BOUDIN, Circuit Judge.
This appeal, from a preliminary injunction granted by the district court, concerns the so-called section 8 program for rental assistance. It is a federally funded and supervised rent subsidy program for low-income tenants, but it is administered primarily through local units called public housing authorities or “PHAs.” See 42 U.S.C.A. § 1437f (West Supp.1999). The program was created by a 1974 amendment to the Housing Act of 1937, Housing and Community Development Act of 1974, Pub.L. No. 93-383, Title II, § 201(a), 88 Stat. 633, 662-66, and has been revised since then, importantly in 1998 by the Quality Housing and Work Responsibility Act of 1998, Pub.L. No. 105-276, Title V, § 545,112 Stat. 2518, 2596-604.
A main form of assistance is the certificate or voucher (the latter is the current term),1 which the PHA may issue to certain low-income families, 42 U.S.C.A. § 1437f(o)(4) (West Supp.1999) — defined as families earning at or below 80 percent of the median income in the area, id. § 1437a(b)(2). The voucher program requires the PHA to pay to the family’s landlord the difference between the gross rent or a “payment standard” adopted by the PHA, and a lesser amount paid by the family. See id. § 1437f(o)(2); 24 C.F.R. §§ 982.503, 982.505 (1999). PHAs normally do not have enough funds to subsidize all of the families that meet the financial requirements for assistance. As a result, the common practice is for each PHA to maintain a waiting list for applicants who will receive vouchers if and when existing vouchers are surrendered or appropriations increase. See 24 C.F.R. §§ 982.204, 982.205 (1999).
The vouchers are awarded, by local PHAs, but applicants need not be local residents when they apply, id. § 982.202(b)(1), and they can apply for vouchers from any of the many PHAs in the state (Massachusetts has about 130 local PHAs that administer approximately 40,000 vouchers). Apparently, the PHA may insist that the voucher be used for local rental at the outset, (it is not clear whether the PHAs here so insisted), but a portability provision permits the user to move after twelve months while retaining the subsidy to any area in which a section 8 program is administered. 42 U.S.C.A. § 1437f(r) (West Supp.1999); 24 C.F.R. § 982.353 (1999). There is thus ample incentive for persons to apply to one or more (sometimes many more) PHAs outside communities where they live or work.
[46]*46In 1998, eight suburban PHAs in Eastern Massachusetts — later named as defendants in this case- — determined that their present waiting lists would soon be exhausted and that it would be less expensive to obtain new applicants by a joint advertising program. The PHAs, which are generally coextensive with a local town or city, are located in the towns of Abing-ton, Avon, Bridgewater, Halifax, Holbrook, Middleborough, Pembroke, and Rockland. The PHAs all proposed to hold new, separately conducted lotteries for additional applicants on December 1, 1998. Further, each PHA proposed to give preference to local residents so that “local residents” (defined as those currently living or working in the PHA) would be listed ahead of those applicants currently residing outside the PHA in question. Local preferences in the awarding of section 8 vouchers are explicitly permitted by the governing statute, as amended in 1998, see 42 U.S.C.A. § 1437f(o)(6) (West Supp.1999), although with what qualifications remains to be considered.
One qualification on the ranking of applicants by lottery is the so-called 75 percent rule. This is a requirement adopted in 1998 that, save in specified circumstances not relevant here, 75 percent of the families “initially provided tenant-based assistance under section 1437f ... by a public housing agency in any fiscal year” must be families whose, income is at or below 30 percent of the area median income, 42 U.S.C.A. § 1437n(b)(l) (West Supp.1999), also referred to as “extremely low income families” in HUD’s regulations, 24 C.F.R. § 982.201(b)(2) (1999); id. § 903.7(a)(1)®. This is an income level well below the 80 percent ceiling needed to qualify for the vouchers.
In the public notice given in October 1998, the PHAs stated that applications could be requested, in person or by phone, during prescribed hours (9:30 a.m. to 2:30 p.m.) on October 29 and 30, 1998, for each PHA at a designated office in the respective communities. Applications were required to be returned by hand or postmarked no later than noon on November 17, 1998. The notice said that a separate lottery would be held at each of the eight offices on December 1, 1998, at 1 p.m.
On November 16, 1998, the plaintiffs filed the present action in the federal district court. The four individual plaintiffs are four women, three Hispanic and one African American, who have very low incomes and do not reside or work in any of the eight PHAs named as defendants. A fifth named plaintiff is the Massachusetts Coalition for the Homeless which, according to the complaint, numbers among its members homeless and poor individuals in search of housing, many of whom are African-American or Hispanic. The complaint, framed as a class action, sought injunctive relief against the PHAs on four separate fronts.
