utter lack of specifics on an employer's part does little to sharpen the fact-finder's inquiry and is equally unacceptable. Here, however, MSES does not offer us an "utter lack of specifics," but rather details the numerous problems that produce its general inefficiency--the district court finding sufficient evidence of each problem and of efforts by MSES to combat it. Naturally, the systematic problems proved to exist more credibly spawn disparities in cases such as MSES's, which involve scores of different referrals to scores of different employers, than in cases concerning the hiring practices of a single employer for a single position. MSES need not prove that the profferred reasons actually motivated it. Burdine, 248 U.S. at 254, 101 S.Ct. at 1094. Proving that they did not is a task for Hill, as the plaintiff, in connection with the pretext issue. To overturn a finding as clearly erroneous, the entire evidence must leave us with a "definite and firm conviction that a mistake has been committed". Anderson v. Bessemer City, 470 U.S. 564, 105 S. Ct. 1504, 84 L. Ed. 2d 518 (1985); United States v. Gypsum Co., 333 U.S. 364, 68 S. Ct. 525, 92 L. Ed. 746 (1918). While we would have preferred even greater evidentiary support for MSES's systematic problems, we cannot say the court's subsidiary findings were unsupported or clearly erroneous; therefore, we must accept them.

If in truth an agency's or an employer's verified, detailed and documented inefficiency, absent any discriminatory animus, accounts for results that nonetheless appear at first glance to be the product of discrimination, it would be the height of unfairness to infer fallaciously such a discriminatory animus. We do not mean to condone inefficiency; we simply cannot punish it under Title VII. With solace, we note: the phrase "to raise a shield of inefficiency," no matter how well factored and documented, does not translate automatically as "to fend off a claim of discrimination;" a claimant may show pretext.

To show pretext, Hill relied upon the statistical evidence, but the lack of an experience component and a job preference component render the statistics fatally defective. Pegues at 770. Lest we forget that MSES's stated goal, imposed by EMSII, is to match open jobs with the most qualified applicants, corresponding referral decisions turn heavily upon experience and education. Failure to include these factors would render some applicants ineligible for referral, decreasing the size of the applicant pool upon which the statistical analysis rests and producing misleading numbers.

It is true, as MSES's expert testified, that incorporating these components would only lower the standard deviations presented to us, but by how much we do not know and the evidence does not tell us, particularly given the components' critical nature. Statistically accounting for qualifications might explain the disturbing number of short referrals, specifically those in which MSES referred only whites (supra, p. 1237); but absent such particularized accounting, we cannot distill discriminatory animus from the practice. In her brief, pointing out that the same DOT Codes include traditionally minority jobs (e.g., counter attendant) and traditionally white jobs (e.g., waitresses), Hill accuses MSES of steering. Again, we lack any concrete evidence upon which to base such a conclusion. The Title VII plaintiff must ultimately prove discrimination by a preponderance of the evidence; and, accordingly, such evidentiary gaps weigh against Hill.

The gaps in Hill's statistics do not stand as the sole barrier to her attempt to show pretext. MSES's system, though inefficient, appears facially to be neutral. The percentages of referred and unreferred applicants equalled the overall racial percentages in the 211 and 311 pool: The same is true for out-of-code referral (which slightly favored black applicants). Of course, facial neutrality at the bottom line cannot immunize an agency or employer from proved acts of discrimination against an individual; but, at the same time, we cannot infer discrimination against an individual from statistics displaying neutrality toward the group. Furnco Constr. Corp. v. Waters, 438 U.S. 567, 579-80, 98 S. Ct. 2943, 2950-51, 57 L. Ed. 2d 957 (1978) ("Proof that [a] work force was racially balanced or that it contained a disproportionately high percentage of minority employees is not wholly irrelevant on the issue of intent when that issue is yet to be decided"); cf. Connecticut v. Teal, 457 U.S. 440, 454, 102 S. Ct. 2525, 2534, 73 L. Ed. 2d 130 (Racially balanced promotions at the bottomline did not relieve an employer from the requirement that it show that a written screening test, determining the eligible candidates for promotion and disproportionately disqualifying blacks, was sufficiently job-related).

