*259Mr. Justice Stewart

delivered the opinion of the Court.

This case presents a challenge to the constitutionality of the Massachusetts veterans’ preference statute, Mass. Gen. Laws Ann., ch. 31, § 23, on the ground that it discriminates against women in violation of the Equal Protection Clause of the Fourteenth Amendment. Under ch. 31, § 23,1 all veterans who qualify for state civil service positions must be considered for appointment ahead of any qualifying nonveterans. The preference operates overwhelmingly to the advantage of males.

The appellee Helen B. Feeney is not a veteran. She brought this action pursuant to 42 U. S. C. § 1983, alleging that the absolute-preference formula established in ch. 31, § 23, inevitably operates to exclude women from consideration for the best Massachusetts civil service jobs and thus unconstitutionally denies them the equal protection of the laws.2 The three-judge District Court agreed, one judge dissenting. Anthony v. Massachusetts, 415 F. Supp. 485 (Mass. 1976).3

*260The District Court found that the absolute preference afforded by Massachusetts to veterans has a devastating impact upon the employment opportunities of women. Although it found that the goals of the preference were worthy and legitimate and that the legislation had not been enacted for the purpose of discriminating against women, the court reasoned that its exclusionary impact upon women was nonetheless so severe as to require the State to further its goals through a more limited form of preference. Finding that a more modest preference formula would readily accommodate the State’s interest in aiding veterans, the court declared ch. 31, § 23, unconstitutional and enjoined its operation.4

Upon an appeal taken by the Attorney General of Massachusetts,5 this Court vacated the judgment and remanded the case for further consideration in light of our intervening decision in Washington v. Davis, 426 U. S. 229. Massachusetts v. Feeney, 434 U. S. 884. The Davis case held that a neutral law does not violate the Equal Protection Clause solely because it results in a racially disproportionate impact; instead the disproportionate impact must be traced to a purpose to discriminate on the basis of race. 426 U. S., at 238-244.

Upon remand, the District Court, one judge concurring and one judge again dissenting, concluded that a veterans’ hiring preference is inherently nonneutral because it favors a class from which women have traditionally been excluded, and that *261the consequences of the Massachusetts absolute-preference formula for the employment opportunities of women were too inevitable to have been “unintended.” Accordingly, the court reaffirmed its original judgment. Feeney v. Massachusetts, 451 F. Supp. 143. The Attorney General again appealed to this Court pursuant to 28 U. S. C. § 1253, and probable jurisdiction of the appeal was noted. 439 U. S. 891.

I

A

The Federal Government and virtually all of the States grant some sort of hiring preference to veterans.6 The Massachusetts preference, which is loosely termed an “absolute lifetime” preference, is among the most generous.7 It *262applies to all positions in the State’s classified civil service, which constitute approximately 60% of the public jobs in the State. It is available to “any person, male or female, including a nurse,” who was honorably discharged from the United States Armed Forces after at least 90 days of active service, at least one day of which was during “wartime.” 8 Persons who are deemed veterans and who are otherwise qualified for a particular civil service job may exercise the preference at any time and as many times as they wish.9

*263Civil service positions in Massachusetts fall into two general categories, labor and official. For jobs in the official service, with which the proofs in this action were concerned, the preference mechanics are uncomplicated. All applicants for employment must take competitive examinations. Grades are based on a formula that gives weight both to objective test results and to training and experience. Candidates who pass are then ranked in the order of their respective scores on an “eligible list.” Chapter 31, § 23, requires, however, that disabled veterans, veterans, and surviving spouses and surviving parents of veterans be ranked — in the order of their respective scores — above all other candidates.10

Rank on the eligible list and availability for employment are the sole factors that determine which candidates are considered for appointment to an official civil service position. When a public agency has a vacancy, it requisitions a list of “certified eligibles” from the state personnel division. Under formulas prescribed by civil service rules, a small number of candidates from the top of an appropriate list, three if there is only one vacancy, are certified. The appointing agency *264is then required to choose from among these candidates.11 Although the veterans’ preference thus does not guarantee that a veteran will be appointed, it is obvious that the preference gives to veterans who achieve passing scores a well-nigh absolute advantage.

B

The appellee has lived in Dracut, Mass., most of her life. She entered the work force in 1948, and for the next 14 years worked at a variety of jobs in the private sector. She first entered the state civil service system in 1963, having competed successfully for a position as Senior Clerk Stenographer in the Massachusetts Civil Defense Agency. There she worked for four years. In 1967, she was promoted to the position of Federal Funds and Personnel Coordinator in the same agency. The agency, and with it her job, was eliminated in 1975.

