In the fall of 1985, after years of heavy losses, TWA decided that it had to reduce its labor costs or go under. (TWA finally declared bankruptcy during the pendency of this appeal, but the bankruptcy judge lifted the automatic stay to permit the appeal to be argued and decided.) The measures it took included a 14 percent cut in the wages of its nonunionized employees and a reduction in those employees’ fringe benefits. The reduction included capping paid vacations at 4 weeks a year. Previously, all employees who had more than 16 years of service with the company had been entitled to even more paid vacation — workers who had at least 30 years of service to 7 weeks. We do not understand this to have been a contractual entitlement. It was a practice that the company was free as a matter of contract law to change at any time, provided that the change was prospective — that is, that it did not take away vacation rights that the worker had accrued by working during the period in which the previous policy was in effect. If a worker who under the previous policy was entitled to 1 week’s vacation for every 7 weeks that he worked had worked 35 weeks and thus had earned 5 weeks of vacation that he had not taken when the new policy was announced, the company could not confiscate a week of his vacation. Illinois Wage Payment and Collection Act, Ill.Rev.Stat. ch. 48, M 39m-l et seq.; National Metalcrafters v. McNeil, 784 F.2d 817 (7th Cir.1986). For the future, however, his vacation entitlements would accrue at a slower rate.
Most workers having more than 16 years of service with the same company, and all workers with 30 years or more, are at least 40 years old. So the brunt of the vacation cap inevitably fell mainly on workers within the class protected by the Age Discrimination in Employment Act, 29 U.S.C. §§ 621 et seq., precipitating this suit by members of the protected class whose vacation rights were curtailed by the 1985 cost-cutting program. The district judge granted summary judgment for TWA on the grounds that the plaintiffs had presented no evidence of intentional discrimination (“disparate treatment”) and that their disparate-impact claim was barred by section 4(f)(2) of the Act, which allows an employer “to observe the terms of a bona fide seniority system ... which is not a subterfuge to evade the purposes of this [Act].” 29 U.S.C. § 623(f)(2). 767 F.Supp. 867 (N.D.Ill.1991). TWA defends the district court’s reasoning but adds that the plaintiffs haven’t made out even a prima facie claim of age discrimination.
We have our doubts, unnecessary to resolve, whether the seniority provision is applicable here. The provision was modeled on a provision in Title VII, 42 U.S.C. § 2000e-2(h), where its function is plain— to protect workers’ seniority rights notwithstanding that black and other minority workers will often have been hired more recently than white workers and hence may be disadvantaged by strict adherence to seniority in determining layoffs and rehires. International Brotherhood of Teamsters v. United States, 431 U.S. 324, 349-53, 97 S.Ct. 1843, 1861-64, 52 L.Ed.2d 396 (1977). The provision thus operates as an exemption, a safe harbor. Its function in the age discrimination setting is less clear. The strong positive correlation between seniority and age means that seniority systems usually benefit the class protected by the age discrimination law, in contrast to the tendency of such systems to harm the intended beneficiaries of Title VII. We can, though, imagine a case in which an older worker who had been hired recently was laid off before a younger one because the latter had greater seniority, and in such a case the seniority provision of the statute would protect the younger worker. That hypothetical case is remote from our case. The vacation benefits that TWA curtailed had been keyed to seniority, but by being so keyed they benefited older workers at the expense of younger ones. So TWA cannot argue that when it curtailed those benefits it hurt older workers for the sake of preserving the seniority of younger ones. There is no competing interest of senior workers in this case; the senior, the protected, and the injured are one and the same, as the plaintiffs are at *1163pains to argue because it is the core of their disparate-impact case.
Section 4(f)(2) also provides protection for “bona fide employee benefit plan[s],” and here the function of the exemption (which has no counterpart in Title VII) is clearer. It allows the employer to take account of, and reflect in the compensation package that he offers his employees, the higher costs of certain fringe benefits, such as life and medical insurance, for older workers. Public Employees Retirement System v. Betts, 492 U.S. 158, 109 S.Ct. 2854, 106 L.Ed.2d 134 (1989); Metz v. Transit Mix, Inc., 828 F.2d 1202, 1219-20 (7th Cir.1987) (dissenting opinion). We do not understand TWA to be arguing that a schedule of paid vacations can be regarded in that light.
