Opinion for the Court filed by Senior Circuit Judge WILLIAMS.
Opinion concurring in part and concurring in the judgment filed by Circuit Judge ROGERS.
WILLIAMS, Senior Circuit Judge:
The plaintiffs in this case are two fair housing organizations in New Orleans and five African-American homeowners. They claim that a program to help homeowners rebuild after hurricanes Katrina and Rita has employed a grant formula that violates the antidiscrimination provisions of the Fair Housing Act, particularly 42 U.S.C. § 3604(a). The program is administered by the Office of Community Development (“OCD”) of the State of Louisiana, an entity we treat interchangeably with its predecessor, the Louisiana Recovery Authority [1080]*1080(“LRA”). Defendant Robin Keegan is the Executive Director of the OCD. The U.S. Department of Housing and Urban Development (“HUD”) is a defendant in the suit but has taken no part in this appeal and cross-appeal.
Plaintiffs sought preliminary injunctive relief, which the district court initially denied on the ground that it was barred by state sovereign immunity under the Eleventh Amendment. Plaintiffs appealed. While the appeal was pending, the district court granted the plaintiffs’ later request for narrower injunctive relief. Defendant Keegan cross-appealed. Because it is plain that there are some potential remedies not running afoul of sovereign immunity, we have jurisdiction to address the substantive merits of plaintiffs’ claim (regardless of sovereign immunity’s possible applicability to the injunction initially sought). Doing so, we find that the plaintiffs’ showing on those merits is too weak to give them the likelihood of success required for a preliminary injunction. Accordingly, we affirm the denial of plaintiffs’ initially requested injunction and reverse the grant of their follow-up injunctive claim.1
In 2005 approximately 123,000 owner occupied homes in Louisiana were destroyed or severely damaged by hurricanes Katrina and Rita. Deferred Appendix (“D.A.”) 47. Using federal funds, Louisiana created the Road Home Homeowner Assistance Program to assist Louisianans displaced by the hurricanes. The program’s principal activity has been provision of grants to homeowners who wish to repair or rebuild and reoccupy their damaged homes. As of March 24, 2011, it had given 117,744 Louisiana homeowners almost $8 billion in grants for rebuilding. Office of Community Development, The Homeowner Assistance Program Weekly Situation and Pipeline Report Week 247, 1 (March 29, 2011) (“Weekly Situation and Pipeline Report”), http://www.road21a.org/Docs/pipeline/week 247pipeline.pdf. The program is supported by three special appropriations of Community Development Block Grants by Congress in 2005, 2006, and 2007. Pub.L. No. 109-148, 119 Stat. 2680, 2779-81 (Dec. 30, 2005); Pub.L. No. 109-234, 120 Stat. 418, 472-73 (June 15, 2006); Pub.L. No. 110-116, 121 Stat. 1295, 1343-44 (Nov. 13, 2007) (“2007 Act”).
The program’s grants fall into three categories. Option 1 grants are available to those Louisiana homeowners who plan to repair or rebuild their damaged homes. Homeowners receiving Option 1 grants must agree to a covenant requiring their rebuilt home to be owner-occupied for at least three years (the covenant runs with the house so that any buyer of the house is bound if the grant recipient sells the house). Options 2 and 3 relate to homeowners without plans to rebuild on their original homesites. This case concerns only grants made under Option 1.
OCD has calculated these grants by taking the lesser of the pre-Katrina value of the home and the cost-to-rebuild, subtracting the value of any FEMA or insurance proceeds received by the homeowner, and, if the homeowner carried no home insur[1081]*1081anee, applying a 30% penalty (i.e., 30% of the lower of value or cost-to-rebuild, minus any FEMA grant). D.A. 110. The resulting total is capped at $150,000. The formula has special provision for houses more than half (but not fully) destroyed and for homeowners who must raise the elevation of their homes in order to comply with new regulations. Homeowners with incomes that are no higher than 80% of the “area median income” may receive an Additional Compensation Grant (“ACG”) to cover any difference between the cost-to-rebuild and the amount of their compensation and elevation grants. ACGs were initially subject to a $50,000 limit, but that was removed before the district court heard plaintiffs’ motion for a preliminary injunction. The $150,000 cap applies to any recipient’s total grant, including any elevation grant or ACG. The various formulae used in awarding grants have been developed by the LRA and OCD and approved by HUD.
