POSNER, Circuit Judge.

This suit charges that the J.C. Penney Company, which owns a chain of retail stores, committed sex discrimination in violation of Title VII of the Civil Rights Act of 1964, 42 U.S.C. §§ 2000e et seq. The plaintiff, Diane Colby, is an employee of Penney in Illinois. She is suing on behalf of herself and all similarly situated Penney employees in Illinois. The suit challenges Penney’s “head of household” rule. Until 1971, Penney allowed only its male employees to elect coverage for their spouses under the company’s medical and dental plans. In that year Penney replaced this explicitly sex-based rule with the “head of household” rule. This rule allows any employee, male or female, to elect coverage for the employee’s spouse if the spouse *1122 earns less than the Penney employee. Mrs. Colby and the members of her class are female employees who cannot elect coverage for their spouses because their spouses earn more than they do. She argues that although the head of household rule may not (despite its origins) be intended to discriminate against women, it has that effect. Female employees of Penney tend to be concentrated in low-paying jobs, and male wages are on average higher than female wages; hence a larger fraction of male than female employees of Penney are eligible to elect coverage for their spouses under the head of household rule. At argument Mrs. Colby’s lawyer, staring fixedly at the members of this panel, likened the head of household rule to a rule disqualifying bald persons: the incidence of the rule would be concentrated on one sex even though the rule was not explicitly based on sex. Such “disparate impact” can condemn an employment practice under Title VII, unless some good business justification is shown. See, e.g., Dothard v. Rawlinson, 433 U.S. 321, 97 S.Ct. 2720, 53 L.Ed.2d 786 (1977).

Mrs. Colby’s suit was filed in 1980, but lay dormant for five years awaiting the outcome of a parallel suit brought by the Equal Employment Opportunity Commission in the federal district court in Detroit, Michigan, challenging Penney's head of household rule on a nationwide basis. The EEOC lost that suit late in 1985, and its appeal to the Sixth Circuit is pending. See EEOC v. J.C. Penney Co., 632 F.Supp. 871 (E.D.Mich.1985), appeal docketed, No. 86-1139 (6th Cir. Feb. 18, 1986). When the district court’s decision in the Detroit case was handed down, the district judge in this case had under advisement Penney’s motion for summary judgment, and in March 1986 he granted the motion on the ground that the Detroit decision was stare decisis. The judge described the failure of the EEOC’s suit as “persuasive precedent” but did not discuss the merits of the Detroit decision or of this case. The court noted noncommittally that another suit against Penney’s head of household rule had also failed. See Wambheim v. J.C. Penney Co., 705 F.2d 1492 (9th Cir.1983).

Mrs. Colby has appealed, challenging not only the dismissal of her action but the district court’s earlier ruling refusing (without any statement of reasons) to certify her suit as a class action. Penney defends both the dismissal and the denial of class certification on a variety of grounds, most of which the district court did not address.

We first address Penney’s argument that Mrs. Colby has no standing to maintain this suit, and therefore that it is not within the subject matter jurisdiction of the federal courts. O’Shea v. Littleton, 414 U.S. 488, 494, 94 S.Ct. 669, 675, 38 L.Ed.2d 674 (1974). Penney argues, frivolously, that Mrs. Colby has no standing because she has never applied for spouse coverage. But the whole point of the suit is that she is ineligible; nor does Penney contend otherwise. Second and more substantially, but we think unavailingly, Penney argues that throughout the entire period of Mrs. Colby’s employment with Penney her husband has never lacked insurance coverage at his place of employment, and therefore she has not been harmed by the head of household rule. But that is like saying that if an insurance company cancels your medical insurance policy, you can't complain till you get sick. Mrs. Colby claims to want spouse coverage, and there is no basis in the record compiled thus far in this case to doubt that she is sincere and would derive a benefit from the additional coverage for which she would not have to pay. For aught that appears, Mr. Colby’s coverage by his employer is less generous than Penney’s spouse coverage, even though he has a higher salary than his wife — indeed, maybe he has a higher salary because his employer provides fewer fringe benefits. Or he and his wife may be worried about what would happen if he lost his job. Maybe he would like to quit and find another job and would do so if he had interim coverage, which may be unavailable from his present employer. Although it is possible that this is a nuisance suit and Mrs. Colby has no desire for spouse coverage, her complaint alleges that she does desire it and Penney has not cast enough doubt on the truth of the allegation by its speculations in this court to justify our throwing the case out on jurisdictional grounds or even remanding for a hearing on jurisdiction.

