with whom The Chief Justice, Mr. Justice Stewart, and Mr. Justice Blackmun join, dissenting.
In light of this Court’s recent decisions beginning with Reed v. Reed, 404 U. S. 71 (1971), one cannot say that *225there is no support in our cases for the result reached by the Court. One can, however, believe as I do that careful consideration of these cases affords more support for the opposite result than it does for that reached by the Court. Indeed, it seems to me that there are two largely separate principles which may be deduced from these cases which indicate that the Court has reached the wrong result.
The first of these principles is that cases requiring heightened levels of scrutiny for particular classifications under the Equal Protection Clause, which have originated in areas of the law outside of the field of social insurance legislation, will not be uncritically carried over into that field. This does not mean that the phrase “social insurance” is some sort of magic phrase which automatically mutes the requirements of the equal protection component of the Fifth Amendment. But it does suggest that in a legislative system which distributes benefit payments among literally millions of people there are at least two characteristics which are not found in many other types of statutes. The first is that the statutory scheme will typically have been expanded by amendment over a period of years so that it is virtually impossible to say that a particular amendment fits with mathematical nicety into a carefully conceived overall plan for payment of benefits. The second is that what in many other areas of the law will be relatively low-level considerations of “administrative convenience” will in this area of the law bear a much more vital relation to the overall legislative plan because of congressional concern for certainty in determination of entitlement and promptness in payment of benefits.
The second principle upon which I believe this legislative classification should be sustained is that set forth in our opinion in Kahn v. Shevin, 416 U. S. 351 (1974). The effect of the statutory scheme is to make it easier for widows to obtain benefits than it is for widowers, since the former *226qualify automatically while the latter must show proof of need. Such a requirement in no way perpetuates or exacerbates the economic disadvantage which has led the Court to conclude that gender-based discrimination must meet a different test from other types of classifications. It is, like the property tax exemption to widows in Kahn, a differing treatment which “ 'rest[s] upon some ground of difference having a fair and substantial relation to the object of the legislation.’ ” Id., at 355.
I
Both Weinberger v. Wiesenfeld, 420 U. S. 636 (1975), and Frontiero v. Richardson, 411 U. S. 677 (1973), are undoubtedly relevant to the decision of this case, but the plurality overstates that relevance when it says that these two cases “plainly require affirmance of the judgment of the District Court.” Ante, at 204. The disparate treatment of widows and widowers by this Act is undoubtedly a gender-based classification, but this is the beginning and not the end of the inquiry. In the case of classifications based on legitimacy, and in the case of irrebuttable presumptions, constitutional doctrine which would have invalidated the same distinctions in other contexts has been held not to require that result when they were used within comprehensive schemes for social insurance. The same result should obtain in the case of constitutional principles dealing with gender-based distinctions.
In Levy v. Louisiana, 391 U. S. 68 (1968), the Court held that a Louisiana statute which allowed legitimate but not illegitimate children to recover for the wrongful death of their mother violated the Equal Protection Clause of the Fourteenth Amendment. Another Louisiana statute was challenged on similar grounds in Weber v. Aetna Cas. & Surety Co., 406 U. S. 164 (1972). The statute in Weber was defended on the ground that it did not preclude entirely *227the recovery of workmen’s compensation by illegitimate children, since acknowledged illegitimates were permitted to recover on the same basis as legitimate children. The Court rejected that distinction, however, and held that this statute also violated the Equal Protection Clause.
Two Terms later we held invalid under the Fifth Amendment a portion of the child’s benefits provisions of the Social Security Act. The challenged provision flatly excluded one class of illegitimate children notwithstanding their actual dependence upon a disabled parent, while granting benefits to other classes of illegitimates and to legitimates on the basis of demonstrated or presumed dependence upon such a parent. Jimenez v. Weinberger, 417 U. S. 628 (1974). We relied on our earlier decision in Weber, supra, to reach this result.
Last Term, however, in Mathews v. Lucas, 427 U. S. 495 (1976), we upheld the portion of these same child’s benefits provisions which conclusively presume dependency for all but a specified group of illegitimate children. This use of illegitimacy to define a group required to present proof of dependency was held not to deny equal protection to those singled out.