Pertinently, the complaint charged that the use of the local residency preference in connection with the lottery violated the Equal Protection Clause of the 14th Amendment and various civil rights statutes and regulations; and it also charged that the defendants were threatening to violate the statutory requirement that 75 percent of the vouchers be reserved for extremely low income families. Two other claims made in the complaint (a procedural due process claim and a state law claim) have not been pressed on this appeal, so we do not discuss them further.
On November 30,1998, the district court held a hearing and issued a temporary restraining order, permitting the lotteries to proceed as scheduled but precluding the ranking of the waiting lists based on the residency preferences and the distribution of vouchers based on such re-ordered lists until the court ruled on the preliminary injunction request. The lotteries took place the next day, December 1, 1998. Thereafter, the parties as directed by the district court submitted analyses of both the pre-existing and the new lottery-determined lists. The calculations included predictions about the effect of the local resi[47]*47dency preference on the racial and income make-up of the new lists.
On December 30, 1998, the district court entered a preliminary injunction and a memorandum opinion with supporting reasoning. The district court found that use of the residency preferences created a likelihood that some of the defendant PHAs would violate the 75 percent rule and that others would violate an anti-discrimination provision of the Fair Housing Act, 42 U.S.C. § 3604(a) (1994).2 It based these judgments on analyses of who the voucher recipients in each PHA would probably be (including their poverty levels and racial make-up) if the residency preferences were or were not used.
Based on these analyses, the court enjoined the Abington and Rockland PHAs from distributing vouchers based on the new lists using the residency preference until they presented, and the court approved, a specific plan to assure compliance with the 75 percent rule. The Avon, Holbrook, and Middleborough PHAs were enjoined from distributing vouchers based on the new lists using the residency preferences because of the disparate racial impact of the local-resident preferences. The court did not enjoin the use of the old soon-to-be exhausted lists, and the other three PHAs were not enjoined because they were not yet ready to use their new lists; but the court did order them to give notice to the court before using the new lists.
The district court’s injunction was preliminary and the court ordered a scheduling conference in February 1999 looking toward discovery and further proceedings. The defendant PHAs have appealed from the preliminary injunction. 28 U.S.C. § 1292(a)(1) (1994). In this court, the plaintiffs are supported by the NAACP, Brockton Area Branch as amicus curiae. For reasons now to be explained, we affirm the preliminary injunction so far as it is addressed to the 75 percent rule, remand for further proceedings concerning the disparate impact issue, and in the latter instance permit the district court to maintain its ban on the use of the local residency preferences for a limited time pending further proceedings.
The criteria for the grant of a preliminary injunction are the familiar four: likelihood of success, risk of irreparable harm, the balance of equities and the public interest. See Ross-Simons of Warwick, Inc. v. Baccarat, Inc., 102 F.3d 12, 15 (1st Cir.1996). Although it is sometimes said that a preliminary injunction is reviewed for “abuse of discretion,” the standard of review obviously depends on the issue under consideration. Generally speaking, pure issues of law (e.g., the construction of a statute) are reviewed de novo, findings of fact for clear error, and “judgment calls” with considerable deference depending upon the issue. See Public Serv. Co. of New Hampshire v. Patch, 167 F.3d 15, 22 (1st Cir.1998).
1. On this appeal, the PHAs begin with a set of general attacks on the grant of any relief. They first say that the data that underpinned the district court’s decision on both the 75 percent rule and the disparate impact claims depended on two assumptions: that the number of vouchers to be distributed in each PHA in the next year would equal the number estimated by defendants, and that each of the individuals who achieved high ranking in the applicant lottery would qualify for and accept the voucher when offered. The PHAs assert that neither premise is sound and that therefore it is unreasonable to predict that the 75 percent rule will be violated or that a disparate racial impact will occur.
It is true that these two assumptions were made, that they may turn out to be wrong in individual cases, and that small changes could be significant because of the [48]*48small number of available vouchers involved (e.g., 15 in Rockland). But no projections in this situation can be anything other than approximations, and the defendants offer nothing any better. And, of course, the deviations could make the defendants’ legal position worse rather than better. While in some cases the best approximations may be so speculative that no relief could rationally be based upon them, see Ross-Simons, 102 F.3d at 19, this does not seem to us so extreme a case.