Hill suitably identified the specific employment practice challenged--a referral system utilizing subjective traits and out-of-code referrals--but again, her statistical analysis failed her. Wards Cove at 109 S. Ct. 2124, citing Watson. The figures--the same asserted to demonstrate pretext under the disparate treatment model--do not offer sufficient proof that the challenged practice disproportionately affected a protected group (here, blacks) because, again, these figures neglected the education and experience factors critical to referral decisions. Since the district court considered all the evidence, both statistical and non-statistical, in reaching its conclusion of no discrimination under the disparate impact model, we decline Hill's invitation to remand for any further findings.

Having reached the ultimate issue of discrimination, and having found none, we need go no further. Yet, we cannot permit future Title VII analysis to be tainted by the district court's casual acceptance of the following fallacy: both black and white MSES clerks serviced Hill; therefore, she was not the victim of discrimination, (that is, "Blacks never discriminate against other blacks.") To the argument's self-evident overbreadth, we add that the Supreme Court has rejected it. Castenada v. Partida, 430 U.S. 482, 513, 97 S. Ct. 1272, 1289, 51 L. Ed. 2d 498 (1977).12 We recognize the court's blameless intent simply to explain what inferences may be drawn from an employer's use of subjective criteria (which does not "necessarily" indicate discriminatory animus or impact, regardless of the determining individual's color); but we must be sure that the court's dubious explanatory language does not generate a later, ill-conceived concurrence with the fallacy.

To prevail on a retaliation claim, Hill must prove that she engaged in a protected activity, that an adverse employment action occurred, and that a causal connection between the two existed. McMillan v. Rust College, Inc., 710 F.2d 1112 (5th Cir.1983). Hill contends that her EEOC complaint prompted MSES employees to harass her. Specifically, the employees allegedly stared at her, followed her, prolonged the time she spent waiting for a clerk, relegated her file to black referral clerks, destroyed her identification card, deleted experience data from her 511 form, and criticized her claim of discrimination to a Work Incentive class. All of this, according to Hill, caused her to be hospitalized and to miss referrals. The fact that Hill received numerous referrals and continued services after filing the EEOC complaint did not escape the district court's attention. In short, the magistrate sitting at trial and observing the witnesses judged as more credible the employees' testimony denying Hill's allegations: MSES replaced rather than simply destroyed Hill's identification card; whether MSES employees actually mentioned Hill's claim to the Work Incentive class or, if they did, what exactly they said, is unclear; and no testimony indicated that MSES employees made any statements in retaliation for Hill's EEOC complaint. The above findings not being clearly erroneous, we affirm the court's judgment in MSES's favor.

AFFIRMED.

Inefficiency is racially neutral. It cannot of itself account for racial discrimination. When a person establishes a prima facie case that a state agency has discriminated against her because of her race, I would not accept as a "legitimate nondiscriminatory" reason for the agency's admitted difference in treatment between her and the members of another race that the ineptness of its employees, the inadequacy of its budget, the size of its system, or the bureaucratic nature of the machinery it has devised suffice to negate the discriminatory animus manifested by the decisions, at least partially subjective, that its own employees have made.

The Mississippi State Employment Service is not a disembodied entity on high that functions independently of its employees. It is itself the deus ex machina that puts the actors on the stage and directs them. The allegedly inefficient, regulation-hamstrung, overworked employees whom it has hired and through whom it functions are the MSES. It may not wash its hands of the spots, whether damned or merely discriminatory, by which their actions have stained the MSES. Indeed were its employees merely inefficient, their actions would not have resulted in racial discrimination, for inefficiency would function as much to hamper white as minority applicants in seeking jobs. Yet MSES's actions in fact put stumbling blocks in Ms. Hill's path that were not placed before the paths of white job applicants.