During her 12-year tenure as a public employee, Ms. Feeney took and passed a number of open competitive civil service examinations. On several she did quite well, receiving in 1971 the second highest score on an examination for a job with the Board of Dental Examiners, and in 1973 the third highest on a test for an Administrative Assistant position with a mental health center. Her high scores, however, did not win her a place on the certified eligible list. Because of the veterans’ preference, she was ranked sixth behind five male veterans on the Dental Examiner list. She was not certified, and a lower scoring veteran was eventually appointed. On the 1973 examination, she was placed in a position on the list behind 12 male veterans, 11 of whom had lower scores. Following the other examinations that she took, her name was similarly ranked below those of veterans who had achieved passing grades.

*265Ms. Feeney’s interest in securing a better job in state government did not wane. Having been consistently eclipsed by veterans, however, she eventually concluded that further competition for civil service positions of interest to veterans would be futile. In 1975, shortly after her civil defense job was abolished, she commenced this litigation.

C

The veterans’ hiring preference in Massachusetts, as in other jurisdictions, has traditionally been justified as a measure designed to reward veterans for the sacrifice of military service, to ease the transition from military to civilian life, to encourage patriotic service, and to attract loyal and well-disciplined people to civil service occupations.12 See, e. g., Hutcheson v. Director of Civil Service, 361 Mass. 480, 281 N. E. 2d 53 (1972). The Massachusetts law dates back to 1884, when the State, as part of its first civil service legislation, gave a statutory preference to civil service applicants who were Civil War veterans if their qualifications were equal to those of nonveterans. 1884 Mass. Acts, ch. 320, § 14 (sixth). This tie-breaking provision blossomed into a truly absolute preference in 1895, when the State enacted its first general veterans’ preference law and exempted veterans from all merit selection requirements. 1895 Mass. Acts, ch. 501, § 2. In response to a challenge brought by a male non-veteran, this statute was declared violative of state constitutional provisions guaranteeing that government should be *266for the “common good” and prohibiting hereditary titles. Brown v. Russell, 166 Mass. 14, 43 N. E. 1005 (1896).

The current veterans’ preference law has its origins in an 1896 statute, enacted to meet the state constitutional standards enunciated in Brown v. Russell. That statute limited the absolute preference to veterans who were otherwise qualified.13 A closely divided Supreme Judicial Court, in an advisory opinion issued the same year, concluded that the preference embodied in such a statute would be valid. Opinion of the Justices, 166 Mass. 589, 44 N. E. 625 (1896). In 1919, when the preference was extended to cover the veterans of World War I, the formula was further limited to provide for a priority in eligibility, in contrast to an absolute preference in hiring.14 See Corliss v. Civil Service Comm’rs, 242 Mass. 61, 136 N. E. 356 (1922). In Mayor of Lynn v. Commissioner of Civil Service, 269 Mass. 410, 414, 169 N. E. 502, 503-504 (1929), the Supreme Judicial Court, adhering to the views expressed in its 1896 advisory opinion, sustained this statute against a state constitutional challenge.

Since 1919, the preference has been repeatedly amended to cover persons who served in subsequent wars, declared or *267undeclared. See 1943 Mass. Acts, ch. 194; 1949 Mass. Acts, ch. 642, § 2 (World War II); 1954 Mass. Acts, ch. 627 (Korea); 1968 Mass. Acts, ch. 531, § 1 (Vietnam) ,15 The current preference formula in ch. 31, § 23, is substantially the same as that settled upon in 1919. This absolute preference— even as modified in 1919 — has never been universally popular. Over the years it has been subjected to repeated legal challenges, see Hutcheson v. Director of Civil Service, swpra (collecting cases), to criticism by civil service reform groups, see, e. g., Report of the Massachusetts Committee on Public Service on Initiative Bill Relative to Veterans’ Preference, S. No. 279 (1926); Report of Massachusetts Special Commission on Civil Service and Public Personnel Administration 37-43 (June 15, 1967), and, in 1926, to a referendum in which it was reaffirmed by a majority of 51.9%. See id., at 38. The present case is apparently the first to challenge the Massachusetts veterans’ preference on the simple ground that it discriminates on the basis of sex.16