In Stanley v. Illinois, 405 U. S. 645 (1972), we held that Illinois might not under the equal protection guarantee of the Fourteenth Amendment deny a hearing on parental fitness to an unwed father when such a hearing was granted to all other parents whose custody of their children was challenged. In Cleveland Board of Education v. LaFleur, 414 U. S. 632 (1974), we likewise held invalid school board regulations requiring pregnant school teachers to take unpaid maternity leave commencing four or five months before their expected birth.
Yet, the Term following LaFleur, we decided Weinberger v. Salfi, 422 U. S. 749 (1975), in which a three-judge District Court had held invalid a duration-of-relationship requirement *228for surviving wives in order that they might receive benefits under the Social Security Act. The District Court relied on Stanley and LaFleur, but we declined to extend those cases into the area of a complex social insurance scheme such as this Act, saying:
“We think that the District Court's extension of the holdings of Stanley, Vlandis, and LaFleur to the eligibility requirement in issue here would turn the doctrine of those cases into a virtual engine of destruction for countless legislative judgments which have heretofore been thought wholly consistent with the Fifth and Fourteenth Amendments to the Constitution.” 422 U. S., at 772.
The Court’s recent treatment of gender-based discrimination begins with Reed v. Reed, 404 U. S. 71 (1971), in which the Court invalidated a provision of the Idaho probate code which contained an across-the-board flat preference for men over women as putative administrators of the estate of a decedent. The following Term we relied on the equal protection component of the Fifth Amendment to hold invalid an Act of Congress relating to military pay which allowed a male member of the uniformed services to claim his wife as a dependent without any showing of such a fact, but which required a female member to show that her husband was in fact dependent on her before she could make such a claim. The consequences of spousal dependency were increased fringe benefits which had been provided in an effort to make the uniformed services competitive with business and industry. Frontiero v. Richardson, supra, at 679.
The next Term, however, we refused to invalidate at the behest of a male property taxpayer a provision of Florida law which allowed widows, but not widowers, an exemption from property taxation in the amount of $500. Kahn v. Shevin, 416 U. S., 351 (1974). Weinberger v. Wiesenfeld, decided one year later, relied on Frontiero, supra, in holding invalid a section of the Social Security Act which allowed *229benefits to a surviving widow but flatly denied them to a surviving widower. The Court said:
“Since the gender-based classification of § 402 (g) cannot be explained as an attempt to provide for the special problems of women, it is indistinguishable from the classification held invalid in Frontiero. Like the statutes there, ‘[b]y providing dissimilar treatment for men and women who are . . . similarly situated, the challenged section violates the [Due Process] Clause.’ Reed v. Reed, 404 U. S. 71, 77 (1971).” 420 U. S., at 653.
Two observations about Wiesenfeld are pertinent. First, the provision of the Social Security Act held unconstitutional there flatly denied surviving widowers the possibility of obtaining benefits no matter what showing of need might be made. The section under attack in the instant case does not totally foreclose widowers, but simply requires from them a proof of dependency which is not required from similarly situated widows. Second, Wiesenfeld was decided before either Weinberger v. Salfi, supra, or Mathews v. Lucas, supra. Each of those decisions refused uncritically to extend into the field of social security law constitutional proscriptions against distinctions based on illegitimacy and irrebuttable presumptions which had originated in other areas of the law. While the holding of Wiesenfeld is not inconsistent with Salfi or Lucas, its reasoning is not in complete harmony with the recognition in those cases of the special characteristics of social insurance plans.
II
Those special characteristics arise from the nature of the legislative problem which numerous sessions of Congress have had to face in defining the coverage of the Social Security Act. The program has been participatory from the outset, in the sense that benefits have not been extended to persons *230without at least a close relationship to a person paying into the system during his working life. But Congress did not legislate with the idea that it was fulfilling any narrow contractual obligation owed to the program participant. On the contrary, Congress has continually increased the amounts of benefits paid, and expanded the pool of eligible recipients by singling out additional, identifiable groups having both the requisite relationship to the contributing worker and a degree of probable need which, in the legislative judgment, justifies assistance. It is not difficult to predict some traits of the system emerging from this sort of step-by-step legislative expansion.