In their general critique, the PHAs offer two other arguments. One is that use of the residency preference does not permanently deny vouchers to any qualified person; it merely delays them. But given the limits on funding, the delays apparently ■ stretch into years, so delay may for practical purposes be the equivalent of denial. The other general argument is that the district court’s injunction fails to accord due weight, in assessing the public interest, to Congress’s general support for local preferences. Congress’s view of such local preferences bears on their lawfulness; but if their use is unlawful in these circumstances, injunctive relief would be lawful.
2. This brings us to the PHAs’ specific attacks on the first branch of the injunction, which rests on the statutory requirement that 75 percent of the vouchers be directed to families with incomes at or below 30 percent of the area median income. The district court found that use of the local residency preferences meant that two PHAs (Abington and Rockland) would likely violate this requirement if the local residency preferences were used but that elimination of the residency preferences would ensure that the 75 percent rule would be met. Accordingly, the court ordered that these two PHAs distribute vouchers in accordance with the lottery ranking without providing a preference to residents.
The PHAs’ best argument is that compliance with the 75 percent rule can be achieved by what they call “a far less intrusive means,” namely, by having the local authority pass over an applicant and award the voucher to the next available applicant whenever this is necessary to meet the 75 percent rule. But it appears that in the district court the defendants merely mentioned this remedy as a possibility and, as the court pointed out, proffered no detailed plan to monitor the distribution of vouchers to assure this result. Moreover, the district court mentioned the prospect of vacating the injunction as soon as the PHAs presented it with a “specific plan” to ensure compliance with the 75 percent rule.
After the district court’s judgment and during the pendency of this appeal, the PHAs filed an unopposed motion to supplement the record with recent amendments to their respective plans. The amendments state:
Of the families initially provided tenant based assistance under section 8 by the ... Housing Authority in any fiscal year, not less than 75% shall be families whose incomes do not exceed 30% of the area median income as determined by the Secretary. The local preference policy is subordinate to this provision, and admissions to the program will be monitored to assure compliance.
These amendments are certainly a step in the right direction and, in addition, may indicate the PHAs’ abandonment of another disputed point: in the district court the PHAs argued that the 75 percent rule could be satisfied not by computing the number of section 8 vouchers obtained by extremely poor families in each fiscal year but by including numbers from outside the fiscal year.
It may be that this post-judgment effort at compliance is sufficient, but the plaintiffs should have a chance to express any objections and the district court to resolve them. Perhaps the district court will conclude that more detail is necessary on just how monitoring will be done and how compliance will be assured; or there may be [49]*49other concerns about the adequacy of the amendments. The district court is the proper place for these doubts to be resolved. Board of Educ. of Oklahoma City Pub. Schs. v. Dowell, 498 U.S. 237, 249-50, 111 S.Ct. 630, 112 L.Ed.2d 715 (1991); In re Gulf Aerospace Corp., 449 F.2d 733, 734-35 (5th Cir.1971).
We agree that the first time around, a district court normally ought to use the least intrusive method of ensuring compliance with the law. Swift & Co. v. United States, 196 U.S. 375, 396, 25 S.Ct. 276, 49 L.Ed. 518 (1905); Educational Testing Servs. v. Katzman, 793 F.2d 533, 544-45 (3d Cir.1986). And we have no doubt that on remand the district court will promptly consider the PHAs’ new claim that the preliminary injunction should be withdrawn as to the 75 percent rule because compliance has now been established. If dissatisfied with the result, the denial of a requested modification is immediately appealable. 28 U.S.C. § 1292(a)(1) (1994).
3. The defendants’ most important attack is on the determination by the district judge that the resident preference constituted unlawful racial discrimination under the Fair Housing Act. See note 2, above. They point out that the district court found that there was no evidence of any intent to discriminate, and that the court rested entirely on the alleged disparate racial impact of the preference. They argue that the racial impact of using the preference was inadequately proven or minimal at best, and that in all events the district court erred by disregarding the congres-sionally approved desire to permit PHAs to establish local preferences.
A constitutional attack on racial discrimination requires an intent to discriminate by a state actor, Washington v. Davis, 426 U.S. 229, 239, 96 S.Ct. 2040, 48 L.Ed.2d 597 (1976); but in a number of other specific areas, Congress has gone further b3f statute: it has not only addressed private (as opposed to state) conduct but has in some statutes outlawed actions in particular areas that have a disparate racial impact unless adequately justified. This is well settled in the case of employment discrimination under Title VII of the 1964 Civil Rights Act, see Griggs v. Duke Power Co., 401 U.S. 424, 432-33, 91 S.Ct. 849, 28 L.Ed.2d 158 (1971), and we now similarly read the Fair Housing Act provision underpinning the district court’s decision.