In addition to my inability to accept the proffered excuses as legitimate reasons for the nonreferral of black applicants, I would find the magistrate's findings of subsidiary fact as to pretext clearly erroneous, and I would find that Hill has established a violation of Title VII of the Civil Rights Act of 1964, 42 U.S.C. Sec. 2000e et seq., the Fair Employment Practices Act, by disparate treatment. I therefore respectfully dissent from the majority opinion without reaching the disparate impact or retaliation claims.

The Act forbids racial, religious, and sexual discrimination in employment. In McDonnell Douglas Corp. v. Green1 the Supreme Court set forth the basic allocation of burdens and order of presentation of proof in a Title VII case alleging discriminatory treatment. First, the plaintiff has the burden of proving by the preponderance of the evidence a prima facie case of discrimination. Second, if the plaintiff succeeds in proving the prima facie case, the burden shifts to the defendant "to articulate some legitimate, nondiscriminatory reason for the employee's rejection."2 Third, should the defendant carry this burden, the plaintiff must then have an opportunity to prove by a preponderance of the evidence that the legitimate reasons offered by the defendant were not its true reasons, but were a pretext for discrimination.3 To fulfill its duty of articulating a legitimate non-discriminatory reason, the defendant must clearly set forth, through the introduction of admissible evidence, the reasons for the plaintiff's rejection. The explanation provided must be legally sufficient to justify a judgment for the defendant.4

The magistrate found, and the majority opinion affirms, that Hill established a prima facie case of racial discrimination. The evidence of such discrimination was not merely statistical. From April 1982 to April 1983, the period in which Hill contends she was a victim of discrimination, she sought employment as a waitress or a cashier partially because she could earn more than the minimum wage usually paid fast-food restaurant counter attendants (who, because they not only accept and deliver orders but receive payment are sometimes denominated "cashiers"). As the majority opinion indicates in a footnote,5 on several occasions during the period when Hill sought employment as a cashier or waitress MSES demonstrated an unusual pattern of referrals of applicants for the position of cashier.

Case 1. MSES referred 22 whites for interviews for employment as cashier, of whom 5 were hired. It did not refer a single black applicant. Ten of the 22 whites were "out of code," meaning that "cashier" was not their primary prior occupation and three did not even meet the minimal experience criterion established by the job order. Two had no experience and five lacked the requisite education. "While the order remained open, Hill, who admittedly would have qualified for the job," the majority opinion concedes, "sought referrals, though not specifically asking for this position,"6 as indeed she could hardly have done unless she had specific knowledge of the job order. Instead, "during the period that this order remained open and Hill sought referrals," the MSES did refer her to what it calls "alternate positions,"7 all of them stereotypical for black persons in the deep south: to a motel as a "housekeeper," an euphemism for a maid; to two fast-food chains, Church's and Kentucky Fried Chicken as a "cashier", and to another as a counter helper. These were in effect jobs making beds or serving fried chicken. It scarcely bears mention that MSES does not contend that Hill was unqualified in any way for the job to which only whites were referred.

Case 2. An order for cashiers requesting ten referrals remained open for five weeks. MSES referred only five applicants, four of whom were white. Neither Hill nor any other putatively qualified black was referred for the remaining five interview slots.

Case 3. "[A] job order for emergency room technicians, also open while Hill sought referrals, requested five applicants. MSES referred only one white woman, out of code. Hill, who previously worked at a hospital, was not referred."8 MSES does not contend that Hill's prior hospital experience did not qualify her for at least an interview with the employer, who requested four more persons for interviews than the number actually referred.