One is that the resulting statute, like the process which produced it, extends benefits in a piecemeal fashion. There will be some individuals with needs demonstrably as great as those within a class of qualifying beneficiaries who will nonetheless be treated less favorably than that class. This is because these classes, formulated and reformulated over a period of decades, could not perfectly mirror the abstract definition of equality of need unless Congress were to burden the system with numerous individualized determinations which might frustrate the primary purposes of the Act.
Another characteristic of the Social Security statute which is predictable from the manner of its enactment, is the balance between a desire that payments correlate with degree of need and a recognition that precise correlation is unattainable given the administrative realities of the situation. No one would contend, for example, that all wives of program participants, who are over 62 and entitled to old-age or disability-insurance benefits in their own right equal to no more than one-half of their husband's primary amount, are needy. Nonetheless the administrative problems of determining actual need have led Congress to employ these and factors like them as the determinants of eligibility. 42 U. S. C. § 402 (b)(1) (1970 ed. and Supp. V). The overin*231clusiveness of such categorizations is, in many cases, not only tolerable but Solomonic. For had Congress attempted to distribute program funds in precise accordance with a purpose to alleviate need, it could very well have created a procedural leviathan consuming substantial amounts of those funds in case-by-case determinations of eligibility.
The provisions at issue in this case, relating to widows’ and widowers’ benefits, display all the earmarks of their origins in the oft-repeated process of legislative reconsideration and expansion of beneficiary groups. As originally enacted in 1935, the Social Security Act provided for old-age benefits only to the wage earner. 49 Stat. 623. In 1939, additional provisions were made for benefits to the wage earner’s family, including wives and widows, but not including husbands and widowers. The widow’s benefit was in an amount larger by one-half than that for the wife, and was available notwithstanding the widow’s primary entitlement to benefits in an amount greater than permissible in the case of a wife.1 All things considered, the 1939 amendments *232reflect a legislative judgment that elderly wives and widows of Social Security recipients were needy groups, and that of the two, the plight of widows was especially severe.2 I agree with the plurality’s statement that “[t]here is no indication whatever in any of the legislative history that Congress *233gave any attention to the specific case of nondependent widows, and found that they were in need of benefits despite their lack of dependency . . ." Ante, at 215. But neither is there any reason to doubt that it singled out the group of aged widows for especially favorable treatment, see n. 1, supra, because it saw prevalent throughout that group a characteristically high level of need.
In 1950, Congress created two new categories of old-age and survivors’ insurance benefits—for husbands and widowers. With one exception, these provisions were identical to the sections dealing with wives’ and widows’ benefits. A husband or widower was required additionally to prove that he had been dependent upon his wife for half of his support at the time she became eligible for benefits, or, in the case of the widower, at the time of her death. 64 Stat. 483, 485. This enactment obviously reflected a congressional judgment that there were needy persons in those groups who should properly be able to receive benefits, but that their numbers were not so great as to justify automatic qualification on the basis of age and marriage to a wage-earning wife. Proof of dependence upon the wife for one-half of a husband’s support was adopted as a suitable means of eliminating large numbers of men with independent incomes, while preserving an entitlement to benefits in the cases of those shown to lack substantial means of support apart from funds actually brought in by the wife.
Subsequent amendments have altered the statute somewhat—predictably in the direction of expanded coverage3*234—but as relevant to this case the basic scheme has remained unchanged. The present statutory treatment of widows and widowers would seem to reflect a pair of legislative judgments about the needs of those two groups. The first is that the persons qualifying for spousal benefits are likely to have even more substantial needs after the passing of their spouse. This is indicated both by the increase in benefits to qualifying widows and widowers which now stand at 100% of the primary amount compared with the 50% paid to spouses,4 and by the increase in the amount of primary benefits that a person may separately receive without losing entitlement to benefits under the spouse’s account. While the spouse of a living wage earner loses such entitlement upon receipt of his or her own primary benefits equal to 50% of the wage earner’s primary amount, a surviving spouse does not lose such entitlement until receiving separate benefits equal to 100%.5
The second legislative judgment implicit in the widow’s and widower’s provisions is that widows, as a practical matter, are much more likely to be without adequate means of support than are widowers. The plurality opinion makes much of establishing this point, ante, at 212-217, that the absence of any dependency prerequisite to the award of widow’s benefits reflects a judgment, resting on “administrative convenience,” that dependence among aged widows is frequent enough to justify waiving the requirement entirely. I differ not with the recognition of this administrative convenience *235purpose but with the conclusion that such a purpose necessarily invalidates the resulting classification. Our decisions dealing with social welfare legislation indicate that our inquiry must go further. For rational classifications aimed at distributing funds to beneficiaries under social insurance legislation weigh a good deal more heavily on the governmental interest side of the equal protection balance than they may in other legislative contexts. The “administrative convenience” which is afforded by such classifications in choosing the administrator of a decedent’s estate, see Reed v. Reed, 404 U. S. 71 (1971), is significantly less important to the effectiveness of the legislative scheme than is the “convenience” afforded by classifications in administering an Act designed to provide benefits to millions upon millions of beneficiaries with promptness and certainty. For this reason, the plurality errs in merely dispatching this statute with an incantation of “administrative convenience.” It should go further and consider the governmental interest advanced by the statutory classification in a social insurance statute such as this, in light of the claimed injury to appellee.