We turn now to the question whether the suit was properly dismissed on the merits. The ground relied on by the district court was stare decisis. That means deciding a case in accordance with what *1123 has been decided previously in other, similar cases (similar in the sense of not being legally distinguishable); and in a literal sense that is what the district judge did. But the doctrine is more complex than the summary definition just offered, and was misapplied. The distinction essential to understanding the doctrine is between the persuasiveness and the authority of a previous decision. Any decision may have persuasive force, and invite — indeed compel— the careful and respectful attention of a court confronted with a similar case. But unless the earlier decision is authoritative, the court that decides the later case does not discharge its judicial responsibilities adequately by merely citing the earlier decision and following it without so much as indicating agreement with it, let alone analyzing its merits. That is what the district court did here; it dismissed this case because the district court in Detroit had dismissed a similar case, not because the reasoning in the Detroit decision was persuasive. The district judge in this case did say that the “EEOC’s action ... does offer persuasive precedent,” but taken in conjunction with his earlier statement that “the doctrine of stare decisis can serve to preclude the possibility that plaintiff can prevail in this case,” and his failure to discuss the merits of the Detroit decision or of the present case, the reference to “persuasive precedent” appears intended to mean only that the dismissal of the EEOC’s suit entitled the judge in this case to treat the Detroit decision as stare decisis.

That would be a proper application of stare decisis only if the Detroit decision were authoritative, which is to say, had weight independent of its persuasive power. Whether a decision is authoritative depends on a variety of factors, of which the most important is the relationship between the court that decided it and the court to which it is cited later as a precedent. The simplest relationship is hierarchical: the decisions of a superior court in a unitary system bind the inferior courts. The most complex relationship is between a court and its own previous decisions. A court must give considerable weight to those decisions unless and until they have been overruled or undermined by the decisions of a higher court, or other supervening developments, such as a statutory overruling. But it is not absolutely bound by them, and must give fair consideration to any substantial argument that a litigant makes for overruling a previous decision.

One can hardly speak of stare decisis at all with regard to parallel court systems. We are bound to follow a decision of the Supreme Court unless we are powerfully convinced that the Court would overrule it at the first opportunity. Olson v. Paine, Webber, Jackson & Curtis, Inc., 806 F.2d 731 (7th Cir.1986). In cases where the rule of decision is the law of some state, we owe the same deference to the highest court of that state. Erie R.R. v. Tompkins, 304 U.S. 64, 78-80, 58 S.Ct. 817, 822-23, 82 L.Ed. 1188 (1938). But we can give decisions of the House of Lords or the Cour de Cassation as little weight as we like (putting aside complications introduced by conflict of laws principles, which might make foreign law controlling, in just the same way that state law controls the substantive issues in a diversity suit). We have an intermediate obligation to our sister federal courts of appeals. Bearing in mind the interest in maintaining a reasonable uniformity of federal law and in sparing the Supreme Court the burden of taking cases merely to resolve conflicts between circuits, we give most respectful consideration to the decisions of the other courts of appeals and follow them whenever we can. Our district judges should, of course, do likewise with regard to such decisions, as we noted in Richards v. Local 134, Int’l Brotherhood of Electrical Workers, 790 F.2d 633, 636 (7th Cir.1986). But neither this court nor the district courts of this circuit give the decisions of other courts of appeals automatic deference; we recognize that, within reason, the parties to cases before us are entitled to our independent judgment. For a good discussion see 1 B Moore’s Federal Practice ¶ 4.02[1], at pp. 14-16 (2d ed. 1984).

*1124 A posture somewhere in between some deference and complete deference is proper when cases in different circuits challenge the same practice of the same defendant, particularly if different outcomes would place the defendant under inconsistent obligations. If Penney had lost in Detroit, Mrs. Colby might have been able to use the judgment there to bar Penney from defending itself in this case. See Parklane Hosiery Co. v. Shore, 439 U.S. 322, 99 S.Ct. 645, 58 L.Ed.2d 552 (1979) (“offensive” use of collateral estoppel). A judgment for Penney cannot be used against her in the same way, because she was not a party to the Detroit case; but the fact that the Detroit case involved the identical issue in a lawsuit brought against the identical defendant by a plaintiff having the same interests as Mrs. Colby is a reason for the district court in the present case to have given serious consideration to the Detroit decision — but not a reason to invoke stare decisis and give the decision complete deference automatically, as the court appears to have done.

