This interlocutory appeal arises from lawsuits originally filed by the States of California and New Hampshire in their respective state courts against corporations that manufactured, refined, marketed, or distributed gasoline containing methyl tertiary butyl ether (“MTBE”). Refiners have added MTBE to some gasoline since the late 1970s in order to enhance its octane content. The use of MTBE significantly increased after 1990, when Congress established the Reformulated Gasoline Program (“RFG Program”) as part of its amendments to the Clean Air Act (“CAA”). See 42 U.S.C. § 7545(k). Until 2005, when the Clean Air Act was again amended, the RFG Program required the use of reformulated gasoline containing at least 2 percent chemical oxygen by weight in certain metropolitan areas with high summertime smog levels. See id. § 7645(k)(2)(B) (2000). States with less severe smog problems may opt in to the RFG Program. Id. § 7545(k)(6). Pursuant to regulations promulgated by the Environmental Protection Agency (“EPA”) in 1991, MTBE is one of several different oxygenates that may be used to certify gasoline as reformulated. 40 C.F.R. § 80.46(g)(2)(i). These provisions of the RFG Program took effect in 1995. 42 U.S.C. § 7545(k)(5). The CAA amendments also created the Oxygenated Fuels Program (“Oxyfuel Program”), which requires gasoline containing 2.7 percent oxygen by weight during winter months in areas that do not meet national air quality standards for carbon monoxide. 42 U.S.C. § 7545(m)(2).
The complaints allege that MTBE contaminated public drinking water supplies through discharges, leaks, overfills, and spills from gasoline delivery facilities, as well as from the release of gasoline in certain consumer and commercial activities. When released into the environment, MTBE can render water undrinkable by giving it a foul taste and odor, and, at high concentrations, it poses a risk to human health. Because of its chemical properties, *115MTBE dissolves easily in water, does not biodegrade, and can disperse quickly through a water supply, reaching areas far from its initial release. This makes MTBE more difficult and expensive to remove from groundwater than other gasoline constituents that also pose a threat to the environment and human health. The complaints allege various theories of liability, under the laws of California and New Hampshire, to hold the defendants responsible for the damage caused by MTBE.
These two cases are among scores of related actions removed from state court and transferred to the Southern District of New York by the Judicial Panel on Multi-district Litigation pursuant to MDL No. 1358, In re Methyl Tertiary Butyl Ether (“MTBE”) Products Liability Litigation. Before considering California’s and New Hampshire’s motions to remand, the district court had already denied similar motions by plaintiffs in other related cases. In two decisions, the district court held that the cases were removable under either (1) the federal officer removal statute, 28 U.S.C. § 1442, because the defendants had acted at the direction of a federal agency in adding MTBE to gasoline, In re Methyl Tertiary Butyl Ether (“MTBE”) Prods. Liab. Litig., 342 F.Supp.2d 147, 156-58 (S.D.N.Y.2004) (“MTBE III”), or (2) the bankruptcy removal statute, 28 U.S.C. § 1452, because defendant Texaco, Inc. (now ChevronTexaco Corp.) had earlier filed for and been discharged from bankruptcy, In re Methyl Tertiary Butyl Ether (“MTBE”) Prods. Liab. Litig., 341 F.Supp.2d 386, 413-14 (S.D.N.Y.2004) (“MTBE V”). In the latter case, the district court also held that subject matter jurisdiction could not be predicated on the basis of a substantial federal question or complete preemption. Id. at 402-03, 406-11.
The district judge asked California and New Hampshire to brief the then-undecided question of sovereign immunity. She then held that “the removal of cases filed by State Plaintiffs does not violate principles of sovereign immunity.” In re Methyl Tertiary Butyl Ether (“MTBE”) Prods. Liab. Litig., 361 F.Supp.2d 137, 148 (S.D.N.Y.2004) (“MTBE VI”). This interlocutory appeal followed. Two issues are presented: (1) whether principles of sovereign immunity are violated when a state plaintiff voluntarily prosecutes a claim in state court and the action is removed from state to federal court pursuant to a statute that expressly authorizes removal; and (2) if not, whether the district court had subject matter jurisdiction over this matter under the federal officer removal statute, 28 U.S.C. § 1442, the bankruptcy removal statute, 28 U.S.C. § 1452, or some other ground. Because the subject matter jurisdiction of the district court, unlike the issue of sovereign immunity, is not normally the proper subject of an interlocutory appeal, we must also determine whether we may even reach the second issue.
DISCUSSION
We review the district court’s legal conclusions de novo. Cooper v. New York State Office of Mental Health, 162 F.3d 770, 773 (2d Cir.1998) (sovereign immunity); Bechtel v. Competitive Techs, Inc., 448 F.3d 469, 471 (2d Cir.2006) (subject matter jurisdiction).
I. Sovereign Immunity
The Eleventh Amendment provides that “[t]he Judicial power of the United States shall not be construed to extend to any suit in law or equity, commenced or prosecuted against one of the United States by Citizens of another State, or by Citizens or Subjects of any Foreign State.” U.S. Const, amend. XI. The question present*116ed here is whether a suit filed by a state in its own courts, and then removed to federal court, is a suit “commenced or prosecuted against” that state.
Early in its history, the Supreme Court answered that question in the negative. Speaking through Chief Justice Marshall, the Court held in Cohens v. Virginia, 19 U.S. (6 Wheat.) 264, 5 L.Ed. 257 (1821), that the character of the parties to a lawsuit does not change when the forum in which the suit is heard does:
To commence a suit, is to demand something by the institution of process in a Court of justice; and to prosecute the suit, is, according to the common acceptation of language, to continue that demand. By a suit commenced by an individual against a State, we should understand process sued out by that individual against the State, for the purpose of establishing some claim against it by the judgment of a Court; and the prosecution of that suit is its continuance. Whatever may be the stages of its progress, the actor is still the same.... If a suit, brought in one Court, and carried by legal process to a supervising Court, be a continuation of the same suit, then this suit is not commenced nor prosecuted against a State. It is clearly in its commencement the suit of a State against an individual, which suit is transferred to [federal] Court, not for the purpose of asserting any claim against the State, but for the purpose of asserting a constitutional de-fence against a claim made by a State.