But even if section 4(f)(2) does not shield TWA from liability for curtailing benefits, the company’s alternative ground — that the plaintiffs have not made out even a prima facie case of disparate impact — is solid. We need not speculate on how much of the theory of disparate impact, viewed as a judicial doctrine of general applicability to discrimination cases, survives Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 109 S.Ct. 2115, 104 L.Ed.2d 733 (1989) (a question we discussed in Allen v. Seidman, 881 F.2d 375 (7th Cir.1989); Ortega v. Safeway Stores, Inc., 943 F.2d 1230, 1244-45 (10th Cir.1991), and Newark Branch v. Town of Harrison, 940 F.2d 792, 803-04 (3d Cir.1991)); or what, if any, bearing section 105 of the Civil Rights Act of 1991, affirming disparate impact, has on that question; or whether the new civil rights act has any possible bearing on an age discrimination suit, cf. Tyler v. Bethlehem Steel Corp., 958 F.2d 1176, 1181-82 (2d Cir.1992), since it does not amend the Age Discrimination in Employment Act (save for an irrelevant modification of the limitation period); or whether disparate impact has ever been a viable theory of age discrimination, as questioned in Davidson v. Board of Governors, 920 F.2d 441, 444 (7th Cir.1990), and Metz v. Transit Mix, Inc., supra, 828 F.2d at 1220 (dissenting opinion), and in a student-written Note, “Age Discrimination and the Disparate Impact Doctrine,” 34 Stan.L.Rev. 837 (1982), which finds a safe harbor in the provision (section 4(f)(1)) that allows employers to base differential compensation on “reasonable factors other than age.” The weight of authority, illustrated by EEOC v. Bordens, Inc., 724 F.2d 1390, 1394-95 (9th Cir.1984), and Leftwich v. Harris-Stowe State College, 702 F.2d 686, 690 (8th Cir.1983), is contra — is that disparate impact is a viable doctrine under the age discrimination law — and it is at least clear that an employer cannot be permitted to use a proxy for age, such as having grey hair, to evade the prohibition of intentional discrimination, Metz v. Transit Mix, Inc., supra, 828 F.2d at 1220 (dissenting opinion); but such use is really a species of intentional discrimination, which disparate impact is not, or at least not quite. These are wonderful conundra but it is enough for our decision today that this case makes no sense in disparate impact terms.
The thinking behind the concept of disparate impact is that if an employment practice bears disproportionately against the members of a protected group, the employer ought to be required to justify it. Dothard v. Rawlinson, 433 U.S. 321, 97 S.Ct. 2720, 53 L.Ed.2d 786 (1977). So if he fixes a minimum height which just happens — perhaps entirely accidentally, innocently — to exclude almost all women but few men, he ought to be required to justify the rule by reference to nondiscriminatory factors, such as the needs of the job. Id. No one until this case had suggested pushing this principle to the point of deeming changes in compensation, made in response to business adversity, suspect merely because the impact of the changes might not be — almost certainly would not be — random with respect to age. Cf. Metz v. Transit Mix, Inc., supra, 828 F.2d at 1210. A company that for legitimate business reasons decides to cut wages across the board, or to cut out dental insurance, or to curtail the use of company ears is not required to conduct a study to determine the impact of the measure on employees grouped by age and if it is nonrandom to prove that the same amount of money could not have been saved in some different fashion. As the *1164plaintiffs’ counsel himself emphasized to us at argument, virtually all elements of a standard compensation package are positively correlated with age. The 14 percent wage cut that the plaintiffs do not challenge unquestionably hit TWA’s older workers harder, because their wages are on average higher. The elimination of dental insurance, which the plaintiffs also do not challenge, likewise must have hit the older workers harder because (after one reaches adulthood, anyway) dental problems tend to increase with age.
The plaintiffs’ counsel admitted that on their view of the law TWA cannot as a practical matter reduce its paid vacations to a maximum of 4 weeks, generous as that maximum is, because it would mean slashing 3 weeks off the vacations of the most senior workers, a result impossible to accomplish on an age-neutral basis. TWA cannot lop 3 weeks off the vacations of its most junior workers, because they are entitled to only 2 weeks. Even if a proportional reduction is considered nondisparate (but why should it be? A week is a week), there is no practical way to achieve it. A % reduction in the paid vacations of the junior employees (the proportional reduction necessary to knock the senior employees down to 4 weeks) would leave those employees with only 6 days of paid vacation a year and the plaintiffs’ counsel concedes that TWA cannot give its junior employees less than 2 weeks and hope to retain them as employees; and it cannot give them 2 weeks without leaving the senior employees with more than 4. The next tier of employees, by seniority, get 3 weeks of vacation. Cut them by % and they will have less than 2 weeks. There is no way to bring the most senior employees down to 4 weeks that will maintain a rational system of paid vacations for all employees (a minimum of 2 weeks vacation for all) yet not cut the more senior workers’ vacations more than the less senior ones’ proportionately as well as absolutely. There is something wrong with an interpretation of the Age Discrimination in Employment Act that forbids a bankrupt corporation to adopt a companywide policy of limiting paid vacations to 4 weeks a year and that would as a practical matter forbid all firms to reduce wages or fringe benefits in periods of adversity.