Plaintiffs brought suit against HUD and Keegan on behalf of all African-American homeowners in New Orleans who will have participated in the Road Home program by the first day of trial and had their grant amounts calculated on the basis of prestorm value of their homes. The parties agreed to defer briefing on class certification until after commencement of discovery, D.A. 38-39, and there has been no ruling on class certification. Plaintiffs alleged that the Option 1 formula violates the Fair Housing Act, 42 U.S.C. §§ 3604(a), 3605(a), 3608(d), (e)(5), and the Housing and Community Development Act of 1974, 42 U.S.C. § 5304(b)(2). Their theory is that the formula’s use of pre-Katrina house values as a grant ceiling had a disparate impact on African-American homeowners, who tend to live in New Orleans neighborhoods with lower property values than those in predominately white neighborhoods.
Factually, the disparate impact claim relies primarily on a 2008 study by PolicyLink, using data on the Road Home program provided by the LRA and the Louisiana Housing Finance Agency. Plaintiffs focus on a “resource gap” found by the study, which it defined as the difference between total resources available for rebuilding and the cost of rebuilding. PolicyLink calculated resources for rebuilding as including all grants under the Road Home program, plus money received from FEMA and the homeowner’s insurance proceeds. PolicyLink, A Long Way Home: the State of Housing Recovery in Louisiana 2008 (“PolicyLink”), at 39, http://policylink.info/threeyearslater/. For rebuilding costs, it used the same formulae that LRA used in calculating grant awards.
The study found that although African-American Option 1 grant recipients received larger average Road Home grants than white recipients, the “resource gap” was $39,082 on average for African-American grant recipients, compared with $30,863 on average for white grant recipients. PolicyLink at 42. The study suggested that much of the “resource gap” was driven by the element of the formula using home values as a ceiling on grants. Among homeowners who were paid under the prong of the formula looking to preKatrina home values, the average resource gap was nearly $50,000, compared to about $14,000 for homeowners who were paid under the formula’s cost-to-rebuild prong. PolicyLink at 43. The study also noted that “African American households and low-income households had lower prestorm home values than the average for closed
In October 2009 the state of Louisiana (with the consent of HUD) removed the $50,000 ceiling on ACGs for low income homeowners. D.A. 236. As a result homeowners whose income was no higher than 80% of the median became eligible for total grants up to the lesser of the full cost-to-rebuild or $150,000. Because the new formula made low income homeowners eligible for grants up to the cost-to-rebuild, or $150,000, regardless of the pre-Katrina value of their homes, it effectively removed, for low income homeowners, the part of the grant formula that considered homes’ pre-Katrina value. In effect, low income homeowners received the relief requested by plaintiffs. As of May 6, 2010, OCD had paid 9,301 grant recipients $326.5 million in additional ACGs. D.A. 193.
In June 2010 plaintiffs requested a temporary restraining order (“TRO”) and preliminary injunction that would enjoin Keegan “from taking any action to obligate, reallocate and/or spend any funds that, as of June 2, 2010, remain available to the LRA’s Road Home Homeowner Assistance program, and ... from taking any action to submit any request to the U.S. Department of Housing and Urban Development ... that would result in or cause further obligation, award, and/or disbursement of funds.... ” D.A. 68. The purpose of this preliminary injunction would have been to preserve a source of funds for the additional grants that plaintiffs (and the class of homeowners they purport to represent) would be entitled to if their lawsuit were to succeed. The proposed injunction would thwart LRA’s plans to use at least $100 million of available funds on a construction lending program to support loans to Road Home grant recipients who for various reasons had insufficient funds to rebuild and difficulty obtaining credit. D.A. 230-31.
The district court denied plaintiffs’ motion. Although it found that the plaintiffs were likely to be able to make a prima facie showing of disparate impact, it concluded that state sovereign immunity under the Eleventh Amendment, as interpreted by Edelman v. Jordan, 415 U.S. 651, 94 S.Ct. 1347, 39 L.Ed.2d 662 (1974), barred the relief. While Edelman finds “prospective” relief permissible under the Eleventh Amendment, it bars “retroactive” injunctive relief as an impingement on state sovereign immunity. See Edelman, 415 U.S. at 677, 94 S.Ct. 1347. Although the injunction sought would only hold payments in abeyance, plaintiffs sought it in aid of their claim to have grant awards recalculated for Option 1 applicants who had already been paid, and the district court ruled that a mandate of such payments would “effectively constitute an award of damages” for a past breach of a legal duty. Greater New Orleans Fair Housing Action Center v. U.S. Dep’t of Housing & Urban Development, 723 F.Supp.2d 1, 9 (D.D.C.2010).