It is quite true that the Fair Housing Act provision in question uses language that could be thought to refer simply to intentional discrimination, as it speaks of “makfing] unavailable ... a dwelling to any person because of race ....” Id. (emphasis added). In Village of Arlington Heights v. Metropolitan Housing Dev’t Corp., 429 U.S. 252, 271, 97 S.Ct. 555, 50 L.Ed.2d 450 (1977), the Supreme Court expressly declined to decide whether the FHA embraces cases where such intent is lacking but there exists a disparate racial impact and lack of justification. But the consensus among the circuits that have discussed this issue in the housing context is that the Fair Housing Act prohibits actions that have an unjustified disparate racial impact,3 and we find their reasoning persuasive.
However, under statutes like Title VII and Title VIII, merely to show a disparate racial impact is normally not enough to condemn: a vast array of measures, from war-making and the federal budget to local decisions on traffic and zoning, may have a disparate impact. Thus, practically all of the case law, both in employment and [50]*50housing, treats impact as doing no more than creating a prima, facie case, forcing the defendant to proffer a valid justification.4 In this case we are satisfied about the likelihood of disparate racial impact but more doubtful as to the district court’s resolution of the issue of justification, and we address each in turn.
As to impact, the defendant PHAs’ main challenge is to the district court’s reliance on the EEOC’s so-called four-fifths formula. This formula, used primarily in employment cases, is triggered where the selection rate for any race is less than four fifths the rate for the group with the highest selection rate. See 29 C.F.R. § 1607.4(D) (1999). The Supreme Court has said that no single test controls in measuring disparate impact, cf. Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 995-96 n. 3, 108 S.Ct. 2777, 101 L.Ed.2d 827 (1988), and we have approved use of the four-fifths rule as a pertinent benchmark in the employment context, see Boston Police Superior Officers Fed’n v. City of Boston, 147 F.Bd 13, 21 (1st Cir.1998).
Principally, the defendants argue (with some support) that this four-fifths formula — like statistical formulas in general— becomes less reliable when the numbers are small, see Fudge v. City of Providence Fire Dep’t, 766 F.2d 650, 658 n. 10 (1st Cir.1985); and the numbers in each of the PHAs in this case are fairly small. However, the plaintiffs maintain that the formula still provides some evidence of discriminatory effect, which may be enhanced by expert evidence once they have had a chance to develop it.5 Given that the issue of impact is fact-bound and the standard (likelihood of success) is lenient at this stage, we decline to disturb the district court’s preliminary finding of disparate impact. Cf. University of Texas v. Camenisch, 451 U.S. 390, 395, 101 S.Ct. 1830, 68 L.Ed.2d 175 (1981).
The more difficult question is justification. Although the plaintiffs say that virtually nothing can justify a disparate racial impact, this can hardly be so. As already noted, disparate racial effects occur all the time from all manner of government and private actions; and Title VII, the employment counterpart to the Fair Housing Act provision in question, makes such impact merely the basis for requiring justification. See 42 U.S.C. § 2000e-2(k)(l)(A)(i) (1994). Indeed, while Congress’s amendment of the law after Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 109 S.Ct. 2115, 104 L.Ed.2d 733 (1989), stiffened the employer’s burden of justification, it left intact the principle that disparate impact merely creates a prima fa-cie case. See note 4, above.
The district court itself assumed that a preference for local residents was a rational countervailing governmental objective; but it further believed that it was entitled to “balance” this objective against a showing of disparate impact and decide which in the case at hand was more important. Without much discussion, the court then said that the harm from disparate impact was the decisive factor and that the defendants had failed to show that there was no other way to achieve that local preference interest with “less discriminatory effect.” On the facts before us, it is not easy to imagine how a less restrictive alternative would work,6 but the question of “balancing” raises a difficult and important issue.
[51]*51In light of Griggs and the similarity of the statutes, it is a fair reading of the Fair Housing Act’s “because of race” prohibition to ask that a demonstrated disparate impact in housing be justified by a legitimate and substantial goal of the measure in question; but beyond that, we do not think that the courts’ job is to “balance” objectives, with individual judges deciding which seem to them more worthy. True, one circuit court decision did refer to balancing, see Metropolitan Housing Dev’t Corp. v. Village of Arlington Heights, 558 F.2d 1283, 1290-94 (7th Cir.1977), cert. denied, 434 U.S. 1025, 98 S.Ct. 752, 54 L.Ed.2d 772 (1978), but the few later circuit court decisions on point come closer to a simple justification test, see Rizzo, 564 F.2d at 148 — 49; Huntington Branch, NAACP, 844 F.2d at 939-40, and we think this is by far the better approach.
