PATRICK E. HIGGINBOTHAM, Circuit Judge:

This employment discrimination class action has been in the federal courts for eighteen years, a captive to large changes in the controlling law. It now makes its third appearance before this court. On behalf of himself and similarly situated class members, Thomas Johnson appeals the grant of summary judgment in favor of Uncle Ben's, Inc. We affirm.

I.

Thomas Johnson, an employee at a rice-processing plant owned by Uncle Ben's, Inc., filed this suit in 1974. The complaint alleged that, commencing in March 1972, UBI discriminated against him and similarly situated Black and Mexican-American employees in violation of 42 U.S.C. § 1981. He amended the complaint in 1975 to add a claim under Title VII of the Civil Rights Act of 1964, 42 U.S.C. § 2000e et seq.

The district court certified a class of Black and Mexican-American persons who have been employed or may in the future be employed by UBI. The case was tried to the bench from October 3 until October 21, 1977. At the conclusion of Johnson's direct case, the district court dismissed all claims except discrimination in the promotion of Black employees. UBI then called its personnel manager and three expert witnesses. At the conclusion of Johnson's direct examination of an expert rebuttal witness, the district court granted judgment in favor of UBI.

The first district court opinion held that the proportion of Blacks to whites in each job title at UBI should be compared to the ratio of Black to white workers in comparable jobs in the Houston Standard Metropolitan Statistical Area. Finding that the ratio of Black to white workers in each job title at UBI was similar to the proportion of Black to white workers in comparable jobs in the Houston SMSA, the district court held that UBI had not violated Title VII. Johnson v. Uncle Ben's, Inc., 628 F.2d 419, 425 (5th Cir. 1980).

We in turn vacated and remanded for further findings, holding that workers employed in similar jobs in the Houston SMSA were not necessarily the benchmark qualified applicant pool. Johnson v. Uncle Ben's, Inc., 628 F.2d 419 (5th Cir. 1980). We stated:

"If [UBI] hires laterally, the relevant comparison is to the general or qualified outside labor force. If Uncle Ben's fills jobs by promotion, the relevant comparison, as we recognized in James v. Stockham Valves Fittings Co., 559 F.2d [310] at 331, 341 [(5th Cir. 1977)], is the company's internal work force. The applicability of James in any given case turns on whether vacancies in non-entry level positions are or could be filled by promotion. If the vacant positions ordinarily are filled by lateral hires or hiring from among graduates of relevant educational programs, then the rigid James rule is inapplicable."

Johnson I, 628 F.2d at 425. We remanded for findings regarding "how many of those 394 employees [at UBI] hold jobs that ordinarily cannot be filled by promotion." Id. This district court was instructed to "determine the number of Uncle Ben's jobs that were filled by promotion and the number that were filled by hiring from outside of the Uncle Ben's work force." Id. at 426.

The Supreme Court, however, vacated Johnson I and remanded the case for reconsideration in light of its decision in Texas Dep't of Community Affairs v. Burdine, 450 U.S. 248, 101 S.Ct. 1089, 67 L.Ed.2d 207 (1981). Uncle Ben's, Inc. v. Johnson, 451 U.S. 902, 101 S.Ct. 1967, 68 L.Ed.2d 290 (1981). On remand, we held that Burdine was inapplicable to this disparate impact case and again remanded to the district court for further proceedings as stated in Johnson I. Johnson v. Uncle Ben's, Inc., 657 F.2d 750 (5th Cir. 1981).

Judge Sterling, who originally tried this case and issued the first district court opinion reviewed in Johnson I, died while this case was pending. The case was then assigned to Judge Hughes. On May 2, 1991, Judge Hughes granted summary judgment in favor of UBI. In his opinion, Judge Hughes stated that judgment for UBI was appropriate because Johnson had failed to make a prima facie case of disparate impact and had not stated an actionable claim under § 1981.

The district court held that Johnson proved only that there was a "high percentage of Black employees at Uncle Ben [sic] in low-level jobs versus a low percentage of minority employees in high level jobs." Because Johnson failed to prove that low level employees were the appropriate pool of qualified persons in the relevant labor market, he failed to prove any disparate impact.

