concurring in the judgment.
The District Court’s preliminary injunction remains reviewable because of its continuing effect on the city’s personnel policies. That injunction states that the city may “not apply the seniority policy proposed insofar as it will decrease the percentage of black [persons] in the Memphis Fire Department.”1 Thus, if the city faces a need to lay off Fire Department employees in the future, it may not apply its seniority system. I cannot say that the likelihood that the city will once again face the need to lay off Fire Department employees is so remote that the city has no stake in the outcome of this litigation.2
In my judgment, the Court’s discussion of Title VII is wholly advisory. These cases involve no issue under Title VII; they involve only the administration of a consent decree. The District Court entered the consent decree on April 25, 1980, after having given all parties, including all of the petitioners in this Court, notice and opportunity to object to its entry. The consent decree, like any other final judgment of a district court, was immediately appealable. See Carson v. American Brands, Inc., 450 U. S. 79 (1981). No appeal was taken. Hence, the consent decree became a final judgment binding upon those who had had notice and opportunity to *591object; it was and is a legally enforceable obligation. If the consent decree justified the District Court’s preliminary injunction, then that injunction should be upheld irrespective of whether Title VII would authorize a similar injunction.3 Therefore, what governs these cases is not Title VII, but the consent decree.4
There are two ways in which the District Court’s injunction could be justified. The first is as a construction of the consent decree. If the District Court had indicated that it was merely enforcing the terms of the consent decree, and had given some indication of what portion of that decree it was interpreting, I might be hard pressed to consider the entry of the injunction an abuse of discretion. However, the District Court never stated that it was construing the decree, nor did it provide even a rough indication of the portion of the decree on which it relied. There is simply nothing in the record to justify the conclusion that the injunction was based on a reasoned construction of the consent decree.5
*592The second justification that could exist for the injunction is that the District Court entered it based on a likelihood that it would modify the decree, or as an actual modification of the decree.6 As Justice Blackmun explains, post, at 607, 610-611, modification would have been appropriate if respondents had demonstrated the presence of changed circumstances. However, the only “circumstance” found by the District Court was that the city’s proposed layoffs would have an adverse effect on the level of black employment in the fire department. App. to Pet. for Cert, in No. 82-206, pp. A73-A76. This was not a “changed” circumstance; the percentage of blacks employed by the Memphis Fire Department at the time the decree was entered meant that even then it was apparent that any future seniority-based layoffs would have an adverse effect on blacks. Thus the finding made by the District Court was clearly insufficient to support a modification of the consent decree, or a likelihood thereof.
Accordingly, because I conclude that the District Court abused its discretion in entering the preliminary injunction at issue here, I concur in the judgment.
See also n. 6, infra. There were actually three injunctive orders entered by the District Court, each applying to different positions in the Memphis Fire Department. All use substantially the same language.
In this respect, this litigation is similar to Los Angeles v. Lyons, 461 U. S. 95, 100-101 (1983). There, an injunction against the use of choke-holds by the city’s police department was held not to be moot despite the fact that the police board had instituted a voluntary moratorium of indefinite duration on chokeholds, since the likelihood that the city might one day wish to return to its former policy was not so remote as to moot the case. See also Carroll v. Princess Anne, 393 U. S. 175, 178-179 (1968).
The Court seems to suggest that a consent decree cannot authorize anything that would not constitute permissible relief under Title VII. Ante, at 578-579. I share Justice Blackmun’s doubts as to whether this is the correct test. See post, at 611, n. 9, 614-616. The provisions on which the Court relies, 42 U. S. C. §§ 2000e-2(h) and 2000e-5(g), merely state that certain seniority arrangements do not violate Title VII, and define the limits of appropriate relief for a Title VII violation, respectively. They do not place any limitations on what the parties can agree to in a consent decree. The Court does not suggest that any other statutory provision was violated by the District Court. The Court itself acknowledges that the administration of a consent decree must be tested by the four corners of the decree, and not by what might have been ordered had respondents prevailed on the merits, ante, at 574, which makes its subsequent discussion of Title VII all the more puzzling.
If the decree had been predicated on a finding that the city had violated Title VII, the remedial policies underlying that Act might be relevant, at least as an aid to construction of the decree. But since the settlement expressly disavowed any such finding, the Court’s exposition of Title VII law is unnecessary.
Justice Blackmun explains, post, at 607-610, how the consent decree could be construed to justify the injunction. I find nothing in the record *592indicating that this is the theory the District Court actually employed. While I recognize that preliminary injunction proceedings are often harried affairs and that district courts need substantial leeway in resolving them, it nevertheless remains the case that there must be something in the record explaining the reasoning of the District Court before it may be affirmed. That is the purpose of Federal Rule of Civil Procedure 65(d)’s requirement that “[ejvery order granting an injunction and every restraining order shall set forth the reasons for its issuance . . . .”
It seems likely that this second justification was the actual basis for the entry of the injunction. The District Court's phrasing of the question it faced was whether “it should exercise its authority to modify a Consent Decree,” App. to Pet. for Cert. A73. The focus of the Court of Appeals’ opinion reviewing the preliminary injunction was the “three grounds upon which a consent decree may later be modified,” 679 F. 2d 541, 560 (CA6 1981). Most important, the practical effect of the District Court’s action indicates that it should be treated as a modification. Until it is reviewed, it will effectively govern the procedure that the city must follow in any future layoffs, and that procedure is significantly different from the seniority system in effect when the consent decree was negotiated and signed. *593Mootness. “The usual rule in federal cases is that an actual controversy must exist at stages of appellate or certiorari review, and not simply at the date the action is initiated.” Roe v. Wade, 410 U. S. 113, 125 (1973). In the absence of a live controversy, the constitutional requirement of a “case” or “controversy,” see U. S. Const., Art. Ill, deprives a federal court of jurisdiction. Accordingly, a case, although live at the start, becomes moot when intervening acts destroy the