Where different outcomes would place the defendant under inconsistent legal duties, the case for the second court’s not going into conflict with the first is particularly strong. A conflict would place the defendant in an impossible position unless the Supreme Court agreed to hear the case, which it might be reluctant to do if the conflicting decisions, however excruciating for the defendant, raised no issue of general significance — yet might feel obliged to do anyway. See, e.g., GTE Sylvania, Inc. v. Consumers Union of the United States, Inc., 445 U.S. 375, 100 S.Ct. 1194, 63 L.Ed.2d 467 (1980). The second condition for greater deference — different outcomes would place the defendant under inconsistent obligations — is missing in this case, however. And even in that situation, district judges in this circuit must not treat decisions by other district judges, in this and a fortiori in other circuits, as controlling, unless of course the doctrine of res judicata or of collateral estoppel applies. Such decisions will normally be entitled to no more weight than their intrinsic persuasiveness merits. The reasons we gave for giving some though not controlling weight to decisions of other federal courts of appeals do not apply to decisions of other district courts, because the responsibility for maintaining the law’s uniformity is a responsibility of appellate rather than trial judges and because the Supreme Court does not assume the burden of resolving conflicts between district judges whether in the same or different circuits. Federal district judges in Detroit do not make law that is binding on federal district judges in Chicago. In light of this bedrock principle it becomes almost a detail that the EEOC was proceeding on a different theory than the plaintiff in this case (intentional discrimination there versus disparate impact here) and that the Detroit decision is not final, since it has been appealed and the appeal is pending.

The Detroit decision and even more clearly the Ninth Circuit decision have no preclusive effect in the present case because of res judicata or collateral estoppel, as the district judge in this case correctly held en route to his unfortunate application of stare decisis, but as Penney contests. Mrs. Colby was not a party to either of those suits; she has never had her day in court. It is true as Penney points out that the EEOC in the Detroit case was seeking nationwide relief against Penney’s head of household rule and that, if the EEOC had won, its victory would have given Mrs. Colby and the members of her class all that she seeks by the present suit. Penney argues that in effect Mrs. Colby was a member of a class represented by the EEOC and that not having opted out she is bound by the outcome of the class action. We disagree, however, apart from the fact that since the appeal from the Detroit decision is pending, that decision is not final for purposes of allowing a defense of res judicata to be sustained in the present suit.

Mrs. Colby relies heavily on General Telephone Co. v. EEOC, 446 U.S. 318, 100 S.Ct. 1698, 64 L.Ed.2d 319 (1980), where the Supreme Court held that the EEOC, when it brings a suit seeking injunctive relief and *1125 backpay for a group of employees, is not required, and cannot be compelled, to seek class certification under Rule 23. Such a certification would bar additional litigation by members of the class unless they opted out and the certification was under a provision of Rule 23 that permits opting out. But the Court said that Rule 23 “has no application to” a suit by the EEOC. Id. at 324, 100 S.Ct. at 1703. The Court was “unconvinced that it would be consistent with the remedial purpose of the statutes to bind all ‘class’ members with discrimination grievances against an employer by the relief obtained under an EEOC judgment or settlement against the employer.” Id. at 333, 100 S.Ct. at 1707. It is true that this passage assumes that the EEOC has won rather than lost the suit, at least to the extent of obtaining a settlement, but we cannot see what difference this might make to the Court’s holding. If as in this case the EEOC loses, still the persons on whose behalf it sued cannot be bound by the judgment as if they had been parties to the EEOC’s suit or members of a class in a class action certified under Rule 23.

