concurring in the judgment.
Although my conclusion is the same, my appraisal of the relevant discrimination and my reasons for concluding that it is unjustified, are somewhat different from those expressed by Mr. Justice Brennan.
First, I agree with Mr. Justice Rehnquist that the constitutional question raised by this plaintiff requires us to focus on his claim for benefits rather than his deceased wife’s tax obligation. She had no contractual right to receive benefits or to control their payment; moreover, the payments are not a form of compensation for her services.1 At the same salary *218level, all workers must pay the same tax, whether they are male or female, married or single, old or young, the head of a large family or a small one. The benefits which may ultimately become payable to them or to a wide variety of beneficiaries—including their families, their spouses, future spouses, and even their ex-wives—vary enormously, but such variations do not convert a uniform tax obligation into an unequal one. The discrimination against this plaintiff would be the same if the benefits were funded from general revenues. In short, I am persuaded that the relevant discrimination in this case is against surviving male spouses, rather than against deceased female wage earners.2
Second, I also agree with Mr. Justice Rehnquist that a classification which treats certain aged widows3 more favorably than their male counterparts is not “invidious.” Such a classification does not imply that males are inferior to females, cf. Mathews v. Lucas, 427 U. S. 495, 516 (Stevens, J., dissenting) ; does not condemn a large class on the basis of the misconduct of an unrepresentative few, cf. Craig v. Boren, 429 U. S. 190, 211 (Stevens, J., concurring); and does not add to the burdens of an already disadvantaged discrete minority. *219Cf. Hampton v. Mow Sun Wong, 426 U. S. 88, 102. It does, however, treat similarly situated persons differently solely because they are not of the same sex.
Third, Mr. Justice Rehnquist correctly identifies two hypothetical justifications for this discrimination that are comparable to those the Court found acceptable in Mathews v. Lucas, supra, and Kahn v. Shevin, 416 U. S. 351. Neither the “administrative convenience” rationale of Lucas, nor the “policy of cushioning the financial impact of spousal loss upon the sex for which that loss imposes a disproportionately heavy burden,” Kahn v. Shevin, supra, at 355, can be described as wholly irrational. Nevertheless, I find both justifications unacceptable in this case.
The administrative-convenience rationale rests on the assumption that the cost of providing benefits to nondependent widows is justified by eliminating the burden of requiring those who are dependent to establish that fact. Mr. Justice Rehnquist’s careful analysis of the relevant data, see post, at 238-239, n. 7, demonstrates that at present only about 10% of the married women in the relevant age bracket are nondependent. Omitting any requirement that widows establish dependency therefore expedites the processing of about 90% of the applications. This convenience must be regarded as significant even though procedures could certainly be developed to minimize the burden.4
But what is the offsetting cost that Congress imposed on the Nation in order to achieve this administrative convenience? Assuming that Congress intended only to benefit dependent spouses, and that it has authorized payments to *220nondependent widows to save the cost of administering a dependency requirement for widows, it has paid a truly staggering price for a relatively modest administrative gain: The cost of payments to the hundreds of thousands of widows who are not within the described purpose of the statute is perhaps $750 million a year.5 The figures for earlier years were presumably smaller, but must still have been large in relation to the possible administrative savings. It is inconceivable that Congress would have authorized such large expenditures for an administrative purpose without the benefit of any cost analysis, or indeed, without even discussing the problem. I am therefore convinced that administrative convenience was not the actual reason for the discrimination.6
*221It is also clear that the disparate treatment of widows and widowers is not the product of a conscious purpose to redress the “legacy of economic discrimination” against females. Kahn v. Shevin, supra, at 359 (Brennan, J., dissenting). The widows who benefit from the disparate treatment are those who were sufficiently successful in the job market to become nondependent on their husbands. Such a widow is the least likely to need special benefits. The widow most in need is the one who is “suddenly forced into a job market with which she is unfamiliar, and in which, because of her former economic dependency, she will have fewer skills to offer.” 416 U. S., at 354. To accept the Kahn justification we must presume that Congress deliberately gave a special benefit to those females least likely to have been victims of the historic discrimination discussed in Kahn. Respect for the legislative process precludes the assumption that the statutory discrimination is the product of such irrational lawmaking.