Plaintiffs appealed the denial of this first preliminary injunction motion, arguing that the district court erred in its analysis of the implications of Edelman. They also filed a motion in the district court for a second TRO and preliminary injunction. This alternative relief was to bar OCD from using the house value element of the existing grant formula in any future grants, relief that did not, under the dis[1083]*1083trict court’s prior ruling, run afoul of Edelman’s bar on retrospective relief. Little was at stake, however. In her opposition to this second motion, defendant Keegan pointed out that as of July 29, 2010 there were only 141 African-American Option 1 grant applicants in Orleans Parish who were not eligible for an ACG and had not yet received any grants. Furthermore, of these 141 applications several did not meet program eligibility requirements, 54 had not replied to correspondence from OCD to indicate their continuing interest in the program for a substantial period of time, and over 80 had various deficiencies in their documentation. D.A. 856. Because the deadline for submitting initial Option 1 grant applications has already passed, there will not be any new applicants. 2007 Act, 121 Stat. at 1343, § 159(a).
The district court granted this second motion; because the second requested injunction would affect only the awards of future grantees, the relief sought would be permissible under Edelman and Ex Parte Young, 209 U.S. 123, 28 S.Ct. 441, 52 L.Ed. 714 (1908). Greater New Orleans Fair Housing Action Center v. U.S. Dep’t of Housing & Urban Development, 723 F.Supp.2d 11, 13 (D.D.C.2010). As before, it found that plaintiffs had shown a likelihood of success on the merits of their disparate impact claim, and in addition found the other elements of the conventional preliminary injunction formula tilting in plaintiffs’ favor. Id. at 13-14.
Defendant Keegan filed a cross-appeal of the district court’s ruling on the second motion for a TRO and preliminary injunction. She argues, among other things, that plaintiffs do not have standing and that the district court did not have jurisdiction to enter this preliminary injunction because jurisdiction over the matter had shifted to the court of appeals.
While the appeal and cross-appeal were pending before us, plaintiffs filed a motion for an injunction pending appeal. We granted that motion and enjoined Keegan “from committing Road Home funds to any new projects, such as the proposed construction lending program, pending disposition of this appeal.” Greater New Orleans Fair Housing Action Center v. U.S. Dep’t of Housing & Urban Development, No. 10-5257 (D.C.Cir. Sept. 22, 2010) (order granting injunction pending appeal). This ruling, in effect, gave plaintiffs temporarily the relief they sought in the first motion for a preliminary injunction.
In ruling on a preliminary injunction a key issue — often the dispositive one — is whether the movant has shown a substantial likelihood of success on the merits. See Munaf v. Geren, 553 U.S. 674, 690, 128 S.Ct. 2207, 171 L.Ed.2d 1 (2008). As we’ve seen, the district court resolved that issue against plaintiffs on the ground that the relief sought would violate Louisiana’s sovereign immunity. It ruled that requiring Louisiana to adjust the award of grant recipients who had already been paid under the Road Home program would be retrospective relief of the type barred by Edelman. The issue is in fact extremely complex, for several reasons: (1) Although the Road Home program is administered by a state agency, it is funded by the federal government; (2) Although Congress initially allowed the state agency a fairly wide discretion to deploy the funds— subject to HUD approval — in response to the Katrina and Rita disasters, the 2007 Act explicitly required them to be used “solely for the purpose of covering costs associated with otherwise uncompensated but eligible claims that were filed on or before July 31, 2007, under the Road Home program.” 2007 Act, 121 Stat. at 1343, § 159(a); (3) The OCD and defendant Keegan have changed the Road Home formula over time, giving additional [1084]*1084sums to prior recipients (leading plaintiffs to divide the grants into categories they call “initial” and “final”). And there is no precedent in our circuit applying Edelman to comparable facts. Other circuits have adopted greatly varying interpretations of Edelman where a state’s discretion over the funds at issue is limited. Compare Vaqueria Tres Monjitas Inc. v. Irizarry, 587 F.3d 464, 478-80 (1st Cir.2009) (holding that Edelman does not bar retroactive relief paid from special state trust fund), reh’g denied Vaqueria Tres Monjitas Inc. v. Irizarry, 600 F.3d 1 (1st Cir.2010); Bennett v. White, 865 F.2d 1395, 1408 (3rd Cir.1989) ' (permitting retrospective relief under Edelman to the extent that any funds paid by the state will be reimbursed by the United States), and Brown v. Porcher, 660 F.2d 1001, 1006-07 (4th Cir.1981) (holding that Edelman does not bar retrospective relief when the award is not paid from a state’s general revenue fund), with Paschal v. Jackson, 936 F.2d 940, 944 (7th Cir.1991) (declining to inquire as to the source of state funds), and Esparza v. Valdez, 862 F.2d 788, 794 (10th Cir.1988) (rejecting a rule that would look to whether “relief would be paid out of a general or special revenue fund”).