Id. at 408-09. As a result, the Chief Justice concluded, “[t]he [Eleventh] amendment ... extended to suits commenced or prosecuted by individuals, but not to those brought by States.” Id. at 407. While the particular issue in Cohens was whether the Eleventh Amendment barred Supreme Court review of a judgment obtained by a state against an individual in state court, the holding is equally applicable here.
Nevertheless, state sovereign immunity is broader than, and independent of, the immunity provided by the Eleventh Amendment. See Alden v. Maine, 527 U.S. 706, 713, 119 S.Ct. 2240, 144 L.Ed.2d 636 (1999); Idaho v. Coeur d’Alene Tribe of Idaho, 521 U.S. 261, 267, 117 S.Ct. 2028, 138 L.Ed.2d 438 (1997). Indeed, since Hans v. Louisiana, 134 U.S. 1, 10 S.Ct. 504, 33 L.Ed. 842 (1890), the Supreme Court has recognized that state sovereign immunity is not confined by the text of the Eleventh Amendment, but is part of the constitutional structure. California and New Hampshire argue that these broader principles are offended when a suit voluntarily prosecuted by a state is removed to federal court.
Specifically, California argues that Hans raises a presumption against proceedings or suits that were “anomalous and unheard of’ at the time the Constitution was adopted, and that the removal of cases brought by states from state courts was unheard of at that time. Cal. Br. at 15-20. The holding in Hans, however, was that “[t]he suability of a state, without its consent, was a thing unknown to the law.” 134 U.S. at 16, 10 S.Ct. 504. While it is true that, prior to the adoption of the Constitution, removal of a case from state court was “unheard of,” this was because “each State had complete and exclusive authority to administer by its courts all the law, civil and criminal, which existed within its borders. Its judicial power extended over every legal question that could arise.” Tennessee v. Davis, 100 U.S. 257, 267, 25 L.Ed. 648 (1880). In other words, the constitutional framework for removal — including the existence of a federal judiciary — did not exist.
Nevertheless, there has been some provision for the removal of cases from *117state to federal courts ever since the original Judiciary Act of 1789. This strong historical foundation has combined with inferences drawn from Article III, Section 2 of the Constitution— the grant to Congress of the power to regulate the business of the federal courts — to put the constitutionality of removal jurisdiction beyond question.
14B Charles Alan Wright, Arthur R. Miller, & Edward H. Cooper, Federal Practice and Procedure § 3721, at 288 (3d ed.1998). Indeed, the first federal officer removal statute expressly contemplating the removal of suits brought by states— the predecessor to 28 U.S.C. § 1442 — was enacted during the War of 1812 and signed into law by James Madison. 3 Stat. 195, 198-99 (1815); see generally Willingham v. Morgan, 395 U.S. 402, 405-06, 89 S.Ct. 1813, 23 L.Ed.2d 396 (1969) (tracing the long history of the modern federal officer removal statute).
In Tennessee v. Davis, the Supreme Court expressly upheld the constitutionality of one of the early incarnations of the federal officer removal statute. 100 U.S. at 261. Davis, an internal revenue collector for the federal government, was fired upon while attempting to seize an illegal distillery under the authority of the federal revenue laws. He returned fire, killing one man, and was subsequently indicted for murder by a state grand jury. Davis sought to have his criminal prosecution removed to federal court under a federal statute permitting the removal of “any civil suit or criminal prosecution ... commenced in any court of a State against any officer appointed under, or acting by authority of, any revenue law of the United States, now or hereafter enacted, or against any person acting by or under authority of any such officer, on account of any act done under color of his office or of any such law, or on account of any right, title, or authority claimed by such officer or other person under any such law.” Id. (internal quotation marks omitted).
The Supreme Court held that the case was properly removed. Addressing the argument that this violated the state’s sovereignty, the Court found no offense in requiring the state, in these limited circumstances, to prosecute its case in a federal forum:
The argument so much pressed upon us, that it is an invasion of the sovereignty of a State to withdraw from its courts into the courts of the general government the trial of prosecutions for alleged offences against the criminal laws of a State, even though the defence presents a case arising out of an act of Congress, ignores entirely the dual character of our government. It assumes that the States are completely and in all respects sovereign. But when the national government was formed, some of the attributes of State sovereignty were partially, and others wholly, surrendered and vested in the United States. Over the subjects thus surrendered the sovereignty of the States ceased to extend. Before the adoption of the Constitution, each State had complete and exclusive authority to administer by its courts all the law, civil and criminal, which existed within its borders. Its judicial power extended over every legal question that could arise. But when the Constitution was adopted, a portion of that judicial power became vested in the new government created, and so far as thus vested it was withdrawn from the sovereignty of the State. Now the execution and enforcement of the laws of the United States, and the judicial determination of questions arising under them, are confided to another sovereign, and to that extent the sovereignty of the State is restricted. The removal of cases arising under those laws, from State into Feder*118al courts, is, therefore, no invasion of State domain.
Id. at 266-67. This holding is a complete answer to the argument that the Hans presumption precludes removal here.
More recently, the Supreme Court has reaffirmed the proposition that “ ‘the constitutional powers of Congress to authorize the removal of criminal cases for alleged offences against State laws from State courts to the ... courts of the United States, when there arises a Federal question in them, is as ample as its power to authorize the removal of a civil case.’ ” Mesa v. California, 489 U.S. 121, 128, 109 S.Ct. 959, 103 L.Ed.2d 99 (1989) (quoting Davis, 100 U.S. at 271). Other Supreme Court cases support this conclusion. In Ames v. Kansas, 111 U.S. 449, 4 S.Ct. 437, 28 L.Ed. 482 (1884), the Supreme Court considered the removal to federal court of a civil action to block a corporate merger filed by a state, in its state court, against corporations and corporate directors. The Court framed the question before it in a way clearly applicable to the present cases:
The ... question we have to consider is, therefore, whether suits cognizable in the courts of the United States on account of the nature of the controversy, and which need not be brought originally in the [S]upreme [C]ourt, may now be brought in or removed to the circuit courts without regard to the character of the parties. All admit that the act does give the requisite jurisdiction in suits where a State is not a party, so that the real question is, whether the Constitution exempts the States from its operation.