The district court also found that Johnson failed to prove that any specific employment practice had a disparate impact upon the rate of Black promotion and that UBI had, in any event, rebutted any prima facie case by producing legitimate business reasons for its employment practices. Finally, relying on Patterson v. McLean Credit Union, 491 U.S. 164, 109 S.Ct. 2363, 105 L.Ed.2d 132 (1989), the district court rejected Johnson's § 1981 claim, finding that the claim did not rest on discrimination in the formation of a new employment contract.

The trial evidence consists largely of statistics concerning placement of Black and white employees at UBI's two processing plants and administrative offices in Houston, Texas. UBI's workforce is organized into three categories — plant workers paid an hourly wage, office workers paid an hourly wage, and salaried personnel. Each group is, in turn, subdivided into "zones," each zone representing a wage or salary range.

Johnson presented undisputed statistical evidence that Black employees were generally clustered in the bottom job zones within each of the three job categories, while the top job zones in each category were filled by white employees. Black employees comprise 95.3% of the workforce in the three lowest plant hourly job zones, holding jobs as porters, warehousemen, packers, fork lift operators, fumigators, bran hull helpers, rough rice helpers, and mill helpers. However, white employees held all of the highest two plant hourly job zones, including maintenance first class, boiler operator, and miller first class. The patterns were similar in office hourly and salaried positions. That is, Blacks were in the lowest office job zones, such as cafe porter, junior file clerk, and cook, and lowest salaried positions, including microbiology analyst and accountant. Whites held jobs in the higher zones in both office and salaried categories, including stenographer, export service clerk, receptionist, and computer operator and most of the salaried managerial and supervisory positions.

Johnson did not deny that promotion across category lines, while possible, was unusual. Generally, workers were promoted only to the top of the job category in which they start their employment. However, the parties fiercely disputed the lines of promotion within each of the three job categories. UBI argued at trial and on appeal that workers qualified to hold jobs in the lower zones of a job category were not necessarily qualified to hold higher jobs in the same category.

Johnson replied that the court should look to lower job zones as the qualified applicant pool for the higher job zones because the natural line of progression was a low level entry followed by gradual promotions through the job zones rising through the plant, office, or salaried hierarchy. He argues that Black employees were not promoted at the same rate as whites. Black workers entered at a low level and stayed there, stopped by a glass ceiling of race discrimination.

Johnson offered data showing that most jobs at UBI were filled through promotion. According to Johnson's undisputed evidence, in March of 1972, 65.3% of all the salaried positions, 53.6% of the office hourly positions, and 91.0% of the plant hourly positions at UBI were filled by promotion and not by initial hire from outside the UBI workforce. Johnson's data, however, said little about which jobs at UBI were filled by promotion or, more importantly, from which jobs different UBI jobholders were promoted.

II. Johnson's Title VII Claim

Johnson contends that three of UBI's employment practices — tests, formal educational requirements, and subjective promotion decisions by supervisors — had the effect of denying promotions to a disparate proportion of Black employees. For his prima facie case of disparate impact, Johnson "need[ed] only show that the facially neutral employment standards operate more harshly on one group than another." Carpenter v. Stephen F. Austin State University, 706 F.2d 608, 621 (5th Cir. 1983). This initial burden included proof of a specific practice or set of practices resulting in a significant disparity between the proportion of Black employees at UBI and the proportion of Blacks in the pool of qualified applicants. Cox v. City of Chicago, 868 F.2d 217, 220 (7th Cir. 1989).