By contrast, we find the substantive merits relatively easy. But we can reach them only if permitted to do so by Steel Co. v. Citizens for a Better Environment, 523 U.S. 83, 118 S.Ct. 1003, 140 L.Ed.2d 210 (1998), which generally gives an absolute priority to jurisdictional questions. We conclude that Steel Co. presents no obstacle. We can perceive no good reason that Louisiana’s sovereign immunity might preclude all relief. Although defendant Keegan makes several wholly implausible arguments to the effect that the second motion for a preliminary injunction is barred by the Eleventh Amendment, these arguments rest on a fundamental confusion over the significance of Ex Parte Young, and they read Idaho v. Coeur d’Alene Tribe of Idaho, 521 U.S. 261, 117 S.Ct. 2028, 138 L.Ed.2d 438 (1997), as if it effectively overruled Ex Parte Young. At a minimum, it is clear that under Edelman the district court has jurisdiction to order LRA to use a different formula for future grantees. Thus, unlike the typical jurisdictional issue, the Edelman issue is not one that calls into question the authority of the court to hear plaintiffs’ claim on the merits. So it is clear that regardless of how the Eleventh Amendment might apply to the first preliminary injunction, the district court properly exercised jurisdiction over both the case as a whole and plaintiffs’ motion for a preliminary injunction, and that we may do the same.
We note for the record that, apart from situations like this where the only jurisdictional obstacle is to particular varieties of relief, there is considerable uncertainty about sequencing in the Eleventh Amendment context. The Supreme Court has found it permissible, “indeed appropriate,” to decide the issue of whether a statute provides for suits against the states before determining whether, if it did, the Eleventh Amendment would permit such a suit. Vermont Agency of Natural Resources v. United States ex rel. Stevens, 529 U.S. 765, 780, 120 S.Ct. 1858, 146 L.Ed.2d 836 (2000). It reasoned that the statutory issue was “logically antecedent to the existence of the Eleventh Amendment question,” id. (internal quotations omitted), and that the court, in resolving that issue would not “pronounce upon any issue, or upon the rights of any person, beyond the issues and persons that would be reached under the Eleventh Amendment inquiry anyway” (as it might in resolving questions such as whether the statute permitted private causes of action generally). Id. Thus the “combination of logical priority and virtual coincidence of scope makes it possible, and indeed appropriate, to decide the statutory issue first.” Id. at 779-80, 120 [1085]*1085S.Ct. 1858. See also United States ex rel. Long v. SCS Business & Technical Institute, Inc. 173 F.3d 890, 893-98 (D.C.Cir.1999). But at least one circuit is ready to consider merits before passing on an Eleventh Amendment issue on the grounds that state sovereign immunity, unlike ordinary subject matter jurisdiction issues such as standing, may be waived by the state and, if not raised by the state, may but need not be considered by the court sua sponte. Parella v. R.I. Employees’ Retirement System, 173 F.3d 46, 53-57 (1st Cir.1999); see also Greenless v. Almond, 277 F.3d 601, 607-08 (1st Cir.2002); cf. United States ex rel. Long, 173 F.3d at 893-94 (noting several Supreme Court hints that the Eleventh Amendment was no more than “quasi-jurisdictional”). But see Vermont Agency, 529 U.S. at 779, 120 S.Ct. 1858 (noting that personal jurisdiction “is an essential element of the jurisdiction of a district court, without which the court is powerless to proceed to an adjudication”). But because here the Eleventh Amendment is at most a barrier to some possible remedies and the district court’s jurisdiction over the case as a whole is clear, we can consider the merits first, without sorting out the sequencing issue for other contexts.