Case 4. One day after she had filed her EEOC complaint, Hill again sought a job referral as a waitress. Instead she was referred to a Ramada Inn as a "housekeeper." Upon arrival she found that the Inn had filed a request for a waitress with the MSES. "That same day, at the direction of Hill's attorney, a white female employee of a legal services organization sought referrals for waitress jobs with MSES, and the clerk, different than [sic] the clerk with whom Hill spoke [sic], provided her two waitress referrals."9

We, as well as other courts, have recognized that subjective selection processes tend to facilitate the consideration of discriminatory criteria.10 While the MSES admits that the employees with whom Hill had spoken did use subjective criteria as well as job codes in making referrals, the job referrers denied racial considerations in referral selection. Instead, the MSES and its employees presented a united front: the problems merely reflected employee error. The coincidence of such employee error consistently operating to the detriment of Hill strains credulity. A cashier may sometimes be "short," that is, she may sometimes turn in less cash than her register records, but if this is due to error, then she ought also, at least on some occasions, to be "over," having more cash than the register records. There is reason to suspect the employee who is always short and never over. Moreover, Hill's statistical expert testified to a significant statistical deviation in referring blacks. Even the MSES's statistical expert conceded that Hill's data demonstrated that the MSES system has a statistically significant disparate effect on black applicants, although he rejected the magnitude of the measure.

The majority refers to the MSES "system" as racially neutral. But verbal neutrality does not suffice to explain discrimination in application. The MSES has not shown a single instance in which its inefficient system favored black employees or in which the failure to refer employees to jobs resulted from difficulty in reaching Hill--or from overwork--or from red tape. Presumably so inefficient a system would sometimes result in the neglect or oversight of white applicants. The evidence shows that the MSES system is inefficient only when it fails to refer black employees.

The majority opinion does not even attempt to define what is meant by a legitimate nondiscriminatory reason. It contents itself with the conclusion that an employer's inefficiency, limited budget, self-manufactured red tape, and the excessive work load it places on its own employees, suffice. I submit that none of these alone nor all together constitute legitimate reasons for discrimination. Surely the word "legitimate" must mean more than simply non-criminal or not forbidden by law, for we would not as a matter of course accept either a criminal act or one otherwise forbidden by law as sufficient to justify discrimination. Instead, a "legitimate" reason must be one that is justifiable in view of the purposes of the Fair Employment Practices Act.

The Supreme Court has given support to this interpretation, stating that "[t]he broad overriding interest, shared by employer, employee, and consumer, is efficient and trustworthy workmanship assured through fair and ... neutral employment and personnel decisions."11 A legitimate reason must be in some degree related to job efficiency, the employer's needs, or the applicant's qualifications. That, to me, is implicit, though never expressed, in Texas Department of Community Affairs v. Burdine.12 Oversight of an applicant is not a legitimate reason, for, like inefficiency, it does not serve any justifiable employment purpose. Thus, in Roberts v. Gadsden Memorial Hospital13 the employer responded to a prima facie case of disparate treatment by claiming that he had simply never considered Roberts for the available position and therefore could not have been guilty of discriminatory intent. Instead, he had hired someone with whom he was friendly on a social basis. The court held that the employer's failure to consider Roberts was itself disparate treatment and therefore could not constitute a defense:

where ... an employer has reason to know that an employee is qualified for a position and that the employee might desire to be considered for that position, the employer's failure to consider the employee for promotion is not a legitimate, nondiscriminatory reason sufficient to rebut an inference of intentional disparate treatment.14

Cant by an employer and homilies against discrimination delivered to its employees do not suffice to explain the discriminatory effect of its employees' practices. If its employees' actions result in racial discrimination, their malfeasance is not sufficient to rebut the inference of racial discrimination already demonstrated by the plaintiff in establishing her prima facie case. The MSES is itself responsible for its employees' actions.

The need for efficiency constitutes a defense in disparate-treatment cases. It would be ironic if the employer's claim of its own inefficiency also served as a defense.

This ought to dispose of the defense to the disparate-treatment case. Even if it does not, I submit that the magistrate was clearly erroneous in concluding that the reasons advanced by MSES were not a pretext. The MSES, acting through its employees--the only way in which it could act--intentionally discriminated against Hill.

For these reasons, I would reverse the judgment, render judgment in Hill's favor on her discrimination claims and remand the case to the district court to determine the appropriate remedy.