D

The first Massachusetts veterans’ preference statute defined the term “veterans” in gender-neutral language. See *2681896 Mass. Acts, ch. 517 § 1 (“a person” who served in the United States Army or Navy), and subsequent amendments have followed this pattern, see, e. g., 1919 Mass. Acts, ch. 150, § 1 (“any person who has served . . .”); 1954 Mass Acts, ch. 627, § 1 (“any person, male or female, including a nurse”). Women who have served in official United States military units during wartime, then, have always been entitled to the benefit of the preference. In addition, Massachusetts, through a 1943 amendment to the definition of “wartime service,” extended the preference to women who served in unofficial auxiliary women’s units. 1943 Mass. Acts, ch. 194.17

When the first general veterans’ preference statute was adopted in 1896, there were no women veterans.18 The statute, however, covered only Civil War veterans. Most of them were beyond middle age, and relatively few were actively competing for public employment.19 Thus, the impact of *269the preference upon the employment opportunities of non-veterans as a group and women in particular was slight.20

Notwithstanding the apparent attempts by Massachusetts to include as many military women as possible within the scope of the preference, the statute today benefits an overwhelmingly male class. This is attributable in some measure to the variety of federal statutes, regulations, and policies that have restricted the number of women who could enlist in the United States Armed Forces,21 and largely to the simple *270fact that women have never been subjected to a military draft. See generally Binkin and Bach 4-21.

When this litigation was commenced, then, over 98% of the veterans in Massachusetts were male; only 1.8% were female. And over one-quarter of the Massachusetts population were veterans. During the decade between 1963 and 1973 when the appellee was actively participating in the State’s merit selection system, 47,005 new permanent appointments were made in the classified official service. Forty-three percent of those hired were women, and 57% were men. Of the women appointed, 1.8% were veterans, while 54% of the men had veteran status. A large unspecified percentage of the female appointees were serving in lower paying positions for which males traditionally had not applied.-22 *271On each of 50 sample eligible lists that are part of the record in this case, one or more women who would have been certified as eligible for appointment on the basis of test results were displaced by veterans whose test scores were lower.

At the outset of this litigation appellants conceded that for “many of the permanent positions for which males and females have competed” the veterans’ preference has “resulted in a substantially greater proportion of female eligibles than male eligibles” not being certified for consideration. The impact of the veterans’ preference law upon the public employment opportunities of women has thus been severe. This impact lies at the heart of the appellee’s federal constitutional claim.

II

The sole question for decision on this appeal is whether Massachusetts, in granting an absolute lifetime preference to veterans, has discriminated against women in violation of the Equal Protection Clause of the Fourteenth Amendment.

A

The equal protection guarantee of the Fourteenth Amendment does not take from the States all power of classification. Massachusetts Bd. of Retirement v. Murgia, 427 U. S. 307, 314. Most laws classify, and many affect certain groups *272unevenly, even though the law itself treats them no differently from all other members of the class described by the law. When the basic classification is rationally based, uneven effects upon particular groups within a class are ordinarily of no constitutional concern. New York City Transit Authority v. Beazer, 440 U. S. 568; Jefferson v. Hackney, 406 U. S. 535, 548. Cf. James v. Valtierra, 402 U. S. 137. The calculus of effects, the manner in which a particular law reverberates in a society, is a legislative and not a judicial responsibility. Dandridge v. Williams, 397 U. S. 471; San Antonio School Dist. v. Rodriguez, 411 U. S. 1. In assessing an equal protection challenge, a court is called upon only to measure the basic validity of the legislative classification. Barrett v. Indiana, 229 U. S. 26, 29-30; Railway Express Agency v. New York, 336 U. S. 106. When some other independent right is not at stake, see, e. g., Shapiro v. Thompson, 394 U. S. 618, and when there is no “reason to infer antipathy,” Vance v. Bradley, 440 U. S. 93, 97, it is presumed that “even improvident decisions will eventually be rectified by the democratic process . . . .” Ibid.

Certain classifications, however, in themselves supply a reason to infer antipathy. Race is the paradigm. A racial classification, regardless of purported motivation, is presumptively invalid and can be upheld only upon an extraordinary justification. Brown v. Board of Education, 347 U. S. 483; McLaughlin v. Florida, 379 U. S. 184. This rule applies as well to a classification that is ostensibly neutral but is an obvious pretext for racial discrimination. Yick Wo v. Hopkins, 118 U. S. 356; Guinn v. United States, 238 U. S. 347; cf. Lane v. Wilson, 307 U. S. 268; Gomillion v. Lightfoot, 364 U. S. 339. But, as was made clear in Washington v. Davis, 426 U. S. 229, and Arlington Heights v. Metropolitan Housing Dev. Corp., 429 U. S. 252, even if a neutral law has a disproportionately adverse effect upon a racial minority, it is unconstitutional under the Equal Protection Clause only if that impact can be traced to a discriminatory purpose.