In evaluating likelihood of success on the substantive merits, we review the district court’s legal conclusions de novo. Davis v. PBGC, 571 F.3d 1288, 1291 (D.C.Cir.2009). Contrary to the district court, we conclude that plaintiffs’ disparate impact claim lacks merit.
Section 3604(a) of the Fair Housing Act makes it unlawful “To refuse to sell or rent after the making of a bona fide offer, or to refuse to negotiate for the sale or rental of, or otherwise make unavailable or deny, a dwelling to any person because of race, color, religion, sex, familial status, or national origin.” 42 U.S.C. § 3604(a). We have not decided whether this permits disparate impact claims. 2922 Sherman Avenue Tenants’ Association v. District of Columbia, 444 F.3d 673, 679 (D.C.Cir.2006). The argument that it does rests largely on the point that the language is parallel to that of Title VII of the Civil Rights Act of 1964, 42 U.S.C. § 2000e-2, under which a plaintiff can prevail by showing either disparate impact or disparate treatment. See 2922 Sherman Avenue Tenants’ Association, 444 F.3d at 679. Each of the eleven circuits that have resolved the matter has found the disparate impact theory applicable under the Fair Housing Act. Id. As the matter has not been briefed and its resolution is unnecessary to the outcome, however, we shall, as in 2922 Sherman Avenue Tenants’ Association, simply assume the availability of the disparate impact theory. We note, however, that no court has applied disparate impact theory to the formulae used administering a grant program.
Similarly, we need not decide as between the two leading tests for applying disparate impact, the four-factor balancing test of the Seventh Circuit, Metropolitan Housing Development Corp. v. Village of Arlington Heights, 558 F.2d 1283, 1290 (7th Cir.1977), or the burden-shifting test of the Second Circuit, Huntington Branch, NAACP v. Town of Huntington, 844 F.2d 926, 935-36 (2nd Cir.1988), aff'd on other grounds, 488 U.S. 15, 18, 109 S.Ct. 276, 102 L.Ed.2d 180 (1988). As in 2922 Sherman Avenue Tenants’ Association, the success of plaintiffs’ claim doesn’t turn on the details of the legal test; either approach requires proof of disproportionate impact, measured in some plausible way, 444 F.3d at 679-80, and plaintiffs have not offered such proof.
When presenting a disparate impact claim, a plaintiff must generally “demonstrate with statistical evidence that the [1086]*1086practice or policy has an adverse effect on the protected group.” Garcia v. Johanns, 444 F.3d 625, 633 (D.C.Cir.2006) (evaluating a disparate impact claim under the Equal Credit Opportunity Act); see also Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 987, 108 S.Ct. 2777, 101 L.Ed.2d 827 (1988). “The correct inquiry is whether the policy in question had a disproportionate impact on the minorities in the total group to which the policy was applied.” Betsey v. Turtle Creek Associates, 736 F.2d 983, 987 (4th Cir.1984); see also Darensburg v. Metropolitan Transportation Commission, No. 09-15878, 636 F.3d 511, 2011 WL 540592 (9th Cir.2011) (evaluating a disparate impact claim alleging discriminatory allocation of mass transit funds). In this case, the policy — the grant formula that called for using the lesser of preKatrina home value and cost to repair— was applied to all applicants for Option 1 Road Home grants. The question, then, is whether the formula as a whole had a disparate impact on African-American grant applicants.
We pause to consider examples of what is precluded by a focus on the effects of the formula as a whole. In any state where African-American and white homeowners have significantly different economic profiles, it will presumably be the case that particular elements of a complex formula such as that of the Road Home program will have a disproportionate negative impact on African-Americans, an impact potentially offset by other elements of the formula. For example, assuming that use of the pre-Katrina value of home as one element of the formula favors white homeowners on average, the PolicyLink study invoked by plaintiffs tells us that African-American homeowners received less money from insurance on average, PolicyLink at 38, so that the formula’s deduction of insurance proceeds from the grant appears to favor African-Americans. Similarly, the $150,000 cap on total grants would seem to disfavor wealthier (and therefore, according to the PolicyLink study, disproportionately white) grant recipients. It seems unlikely that an agency could fashion a grant formula in this context without some element having a disproportionate impact on a protected group. (For the moment we defer consideration of the consequences if, as under Title VII, whites are also acceptable plaintiffs.)