Statistical disparities between the relevant labor pool and UBI's workforce are not sufficient. Pouncy v. Prudential Ins. Co. of America, 668 F.2d 795, 800-801 (5th Cir. 1982). A plaintiff must offer evidence "isolating and identifying the specific employment practices that are allegedly responsible for any observed statistical disparities." Wards Cove Packing Co., Inc. v. Atonio, 490 U.S. 642, 109 S.Ct. 2115, 2124, 104 L.Ed.2d 733 (1989) (quoting Watson v. Fort Worth Bank and Trust, 487 U.S. 977, 108 S.Ct. 2777, 2788, 101 L.Ed.2d 827 (1988) (plurality opinion)). Johnson must also "offer statistical evidence of a kind and degree sufficient to show that the practice in question has caused the exclusion of applicants for jobs and promotions because of their membership in a protected group." Watson, 108 S.Ct. at 2788-89; Pouncy, 668 F.2d at 801 ("The disparate impact model requires proof of a causal connection between a challenged employment practice and the composition of the work force"). Absent "a systematic analysis of the racial effects of all promotional criteria for each rank," Black Fire Fighters Ass'n. v. City of Dallas, Texas, 905 F.2d 63, 63 (5th Cir. 1990), Johnson cannot establish a prima facie case of disparate impact.

Johnson challenges three employment practices: UBI's "educational requirements," UBI's "subjective system of promotion," and UBI's use of irrelevant employment tests. Johnson was able to do little more than describe the content or application of the requirements. He failed a fortiori to show the specific effect that each had on Black promotions.

As evidence of UBI's educational requirements, Johnson cites the testimony of Dr. Richard Jeanneret, an industrial psychologist and UBI's expert witness. Jeanneret testified concerning his study of 119 job titles at UBI, in which he interviewed UBI's employees, observed work at UBI's facilities, and studied various job descriptions. At trial, he testified about the level of education that he believed UBI employees would need to perform different jobs successfully. Jeanneret did not explain in detail which jobs at UBI required which levels of education. He simply described the number of jobs at UBI that required a high school or college degree "or equivalent experience."

Jeanneret also did not purport to testify concerning the educational levels that UBI actually required. Rather, he testified only about the skills he believed UBI's employees ought to have. He conceded that UBI's own job descriptions did not contain "written educational requirements" and that he was testifying from his "expertise as opposed to some requirement that is imposed at Uncle Ben's." Jeanneret testified that UBI's "posting notices" announcing "something about education or an education related item" such as "training in . . . chemistry or math," but he did not testify at any time that UBI actually required employees to have any degrees or formal education level for promotion to any UBI job.

Dr. Jeanneret's testimony, therefore, does not compel the conclusion that UBI had formal educational prerequisites for promotion. Aside from Dr. Jeanneret's testimony, Johnson relies on his own testimony to establish that educational requirements existed at UBI. He testified that a supervisor told him that "you don't have the science background or the academic background to satisfy the needs of the job [to which Johnson sought promotion]" (emphasis added). However, even if the district court credited Johnson's testimony, that testimony indicates at most that a supervisor told Johnson that he lacked necessary "background" in science, not that Johnson lacked a formal degree or other specific educational prerequisite to promotion. According to Johnson's own testimony, the supervisor simply informed Johnson that "you don't have the skills that we are looking for." This testimony indicates at most that UBI required some unspecified level of scientific training.

In contrast to the testimony of Dr. Jeanneret and Johnson himself, UBI's personnel director, Herman Koehn, presented specific testimony that UBI did not require any particular level of education for most of the jobs at UBI. Koehn testified that "[w]e don't make an evaluation on whether [one] finished high school or not in terms of [whether one will] be offered a job." According to Koehn, he "never thought about job requirements in terms of high school or no high school." Rather than rely on formal education, Koehn testified that UBI

"would focus on the job and the ability to make numerical calculations and reading and writing. And [this would] not necessarily [be] reflected upon the number of years at school. It would be what they had learned and what they were able to do through displaying what they can do on the job."

Koehn also noted that there were supervisors who had never obtained a college degree. Koehn admitted that food supervisors in research and development had to have "knowledge in the sciences," but he denied that this knowledge required a "specific degree." Rather, the supervisor in research and development needed an educational background in the "physical sciences, chemistry, courses that relate to . . . food science." Koehn also stated that microbiology analysts ought to have "academic training in microbiology," but, again, he did not specify the level of training expected.

In short, Johnson presented frail evidence concerning the differing educational backgrounds that UBI required for different jobs and presented no evidence whatsoever concerning how many Black employees failed to meet UBI's requirements. Johnson contends that any educational requirements, regardless of their content, would "by definition" have a disparate impact on Black promotion rates, because Blacks in general tend to have less education than whites. To support this argument, Johnson cites national data from the 1970 U.S. census in his brief on appeal.