*273Classifications based upon gender, not unlike those based upon race, have traditionally been the touchstone for pervasive and often subtle discrimination. Caban v. Mohammed, 441 U. S. 380, 398 (Stewart, J., dissenting). This Court’s recent cases teach that such classifications must bear a close and substantial relationship to important governmental objectives, Craig v. Boren, 429 U. S. 190, 197, and are in many settings unconstitutional. Reed v. Reed, 404 U. S. 71;, Frontiero v. Richardson, 411 U. S. 677; Weinberger v. Wiesenfeld, 420 U. S. 636; Craig v. Boren, supra; Califano v. Goldfarb, 430 U. S. 199; Orr v. Orr, 440 U. S. 268; Caban v. Mohammed, supra. Although public employment is not a constitutional right, Massachusetts Bd. of Retirement v. Murgia, supra, and the States have wide discretion in framing employee qualifications, see, e. g., New York City Transit Authority v. Beazer, supra, these precedents dictate that any state law overtly or covertly designed to prefer males over females in public employment would require an exceedingly persuasive justification to withstand a constitutional challenge under the Equal Protection Clause of the Fourteenth Amendment.

B

The cases of Washington v. Davis, supra, and Arlington Heights v. Metropolitan Housing Dev. Corp., supra, recognize that when a neutral law has a disparate impact upon a group that has historically been the victim of discrimination, an unconstitutional purpose may still be at work. But those cases signaled no departure from the settled rule • that the Fourteenth Amendment guarantees equal laws, not equal results. Davis upheld a job-related employment test that white people passed in proportionately greater numbers than Negroes, for there had been no showing that racial discrimination entered into the establishment or formulation of the test. Arlington Heights upheld a zoning board decision that tended to perpetuate racially segregated housing patterns, *274since, apart from its effect, the board’s decision was shown to be nothing more than an application of a constitutionally neutral zoning policy. Those principles apply with equal force to a case involving alleged gender discrimination.

When a statute gender-neutral on its face is challenged on the ground that its effects upon women are disproportionably adverse, a twofold inquiry is thus appropriate. The first question is whether the statutory classification is indeed neutral in the sense that it is not gender based. If the classification itself, covert or overt, is not based upon gender, the second question is whether the adverse effect reflects invidious gender-based discrimination. See Arlington Heights v. Metropolitan Housing Dev. Corp., supra. In this second inquiry, impact provides an “important starting point,” 429 U. S., at 266, but purposeful discrimination is “the condition that offends the Constitution.” Swann v. Charlotte-Mecklenburg Board of Education, 402 U. S. 1, 16.

It is against this background of precedent that we consider the merits of the case before us.

Ill

A

The question whether ch. 31, § 23, establishes a classification that is overtly or covertly based upon gender must first be considered. The appellee has conceded that ch. 31, § 23, is neutral on its face. She has also acknowledged that state hiring preferences for veterans are not per se invalid, for she has limited her challenge to the absolute lifetime preference that Massachusetts provides to veterans. The District Court made two central findings that are relevant here: first, that ch. 31, § 23, serves legitimate and worthy purposes; second, that the absolute preference was not established for the purpose of discriminating against women. The appellee has thus acknowledged and the District Court has thus found *275that the distinction between veterans and nonveterans drawn by ch. 31, § 23, is not a pretext for gender discrimination. The appellee’s concession and the District Court’s finding are clearly correct.

If the impact of this statute could not be plausibly explained on a neutral ground, impact itself would signal that the real classification made by the law was in fact not neutral. See Washington v. Davis, 426 U. S., at 242; Arlington Heights v. Metropolitan Housing Dev. Corp., supra, at 266. But there can be but one answer to the question whether this veteran preference excludes significant numbers of women from preferred state jobs because they are women or because they are nonveterans. Apart from the facts that the definition of “veterans” in the statute has always been neutral as to gender and that Massachusetts has consistently defined veteran status in a way that has been inclusive of women who have served in the military, this is not a law that can plausibly be explained only as a gender-based classification. Indeed, it is not a law that can rationally be explained on that ground. Veteran status is not uniquely male. Although few women benefit from the preference, the nonveteran class is not substantially all female. To the contrary, significant numbers of nonveterans are men, and all nonveterans — male as well as female — are placed at a disadvantage. Too many men are affected by ch. 31, § 23, to permit the inference that the statute is but a pretext for preferring men over women.