With possible qualifications, Congress might go further and determine that disparate impact alone should condemn a program or action, and an authorized agency might make the same choice; indeed, in regulations mentioned below, HUD may have gone further than the “because of race” statute itself. But to have federal judges make such policy choices is essentially to impose on them the job of making decisions that are properly made by Congress or its executive-branch delegates; and the balancing approach is in tension with the course taken by the Supreme Court and Congress under Title VII where a standard of justification is constructed and applied.
Such a standard is doubtless more difficult to construct here because a single criterion — like the relationship of the test to job performance used under Title VII— is hardly possible. But this case does not involve fine shadings. The PHAs say that they are local agencies seeking to assist local constituents first. Preferences for local residents have a considerable history, cf. County Bd. of Arlington County, Virginia v. Richards, 434 U.S. 5, 6-7, 98 S.Ct. 24, 54 L.Ed.2d 4 (1977); Lai v. New York City Gov’t, 163 F.3d 729, 731 (2d Cir.1998), and by the 1998 amendment Congress itself endorsed the use of locally determined preferences in distributing section 8 vouchers.7 It is hard not to treat Congress’s own enactment as justification enough to satisfy a statutory impact discrimination claim of the kind before us.
We thus conclude that (absent intentional discrimination), the residency preference does not violate the “because of race” provision of the Fair Housing Act standing alone. But in the district court the plaintiffs offered other legal bases for challenging the residency preference based on racial effect. Although those issues were not resolved by the district court, some initial mention of them is needed to determine whether the injunction as to the three PHAs affected by this branch of the injunction should be continued or vacated pending further proceedings.
4. The alternative grounds all appear to grow out of statutory and other provisions that impose obligations on HUD, the PHAs or both “affirmatively” to further “fair housing.” Further, HUD has adopted regulations specifically directed at the use of local preferences, at least a few of which are more precise than the “affirmatively further” language but may arguably rest on HUD’s own power affirmatively to foster fair housing. We begin with the general obligations and then turn to more specific ones.
We put to one side for the present the obligations of HUD itself, since this is a suit directed not against the Secretary but [52]*52against the PHAs.8 This leaves at least one pertinent statute directed to the PHAs themselves. The statute is a new provision that requires the PHAs to file annual plans and certify each year that their plans will affirmatively further fair housing, 42 U.S.C.A. § 1437c-l(d)(15) (West Supp. 1999). But — passing the question whether this requirement applied to the 1998 lotteries9 — the plan and certificate are to be reviewed by the Secretary, id. § 1437c-l(i), so it is unclear to what extent the adequacy of a certification is open to de novo judicial review.
Finally, HUD has a series of regulations, both predating and postdating the 1998 statute, that may bear on this matter. One set require the PHAs affirmatively to further fair housing objectives. See 24 C.F.R. § 5.410(i)(2) (1997); 24 C.F.R. §§ 903.7(o), 982.53(b), (c) (1999). Another set of regulations deals directly with residency preferences and includes requirements for prior HUD permission, 24 C.F.R. § 982.208(b) (1996); see also 24 C.F.R. § 5.410(h)(2) (1999), and various tests for such preferences, 24 C.F.R. § 982.207(b) (1999). Which of these provisions applies, what they mean, and how they may be affected by the 1998 statute remain to be addressed by the district court. Cf 64 Fed.Reg. 56,894, 56,900 (Oct. 21, 1999) • (discussing the effect of the statute).
This is a mare’s nest of problems, and nothing we have said should be regarded as more than a preliminary foray. However, despite our disagreement with the district court’s “balancing” rationale under the Fair Housing Act’s “because of race” provision, it is not clear to us whether the residency preferences at issue here will survive on remand. For that reason, we think that the district court is entitled to preserve the status quo for a limited period while it gives the plaintiffs a chance to develop alternative arguments not yet addressed by the district court. See SEC v. Lehman Bros., Inc., 157 F.3d 2, 9 (1st Cir.1998).
Accordingly, the injunction as to Avon, Holbrook and Middleborough is affirmed insofar as it prohibits these PHAs from distributing vouchers based upon the new lottery-determined lists using residency preferences; but within 90 days the prohibition must either be vacated by the district court or reinstituted based upon the grounds not yet considered by the district court, either side being able to appeal from an unfavorable disposition. The injunction against Abington and Rockland is affirmed insofar as it prevents them from distributing vouchers using the residency preference unless and until there is a satisfactory plan presented to and approved by the district court to ensure compliance with the 75 percent rule.
It is so ordered.