For the same reason, attention to the formula’s effects as a whole bars our looking at the effects of the grant formula only in Orleans Parish. If the economic profiles of racial groups differ from parish to parish in the parts of Louisiana affected by hurricanes Katrina and Rita, then a grant formula that has non-disparate racial effects for Louisiana as a whole can easily have a disparate impact on African-American residents in at least some individual parishes (not to mention smaller geographic units). To allow plaintiffs to pick a special subset of the affected localities to test for disparate impact would, just like allowing them to single out of the effects of a single formula element, expose almost any grant formula to litigation. Although plaintiffs focus much of their case on Orleans Parish, we must consider the impact on Louisiana as a whole.
Apart from these obvious limits, it is hard to see just how one may affirmatively identify a sound benchmark for assessing the disparateness of a grant formula’s impact (or, to put the problem slightly differently, to say what a non-disparate impact would look like). Plaintiffs urge us to focus on the “resource gap” facing Option 1 grant recipients, that is, the difference between the cost-to-rebuild and the “total resources” of an applicant — the latter being defined (by them and the PolicyLink study) as the grantees’ total receipts from insurance, FEMA and the Road Home program. They point to the PolicyLink [1087]*1087finding of a “resource gap” between African-American and white grant recipients of slightly over $8,000 ($39,082 on average for African-American grant recipients, compared with an average of $30,863 for white grant recipients). PolicyLink at 42.
Even if we accepted “resource gap” as the appropriate measure, which we do not, plaintiffs’ evidence is problematic at best. The PolicyLink study predates the removal of the $50,000 cap on ACGs (which in turn predated the district court injunction proceeding), creating what OCD calls the Additional ACG program. That removal has had considerable impact. By May 6, 2010 (before the district court ruled on the first motion for a TRO and preliminary injunction), OCD had paid 9,301 grant recipients $326.5 million in Additional ACGs. D.A. 193. According to the most recent data, OCD has now disbursed $466,313,036.52 to 13,267 grant recipients. Weekly Situation and Pipeline Report at 5. This is more than 10% of Option 1 grants to low income recipients and represents an average Additional ACG of over $35,000. Id. We do not know the fraction of these grants received by African-Americans. But if African-American grant recipients are more likely to qualify as low income and are more likely to face a resource gap, as the PolicyLink study suggests, they would be eligible substantially more often, and for substantially larger average Additional ACGs, than would white recipients. Thus removal of the ceiling on ACGs casts grave doubt on plaintiffs’ claim even under their “resource gap” theory.
At oral argument, plaintiffs’ counsel responded to the removal of the $50,000 cap by arguing that it would have no effect among recipients altogether ineligible for ACGs, i.e., those with incomes higher than 80% of “area median income.” But this sort of analytical cherry-picking seems to us no more valid than the same practice that we rejected as applied to individual elements of the formula or to geographic units.
In any event, we find the “resource gap” an inappropriate benchmark. The “resource gap” measures neither the resources a recipient receives under the Road Home program, nor a grant recipient’s ability to obtain housing, nor the extent to which the Road Home program compensates a recipient for hurricane-inflicted losses. It measures only the extent to which grant receipts, FEMA receipts and insurance receipts fall short of the cost-to-rebuild a house of the size that the grant recipient owned before the hurricanes. This is a completely artificial metric. It represents an unprincipled mishmash of considerations in tension with one another. The “resource gap” metric gestures at measuring need by calculating shortfalls from the cost of rebuilding rather than size of the total grant. But the cost-to-rebuild element rests on the size of applicant’s former house, rather than assuming (as would an egalitarian concern for need) an equal cost of housing for all. And the treatment of insurance seems completely anomalous. It makes homeowners who bothered to insure their homes appear to be the recipients of some sort of government munificence, while in fact their grants have been reduced on account of their precautions (a reduction presumably made because OCD’s resources were finite and because the justifications for the “collateral source” rule are absent here).