The concept of disparate impact was developed for the purpose of identifying situations where, through inertia or insensitivity, companies were following policies that gratuitously — needlessly—although not necessarily deliberately, excluded black or female workers from equal employment opportunities. Griggs v. Duke Power Co., 401 U.S. 424, 91 S.Ct. 849, 28 L.Ed.2d 158 (1971); Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 987, 108 S.Ct. 2777, 2785, 101 L.Ed.2d 827 (1988). Often these were policies that had been adopted originally for discriminatory reasons and had not been changed when the employer ceased deliberately discriminating — if he had; for another way of looking at the disparate impact approach is that it is primarily intended to lighten the plaintiff's heavy burden of proving intentional discrimination after employers learned to cover their tracks.
Across-the-board cuts in wages and fringe benefits necessitated by business downturns or setbacks are a far cry from the situations that brought the theory into being. These cuts are not a legacy of deliberate discrimination on grounds of age; they are not the product of inertia or insensitivity. They are an unavoidable response to adversity. Their adverse impact on older workers is unavoidable too, because it is impossible to reduce the costs of fringe benefits without making deeper cuts in the benefits of older workers, simply because, by virtue of being older, they have greater benefits. In this situation no inference of discrimination, whether deliberate or merely inadvertent (but avoidable), can be drawn from the greater impact of curtailing employees’ benefits on older workers. Otherwise every time a company tried to reduce its labor costs the federal courts would be dragged in and asked to redesign the reduction so as to shift the burden to some unprotected class of workers — a shift that might incidentally be impossible to accomplish, as it might drive all the younger *1165workers out of the company (the 6-days-a--year vacation problem).
An argument for the judicial assumption of such a role might be teased out of dicta in cases that apply the disparate impact approach to age cases, see, e.g., Abbott v. Federal Forge, Inc., 912 F.2d 867, 876 (6th Cir.1990), or drawn by analogy from racial discrimination cases, such as Watson v. Fort Worth Bank & Trust, supra, 487 U.S. at 988, 108 S.Ct. at 2785, that involve different considerations. But it would not be a good argument, given the limited competence of federal judges to redesign corporate compensation packages. And notice the ratchet in the plaintiffs' theory. On the plaintiffs' view, a benefit that grows with age, as wages and paid vacations do, becomes locked in. If the company encounters adversity, it must balance its books on the backs of the younger workers-cut their wages, their vacations, lay them off, whatever. So business reverses will make the disparity in compensation between younger and older workers grow ever wider in favor of the older.
We could accept the plaintiffs' approach to the extent of finding a disparate impact, and then rule that the employer's business justification was compelling. But that would not be a satisfactory approach. It would mean that every time an employer made an across-the-board cut in wages or benefits he was prima facie violating the age discrimination law. Practices so tenuously related to discrimination, so remote from the objectives of civil rights law, do not reach the prima facie threshold.
Alternatively, however, the plaintiffs argue that having set up a voluntary early-retirement plan designed (as such plans are, of course) to induce its older, and therefore generally more expensive, workers to leave voluntarily, TWA cut the vacations of the older workers as a way of prodding them to take early retirement under the plan. This would be deliberate discrimination, and clearly actionable, Bartman v. Allis-Chalmers Corp., 799 F.2d 311, 314-15 (7th Cir.1986), but the suggestion that this is what happened is implausibly Machiavellian. An extra 3 weeks of vacation a year is very nice, but it is not likely to be the decisive factor in a career employee's decision about whether to give up a paying job for the modest benefits of early retirement. But plausible or not, it is a theory that, as the district judge found, the plaintiffs have been unable to substantiate even to the limited degree that they must do to entitle them to a trial. Their strongest evidence is a statement by one of TWA's officers that at a meeting on the vacation cap "I did remember someone saying that if it really is offensive to them [i.e., older workers], they do have early retirement options." Standing alone, unexplained, this is no evidence that the vacation cap was adopted in order to drive older workers to take early retirement. Everyone knew that the vacation cap would pinch the most senior workers the hardest and that they might squawk-why shouldn't they? But they would have the consolation of being able to take early retirement if they really wanted the additional leisure. That is all the quoted remark says. To construe it as a suggestion that TWA adopted the vacation cap in the hope the older workers would be so offended that they would take early retirement is sheer conjecture. We do not know who made the statement or even whether it was made before or after the cap was adopted. No jury would be permitted to find intentional age discrimination on the basis of this statement. There is not even evidence that any older worker took early retirement because of the vacation cap.
AFFIRMED.