The national population, however, is not the qualified labor pool against which UBI's workforce should be compared. The effect of educational requirements on the ability of Blacks in the national population to get promotions at UBI has little relevance. New York Transit Authority v. Beazer, 440 U.S. 568, 584-87, 99 S.Ct. 1355, 1365-67, 59 L.Ed.2d 587 (1979) (statistics showing that 63%-65% of methadone users in New York City's public programs were Black or Hispanic does not show that a disproportionate number of Black or Hispanic Transit Authority employees were dismissed for using methadone). The question is whether and how specific educational requirements affected UBI employees seeking promotions. It is not obvious that Black UBI employees in the pool of employees qualified for promotion to higher levels would not have the skills or education allegedly required for promotion.

In short, there was little record evidence of the effects of educational requirements on Black promotion rates from the qualified applicant pool — employees at UBI. This is not to say that UBI's entire internal workforce constituted the appropriate statistical pool against which the proportion of Black employees at UBI should be measured. Assuming without deciding that some job zones at UBI should be compared with other lower UBI job zones, we find a complete absence of evidence that UBI employees were barred by educational requirements from reaching higher levels of employment at UBI. The district court did not clearly err in finding that Johnson failed to show that these alleged educational requirements affected Black promotion.

Johnson also argues that UBI allowed its supervisors to make promotion decisions subjectively and that this practice resulted in a disparity between the promotion rates of Black and white employees. However, "an employer's policy of leaving promotion decisions to the unchecked discretion of lower level supervisors should itself raise no inference of discriminatory conduct." Watson, 108 S.Ct. at 2786. See also Pouncy, 668 F.2d at 801-02. Johnson has not offered any evidence that Blacks' allegedly smaller number of promotions was causally related to this subjectivity. This cannot suffice to establish a prima facie case. Wards Cove, 109 S.Ct. at 2124-25.

Finally, Johnson refers to UBI's "use of invalidated employment tests" as one challenged employment practice that had a disparate impact on Black employees seeking promotions. There was testimony that UBI had used three different written tests to evaluate job applicants: (1) a typing test for jobs requiring typing; (2) an arithmetic aptitude test for clerical jobs requiring calculation such as statistical clerk; and (3) a "mental adaptability test," which purported to test basic reading and math skills. The last test was apparently discontinued sometime between 1971 and 1973.

Johnson presented no evidence of the effects of these tests on Black promotions. There was no testimony that Blacks performed more poorly on these tests than whites or that any Black employee was denied a promotion as a result of his performance on these tests. Indeed, Ethylene Burks, one of Johnson's witnesses and the only witness to testify about a Black employee's performance on the mental adaptability test, stated that the employee achieved a high score of 90 on the test. Burks also testified that achievement of any particular test score was not a prerequisite for promotion and that test scores were only one factor among many that a supervisor might consider. Given the dearth of evidence on the effects of the various tests on Black promotion rates, we conclude that Johnson failed to establish any causal nexus between the scores and the alleged disparate impact.

Johnson contends that this court's earlier decision in Johnson I precludes the district court from finding that he had failed to make a prima facie case. According to Johnson, the Johnson I court remanded for the narrow purpose of determining whether the appropriate pool of qualified applicants constituted the entire workforce of UBI or the population of people holding jobs similar to those at UBI in the Houston Standard Metropolitan Statistical Area. Johnson also argues that, if most jobs at UBI were filled through promotion, then, under Johnson I, the district court was required to find that Johnson had succeeded in establishing a prima facie case of disparate impact.

We need not determine whether or not the district court's findings went beyond the mandate of the Johnson I court. Assuming arguendo that they did, we find that intervening Supreme Court decisions justified such a departure. The "mandate rule" is "a specific application of the `law of the case' doctrine." Piambino v. Bailey, 757 F.2d 1112, 1120 (11th Cir. 1985). Under this rule, the district court must follow an appellate decision on an issue in all subsequent trial proceedings unless the presentation of new evidence or an intervening change in the controlling law dictates a different result or if the appellate decision is clearly erroneous and, if implemented, would work an egregious result. Falcon v. General Telephone Co., 815 F.2d 317, 320 (5th Cir. 1987).