An alternative focus would be the total value of Road Home grants received, or the value of Road Home grants as a proportion of a grantee’s uncompensated losses. But on this criterion, the plaintiffs’ own evidence is fatal to their claim. It indicates that African-American grant recipients on average have received larger total awards than did white grant recipi[1088]*1088ents. PolicyLink at 38 (“African American and elderly households received higher average grants because a higher percentage of them received additional compensation grants, which were available to low-income individuals.”). If anything, this discrepancy has become greater since the PolicyLink study, by virtue of removal of the $50,000 cap on ACGs.
In short, then, plaintiffs’ facts are at best sketchy even on the implausible resource-gap theory, and do not begin to show a violation on the size-of-grant metric.
Alternative benchmarks are clearly conceivable. In rejecting the resource-gap theory and in tentatively applying the size-of-grant theory, we do not mean to endorse size-of-grant as necessarily suitable. Choice of a benchmark is further complicated by uncertainty whether one need consider only the impact on minority groups. As Title VII permits white employees to bring job discrimination claims, McDonald v. Santa Fe Trail Transportation Co., 427 U.S. 273, 279-80, 96 S.Ct. 2574, 49 L.Ed.2d 493 (1976), white grant recipients might, on the size-of-grant standard, be able to make a prima facie case of disparate impact. See also Jordan v. Khan, 969 F.Supp. 29, 30-31 (N.D.Ill.1997) (permitting white plaintiff to bring racial discrimination claim under the Fair Housing Act); Miller v. Toume Oaks East Apartments, 797 F.Supp. 557, 561 (E.D.Tex.1992) (same). In the face of this legal uncertainty, adoption of the size-of-grant benchmark would put OCD — and other agencies trying to develop formulae for comparable grants — in a damned-if-you-do, damned-if-you-don’t quandary.
Thus the data offered by plaintiffs to show a disparate impact fall short. The numbers by no means establish that the OCD formula actually in effect at the time of the district court proceeding created results amounting to a violation even under plaintiffs’ resource-gap theory, which in any event we reject. On the less implausible size-of-grant standard, the evidence flatly contradicts their claim. As a result, plaintiffs have not shown a likelihood of success on the merits. While the criteria governing grant of a preliminary injunction include other elements (the balance of harms to the parties and the effect on the public interest), see, e.g., Davis v. PBGC, 571 F.3d at 1291, and we review the district court’s balance of factors for abuse of discretion, id., there is no need to remand to the district court. When a plaintiff has not shown a likelihood of success on the merits, there is no need to consider the remaining factors. See Arkansas Dairy Cooperative Ass’n, Inc. v. U.S. Dep’t of Agriculture, 573 F.3d 815, 832 (D.C.Cir.2009); Apotex, Inc. v. FDA 449 F.3d 1249, 1253 (D.C.Cir.2006).
We therefore affirm the district court’s denial of the initially requested injunction.
We turn now to plaintiffs’ second motion for a TRO and preliminary injunction. The district court granted this motion several weeks after plaintiffs’ notice of appeal of its first injunction ruling. Defendants appeal this decision on several grounds, including jurisdictional ones — that plaintiffs lack standing and that their appeal of the denial of the first motion for a TRO and preliminary injunction divested the district court of jurisdiction over the matter. Under Steel Co. we would normally be required to address these jurisdictional questions before ruling on the substantive merits, but this case fits well within an exception recognized by Steel Co. If the outcome on the merits is “foreordained” by another ruling, such that “the pretermission of the jurisdictional question [is not used] as a device for reaching a question of law that otherwise would have gone unaddressed,” then a court may turn to [1089]*1089the merits without considering the jurisdictional issue. Steel Co., 523 U.S. at 98, 118 S.Ct. 1003. In ruling on the first motion for a preliminary injunction, we found that plaintiffs’ likelihood of success on the merits of their underlying claim was slim. Although the scope of relief requested differs, the second motion implicates the exact same disparate impact claim and thus shares the first motion’s fatal defect. And, as noted above, when a plaintiff has not shown a likelihood of success on the merits, we need not consider the other factors. Our ruling on the first motion is therefore fully dispositive of the second motion. Accordingly, we reverse the district court’s grant of the second motion.
As set forth in our March 29, 2011 judgment, we affirm the district court’s denial of plaintiffs’ first motion for a TRO and preliminary injunction and reverse the district court’s grant of plaintiffs’ second motion. We also, as previously ordered, vacate the injunction we had issued pending appeal. Finally, we remand the case for proceedings not inconsistent with this opinion.
2. PolicyLink appears to use the term "closed” for those who had received their Road Home grants.