Moreover, as the District Court implicitly found, the purposes of the statute provide the surest explanation for its impact. Just as there are cases in which impact alone can unmask an invidious classification, cf. Yick Wo v. Hopkins, 118 U. S. 356, there are others, in which — notwithstanding impact — the legitimate noninvidious purposes of a law cannot be missed. This is one. The distinction made by ch. 31, § 23, is, as it seems to be, quite simply between veterans and nonveterans, not between men and women.

*276B

The dispositive question, then, is whether the appellee has shown that a gender-based discriminatory purpose has, at least in some measure, shaped the Massachusetts veterans' preference legislation. As did the District Court, she points to two basic factors which in her view distinguish ch. 31, § 23, from the neutral rules at issue in the Washington v. Davis and Arlington Heights cases. The first is the nature of the preference, which is said to be demonstrably gender-biased in the sense that it favors a status reserved under federal military policy primarily to men. The second concerns the impact of the absolute lifetime preference upon the employment opportunities of women, an impact claimed to be too inevitable to have been unintended. The appellee contends that these factors, coupled with the fact that the preference itself has little if any relevance to actual job performance, more than suffice to prove the discriminatory intent required to establish a constitutional violation.

1

The contention that this veterans’ preference is “inherently nonneutral” or “gender-biased” presumes that the State, by favoring veterans, intentionally incorporated into its public employment policies the panoply of sex-based and assertedly discriminatory federal laws that have prevented all but a handful of women from becoming veterans. There are two serious difficulties with this argument. First, it is wholly at odds with the District Court’s central finding that Massachusetts has not offered a preference to veterans for the purpose of discriminating against women. Second, it cannot be reconciled with the assumption made by both the appellee and the District Court that a more limited hiring •preference for veterans could be sustained. Taken together, these difficulties are fatal.

To the extent that the status of veteran is one that few *277women have been enabled to achieve, every hiring preference for veterans, however modest or extreme, is inherently gender-biased. If Massachusetts by offering such a preference can be said intentionally to have incorporated into its state employment policies the historical gender-based federal military personnel practices, the degree of the preference would or should make no constitutional difference. Invidious discrimination does not become less so because the discrimination accomplished is of a lesser magnitude.23 Discriminatory intent is simply not amenable to calibration. It either is a factor that has influenced the legislative choice or it is not. The District Court’s conclusion that the absolute veterans’ preference was not originally enacted or subsequently reaffirmed for the purpose of giving an advantage to males as such necessarily compels the conclusion that the State intended nothing more than to prefer “veterans.” Given this finding, simple logic suggests that an intent to exclude women from significant public jobs was not at work in this law. To reason that it was, by describing the preference as “inherently nonneutral” or “gender-biased,” is merely to restate the fact of impact, not to answer the question of intent.

To be sure, this case is unusual in that it involves a law that by design is not neutral. The law overtly prefers veterans as such. As opposed to the written test at issue in Davis, it does not purport to define a job-related characteristic. To the contrary, it confers upon a specifically described group — perceived to be particularly deserving — a competitive headstart. But the District Court found, and the appellee has not disputed, that this legislative choice was legitimate. The basic distinction between veterans and nonveterans, having been found not gender-based, and the goals of the *278preference having been found worthy, ch. 31 must be analyzed as is any other neutral law that casts a greater burden upon women as a group than upon men as a group. The enlistment policies of the Armed Services may well have discriminated on the basis of sex. See Frontiero v. Richardson, 411 U. S. 677; cf. Schlesinger v. Ballard, 419 U. S. 498. But the history of discrimination against women in the military is not on trial in this case.

2

The appellee’s ultimate argument rests upon the presumption, common to the criminal and civil law, that a person intends the natural and foreseeable consequences of his voluntary actions. Her position was well stated in the concurring opinion in the District Court:

“Conceding . . . that the goal here was to benefit the veteran, there is no reason to absolve the legislature from awareness that the means chosen to achieve this goal would freeze women out of all those state jobs actively sought by men. To be sure, the legislature did not wish to harm women. But the cutting-off of women’s opportunities was an inevitable concomitant of the chosen scheme — as inevitable as the proposition that if tails is up, heads must be down. Where a law’s consequences are that inevitable, can they meaningfully be described as unintended?” 451 F. Supp., at 151.

This rhetorical question implies that a negative answer is obvious, but it is not. The decision to grant a preference to veterans was of course “intentional.” So, necessarily, did an adverse impact upon nonveterans follow from that decision. And it cannot seriously be argued that the Legislature of Massachusetts could have been unaware that most veterans are men. It would thus be disingenuous to say that the adverse consequences of this legislation for women were unintended, in the sense that they were not volitional or in the sense that they were not foreseeable.