If the Johnson I court held that a disparity between the proportion of Blacks in UBI's workforce and the relevant labor pool of qualified applicants together with the use of challenged employment practices were sufficient to establish a prima facie case of disparate impact, it has been contradicted by the Supreme Court's decision in Wards Cove, the Supreme Court's plurality opinion in Watson, and this court's decision in Pouncy. As we have explained, Johnson must identify a causal nexus between a specific employment practice and a disparity in Black promotions. The district court did not err in following Wards Cove and requiring evidence that the particular challenged practices caused a disparity in Black promotions.

III. Johnson's § 1981 Claim

Citing Patterson v. McLean Credit Union, 491 U.S. 164, 109 S.Ct. 2363, 105 L.Ed.2d 132 (1989), the district court held that Johnson's allegations of intentional discrimination were not actionable under 42 U.S.C. § 1981, because Johnson's allegations concerned "post-formation conduct of the employment relationship, rather than . . . the making or enforcing of a new contract." The district court found that Johnson's evidence of discrimination was based entirely on discrimination in "wage increases" and in promotions within each of the three basic job categories. The district court held that movement within each of the three categories — plant hourly, office hourly, and salaried — did not work sufficient change in the employer-employee relationship under Patterson.

Patterson requires discriminatory "conduct at the initial formation of the contract" or "conduct which impairs the right to enforce contract obligations through legal process." Patterson, 109 S.Ct. at 2374. Discriminatory denials of promotion do not state a claim under § 1981 unless the promotion denied to the plaintiff "rises to the level of an opportunity for a new and distinct relation between employee and employer." Id. at 2377.

Determining whether a promotion would create a "new and distinct relation" requires a fact-specific examination into employee's duties, pay, and responsibility before and after the promotion. Harrison v. Associates Corp. of North America, 917 F.2d 195, 198 (5th Cir. 1990). The inquiry does not lend itself to blanket prescriptions. At the least, "[R]outine increases in salary and responsibility which are clearly part of an original contract of employment" do not signal a new employment relation. Harrison, 917 F.2d at 198. "It would be very odd to regard each rung on the career ladder as a different employment relation." McKnight v. General Motors Corp., 908 F.2d 104, 110 (7th Cir. 1990).

Johnson presented little evidence of the precise nature of the promotions assertedly denied its class members. Rather, Johnson urged that class members were denied promotion "from hourly-paid positions to salaried positions and from nonsupervisory positions to supervisory positions." Johnson's anecdotal evidence of specific attempts to obtain promotions showed that the promotions involved routine upward movement by one or two job zones within a single job category — plant, office, or salaried. In most cases, both the pay raise and the change in responsibilities were modest, involving no assumption of supervisory responsibility or change from wage payment to payment of salary.

For instance, Ida Johnson, a junior file clerk (office job zone two), applied for the position of traffic clerk (office job zone four). Both jobs were essentially non-supervisory, clerical positions paid by the hour, the latter being distinguished primarily by the new duty of typing. Likewise, Marie Horner testified that a typist, Brenda Smith, was denied a promotion to the position of office receptionist — again, a move of two zones from one non-supervisory, office-hourly position to another. Zachary Perkins was denied a promotion from steeper-cooker (plant zone four) to dryer operator (plant zone seven). Both were non-supervisory positions involving the operation of plant machinery, and Perkins testified that steeper-cooker operators were normally promoted to dryer operator as a matter of course.

Two class members present a closer case. Clyde Cobb and Johnson himself sought and were denied promotions from non-supervisory jobs in salaried job zone seven (the lowest salaried job zone) to a supervisory position. As the promotion sought was from a non-supervisory position to a supervisory position, there is not a complete absence of evidence that the promotion involved a new employment relation: changes in supervisory status are relevant to determining whether a promotion creates a new and distinct relation under Patterson. Sitgraves v. Allied-Signal, Inc., 953 F.2d 570, 574 (9th Cir. 1992).