But in invoking General Telephone, Mrs. Colby may seem to be tilting at windmills. The Detroit suit never was certified as a class action; hence it could not bar her suit by virtue of Rule 23 even if that suit could have been certified as a class action. The ground for Penney’s res judicata defense is not Rule 23; it is the doctrine of “virtual representation.” (An unfortunate term: the British used it in answer to the colonists’ complaint that they had no representation in the British Parliament.) Under this doctrine, even if a party to the second suit, such as Colby, was not a party to the first suit (the EEOC’s suit), or in privity with that party (e.g., an assignee), still if the party to the first suit was an adequate representative of the other’s interests, the other may be barred by the judgment in that suit. See, e.g., Southwest Airlines Co. v. Texas Int’l Airlines, Inc., 546 F.2d 84, 97-101 (5th Cir.1977). We dealt with the general problem in a recent decision, Secretary of Labor v. Fitzsimmons, 805 F.2d 682, 692-94 (7th Cir.1986) (en banc). Although no uniform pattern has emerged from the cases, General Telephone reinforces our decision in Eirhart v. Libbey-Owens-Ford Co., 616 F.2d 278, 282-83 (7th Cir.1980), which held (before General Telephone) that a consent decree in an EEOC suit does not foreclose subsequent suits by the individuals on whose behalf the EEOC had sued; “virtual representation,” whatever its vitality elsewhere, does not apply to Title VII. Salinas v. Roadway Express, Inc., 735 F.2d 1574, 1580 (5th Cir.1984), is a similar case, with a similar result. This part of Eirhart is only an alternative holding, and it was not a case the EEOC had lost; still, we have found no decision that applies the doctrine of “virtual representation” to Title VII, and to do so would be inconsistent with the spirit though not the letter of General Telephone. The ground on which the Supreme Court held Rule 23 inapplicable was that Congress would not have wanted the judgment in the EEOC’s case to foreclose private suits, and that ground is as applicable to an attempt to achieve foreclosure through the judge-made doctrine of virtual representation as it is to an attempt to do the same thing through Rule 23, which is just another way of skinning the same cat.

Next Penney argues that even if the present suit is not barred, we should affirm dismissal simply because the suit has no possible merit. Penney says that we held in American Nurses’ Ass’n v. Illinois, 783 F.2d 716 (7th Cir.1986), that where the sexual discrimination challenged under Title VII is a difference in compensation and the plaintiff is proceeding only under a theory of disparate impact, her suit must fail; a difference in compensation that merely affects the sexes differently does not violate Title VII. Penney asks us — quite legitimately — to invoke stare decisis and stand by our decision in American Nurses’ Ass’n. We would respond favorably to this request if we thought that the decision had held what Penney says it did. The relevant issue there was whether “comparable worth” is a permissible theory *1126 for a Title VII suit, and we held it was not. The plaintiffs had tried to base liability on the fact that their employer paid higher wages to workers in job classifications predominantly occupied by men than to workers in job classifications predominantly occupied by women, though it paid the same wages to men and women within each classification. It was not a matter of the employer’s having some rule, test, or criterion for hiring or assignments, which if it was discriminatory in effect (“disparate impact”) might be challenged on that ground. The challenge was to the wage differences among the different job classifications. Since women were not barred from seeking employment in the higher-paying job classifications, the wage difference reflected the relative market value of the jobs, not of men and women. And this meant that there was no feasible way for a court to correct the discrimination. Unlike a case where men and women are paid different wages for the same work, so that the discrimination can be cured by ordering the employer to pay them the same wages, or are prevented from competing for jobs, so that the discrimination can be eliminated by lifting the bar, the court in a “pure” comparable worth case (i.e., one where women are not steered by the employer into the lower-paying job classifications) would have to fix the wage differential between different job classifications; for no one supposed that every employee of the State of Illinois should be paid the identical wage. Nor was it apparent how women would benefit from a judicial order that reduced wages in traditional “men’s jobs” and raised them in traditional “women’s jobs.” The order would both hurt women who held “men’s jobs” and encourage more men to compete for jobs at present held by women. Finally, by divorcing compensation from market wages, such an order could seriously impair the efficiency of labor markets. All things considered, we could find no basis for thinking that Congress wanted the courts to get involved in the comparable worth question.

The present case is different. Penney does not argue that the market constrains it to provide a benefit package in which entitlement to spouse coverage ceases when the spouse is paid more than the Penney employee. Nor would a judicial order abrogating this “head of household” rule be difficult to administer, or require the district court to determine the “proper” wage differential between different job classifications. If it is true as Mrs. Colby argues (and we do not understand Penney to be denying the point) that the result of the rule is to give most men but very few women the option of spouse coverage, then it does seem to bear a family resemblance to a rule that is written in terms of bald persons or persons above a certain height or certain weight; the attribute that triggers application of the rule (here, relative spousal wages) correlates with sex. All the plaintiff need show to make out a prima facie case of disparate impact is that the ostensibly neutral criterion (like height or weight or baldness) excludes a disproportionate fraction of a favored group, such as blacks or women. See, e.g., Dothard v. Rawlinson, supra, 433 U.S. at 329, 97 S.Ct. at 2726.