Id. at 470. Relying on Cohens v. Virginia, the Court rejected the state’s claim. Instead, it held that “[t]he argument would have great force if urged to prove that this court could not establish the demand of a citizen upon his State, but is not entitled to the same force, when urged to prove that this court cannot inquire whether the Constitution or laws of the United States protect a citizen from a prosecution instituted against him by a State.” Id. (internal quotation marks omitted).
Somewhat more recently, in Illinois v. City of Milwaukee, 406 U.S. 91, 92 S.Ct. 1385, 31 L.Ed.2d 712 (1972), the Court held that a federal district court, not the Supreme Court, had original jurisdiction over a suit brought by a state against political subdivisions of another state. Sovereign immunity was not at issue, but the Court nevertheless found Ames “controlling” and held that when a state sues a party that is not also a state, those suits “ ‘may now be brought in or removed to the Circuit Courts (now the District Courts) without regard to the character of the parties.’ ” Id. at 101, 92 S.Ct. 1385 (quoting Ames, 111 U.S. at 470, 4 S.Ct. 437). While Ames and City of Milwaukee arose under circumstances that differ from the present cases, their express holdings that cases in which a state is the plaintiff may be brought in or removed to federal court are equally applicable here. See California ex rel. Lockyer v. Dynegy, Inc., 375 F.3d 831, 846-47 (9th Cir.2004).
Of course, Davis and Ames must be read in the context of other cases recognizing a state’s broad right to sovereign immunity. While much Eleventh Amendment jurisprudence traditionally has been focused on concerns that a judgment against a state would require it to pay money damages from the public fisc, interfere with public administration, or compel a state to act or restrain from acting, see Pennhurst State Sch. & Hosp. v. Halderman, 465 U.S. 89, 101 n. 11, 104 S.Ct. 900, 79 L.Ed.2d 67 (1984), many of the Supreme Court’s recent decisions on sovereign immunity have placed particular emphasis on the dignity and respect owed *119states as separate sovereigns in our federal system, see, e.g., Alden, 527 U.S. at 709, 119 S.Ct. 2240 (concluding that congressional authorization of suits against states in state courts would be offensive to the “respect and dignity due them as residuary sovereigns”); Coeur d’Alene Tribe, 521 U.S. at 268, 117 S.Ct. 2028 (observing that suits against states in federal court jeopardize “the dignity and respect afforded a State”); Seminole Tribe of Florida v. Florida, 517 U.S. 44, 58, 116 S.Ct. 1114, 134 L.Ed.2d 252 (1996) (stating that the Eleventh Amendment serves to avoid the “indignity of subjecting a State to the coercive process of judicial tribunals at the instance of private parties” (internal quotation marks omitted)); but see Cohens, 19 U.S. at 406 (“We must ascribe the [Eleventh] amendment ... to some other cause than the dignity of a State.”).
All of these recent opinions, however, concerned suits against states that forced them to defend actions prosecuted by private parties. Suits commenced by states stand on a different footing. The removal of the cases here was the result of the voluntary acts of California and New Hampshire in commencing the lawsuits against the defendants. Once having done so, these states subjected themselves to all of the rules and consequences attendant to that decision. Indeed, in the foreign sovereign immunity context, there is authority for the proposition that “when the sovereign seeks recovery, it [is] subject to legitimate counterclaims against it.” Banco Nacional de Cuba v. Sabbatino, 376 U.S. 398, 438, 84 S.Ct. 923, 11 L.Ed.2d 804 (1964); accord Nat’l City Bank of New York v. Rep. of China, 348 U.S. 356, 364, 75 S.Ct. 423, 99 L.Ed. 389 (1955) (“It is recognized that a counterclaim based on the subject matter of a sovereign’s suit is allowed to cut into the doctrine of immunity.”). Because the defendants have not asserted counterclaims, we need not decide whether these holdings also apply to the states in our federal system. Suffice it to say, the removal of these cases will not subject the states to the indignity or fiscal pain of being subject to a money judgment or an injunction — the core concerns underlying the sovereign immunity to which they are entitled.
Our holding today that sovereign immunity does not preclude the removal to federal court of a suit filed by a state plaintiff in state court is consistent with that reached by the vast majority of courts to have considered the issue. In Dynegy, California brought an action against several energy companies alleging violations of state law, and the defendants removed the case to federal court. The state moved to remand on the basis of sovereign immunity. In a carefully researched and reasoned opinion, the Ninth Circuit concluded, as we do today, that “history gives little indication that sovereign immunity was ever intended to protect plaintiff states. Rather, it plainly understands sovereign immunity as protection from being sued.” 375 F.3d at 847; accord City & County of San Francisco v. PG & E Corp., 433 F.3d 1115, 1123 (9th Cir.2006). The Tenth Circuit reached the same conclusion in a state tort action removed to federal court, though it confined its discussion to the Eleventh Amendment, not broader principles of sovereign immunity. See Oklahoma ex rel. Edmondson v. Magnolia Marine Transp. Co., 359 F.3d 1237, 1239-40 (10th Cir.2004) (holding that “the Eleventh Amendment’s abrogation of federal judicial power ‘over any suit ... commenced or prosecuted against one of the United States’ does not apply to suits commenced or prosecuted by a State”). The Fifth Circuit and the Federal Circuit have reached similar conclusions, albeit in cases presenting different procedural postures. See Huber, Hunt & Nichols, Inc. v. Archi*120tectural Stone Co., 625 F.2d 22, 24 n. 6 (5th Cir.1980) (stating, in a case brought against the State of Louisiana, that “[o]f course, the eleventh amendment is inapplicable where a state is a plaintiff.... ”); Regents of the Univ. of California v. Eli Lilly & Co., 119 F.3d 1559, 1564 (Fed.Cir.1997) (denying Eleventh Amendment challenge to transfer of ease brought by a state from one federal district court to another). Numerous district court decisions have also held that the Eleventh Amendment and/or state sovereign immunity does not bar removal of cases filed by a state plaintiff. See, e.g., Virginia v. Bulgartabac Holding Group, 360 F.Supp.2d 791, 796 (E.D.Va.2005); In re Rezulin Prods. Liab. Litig., 133 F.Supp.2d 272, 297 (S.D.N.Y.2001); In re Texas, 110 F.Supp.2d 514, 530-31 (E.D.Tex.2000); Regents of Univ. of Minnesota v. Glaxo Wellcome, Inc., 58 F.Supp.2d 1036, 1040 (D.Minn.1999); Kansas ex rel. Stovall v. Home Cable, Inc., 35 F.Supp.2d 783, 789 (D.Kan.1998); Vermont v. Oncor Commc’ns, Inc., 166 F.R.D. 313, 321 (D.Vt.1996); South Dakota State Cement Plant Comm’n v. Wausau Underwriters Ins. Co., 778 F.Supp. 1515, 1522 (D.S.D.1991).