III
Whatever his actual needs, Goldfarb would, of course, have no complaint if Congress had chosen to require proof of dependency by widows as well as widowers, or if it had simply refrained from making any provision whatever for benefits to surviving spouses. “A legislature may address a problem 'one step at a time,’ or even 'select one phase of one field and apply a remedy there, neglecting the others.’ Williamson v. Lee Optical Co., 348 U. S. 483, 489 (1955).” Jefferson v. Hackney, 406 U. S. 535, 546 (1972); Dandridge v. Williams, 397 U. S. 471, 487 (1970). See Geduldig v. Aiello, 417 U. S. 484, 495 (1974); Lindsley v. Natural Carbonic Gas Co., 220 U. S. 61 (1911). Any claim which he has must therefore turn upon the alleged impropriety of giving benefits to widows *236without requiring them to make the same proof of dependency required of widowers. Yet, in the context of the legislative purpose, this amounts not to exclusion but to overinclusiveness for reasons of administrative convenience which, if reasonably supported by the underlying facts, is not offensive to the Equal Protection Clause in social welfare cases.
A close analogue to this case is presented by our decision last Term in Mathews v. Lucas, 427 U. S. 495 (1976). The plaintiffs there challenged the OASDI provisions for children's benefits, which require no proof of dependency by legitimate children or certain categories of illegitimates,6 but which demand that other illegitimates show dependency by proof that their father lived with them or contributed to their support prior to his death. After first stating that this classification based on legitimacy does not demand “our most exacting scrutiny," id., at 506, the Court concluded that a general requirement of dependency “at the time of death is not impermissibly discriminatory in providing only for those children for whom the loss of the parent is an immediate source of the need.” Id., at 507. It then upheld the waiver of the dependency proof requirement for legitimates and certain others, by the following reasoning:
“The basis for appellees' argument is the obvious fact that each of the presumptions of dependency renders the class of benefit-recipients incrementally overinclusive, in the sense that some children within each class of *237presumptive dependents are automatically entitled to benefits under the statute although they could not in fact prove their economic dependence upon insured wage earners at the time of death. We conclude that the statutory classifications are permissible, however, because they are reasonably related to the likelihood of dependency at death.
“Congress’ purpose in adopting the statutory presumptions of dependency was obviously to serve administrative convenience. While Congress was unwilling to assume that every child of a deceased insured was dependent at the time of death, by presuming dependency on the basis of relatively readily documented facts, such as legitimate birth, or existence of a support order or paternity decree, which could be relied upon to indicate the likelihood of continued actual dependency, Congress was able to avoid the burden and expense of specific case-by-case determination in the large number of cases where dependency is objectively probable. Such presumptions in aid of administrative functions, though they may approximate, rather than precisely mirror, the results that case-by-case adjudication would show, are permissible under the Fifth Amendment, so long as that lack of precise equivalence does not exceed the bounds of substantiality tolerated by the applicable level of scrutiny. See Weinberger v. Salfi, 422 U. S., 749, 772 (1975).
“Applying these principles, we think that the statutory classifications challenged here are justified as reasonable empirical judgments that are consistent with a design to qualify entitlement to benefits upon a child’s dependency at the time of the parent’s death.” 427 U. S., at 508-509, 510.