However, we find that the record evidence concerning the promotions sought by Cobb and Johnson is insufficient to create a genuine fact question. Attainment of supervisory status does not alone create a new and distinct employment relation. Partee v. Metropolitan School District of Washington Township, 954 F.2d 454, 457 (7th Cir. 1992); Mozee v. American Commercial Marine Service Co., 940 F.2d 1036, 1051-55 (7th Cir. 1991). Aside from the supervisory status of the jobs sought by Cobb and Johnson, the other evidence suggested that the promotions would not create a new employment relation. Dr. Jeanneret's undisputed testimony about the supervisory positions was that they could only be filled through promotion from lower-zoned positions. Such testimony indicates that the positions were simply rungs on a career ladder, not new employment contracts. Malhotra v. Cotter Co., 885 F.2d 1305, 1311 (7th Cir. 1989).

In any case, aside from the fact that Johnson and Cobb sought supervisory positions, Johnson has not pointed to specific record evidence that the promotions sought by Cobb and Johnson would create new employment relations. Given that the change from a non-supervisory to a supervisory position does not suffice by itself to create a new employment relation, Johnson has not carried his summary judgment burden. Celotex Corp. v. Catrett, 477 U.S. 317, 106 S.Ct. 2548, 2553, 91 L.Ed.2d 265 (1986); Skotak v. Tenneco Resins, Inc., 953 F.2d 909, 915-17 (5th Cir. 1992).

Johnson's general contention that the promotions in this case involved "new and distinct relations" sits uneasily with his contention at trial that UBI's job zones represented routine steps in a sequential hierarchy in which work in each job zone gave the qualifications needed for the duties of the next zone. With such a natural progression, promotion within a single job category and across only one or two pay zones is not likely to create a new employment contract. To the contrary, they appear to be a fulfillment of expectations implicit in the original employment contract. Johnson strenuously argued at trial that the zone-by-zone promotion was simply the ordinary progression of a UBI employee. It is difficult to accept that proposition and also the proposition that each promotion represented a "new and distinct relation." Carter v. South Central Bell, 912 F.2d 832, 840 (5th Cir. 1990). The district court did not err in concluding that under the undisputed evidence there was no "new and distinct relation between employee and employer" within the meaning of Patterson.

IV. Retroactivity of the Civil Rights Act of 1991

In a Rule 28(j) letter sent to the clerk of this court four days after the enactment of the Civil Rights Act of 1991, Johnson argued that the Civil Rights Act of 1991 ought to be applied to this case retroactively. The Civil Rights Act of 1991, 42 U.S.C. § 2000e-2(k), however, did not alter the "particularity" aspect of Wards Cove as applied in this case. The application of the Act has no effect on our disposition of Johnson's Title VII disparate impact claim, and we need not address whether the Act's provisions affecting Title VII disparate impact claims are retroactive.

The Act would, however, affect the disposition of Johnson's § 1981 claim. Section 101(2)(b) of the Act construes § 1981 to include

"the making, performance, modification, and termination of contracts, and the enjoyment of all benefits, privileges, terms, and conditions of the contractual relationship."

42 U.S.C. § 1981(b). Under § 1981 as amended by the Act, racial harassment and other discrimination in an employment relation occurring after contract formation is actionable. If the Act applies to this case, the district court erred in dismissing Johnson's § 1981 action on the ground that the discrimination did not occur during the formation of a new employment relation.

We must determine whether § 101 of the Act amending § 1981 applies retroactively to cases pending when the Act was enacted. We have not previously addressed the issue. Three circuits and the Equal Employment Opportunity Commission have done so. Luddington v. Indiana Bell Telephone Co., 966 F.2d 225 (7th Cir. 1992) (Posner, J.); Fray v. Omaha World Herald Co., 960 F.2d 1370 (8th Cir. 1992); Mozee v. American Commercial Marine Service Co., 963 F.2d 929 (7th Cir. 1992); Vogel v. City of Cincinnati, 959 F.2d 594 (6th Cir. 1992); EEOC Notice No. 915.002 (December 27, 1991). All have found that the Act does not apply retroactively to conduct occurring before the effective date of the Act.