We are aware of only two cases to have concluded otherwise. In California v. Steelcase Inc., 792 F.Supp. 84 (C.D.Cal.1992), the district court held that the Eleventh Amendment barred removal of an action filed by a state plaintiff because it rendered the state “an involuntary party to an action in federal court.” Id. at 86. Moore ex rel. State of Mississippi v. Abbott Laboratories, Inc., 900 F.Supp. 26 (S.D.Miss.1995), reached the same conclusion, relying entirely on the reasoning of Steelcase. Id. at 30. In their brief analy-ses, however, both courts ignored the text of the Eleventh Amendment insofar as it applied only to cases “commenced or prosecuted against” a state, as well as the voluntary nature of the states’ continued participation in those cases. Moreover, whatever force these cases had is undermined by the fact that Steelcase has been overruled by the Ninth Circuit. See Dynegy, 375 F.3d at 849 n. 15.
Our holding is not inconsistent with Thomas v. FAG Bearings Corp., 50 F.3d 502 (8th Cir.1995), upon which both states rely. In that case, a defendant corporation, sued in federal court under the Comprehensive Environmental Response, Compensation, and Liability Act, the Resource Conservation and Recovery Act (“CERC-LA”), and common law causes of action, sought to join the Missouri Department of Natural Resources as an involuntary party plaintiff based on the agency’s prior announcement that it intended to sue the defendant for the same equitable relief as the citizen plaintiffs. In so doing, the defendant hoped to avoid the possibility that two separate lawsuits would result in multiple or inconsistent judgments. Id. at 504. The Eighth Circuit found that join-der would violate the Eleventh Amendment because it would force the state — to its prejudice — to undertake proceedings against the defendant at a time and place not of its own choosing. Id. at 505. Specifically, the Eighth Circuit observed that the state agency had “not yet completed the regulatory process required by CERC-LA and the Missouri environmental statute.” Id. at 505 n. 10. Premature litigation, it wrote, could prejudice the state in a variety of ways, e.g., by barring it from bringing claims under other environmental statutes in later litigation and by limiting the costs it could seek to recover from the defendant. Id. at 505 n. 10, 506. These consequences would strike at the very heart of the Eleventh Amendment by undermining the state’s asserted “autonomy in decision-making” and harming the public fisc. Id. at 505-06.
*121The present cases, however, do not implicate the significant concerns that motivated the holding in Thomas. Unlike the state agency in Thomas, which had not yet decided to bring suit against the defendant — indeed, it had not even completed its investigation into the polluted site at issue — here, California and New Hampshire have each made and acted upon the decision to commence a lawsuit. This voluntary act subjects them to the consequences that Congress may legitimately attach to such an action. Thus, we conclude that sovereign immunity does not bar the removal of these state-commenced actions to federal court. The question we turn to then is whether the removal of these cases has, in fact, been authorized by Congress.
II. Removal Jurisdiction
A. Jurisdiction to Hear Appeal
While the district court’s denial of the state plaintiffs’ claims of sovereign immunity is appealable under the collateral order doctrine, Puerto Rico Aqueduct & Sewer Auth. v. Metcalf & Eddy, Inc., 506 U.S. 139, 147, 113 S.Ct. 684, 121 L.Ed.2d 605 (1993), the denial of a motion to remand is generally not the proper subject of an interlocutory appeal, see 28 U.S.C. § 1291; Chicago, Rock Island & Pac. R.R. Co. v. Stude, 346 U.S. 574, 578, 74 S.Ct. 290, 98 L.Ed. 317 (1954); Wilkins v. Am. Export-Isbrandtsen Lines, Inc., 401 F.2d 151, 151 (2d Cir.1968). Nevertheless, California and New Hampshire ask us to rule that the district court erred in holding that these cases were removable under the federal officer and bankruptcy removal statutes.
We conclude that review of this question is required pursuant to our independent obligation to satisfy ourselves of the jurisdiction of this court and the court below. “On every writ of error or appeal, the first and fundamental question is that of jurisdiction, first, of this court, and then of the court from which the record comes. This question the court is bound to ask and answer for itself, even when not otherwise suggested, and without respect to the relation of the parties to it.” Great S. Fire Proof Hotel Co. v. Jones, 177 U.S. 449, 453, 20 S.Ct. 690, 44 L.Ed. 842 (1900); accord Steel Co. v. Citizens for a Better Env’t, 523 U.S. 83, 94-95, 118 S.Ct. 1003, 140 L.Ed.2d 210 (1998); Arizonans for Official English v. Arizona, 520 U.S. 43, 73, 117 S.Ct. 1055, 137 L.Ed.2d 170 (1997); FW/PBS, Inc. v. City of Dallas, 493 U.S. 215, 231, 110 S.Ct. 596, 107 L.Ed.2d 603 (1990); Bender v. Williamspori Area Sch. Dist., 475 U.S. 534, 541, 106 S.Ct. 1326, 89 L.Ed.2d 501 (1986).
This obligation is well-founded. “Subject-matter limitations on federal jurisdiction serve institutional interests. They keep the federal courts within the bounds the Constitution and Congress have prescribed. Accordingly, subject-matter delineations must be policed by the courts on their own initiative even at the highest level.” Ruhrgas AG v. Marathon Oil Co., 526 U.S. 574, 583, 119 S.Ct. 1563, 143 L.Ed.2d 760 (1999). Indeed, we have often taken it upon ourselves to determine whether removal jurisdiction existed even where that issue was not itself appealed. See, e.g., Barbara v. New York Stock Exch., Inc., 99 F.3d 49, 53 (2d Cir.1996) (“Although neither party raises this issue on appeal, we are under an obligation to do so sua sponte.”); Mignogna v. Sair Aviation, Inc., 937 F.2d 37, 40 (2d Cir.1991) (“The parties have not raised the issue of removal jurisdiction on this appeal, but it is our obligation to do so sua sponte.”); see also Sykes v. Texas Air Corp., 834 F.2d 488, 492 n. 16 (5th Cir.1987) (“When the district court decides to retain a case in the face of arguments that it lacks jurisdic*122tion, the decision itself is technically unre-viewable; but of course the appellate court reviewing any other aspect of the case must remand for dismissal if the refusal to remand was wrong, i.e., if there is no federal jurisdiction over the case.”).