The same reasoning should control in the case before us. *238As in Lucas, Congress has here adopted a test of dependency as a reasonable surrogate for proof of actual need. In Lucas, legitimates and certain others were not required to satisfy that test because, in the legislative view, there was a sufficiently high rate of dependency among those groups to make the requirement of actual proof administratively counterproductive. Here the dependency test was not imposed upon widows, apparently on a similar belief that the actual rate of dependency was sufficiently high that a requirement of proof would create more administrative expense than it would save in the award of benefits.7
*239IV
Perhaps because the reasons asserted for “heightened scrutiny” of gender-based distinctions are rooted in the fact that women have in the past been victims of unfair treatment, see Frontiero v. Richardson, 411 U. S., at 684-688, the plurality says that the difference in treatment here is not only between a widow and a widower, but between the respective deceased spouses of the two. It concludes that wage-earning wives are deprived “ ‘of protection for their families which men receive as a result of their employment.’ ” Ante, at 206.
But this is a questionable tool of analysis which can be used to prove virtually anything. It might just as well have been urged in Kahn v. Shevin, 416 U. S. 351 (1974), where we upheld a Florida property tax exemption redounding to the benefit of widows but not widowers, that the real discrimination was between the deceased spouses of the respective widow and widower, who had doubtless by their contributions to the family or marital community helped make possible the acquisition of the property which was now being disparately taxed.
*240Since the claim to social security benefits is noncontractual in nature, see Flemming v. Nestor, 363 U. S. 603 (1960), the contributions of the deceased spouse cannot be regarded as creating any sort of contractual entitlement on the part of either the deceased wife or the surviving husband. Here the female wage earner has gotten the degree of protection for her family which Congress was concerned to extend to all. Neither she nor her surviving husband has any constitutional claim to more, simply because Congress has chosen, for administrative reasons, to give benefits to widows without requiring proof of dependency.
Viewed from the perspective of the recipient of benefits, the sections involved here are entirely distinguishable from those which this Court has previously struck down. In Jimenez v. Weinberger, 417 U. S. 628 (1974), the Court invalidated one aspect of the provisions for surviving children’s benefits which were considered in Mathews v. Lucas, 427 U. S. 495 (1976). Those provisions allow legitimate and certain categories of illegitimate children 8 to receive benefits, whether born before or after the onset of the wage earner’s disability. Other illegitimates were entitled to benefits only upon a showing of dependency prior to the disability, and were therefore conclusively denied benefits if born after the wage earner was disabled. Finding a legislative purpose to aid children with needs demonstrated by a dependency relationship to a disabled worker, the Court found equal protection offended by the statute’s denial to some children of any opportunity to prove that they were within that class.
In Weinberger v. Wiesenfeld, 420 U. S. 636 (1975), the Court again invalidated OASDI provisions which denied one group any opportunity to show themselves proper beneficiaries given the apparent statutory purpose. A widow not qualifying for widow’s benefits was entitled to a mother’s benefit if she had in her care a minor child qualifying for a child’s *241benefit, and if she did not receive more than a certain amount of primary benefits in her own right. No such provision was made, however, for a widower in a parallel position. The Court found a purpose in the statute to allow a single parent to stay home and care for the minor child, id., at 648-649, and struck down the denial of benefits to fathers similarly situated. The defect of that statute was its conclusive exception of widowers from the benefited class, solely on the basis of their sex, and in contravention of the legislative purpose to allow parents with deceased spouses to provide personal parental care. There is no plausible claim to be made here that a statutory objective is being thwarted by underinclusiveness of the classes of beneficiaries.
This case is also distinguishable from Frontiero v. Richardson, supra, in the sense that social insurance differs from compensation for work done. While there is no basis for assessing the propriety of a given allocation of funds within a social insurance program apart from an identifiable legislative purpose, a compensatory scheme may be evaluated under the principle of equal pay for equal work done. This case is therefore unlike Frontiero, where the Court invalidated sex discrimination among military personnel in their entitlement to increased quarters allowances on account of marriage, and in the eligibility of their spouses for dental and medical care. These compensatory fringe benefits were available to male employees as a matter of course, but were unavailable to females except on proof that their husbands depended on them for over one-half of their support. Since males got such compensatory benefits even though their wives were not so dependent, females with nondependent husbands were effectively denied equal compensation for equal effort. The same is not true here, where the benefit payments to survivors are neither contractual nor compensatory for work done, and where there is thus no comparative basis for evaluating the propriety of a given benefit apart from the legislative purpose.