*279“Discriminatory purpose,” however, implies more than intent as volition or intent as awareness of consequences. See United Jewish Organizations v. Carey, 430 U. S. 144, 179 (concurring opinion).24 It implies that the decisionmaker, in this case a state legislature, selected or reaffirmed a particular course of action at least in part “because of,” not merely “in spite of,” its adverse effects upon an identifiable group.25 Yet nothing in the record demonstrates that this preference for veterans was originally devised or subsequently re-enacted because it would accomplish the collateral goal of keeping women in a stereotypic and predefined place in the Massachusetts Civil Service.

To the contrary, the statutory history shows that the benefit of the preference was consistently offered to “any person” who was a veteran. That benefit has been extended to women under a very broad statutory definition of the term veteran.26 The preference formula itself, which is the focal *280point of this challenge, was first adopted — so it appears from this record — out of a perceived need to help a small group of older Civil War veterans. It has since been reaffirmed and extended only to cover new veterans.27 When the totality of legislative actions establishing and extending the Massachusetts veterans’ preference are considered, see Washington v. Davis, 426 U. S., at 242, the law remains what it purports to be: a preference for veterans of either sex over nonveterans of either sex, not for men over women.

IV

Veterans’ hiring preferences represent an awkward — and, many argue, unfair — exception to the widely shared view that merit and merit alone should prevail in the employment policies of government. After a war, such laws have been enacted virtually without opposition. During peacetime, they inevitably have come to be viewed in many quarters as undemocratic and unwise.28 Absolute and permanent preferences, as the troubled history of this law demonstrates, have always been subject to the objection that they give the vet*281eran more than a square deal. But the Fourteenth Amendment “cannot be made a refuge from ill-advised . . . laws.” District of Columbia v. Brooke, 214 U. S. 138, 150. The substantial edge granted to veterans by ch. 31, § 23, may reflect unwise policy. The appellee, however, has simply failed to demonstrate that the law in any way reflects a purpose to discriminate on the basis of sex.

The judgment is reversed, and the case is remanded for further proceedings consistent with this opinion.

It is so ordered.

Mr. Justice Stevens,

with whom Mr. Justice White joins,

concurring.

While I concur in the Court’s opinion, I confess that I am not at all sure that there is any difference between the two questions posed ante, at 274. If a classification is not overtly based on gender, I am inclined to believe the question whether it is covertly gender based is the same as the question whether its adverse effects reflect invidious gender-based discrimination. However the question is phrased, for me the answer is largely provided by the fact that the number of males disadvantaged by Massachusetts’ veterans’ preference (1,867,000) is sufficiently large — and sufficiently close to the number of disadvantaged females (2,954,000) — to refute the claim that the rule was intended to benefit males as a class over females as a class.

Mr. Justice Marshall,

with whom Mr. Justice Brennan joins,

dissenting.

Although acknowledging that in some circumstances, discriminatory intent may be inferred from the inevitable or foreseeable impact of a statute, ante, at 279 n. 25, the Court concludes that no such intent has been established here. I cannot agree. In my judgment, Massachusetts’ choice of an absolute veterans’ preference system evinces purposeful *282gender-based discrimination. And because the statutory-scheme bears no substantial relationship to a legitimate governmental objective, it cannot withstand scrutiny under the Equal Protection Clause.

I

The District Court found that the “prime objective” of the Massachusetts veterans’ preference statute, Mass. Gen. Laws Ann., ch. 31, § 23, was to benefit individuals with prior military service. Anthony v. Commonweath, 415 F. Supp. 485, 497 (Mass. 1976). See Feeney v. Massachusetts, 451 F. Supp. 143, 145 (Mass. 1978). Under the Court’s analysis, this factual determination “necessarily compels the conclusion that the State intended nothing more than to prefer 'veterans.’ Given this finding, simple logic suggests than an intent to exclude women from significant public jobs was not at work in this law.” Ante, at 277. I find the Court’s logic neither simple nor compelling.