This obligation is not extinguished because an appeal is taken on an interlocutory basis and not from a final judgment. See Merritt v. Shuttle, Inc., 187 F.3d 263, 268 (2d Cir.1999). Indeed, we relied on Merritt in our order in this case denying Citgo Petroleum Corp.’s motion to limit the scope of the appeal to the issue of sovereign immunity, see Berisha v. BP Amoco Corp., Nos. 04-5974-cv (L), 04-6056-cv (Con) (2d Cir. June 27, 2005), and we do so again today. Because subject matter jurisdiction goes uniquely to the fundamental power of the federal courts to hear a ease, there is no reason why an appellate court should potentially compound an error of the district court by assuming it has jurisdiction. See Steel Co., 523 U.S. at 101-02, 118 S.Ct. 1003 (“For a court to pronounce upon the meaning or the constitutionality of a state or federal law when it has no jurisdiction to do so is, by very definition, for a court to act ultra vires.”). Indeed, there are sound policy reasons for tackling the issue at this stage, as “[i]t is surely in the interest of judicial economy to determine now whether the federal courts continue to have subject matter jurisdiction over [a] controversy [subject to an interlocutory appeal].” Lamar Adver. of Penn, LLC v. Town of Orchard Park, 356 F.3d 365, 372-73 (2d Cir.2004).
Moreover, this obligation is not undermined by the Supreme Court’s holding in Swint v. Chambers County Commission, 514 U.S. 35, 115 S.Ct. 1203, 131 L.Ed.2d 60 (1995). In that case, the Supreme Court held that, when a federal appellate court reviews a district court decision on an interlocutory basis, it may not at the same time review unrelated questions that are not themselves entitled to expedited consideration. Id. at 51, 115 S.Ct. 1203. Specifically, the Court held that, while an order denying a claim of qualified immunity was the proper subject of an interlocutory appeal, another order denying a motion for summary judgment was not an appealable collateral order under the test announced in Cohen v. Beneficial Industrial Loan Corp., 337 U.S. 541, 69 S.Ct. 1221, 93 L.Ed. 1528 (1949). The Court held open the possibility, however, that pendent appellate jurisdiction would be proper if the two rulings were “inextricably intertwined” or if “review of the former decision was necessary to ensure meaningful review of the latter.” Swint, 514 U.S. at 51, 115 S.Ct. 1203. While we have previously suggested that those two phrases mean “essentially the same thing,” Rein v. Socialist People’s Libyan Arab Jamahiriya, 162 F.3d 748, 758 (2d Cir.1998) (Calabresi, J.), we have also left open the possibility that the two phrases are not redundant, see Lamar Adver., 356 F.3d at 371 n. 7.
Nevertheless, there are several reasons why Swint is inapposite. Because the district court in Swint had subject matter jurisdiction over the collateral order on appeal, the Supreme Court did not have occasion to address whether its holding applied to the special subset of cases in which the district court did not have subject matter jurisdiction. Indeed, the issue decided in that case was whether a litigant had the right to appeal the collateral order, not whether the appellate court had the right to address an issue, such as subject matter jurisdiction, on its own initiative. Under these circumstances, we are not inclined to assume that Swint decided a significant issue that it had no reason to consider. See Jenkins v. Dela*123ware, 395 U.S. 213, 216, 89 S.Ct. 1677, 23 L.Ed.2d 253 (1969).
In any event, the present cases satisfy the exception acknowledged in Swint for those circumstances where the appealed interlocutory order cannot be meaningfully reviewed without consideration of a related ruling not itself appealable or where the two issues are inextricably intertwined. Specifically, in Merritt, we held that the review of a collateral order “would be meaningless if the district court was without jurisdiction over that claim in the first instance.” 187 F.3d at 269. This broad holding plainly compels the conclusion that the review of the issue of sovereign immunity would be meaningless if the district court was without subject matter jurisdiction over these actions. More than this, however, the appealable order relating to sovereign immunity and the collateral issue of removal jurisdiction are significantly related. Our holding rejecting the sovereign right asserted by California and New Hampshire to prosecute these cases in courts of their own choosing is predicated on the assumption that federal law authorizes the removal of these cases. Under these circumstances, the ultimate resolution of the sovereign immunity claim turns on whether these cases were properly removed to federal court. The relationship between these two issues is analogous to that between the issues presented in interlocutory appeals by public officials from claims of qualified immunity in cases alleging the violation of a constitutional right. In those cases, we routinely exercise pendent appellate jurisdiction over the issue of whether a constitutional right has been violated, because “the merits of a constitutional claim generally are inextricably intertwined with qualified immunity.... [W]e must determine whether a constitutional right has been violated before deciding whether the right was clearly established.” Demoret v. Zegarelli, 451 F.3d 140, 152 (2d Cir.2006). Similarly, except as an abstract matter, we cannot resolve the defense of sovereign immunity here without determining whether the removal statutes confer subject matter jurisdiction on the district court.