We find the holdings of all other circuits on this issue persuasive. The statutory language and legislative history is inconclusive on the question of retroactive application. Applying a general presumption against retroactive application of substantive laws, we find that § 101 of the Act, 42 U.S.C. § 1981(b), should not be applied to a case pending on appeal that was filed and decided by the trial court before the enactment of § 101 and that arises out of conduct occurring before § 101's enactment.

In determining whether a statute is retroactive, we look first to the language of the statute. The language of the Civil Rights Act of 1991 offers little help. As one court has noted, Congress "dumped the [retroactivity] question into the judiciary's lap without guidance." Luddington at 227. The Act nowhere states that it applies either prospectively or retrospectively. It is silent on the subject, stating only that it "shall take effect upon enactment" — November 21, 1991.

Sections 109(c) and 402(b) of the Act state that the Act should not apply retroactively to certain categories of cases. P.L. No. 102-166, §§ 109(c), 402(b), 105 Stat. 1071-1100. One district court has reasoned that § 402(b) and § 109(c) imply that the statute should generally be applied retroactively. Otherwise, the specific sections withdrawing retrospective application would be "meaningless." Stender v. Lucky Stores, Inc., 780 F. Supp. 1302, 1304-05 (N.D.Cal. 1992).

Stender's reasoning rests too much on negative implication. Congress may have wanted to ensure that certain retroactive applications of the statute were barred without intending to reach any general conclusion about the statute's general retroactive application. Mozee at 933. Several Senators stated as much. Fray, 960 F.2d at 1377. Moreover, attempts to extend the Act explicitly to pending cases failed. President Bush vetoed the Civil Rights Act of 1990, which contained language applying the Act retroactively to pending cases. The Civil Rights Act of 1991 dropped this language and was signed by the President. It may have been that neither the proponents of retroactive application nor the supporters of pure prospectivity could obtain a veto-proof majority concerning the general application of the Act. We do not know, but the relevant point is that the negative implication cannot carry Stender's freight given the swirling confusion surrounding the Act's passage.

Legislative history also sheds little light on whether the Act should apply to pre-enactment conduct. There is little point in reciting speeches made on the floor of Congress concerning retroactivity. These remarks have been summarized before, see, e.g., Fray, 960 F.2d at 1376, and they "contain statements that both favor and disfavor the retroactive application of the 1991 Civil Rights Act to pending cases." Mozee at 934. See also Vogel, 959 F.2d at 598 (noting that Senators Danforth and Kennedy expressed different views concerning retroactivity of Act). We conclude only that members of Congress reached no consensus and left it to the courts to resolve. Luddington at 227; Mojica v. Gannett Co., Inc., 779 F. Supp. 94, 96 (N.D.Ill. 1991).

We are faced with a deliberately ambiguous statute, and we are asked to resolve political questions Congress was not able to answer. This difficulty is not unfamiliar. It is exacerbated by conflicting lines of authority in the Supreme Court's jurisprudence concerning statutory retroactivity. In Bradley v. Richmond School Board, 416 U.S. 696, 716, 94 S.Ct. 2006, 2018, 40 L.Ed.2d 476 (1974), the Supreme Court declared a "general rule that a court is to apply a law in effect at the time it renders its decision." Bradley seems to have adopted this "general rule" "even where the intervening law does not explicitly recite that it is to be applied to pending cases." Id. at 715, 94 S.Ct. at 2018. By contrast, in Bowen v. Georgetown University Hospital, 488 U.S. 204, 109 S.Ct. 468, 471, 102 L.Ed.2d 493 (1988), the Supreme Court held that the Secretary of Health and Human Services could not promulgate retroactive limits on reimbursable Medicare costs. According to Bowen,

"Retroactivity is not favored in the law. Thus, congressional enactments and administrative rules will not be construed to have retroactive effect unless their language requires this result."

Bowen, 109 S.Ct. at 471 (citations omitted). The Supreme Court has acknowledged the "apparent tension" between these two positions, Kaiser Aluminum Chem. Corp. v. Bonjorno, 494 U.S. 827, 110 S.Ct. 1570, 1577, 108 L.Ed.2d 842 (1990), but it has yet to choose between the two presumptions.