That a legislature seeks to advantage one group does not, as a matter of logic or of common sense, exclude the possibility that it also intends to disadvantage another. Individuals in general and lawmakers in particular frequently act for a variety of reasons. As this Court recognized in Arlington Heights v. Metropolitan Housing Dev. Corp., 429 U. S. 252, 265 (1977), “[rjarely can it be said that a legislature or administrative body operating under a broad mandate made a decision motivated solely by a single concern.” Absent an omniscience not commonly attributed to the judiciary, it will often be impossible to ascertain the sole or even dominant purpose of a given statute. See McGinnis v. Royster, 410 U. S. 263, 276-277 (1973); Ely, Legislative and Administrative Motivation in Constitutional Law, 79 Yale L. J. 1205, 1214 (1970). Thus, the critical constitutional inquiry is not whether an illicit consideration was the primary or but-for cause of a decision, but rather whether it had an appreciable role in shaping a given legislative enactment. Where there is *283“proof that a discriminatory purpose has been a motivating factor in the decision, . . . judicial deference is no longer justified.” Arlington Heights v. Metropolitan Housing Dev. Corp., supra, at 265-266 (emphasis added).

Moreover, since reliable evidence of subjective intentions is seldom obtainable, resort to inference based on objective factors is generally unavoidable. See Beer v. United States, 425 U. S. 130, 148-149, n. 4 (1976) (Marshall, J., dissenting) ; cf. Palmer v. Thompson, 403 U. S. 217, 224-225 (1971); United States v. O’Brien, 391 U. S. 367, 383-384 (1968). To discern the purposes underlying facially neutral policies, this Court has therefore considered the degree, inevitability, and foreseeability of any disproportionate impact as well as the alternatives reasonably available. See Monroe v. Board of Commissioners, 391 U. S. 450, 459 (1968); Goss v. Board of Education, 373 U. S. 683, 688-689 (1963); Gomillion v. Lightfoot, 364 U. S. 339 (1960); Griffin v. Illinois, 351 U. S. 12, 17 n. 11 (1956). Cf. Albemarle Paper Co. v. Moody, 422 U. S. 405, 425 (1975).

In the instant case, the impact'of the Massachusetts statute on women is undisputed. Any veteran with a passing grade on the civil service exam must be placed ahead of a non-veteran, regardless of their respective scores. The District Court found that, as a practical matter, this preference supplants test results as the determinant of upper level civil service appointments. 415 F. Supp., at 488-489. Because less than 2% of the women in Massachusetts are veterans, the absolute-preference formula has rendered desirable state civil service employment an almost exclusively male prerogative. 451 F. Supp., at 151 (Campbell, J., concurring).

As the District Court recognized, this consequence follows foreseeably, indeed inexorably, from the long history of policies severely limiting women’s participation in the military.1 *284Although neutral in form, the statute is anything but neutral in application. It inescapably reserves a major sector of public employment to “an already established class which, as a matter of historical fact, is 98% male.” Ibid. Where the foreseeable impact of a facially neutral policy is so disproportionate, the burden should rest on the State to establish that sex-based considerations played no part in the choice of the particular legislative scheme. Cf. Castaneda v. Partida, 430 U. S. 482 (1977); Washington v. Davis, 426 U. S. 229, 241 (1976); Alexander v. Louisiana, 405 U. S. 625, 632 (1972); see generally Brest, Palmer v. Thompson: An Approach to the Problem of Unconstitutional Legislative Motive, 1971 Sup. Ct. Rev. 95, 123.

Clearly, that burden was not sustained here. The legislative history of the statute reflects the Commonwealth’s patent appreciation of the impact the preference system would have on women, and an equally evident desire to mitigate that impact only with respect to certain traditionally female occupations. Until 1971, the statute and implementing civil serv*285ice regulations exempted from operation of the preference any job requisitions “especially calling for women.” 1954 Mass. Acts, ch. 627, §5. See also 1896 Mass. Acts, ch. 517, §6; 1919 Mass. Acts, ch. 150, §2; 1945 Mass. Acts, ch. 725, §2 (e); 1965 Mass. Acts, ch. 53; ante, at 266 nn. 13,14. In practice, this exemption, coupled with the absolute preference for veterans, has created a gender-based civil service hierarchy, with women occupying low-grade clerical and secretarial jobs and men holding more responsible and remunerative positions. See 415 F. Supp., at 488; 451 F. Supp., at 148 n. 9.