The step-by-step evolution of this statutory scheme included a legislative decision to provide benefits for all widows and a separate decision to provide benefits for dependent widowers. Admittedly, each of these separate judgments has *222a rational and benign purpose. But I consider it clear that Congress never focused its attention on the question whether to divide nondependent surviving spouses into two classes on the basis of sex.7 The history of the statute is entirely consistent with the view that Congress simply assumed that all widows should be regarded as “dependents” in some general sense, even though they could not satisfy the statutory support test later imposed on men.8 It is fair to infer that habit, rather than analysis or actual reflection, made it seem acceptable to equate the terms “widow” and “dependent surviving spouse.” That kind of automatic reflex is far different from either a legislative decision to favor females in order to compensate for past wrongs, or a legislative decision that the administrative savings exceed the cost of extending benefits to nondependent widows.
*223I am therefore persuaded that this discrimination against a group of males is merely the accidental byproduct of a traditional way of thinking about females. I am also persuaded that a rule which effects an unequal distribution of economic benefits solely on the basis of sex is sufficiently questionable that “due process requires that there be a legitimate basis for presuming that the rule was actually intended to serve [the] interest” put forward by the Government as its justification. See Hampton v. Mow Sun Wong, 426 U. S., at 103.9 In my judgment, something more than accident is necessary to justify the disparate treatment of persons who have as strong a claim to equal treatment as do similarly situated surviving spouses.
But if my judgment is correct, what is to be said about Kahn v. Shevin? For that case involved a discrimination between surviving spouses which originated in 1885; a discrimination of that vintage cannot reasonably be supposed to have been motivated by a decision to repudiate the 19th century presumption that females are inferior to males.10 It *224seems clear, therefore, that the Court upheld the Florida statute on the basis of a hypothetical justification for the discrimination which had nothing to do with the legislature’s actual motivation. On this premise, I would be required to regard Kahn as controlling in this case, were it not for the fact that I believe precisely the same analysis applies to Weinberger v. Wiesenfeld, 420 U. S. 636.
In Wiesenfeld, the Court rejected an attempt to use “mere recitation of a benign, compensatory purpose” as “an automatic shield,” id., at 648, for a statute which was actually based on “‘archaic and overbroad’ generalization[s],” id., at 643. In Wiesenfeld, as in this case, the victims of the statutory discrimination were widowers. They were totally excluded from eligibility for benefits available to similarly situated widows, just as in this case nondependent widowers are totally excluded from eligibility for benefits payable to nondependent widows. The exclusion in Wiesenfeld was apparently the accidental byproduct of the same kind of legislative process that gave rise to Kahn and to this case. If there is inconsistency between Kahn and Wiesenfeld, as I believe there is, it is appropriate to follow the later unanimous holding rather than the earlier, sharply divided decision. And if the cases are distinguishable, Wiesenfeld is closer on its facts to this case than is Kahn.
For these reasons, and on the authority of the holding in Wiesenfeld, I concur in the Court’s judgment.
For this reason this case is not controlled by Frontiero v. Richardson, 411 U. S. 677.
The contrary analysis in Weinberger v. Wiesenfeld, 420 U. S. 636, 646-647, was not necessary to the decision of that case. See id., at 655 (Rehnquist, J., concurring in result).
In most cases the statutory scheme for the distribution of benefits to the surviving spouses of deceased persons who paid FICA taxes on their earnings does not involve any discrimination on account of sex. Dependent spouses of both sexes are eligible; also, nondependent surviving spouses of both sexes are ineligible if their own social security retirement benefits are as large as those of their deceased spouses. There is, however, a narrow area in which the eligibility of nondependent spouses depends solely on their sex: Those who received between 50% and 75% of their support from their deceased spouses are eligible for benefits if they are female, but not if they are male. Similarly, if their earnings were not covered by the Social Security Act, as was true of the plaintiff in this case, and their earnings were less than 75% of the decedent’s, they are eligible if they are female, but not if they are male. See ante, at 201-202, n. 1.
Dependency in the statutory sense is a clearly defined criterion for eligibility which would have to be applied only once for each applicant. It is a requirement which several other classes of potential beneficiaries are required to meet. Moreover, the requirement would be especially easy to apply since 77% of the women over 55 do not work. (See post, at 238 n. 7.)