Our own decisions straddle the divide between Bowen and Bradley. Some decisions follow Bowen's "general rule barring retroactivity." Sierra Medical Center v. Sullivan, 902 F.2d 388, 392 (5th Cir. 1990). See also Walker v. United States Department of Housing and Urban Development, 912 F.2d 819, 831 (5th Cir. 1990). Other cases, however, follow Bradley's rule that "a change in law while a case is on direct appeal be given affect." See, e.g., Louviere v. Marathon Oil Co., 755 F.2d 428, 430 (5th Cir. 1985).

Forced as we are to choose a canon without the guidance of controlling authority, we find that § 101 should be construed not to apply to cases arising out of conduct occurring prior to the enactment of § 101. We follow the canon that statutes affecting substantive rights "are ordinarily addressed to the future and are to be given prospective effect only." Turner v. United States, 410 F.2d 837, 842 (5th Cir. 1969). See also United States v. Vanella, 619 F.2d 384, 385 (5th Cir. 1980) (quoting Greene v. United States, 376 U.S. 149, 160, 84 S.Ct. 615, 621, 11 L.Ed.2d 576 (1964)) ("`legislation must be considered as addressed to the future, not to the past . . . [and] a retrospective operation will not be given to a statute which interferes with antecedent rights'"). This canon has a lengthy pedigree, see Kaiser Aluminum, 110 S.Ct. at 1579 (Scalia, J., concurring), reflecting obvious and fundamental concerns of fairness and predictability. Luddington at 227.

In Bennett v. New Jersey, 470 U.S. 632, 638-40, 105 S.Ct. 1555, 1559-60, 84 L.Ed.2d 572 (1985), the Court held that substantive provisions of amendments to the 1978 Amendments to the Elementary and Secondary Education Act cannot be applied retroactively to funds expended in 1971-72. In distinguishing Bennett from Bradley, the Supreme Court noted that the rule in Bradley was limited by "another venerable rule of statutory interpretation, i.e., that statutes affecting substantive rights and liabilities are presumed to have only prospective effect." Bennett, 470 U.S. at 639, 105 S.Ct. at 1560. The Bennett Court noted that Bradley concerned allowance of attorney's fees under § 718 of the Emergency School Aid Act, 20 U.S.C. § 1617 — a remedial provision — not substantive obligations or rights under a statute. Id.

Section 101 affects substantive antecedent rights. Under Patterson, § 1981 did not prohibit discrimination in promotions before the enactment of § 101. Section 101 extended § 1981 to such discriminatory conduct. We then presume that § 101 does not apply to conduct that occurred before its enactment, absent clear evidence to the contrary. There is no such clear evidence.

We recognize the apparent anomaly that, at the time of UBI's allegedly discriminatory conduct, Patterson had not yet been decided and, under the decisions of many lower courts, § 1981 applied to racial discrimination in promotions. UBI's reliance on the law announced in Patterson, therefore, may be minimal. Some opinions have argued that, given such minimal reliance, the presumption against retroactivity should not operate. Mozee at 939-40 (Cudahy, J., dissenting); Stender, 780 F. Supp. at 1308; Mojica, 779 F. Supp. at 98.

We are not persuaded. As a matter of law, the rule announced in Patterson applies retroactively to UBI's conduct in 1974. Lavender v. V. B. Transmissions Auto Repair, 897 F.2d 805, 806-07 (5th Cir. 1990). Cf. James B. Beam Distilling Co. v. Georgia, ___ U.S. ___, 111 S.Ct. 2439, 115 L.Ed.2d 481 (1991). UBI is just as entitled to the preservation of its substantive interests under this rule as litigants whose conduct occurred after Patterson was decided. Any other holding would require unwieldy distinctions between classes of litigants based on the degree to which they relied on the legal regime antedating the Civil Rights Act of 1991. We decline to embark on such an inquiry. Luddington at 229-30.

Having decided that § 101 does not apply retroactively to UBI's conduct, it follows that Johnson's § 1981 claims are governed by the Supreme Court's decision in Patterson. As we have explained, we affirm the district court's finding that Patterson bars Johnson's § 1981 claim.

AFFIRMED.