The defendants suggest that we may not determine the jurisdiction of the district court because we do not have appellate jurisdiction “over the entire case.” Defs.’ Br. at 31. While it is true that our jurisdiction is limited to the subject matter of the appeal — we may not, for instance, reach the merits of the case — we unquestionably have jurisdiction to decide whether there is any further impediment to our exercise of jurisdiction over the subject matter of the case and, by extension, the jurisdiction of the court below. The defendants also suggest that we have, on many occasions, “been willing to forego inquiry into the district court’s jurisdiction where the question was not constitutional or where the merits were foreordained, suggesting that such review is not mandatory.” Defs.’ Br. at 33 (citing Dotson v. Griesa, 398 F.3d 156, 177 (2d Cir.2005); In re Arbitration Between Monegasque de Reassurances S.A.M. v. Nak Naftogaz of Ukraine, 311 F.3d 488, 497-98 (2d Cir.2002); Ctr. for Reprod. Law & Policy v. Bush, 304 F.3d 183, 194 (2d Cir.2002); United States v. Miller, 263 F.3d 1, 4 n. 2 (2d Cir.2001); Fama v. Comm’r of Corr. Servs., 235 F.3d 804, 817 n. 11 (2d Cir.2000)). Setting aside the fact that the issue of the propriety of removal cannot so easily be separated from the claims of sovereign immunity, the defendants misstate the import of our decisions. In each of these cases, our holding on the merits directed the dismissal of the complaints— the same result that would have followed had we decided that subject matter jurisdiction was lacking. In the present cases, however, the denial of the plaintiffs’ claims *124of sovereign immunity would not result in the dismissal of the complaints; thus, we have an obligation to determine whether the district court has subject matter jurisdiction to go forward.
In sum, while an order denying a motion to remand may normally be reviewed only after entry of a final judgment, our obligation to remain assured of our jurisdiction and that of the court below remains intact. Therefore, in ruling in this interlocutory appeal, we must determine whether there is a proper basis for removal of these cases.
B. Bases for Removal
The district court recognized two bases for removal of these and other related state court lawsuits to federal court: (1) the federal officer removal statute, for cases in areas covered by the RFG Program and in adjacent “spillover” areas, MTBE III, 342 F.Supp.2d at 156-58; MTBE V, 341 F.Supp.2d at 400, and (2) the bankruptcy removal statute, MTBE V, 341 F.Supp.2d at 414. In determining whether jurisdiction is proper, we look only to the jurisdictional facts alleged in the Notices of Removal. See Colorado v. Symes, 286 U.S. 510, 518-19, 52 S.Ct. 635, 76 L.Ed. 1253 (1932) (“The burden is upon him who claims the removal plainly to set forth by petition made, signed, and unequivocally verified by himself all the facts relating to the occurrence, as he claims them to be, on which the accusation is based.”); see also Mesa, 489 U.S. at 131—34, 109 S.Ct. 959 (requiring averment of a colorable federal defense in the notice of removal). The Supreme Court has held that “statutory procedures for removal are to be strictly construed,” Syngenta Crop Prot., Inc. v. Henson, 537 U.S. 28, 32, 123 S.Ct. 366, 154 L.Ed.2d 368 (2002), and we have held that out of respect for the limited jurisdiction of the federal courts and the rights of states, we must “resolv[eJ any doubts against removability.” Somlyo v. J. Lu-Rob Enters., Inc., 932 F.2d 1043, 1045-46 (2d Cir.1991).
1. Federal Officer Removal
The federal officer removal statute, in relevant part, permits the removal of cases commenced in state court against “any officer (or any person acting under that officer) of the United States or of any agency thereof, sued in an official or individual capacity for any act under color of such office....” 28 U.S.C. § 1442(a)(1). A defendant who is not a federal officer or agency must satisfy three elements to have a suit against it removed under this statute. First, it must show that it is a “person” within the meaning of the statute. Second, it must establish that it was “acting under” a federal officer, which subsumes the existence of a “ ‘causal connection’ between the charged conduct and asserted official authority.” Willingham, 395 U.S. at 409, 89 S.Ct. 1813. Finally, the defendant must raise a colorable federal defense. Mesa, 489 U.S. at 129, 109 S.Ct. 959.
Because it is clear that corporations are “persons” within the meaning of the statute, see 1 U.S.C. § 1; Ryan v. Dow Chem. Co., 781 F.Supp. 934, 946 (E.D.N.Y.1992), we turn to the requirement that the defendants have been “acting under” a federal officer. Although the district court considered this question to be separate from that of the existence of a causal connection between the conduct in question and the federal direction, MTBE III, 342 F.Supp.2d at 154, they tend to collapse into a single requirement: that “the acts that form the basis for the state civil or criminal suit were performed pursuant to an officer’s direct orders or to comprehensive and detailed regulations.” Ryan, 781 F.Supp. at 947. “Critical under the stat*125ute is to what extent defendants acted under federal direction at the time they were engaged in the conduct now being sued upon.” Id. at 946 (citation and internal quotation marks omitted). Removal will not be proper where a private party establishes only that the acts complained of were performed under the “general auspices” of a federal officer. Id. at 947. “Likewise, the mere fact that a corporation participates in a regulated industry is insufficient to support removal absent a showing that the particular conduct being sued upon is closely linked to detailed and specific regulations.” Id.; see also Bakalis v. Crossland Sav. Bank, 781 F.Supp. 140, 145 (E.D.N.Y.1991) (describing the need for some government intervention or control, other than that contemplated by a generally applicable regulatory scheme, as “regulation plus”).
The line between the absence and the presence of “direct control” by a federal officer over a private party is a fine one, depending heavily on the facts of each case and the particular conduct giving rise to the plaintiffs cause of action. For example, in Ryan, Judge Weinstein considered whether seven companies that had manufactured the cancer-causing defoliant Agent Orange for the United States Armed Forces could remove the actions from state to federal court. He first noted that government officers exercised a high degree of control over the defendants’ production and delivery processes. 781 F.Supp. at 938. Judge Weinstein then observed, however, that the defendants were being sued on the theory that Agent Orange was improperly designed and produced, and relied on the assumption that Agent Orange “was a mix of pre-existing chemical formulae that had long been put to domestic commercial use.” Id. at 950. On this basis, he concluded that the defendants had not met the requirements for removal under section 1442:
They are being sued for formulating and producing a product all of whose components were developed without direct government control and all of whose methods of manufacture were determined by the defendants. Although the defendants later produced and delivered Agent Orange under the control of federal officers, these subsequent acts are distinct from the earlier acts of product and manufacturing design being sued upon. The government sought only to buy ready-to-order herbicides, not to cause, control, or prevent the production of the unwanted byproduct, dioxin, which is the alleged cause of plaintiffs’ injuries. The necessary direct and detailed official control over the acts for which the defendants are now being sued is therefore lacking.
Id.