As of 1974, 3,546,000 women received widows’ benefits. (This figure does not include “dually entitled” women who also received benefits on their own social security accounts.) Task Force on Women and Social Security, Women and Social Security: Adapting to a New Era, prepared for the Senate Special Committee on Aging, 94th Cong., 1st Sess., 84 (Comm. Print 1975). Using Mr. Justice Rehnquist’s estimate, 10% of these women, or 354,600, are actually nondependent. The Secretary informs us that the average yearly widower’s benefit is $2,213. Brief for Appellant 5A. Assuming that this figure also applies to widows, a total of $784,729,800 is now being paid to widows who are not actually dependent. Under similar Social Security provisions, 42 U. S. C. §§ 402 (b), (c)(1)(C) (1970 ed. and Supp. V), men but not women whose spouses have retired must prove dependency to qualify for benefits. Calculations based on the same sources and assumptions indicate that each of 270,100 nondependent wives receives $1,168, a total of $315,476,800. Thus, the cost of this administrative convenience amounts to approximately $1 billion each year.
The Secretary appears to concede that this was not the justification. Brief for Appellant 22. Moreover, a 1957 amendment to the statute is inconsistent with this justification. Widow’s benefits were originally not payable to a widow who had lived apart from her husband unless she had been “receiving regular contributions from him toward her support” or unless a court had ordered him to pay support. § 209 (n), 53 Stat. 1378. This provision was retained for widows in 1950 when benefits were extended to dependent widowers. § 216 (h)(2), 64 Stat. 511. The requirement that a widow who had lived separately from her husband receive at least some *221support from him, makes sense if Congress was concerned with the statutory 50% test for dependency; such widows are obviously far less likely to meet that test than widows who had lived with their husbands. But Congress deleted the provision in 1957 and extended benefits to all widows, including those who lived apart from their husbands, with no requirement of support, § 216 (h), 71 Stat. 518. The 1957 amendment is affirmative evidence that Congress intended to provide benefits for all widows regardless of whether they could satisfy the statutory dependency test. It is also noteworthy that elsewhere in the statute Congress indicated its intention to create a presumption of dependency by stating that certain family members are “deemed dependent” under certain circumstances. See §202 (d)(3), 42 U. S. C. §402 (d)(3).
For the reasons stated in Part IV-B of Mr. Justice Brennan’s opinion, the Secretary’s alternative explanation of the statute as being a welfare measure intended to alleviate the poverty of elderly widows is plainly unacceptable.
One indication that the 1939 Act was not the result of a focused decision concerning the needs of nondependent widows vis-à-vis widowers is the breadth of the statutory classification. Under the 1939 Act:
“[C]hildren of covered female workers were eligible for survivors’ benefits only in limited circumstances . . . and no benefits whatever were made available to husbands or widowers on the basis of their wives’ covered employment.” Weinberger v. Wiesenfeld, 420 U. S., at 643-644.
The disqualification of a woman’s surviving children if they had received any support from their father, § 202 (c)(4), 53 Stat. 1365, is particularly difficult to reconcile with the theory that the legislative motive was a conscious desire to remedy sex discrimination.
Similarly, in extending benefits to dependent widowers, Congress made no mention of any determination that nondependent widowers were less needy than nondependent widows, or that nondependent widows deserved greater benefits as a remedy for sex discrimination. See ante, at 216.
The discriminatory feature of the statute can be said to be the fact that women are given the benefit of a broad, vague definition of “dependent” while men are held to a harsh arithmetic standard. This serves to answer the argument that appellee will receive a windfall by a judgment in his favor. Although appellee is not a dependent in the definition applied to widowers, it cannot be said with assurance that he is not a dependent in whatever broad sense Congress had in mind when it classified all widows as dependents.
In the absence of evidence to the contrary, we might presume that Congress had such an interest in mind, see Hampton v. Mow Sun Wong, 426 U. S., at 103, but here that presumption is untenable. Perhaps an actual, considered legislative choice would be sufficient to allow this statute to be upheld, but that is a question I would reserve until such a choice has been made.
This presumption was expressly recognized in the literature of the 19th century. It was this presumption that Mr. Bumble ridiculed when he disclaimed responsibility for his wife’s misconduct. Because a part of his disclaimer is so well known, it may not be inappropriate to quote the entire passage:
“ 'It was all Mrs. Bumble. She would do it,’ urged Mr. Bumble; first looking round to ascertain that his partner had left the room.
“ ‘That is no excuse,’ replied Mr. Brownlow. ‘You were present on the occasion of the destruction of these trinkets, and, indeed, are the more guilty of the two, in the eye of the law; for the law supposes that your wife acts under your direction.
“ ‘If the law supposes that,’ said Mr. Bumble, squeezing his hat emphatically in both hands, ‘the law is a ass—a idiot. If that’s the eye of the *224law, the law’s a bachelor; and the worst I wish the law is, that his eye may be opened by experience—by experience.’ ” C. Dickens, The Adventures of Oliver Twist, c. LI (emphasis added).