Thus, for over 70 years, the Commonwealth has maintained, as an integral part of its veterans’ preference system, an exemption relegating female civil service applicants to occupations traditionally filled by women. Such a statutory scheme both reflects and perpetuates precisely the kind of archaic assumptions about women’s roles which we have previously held invalid. See Orr v. Orr, 440 U. S. 268 (1979); Califano v. Goldfarb, 430 U. S. 199, 210-211 (1977); Stanton v. Stanton, 421 U. S. 7, 14 (1975); Weinberger v. Wiesenfeld, 420 U. S. 636, 645 (1975). Particularly when viewed against the range of less discriminatory alternatives available to assist veterans,2 Massachusetts’ choice of a formula that so severely restricts public employment opportunities for women cannot reasonably be thought gender-neutral. Cf. Albemarle Paper Co. v. Moody, supra, at 425. The Court’s conclusion to the contrary — that “nothing in the record” evinces a “collateral goal of keeping women in a stereotypic and predefined place in the *286Massachusetts Civil Service,” ante, at 279 — displays a singularly myopic view of the facts established below.3

II

To survive challenge under the Equal Protection Clause, statutes reflecting gender-based discrimination must be substantially related to the achievement of important governmental objectives. See Califano v. Webster, 430 U. S. 313, 316-317 (1977); Craig v. Boren, 429 U. S. 190, 197 (1976); Reed v. Reed, 404 U. S. 71, 76 (1971). Appellants here advance three interests in support of the absolute-preference system: (1) assisting veterans in their readjustment to civilian life; (2) encouraging military enlistment; and (3) rewarding those who have served their country. Brief for Appellants 24. Although each of those goals is unquestionably legitimate, the “mere recitation of a benign, compensatory purpose” cannot of itself insulate legislative classifications from constitutional scrutiny. Weinberger v. Wiesenfeld, supra, at 648. And in this case, the Commonwealth has failed to establish a sufficient relationship between its objectives and the means chosen to effectuate them.

With respect to the first interest, facilitating veterans’ transition to civilian status, the statute is plainly overinclusive. Cf. Trimble v. Gordon, 430 U. S. 762, 770-772 (1977); Jimenez v. Weinberger, 417 U. S. 628, 637 (1974). By conferring a permanent preference, the legislation allows veterans to invoke their advantage repeatedly, without regard to their date of discharge. As the record demonstrates, a substantial *287majority of those currently enjoying the benefits of the system are not recently discharged veterans in need of readjustment assistance.4

Nor is the Commonwealth’s second asserted interest, encouraging military service, a plausible justification for this legislative scheme. In its original and subsequent re-enactments, the statute extended benefits retroactively to veterans who had served during a prior specified period. See ante, at 265-267. If the Commonwealth’s “actual purpose” is to induce enlistment, this legislative design is hardly well suited to that end. See Califano v. Webster, supra, at 317; Weinberger v. Wiesenfeld, supra, at 648. For I am unwilling to assume what appellants made no effort to prove, that the possibility of obtaining an ex post facto civil service preference significantly influenced the enlistment decisions of Massachusetts residents. Moreover, even if such influence could be presumed, the statute is still grossly overinclusive in that it bestows benefits on men drafted as well as those who volunteered.

Finally, the Commonwealth’s third interest, rewarding veterans, does not “adequately justify the salient features” of this preference system. Craig v. Boren, supra, at 202-203. See Orr v. Orr, supra, at 281. Where a particular statutory scheme visits substantial hardship on a class long subject to discrimination, the legislation cannot be sustained unless “ ‘carefully tuned to alternative considerations.’ ” Trimble v. Gordon, supra, at 772. See Caban v. Mohammed, 441 U. S. 380, 392-393, n. 13 (1979); Mathews v. Lucas, 427 U. S. 495 (1976). Here, there are a wide variety of less discriminatory means by which Massachusetts could effect its compensatory purposes. For example, a point preference system, such as that maintained by many States and the Federal Government, *288see n. 2, supra, or an absolute preference for a limited duration, would reward veterans without excluding all qualified women from upper level civil service positions. Apart from public employment, the Commonwealth, can, and does, afford assistance to veterans in various ways, including tax abate-ments, educational subsidies, and special programs for needy veterans. See Mass. Gen. Laws Ann., ch. 59, § 5, Fifth (West Supp. 1979); Mass. Gen. Laws Ann., ch. 69, §§ 7, 7B (West Supp. 1979); and Mass. Gen. Laws Ann., chs. 115,115A (West 1969 and Supp. 1978). Unlike these and similar benefits, the costs of which are distributed across the taxpaying public generally, the Massachusetts statute exacts .a substantial price from a discrete group of individuals who have long been subject to employment discrimination,5 and who, “because of circumstances totally beyond their control, have [had] little if any chance of becoming members of the preferred class.” 415 F. Supp., at 499. See n. 1, supra.

In its present unqualified form, the veterans’ preference statute precludes all but a small fraction of Massachusetts women from obtaining any civil service position also of interest to men. See 451 F. Supp., at 151 (Campbell, J., concurring). Given the range of alternatives available, this degree of preference is not constitutionally permissible.

I would affirm the judgment of the court below.