In a subsequent Agent Orange case, a different court found that removal was proper in light of a more developed record that undermined the factual basis for Judge Weinstein’s decision and evidenced the more direct control of federal officers. In Winters v. Diamond Shamrock Chemical Co., 901 F.Supp. 1195 (E.D.Tex.1995), aff'd, 149 F.3d 387 (5th Cir.1998), the district court found that “the Defense Department expressly directed chemical companies to provide a detailed mixture of chemicals known as Agent Orange” and that the defendants were “compelled under threat of criminal sanctions” to deliver it. Id. at 1199. The Fifth Circuit cited those facts in affirming the action’s removal. 149 F.3d at 398; see also Miller v. Diamond Shamrock Co., 275 F.3d 414 (5th Cir.2001). On its own review of the record, the Fifth Circuit also observed that, while the defendant manufacturers had always diluted the active ingredients in Agent Orange with inert chemicals before commercial use, it was the Defense De*126partment that required the defendants to produce Agent Orange comprising only the undiluted active ingredients for use in Vietnam. 149 F.3d at 399. This combination of factors—the government’s detailed specifications, the defendants’ compulsion to provide the product to those specifications, and the government’s oversight over the manufacturing process—caused the court to determine that removal was appropriate. Id. at 399-400. Indeed, relying on these facts, Judge Weinstein would later conclude that his earlier decision in Ryan was wrong and that removal was proper. See In re “Agent Orange” Prod. Liab. Litig., 304 F.Supp.2d 442, 449-50 (E.D.N.Y.2004).
Adopting these standards for determining when a private party is “acting under” a federal officer for the purpose of removal jurisdiction, we now turn to the reasons stated by the district court in sustaining the removal of the complaints here. The district judge began her analysis with three observations. First, the defendants alleged in their Notices of Removal that “the Clean Air Act, and regulations promulgated thereunder, require them to blend oxygenates into gasoline.” MTBE III, 842 F.Supp.2d at 156. Second, “defendants allege that at the time the Clean Air Act was amended and the seven oxygenates were approved for use, both Congress and the EPA were aware that defendants would have to use MTBE in order to comply with the Act’s requirements.” Id. Third, “[although the EPA has identified seven additives that may be used to meet these requirements, MTBE is the only approved oxygenate that is available in quantities sufficient to comply with the Act and regulations.” Id. On this basis, the district judge concluded, “defendants have sufficiently alleged that they added MTBE to gasoline at the direction of the EPA, a federal agency.” Id. We cannot agree.
The conclusion of the district judge is not based on an explicit directive in either the Clean Air Act or its implementing regulations. Significantly, after oral argument in this appeal, the district judge held that “federal law did not require the use of MTBE.” In re Methyl Tertiary Butyl Ether (“MTBE") Prods. Liab. Litig., 457 F.Supp.2d 324, 335 (S.D.N.Y.2006) (denying summary judgment to defendants on issue of preemption). Indeed, the district judge and the defendants acknowledged that the EPA identified six other additives, besides MTBE, that could be blended into reformulated gasoline to meet the requirements imposed by the CAA and the regulations promulgated thereunder. MTBE III, 342 F.Supp.2d at 156. Thus, although the district judge was correct in her first observation—that the defendants alleged they were required to blend oxygenates into gasoline—because this allegation lacks a foundation in law, it is unavailing in supporting removal. Moreover, the complaints here are not predicated simply on the fact that the defendants blended oxygenates into gasoline. Instead, they focus on the use of one particular oxygenate, MTBE, and the harm caused by the release of gasoline containing MTBE into the states’ water supplies.
Under these circumstances, we understand the district court to have found removal proper on the premise that, if the defendants were compelled to use a particular additive—in this case MTBE—be-cause there was no sufficiently available alternative to meet CAA requirements, it would demonstrate that the defendants were acting pursuant to the direct orders of a federal agency, at least where, at the time the regulation was promulgated, the agency was aware of these market forces. We assume for present purposes that such market compulsion would satisfy the federal officer removal statute. Nevertheless, *127the district judge somewhat overstated the language in the Notices of Removal when she observed that the defendants alleged that “Congress and the EPA were aware that defendants would have to use MTBE in order to comply with the Act’s requirements.” MTBE III, 342 F.Supp.2d at 156. The Notices of Removal do not so allege. Nor do they allege that “MTBE is the only approved oxygenate that is available in quantities sufficient to comply with the Act and regulations.” Id. On their faces, the Notices of Removal, which are virtually identical for the present purposes, come closest when they allege that “Congress also knew, when it enacted these oxygenate mandates, that the industry would have to blend MTBE into at least some of the gasoline sold in OFP and RFG areas in order to comply with the program requirements.” N.H. Notice of Removal ¶ 19; Cal. Notice of Removal ¶ 19. The Notices of Removal also allege — without citation to any supporting document — that, at some point in time, the EPA “knew there was not a sufficient capacity of ethanol, on a nationwide basis, nor sufficient infrastructure for ethanol, to comply with the regulation.” N.H. Notice of Removal ¶ 33; Cal. Notice of Removal ¶ 34.
The allegation as to what Congress “knew” is based entirely on the floor statements by a handful of individual legislators rather than explicit findings of fact in the legislation itself. Because such “[fjloor debates reflect at best the understanding of individual Congressmen,” Zuber v. Allen, 396 U.S. 168, 186, 90 S.Ct. 314, 24 L.Ed.2d 345 (1969), they are of little value even in construing ambiguous statutes. The issue here, however, does not turn on the construction of statutory language. Instead, the statements in the floor debates are offered to prove what Congress as a whole understood would be the practical consequences of the statute it was discussing. On this score, they do not constitute competent evidence.
Nevertheless, while we decide this appeal without reliance on the floor debate, the statements made during its course are useful in evaluating the defendants’ arguments that they were victims of a course of action that they were forced by Congress to undertake. Some background is necessary. While the RFG Program required only 2 percent oxygen content for gasoline in areas with too much summertime smog, the Oxyfuel Program imposed a higher standard during the winter months in areas with too much carbon monoxide. The bill introduced in the Senate, passed out of committee and endorsed by the White House, originally provided that gasoline subject to the Oxyfuel Program contain 3.1 percent oxygen, not 2.7 percent oxygen. While ethanol could be used to reach that 3.1 percent threshold, MTBE alone could not. In order to overcome this impediment to the use of MTBE as an oxygenate, and ostensibly to give flexibility to state and local officials, Senator Lautenberg introduced an amendment on the floor reducing the oxygen-content requirement from 3.1 percent to 2.7 percent.