Penney has, however, an argument that a disparate-impact approach cannot be used in any case about fringe benefits. This argument is based on the different wording of the two subsections of Title VII that define forbidden practices by an employer:

It shall be an unlawful employment practice for an employer—

(1) to fail or refuse to hire or to discharge any individual, or otherwise to discriminate against any individual with respect to his compensation, terms, conditions, or privileges of employment, because of such individual’s ... sex ...; or

(2) to limit, segregate, or classify his employees or applicants for employment in any way which would deprive or tend to deprive any individual of employment opportunities or otherwise adversely affect his status as an employee, because of such individual’s ... sex____

42 U.S.C. § 2000e-2(a). The Supreme Court has not yet decided whether subsection (a)(1), like (a)(2), allows a suit based on *1127 disparate impact. See, e.g., Nashville Gas Co. v. Satty, 434 U.S. 136, 144, 98 S.Ct. 347, 352, 54 L.Ed.2d 356 (1977). If not, Mrs. Colby could be in trouble, because denial of a particular and rather minor fringe benefit might not be thought to deprive her of “employment opportunities or otherwise adversely affect [her] status as an employee.” The Court so intimated in Satty. See id. at 145, 98 S.Ct. at 353. So she may be stuck with (a)(1). Surprisingly, we can find only two decisions that actually discuss the question whether (a)(1) allows use of a disparate-impact approach. One is Wambheim, supra, 705 F.2d at 1494, which holds that it does allow this, but gives no reason why; the other is the Detroit case against Penney, see 632 F.Supp. at 877-80, which holds the contrary and which, of course, is now on appeal. A large number of cases assume without discussion that (a)(1) allows disparate impact to be argued. See, e.g., Liberles v. County of Cook, 709 F.2d 1122, 1130 (7th Cir.1983).

There is a powerful textual argument against this assumption. Subsection (a)(1) seems to forbid (prima facie) the use of race, sex, etc. as an employment criterion, which implies disparate treatment as that term is defined in the cases, while subsection (a)(2) seems designed to expand the reach of (a)(1) by forbidding practices having a tendency to harm members of the favored groups, i.e., disparate impact. But the contrary assumption is too well established in the cases to make us eager to reject it; and although Socrates said that the unexamined life is not worth living, not a few judicial doctrines have become well established by assumption rather than analysis. This would not be a good enough answer (especially in light of our earlier discussion of the limits of stare decisis!) if the traditional view were plainly wrong. But in support of it one can point out, first, that the difference in scope between the two subsections is far from clear, so that allowing disparate impact under the second but not the first would create great pressure to interpret the second expansively and would generate a complex jurisprudence revolving around the meaning of “employment opportunities” and “status as an employee”; and, second, that the statute would make little sense if a rule, test, or criterion received careful judicial scrutiny if it affected hiring or promotion but not if it affected compensation.

Penney makes another argument: that the use of disparate impact to challenge a difference in compensation, including fringe benefits, as sex discrimination is barred by the Bennett Amendment to Title VII, 42 U.S.C. § 2000e-2(h). By making the Equal Pay Act the exclusive measure of the employer’s liability for sex discrimination in compensation, the Bennett Amendment allows an employer to pay unequal wages for equal work if the inequality results from “any ... factor other than sex,” 29 U.S.C. § 206(d)(l)(iv). This seems to imply, and the Supreme Court in County of Washington v. Gunther, 452 U.S. 161, 170-71, 101 S.Ct. 2242, 2248, 68 L.Ed.2d 751 (1981), assumed, that a disparate impact — which is to say a discriminatory consequence of a rule or practice having a nondiscriminatory motivation — is not a ground for challenging a practice that is within the scope of the amendment. See also American Nurses’ Ass’n v. Illinois, supra, 783 F.2d at 723. Penney, however, failed to mention the Bennett Amendment in the district court, and we are not disposed to excuse its failure to do so. Since the case must be remanded, we shall leave it to the district judge to decide whether Penney should be forgiven, and allowed to revive its defense based on the amendment.