*242V
The very most that can be squeezed out of the facts of this case in the way of cognizable “discrimination" is a classification which favors aged widows. Quite apart from any considerations of legislative purpose and “administrative convenience” which may be advanced to support the classifition, this is scarcely an invidious discrimination. Two of our recent cases have rejected efforts by men to challenge similar classifications. We have held that it is not improper for the military to formulate “up-or-out” rules taking into account sex-based differences in employment opportunities in a way working to the benefit of women, Schlesinger v. Ballard, 419 U. S. 498 (1975), or to grant solely to widows a property tax exemption in recognition of their depressed plight. Kahn v. Shevin, 416 U. S. 351 (1974). A waiver of the dependency prerequisite for benefits, in the case of this same class of aged widows, under a program explicitly aimed at the assistance of needy groups, appears to be well within the holding of the Kahn case, which upheld a flat $500 exemption to widows, without any consideration of need.
VI
The classification challenged here is “overinclusive” only in the sense that widows over 62 may obtain benefits without a showing of need, whereas widowers must demonstrate need. Because this overinclusion is rationally justifiable, given available empirical data, on the basis of “administrative convenience,” Mathews v. Lucas, supra, is authority for upholding it. The differentiation in no way perpetuates the economic discrimination which has been the basis for heightened scrutiny of gender-based classifications, and is, in fact, explainable as a measure to ameliorate the characteristically depressed condition of aged widows. Kahn v. Shevin, supra, is therefore also authority for upholding it. For both of these reasons, I would reverse the judgment of the District Court.
It is noteworthy that Congress did not simply state generally that immediate family members were entitled to benefits in a certain amount, but set forth several categories of benefits for family members, with unique conditions and benefit amounts attaching to each.
“Wife’s Insurance Benefits” in the amount of one-half the husband’s primary benefit, were to be given to a program participant’s wife if she was over 65, lived with her husband (or received support from him, see 53 Stat. 1378) at the time of filing her application, and was not entitled to primary benefits of her own in an amount equal to or greater than one-half of her husband’s primary amount.
“Widow’s Insurance Benefits” equal to three-fourths the deceased husband’s primary benefit, were made available to an unmarried widow over 65, who lived with the wage earner (or received support from him) at the time of his death, and was not entitled to primary benefits on her own equal to or greater than three-fourths of the husband’s primary amount.
In addition, “Widow’s Current Insurance Benefits” were made available to one failing to qualify for the widow’s benefit solely on account of age, who had in her care a child qualifying for “Child’s Insurance Benefits” *232under still another section of the amended statute. The amendments also provided for “Parent’s Insurance Benefit” and “Lump-Sum Death Payments.” 53 Stat. 1364-1367.
The manner in which these provisions were drafted makes clear that each involved a separate congressional judgment about the most appropriate definition and actual needs of each group.
The Final Report of the Advisory Council on Social Security explained the provision as follows:
“The day of large families and of the farm economy, when aged parents were thereby assured comfort in their declining years, has passed for a large proportion of our population. This change has had particularly devastating effect on the sense of security of the aged women of our country.
“Women as a rule live longer than men. Wives are often younger than their husbands. Consequently, the probabilities are that a woman will outlive her husband. Old-age insurance benefits for the husband, supplemented during his life by an allowance payable on behalf of his wife, fall considerably short, therefore, of providing adequate old-age security.” Hearings on Social Security Act Amendments of 1939 before the House Committee on Ways and Means, 76th Cong., 1st Sess., 31-32 (1939).
Likewise, the House Committee Report described widows over 65, widows with children, orphans, and dependent parents over 65 (to whom the 1939 amendments extended benefits) as the “groups of survivors whose probable need is greatest.” H. R. Rep. No. 728, 76th Cong., 1st Sess., 11 (1939). Thus, there is good reason to suppose that the 1939 enactment of a provision for widow’s benefits was in response to congressional perception of substantial poverty among the large group of aged widows.
The problem persists today in proportions far greater than among the parallel group of aged widowers. In 1974, two out of three poor persons over 65 were women. Four out of five men over 65 were married, but 52% of aged women were widows. Of older women living alone, 33.4% were below the poverty line. 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., 37, 42, 68-69 (Comm. Print 1975).
Among these changes are the lowering of the age of eligibility, the elimination, concerning spouse’s and surviving spouse’s benefits, of any requirement of cohabitation, and the increase in widow’s and widower’s benefits and in permissible primary benefits received in the beneficiary’s own right from 75% to 100% of the wage earner’s primary benefit. Also, additional provision has been made, under each spousal category of benefits, for a divorced spouse who was married to the wage earner for at least 20 years. This Court has recently upheld unanimously the wife’s *234benefits section’s imposition of the minimum-age requirement upon divorced wives with qualifying minor children, while waiving it in the case of undivorced wives caring for such children. Mathews v. De Castro, 429 U.S. 181 (1976).
42 U. S. C. §§ 402 (b)(2), (c)(3), (e)(2)(A), (f)(3)(A) (1970 ed. and Supp. V).
42 U. S. C. §§ 402 (b)(1)(D), (c)(1)(D), (e)(1)(D), (f)(1)(E) (1970 ed. and Supp. V).
Notwithstanding their illegitimacy, children need not demonstrate dependency if entitled to inherit from the insured parent under the state intestacy laws; if the decedent went through a marriage ceremony with the other parent which would have been valid but for a nonobvious legal defect; if the decedent had acknowledged the child in writing; or if he had been decreed to be the child’s father or ordered by a court to support the child because the child was his. 42 U. S. C. §§ 402 (d), 416 (h) (1970 ed. and Supp. V).
There is substantial statistical evidence indicating that the differential treatment of widows and widowers is economically justifiable on the basis of administrative convenience. There is good reason to suppose that few enough aged widows are not, in fact, dependent at the time of their husband's death that the costs of administering the test would exceed the savings resulting from its application. Among married couples throughout our population, 43% of the wives as of 1974 are in the labor force. Bureau of the Census, A Statistical Portrait of Women in the United States 52 (Table 10-9) (Apr. 1976). Among those 43%, wives with husbands over 25 years of age contribute a median of 26.1% of the family income. Ibid. (Table 10-10). This is approximately equal to the 25% maximum contribution one can make and still be statutorily dependent. It thus follows that among the married population as a whole the number of dependent wives is roughly equal to the sum of those who do not work, plus one-half of those who do (since by definition, one-half contribute more and one-half contribute less than the median of 26.1%). That calculation here leads to a conclusion that about 78.5% (57%+21.5%) of all married women are dependent.
With regard to the group of women otherwise qualifying for widow’s benefits, this figure is significantly higher. Whereas the employment rate among women between 20 and 54 is about 56%, the rate for women 55 and over is only 23%. (These figures are derived from data appearing id., at 27-28 (Tables 7-1, 7-2).) Because it is dependency at the time of the working spouse’s death which is relevant under the statute, the work habits of those over 55 are most relevant for determining the actual number of widows who would be excluded by a dependency test. Even if married women over 55 work as often as unmarried women in that group *239(an unlikelihood, given the greater probability that unmarried women will have no alternative means of support), this 23% figure indicates that they work just over one-half as often as the population of all married women (43% of whom work—id., at 52 (Table 10-9)). This suggests that the number of married women over 55 who would satisfy the dependency test is something like 88.5%—the 77% who do not work, plus half of the remaining 23% who do. This nine-tenths correlation appears sufficiently high to justify extension of benefits to the other one-tenth for reasons of administrative convenience.
On the side of widower’s benefits, the incidence of dependent husbands is certainly low enough to justify any administrative expense incurred in screening out those who are not dependent. In 1970, only 2.5% of working wives contributed more than the 75% of the family income which renders the husband dependent. F. Linden, Women: A Demographic, Social and Economic Presentation 34 (1973). Since only 43% of all wives work, the incidence of dependent husbands among all married couples is approximately 1% (.025X.43=.0108).
See n. 6, supra.