We conclude that Colby can attack Penney’s spouse rule under a disparate-impact approach, but of course not all rules that have a disparate impact are unlawful; the defendant can rebut by showing a good business justification for the rule. Although the Ninth Circuit in Wambheim held that Penney had succeeded in doing just that, see 705 F.2d at 1495, Mrs. Colby is not bound by that determination. No evidence on justification has been introduced in this, till recently quiescent, litigation. It may well be as Penney argues that usually the higher-paid spouse has the *1128 more generous provision for coverage of family members, and therefore that the only employees who need spouse coverage are those who earn more than their spouses. But to this the obvious replies, as yet unexplored in this litigation (which may cast the facts in a different light from the Ninth Circuit case), are, first, that under the Penney plans the employee’s dependents are covered regardless of the spouse’s income relative to the employee’s income, and, second, that if it is true that the employee won’t want coverage if his (or her) spouse has a higher income, Penney would not be affected by abrogating the head of household rule. No doubt some employees with higher-paid spouses would elect coverage, because the spouse’s benefit package might be less generous (as we said before, the spouse might have a higher wage in compensation for the lack of generous benefits), or the spouse might be worried about his own (or her own) job security. Why not let employees in those circumstances elect coverage? There may be answers to these questions, but it is not obvious what they are. This case cannot yet be decided on the basis of a valid business justification for the alleged discrimination.

As an original matter we might be inclined to agree that Mrs. Colby’s claim has little to do with the objectives of Title VII. It is hard to see how the head of household rule could hurt women, as opposed to an individual woman (or man, or household) that might have unusual circumstances. A woman in the group represented by Mrs. Colby — the group that consists of the women allegedly hurt by the head of household rule — is by definition married to a man with a higher income. Mrs. Colby wants this woman to be able to obtain a fringe benefit for her husband. She must benefit in some sense from increments to her husband’s income, whether those increments are in the form of insurance or otherwise. That is the essential premise of Mrs. Colby’s standing to sue. Mrs. Colby’s group is almost certainly more prosperous than the women whom the head of household rule entitles to elect spouse coverage— women married to men with lower incomes than they. It may be more prosperous than the third group, the unmarried female employees of Penney. Groups two and three cannot benefit from this suit, and may be harmed, since Penney may fund the increase in benefits to women in Mrs. Colby’s group by reducing benefits for all participants, so that its total outlay is no greater. If so, there will be a redistribution of wealth among the women employees, but no benefit for women as a group; and probably the better-off women will do the best — possibly, as we have said, at the expense of the others. Some male employees will benefit, but by assumption few, and none has bothered to intervene in this suit or bring his own suit.

This type of analysis, however, which seeks to evaluate the impact of a challenged employment practice on the favored group as a whole, seems foreclosed by Connecticut v. Teal, 457 U.S. 440, 455-56, 102 S.Ct. 2525, 2534-35, 73 L.Ed.2d 130 (1982); see also Los Angeles Dept. of Water & Power v. Manhart, 435 U.S. 702, 708, 98 S.Ct. 1370, 1375, 55 L.Ed.2d 657 (1978). Still, we do not mean to prejudge the question, which the district court has not considered yet, whether Mrs. Colby has made out a prima facie case. That is a matter for consideration on remand.

It remains only to consider whether class certification was properly denied. How can we tell? The district court gave no reasons for the denial, and they don’t leap out of the record or Penney’s brief. At first blush this seems a natural case for class treatment under either Rule 23(b)(2) or Rule 23(b)(3). The class of all Penney employees in Illinois who like Mrs. Colby are women whose husbands earn more than they do is numerous, and homogeneous in its interests, and all its members are complaining about the identical practice by the same employer. Maybe Mrs. Colby cannot adequately represent the class because her interest is somewhat tenuous given her husband’s uninterrupted coverage from his own employer, but that is not something to decide in the first instance at the appellate level.

The case should not have been dismissed, and the order of dismissal is reversed. The order denying class certification is vacated, with directions to reconsider it in light of this opinion. Circuit Rule 18 shall apply on remand.

ON PETITION FOR REHEARING

In our opinion we said that J.C. Penney Co. had waived in the district court its arguments based on the Bennett Amendment, but that the court of remand could in its discretion decide to overlook the waiver. Penney’s petition for rehearing makes the sole argument that it in fact *1129 raised the Amendment in the district court; we called for an answer to the petition, and the answer states that the references in the district court record on which Penney relies are too paltry to preserve the issue.

We direct the district court on remand to determine whether or not Penney raised the Bennett Amendment in proper and timely fashion, and to conduct the proceedings on remand in accordance with that determination.

The petition for rehearing is DENIED.