The floor debate on which the defendants rely is derived principally from the colloquy that ensued after Senator Lauten-berg introduced his amendment. Indeed, their evidence of congressional intent is composed largely of statements made by several farm-belt Senators defending the 3.1 percent standard. While Senator Daschle, for example, is quoted by the defendants as saying that he wanted MTBE, along with other oxygenates, “to play a role in achieving a variety of national objectives,” he did so in the context of opposing an amendment that would have encouraged significantly greater use of that additive at the expense of ethanol. See 136 Cong. Rec. S 2280, 2289 (Mar. 7, *1281990). Indeed, any reasonable reading of the floor debate establishes, if anything, that many of the Senators engaged in the debate believed that ethanol production would expand to meet increased demand and that the 3.1 percent standard would operate to the severe detriment of those who manufactured or wished to use MTBE. Otherwise, the Lautenberg Amendment and the debate it engendered were pointless.
More than this, the statements made during the floor debate, if credited, support the premise that many of the defendants actually lobbied Congress for a lower oxygen-content requirement that would make it possible for them to use more MTBE — a chemical created from natural gas and byproducts of the gasoline-refining process — and less ethanol. As Senator Grassley, one of the farm-state legislators, observed:
[B]y dropping the standard down to 2.7 percent, we will actually be cutting ethanol and American farmers out of a major market.
So who will really gain by changing from 3.1 to 2.7 percent? The answer is obvious! Big oil!! Big oil will gain because those companies control the methanol and MTBE. The [bjig oil companies make up much of the membership of the so-called [CJonsumers for [CJompetitive [FJuels which supports this amendment. Big consumers like Shell, Mobil, Exxon, Phillips, Amoco, Chevron, just to name a few.
The debate over this amendment, therefore, really boils down to this: If you want to help big oil maintain its lock on our fuel supply and if you want to weaken the Clean Air Act, then vote to reduce the 3.1-percent requirement down to 2.7 percent.
On the other hand, if you want to improve our air, and provide the Federal Government with some net savings, then vote to retain the 3.1-percent requirement. Because with 3. 1, you are going to encourage the use of ethanol — a proven pollution fighter produced by American farmers, not big oil.
136 Cong. Rec. S 2290, 2295-96 (Mar. 7, 1990) (emphasis added). Likewise, Senator Dole added:
The amendment is an antienvironmen-tal, big oil bailout, and I am surprised anyone would think it should be given serious consideration to a bill which is supposed to help our environment....
Make no mistake, this is a big oil amendment. The major international oil companies control foreign natural gas reservoirs, and they are looking for a market — your gas tank and the gas tanks of your constituents. By voting for the Lautenberg amendment, you will be closing an opportunity for a clean-burning, domestically produced product from our motor fuels market, leaving nothing to compete against oil company-owned feedstocks. If you vote for this amendment, do not complain about high gasoline prices, because you voted against competition.
Id. at 2297. Perhaps the most candid comments on this issue came from Senator Simpson, who explained his vote by stating:
Mr. President, I find myself in disagreement with my fine leader [Senator Dole] and some of my farm State colleagues on this one. I would like to see all types of oxygenates used to reduce carbon monoxide and other pollutants— for I believe they are useful fuel additives and are very necessary in nonat-tainment areas.
We often react to issues based on parochial concerns. Here is such a case. *129I find that I must support the amendment offered by my able colleague, the Senator from New Jersey. There are many small refiners in my State that produce MTBE or methyl tertiary butyl ether and a major corporation has just announced they are planning to build a new MTBE plant in my State and that they will be making MTBE from natural gas. Then we also have the potential to make methanol from coal and natural gas and we have great reserves of those products in my State. I also have farmers to represent and they would like to produce ethanol as well, but not on the same scale as the MTBE producers.
So, I do support the amendment to change the oxygenate number to 2.7. This is assuredly a tough issue, but I must support my State on this one.
Id. at 2299. While the legislation that emerged from the House-Senate Conference suggests that the large oil companies ultimately obtained the result they allegedly sought, namely, the reduction of the oxygenation requirement to 2.7 percent, this proves little more than that they achieved the goal of ensuring that MTBE would be used widely as an oxygenate. It does not establish, however, that the defendants were required to use MTBE or that, if MTBE were not used, ethanol producers would have been unable to meet the expected demand in the few years before the fuel-oxygenation requirements took effect.
Nor are the defendants aided by the allegation that, at some unspecified time, the EPA “knew there was not a sufficient capacity of ethanol, on a nationwide basis, nor sufficient infrastructure for ethanol, to comply with the regulation.” N.H. Notice of Removal ¶ 33; Cal. Notice of Removal ¶ 34. First, apart from what the EPA may have understood about the supply of ethanol at the time the regulations were promulgated, the Notices of Removal do not allege that there was, in fact, an insufficient supply of alternative oxygenates, or that the EPA entertained such an understanding. Second, and more importantly, this allegation does not support the conclusion that had the defendants decided to use ethanol (or some other oxygenate) instead of MTBE, it would have been impossible for the supply of ethanol to have grown to meet that demand before the EPA regulations implementing the CAA amendments took effect. While the defendants allege that “the notices’ allegations of impossibility cover this easily,” Defs.’ Br. at 53 n. 13, we do not find this specific allegation in the Notices of Removal, and the defendants do not cite any particular paragraph where it appears. Nor do they cite any later-filed affidavits which could provide the basis for us to treat the removal petitions as amended to include those allegations. See Willingham, 395 U.S. at 407 n.3, 89 S.Ct. 1813 (treating removal petition as if it had been amended to include information contained in later-filed affidavits). Indeed, although we rely only on the language of the Notices of Removal, the plaintiffs argue persuasively, and without contradiction, that the defendants’ argument that ethanol capacity was insufficient is “a classic example of confusing the cart with the horse: ethanol supplies were low because the oil industry chose not to use it, not the other way around.” N.H. Br. at 28.
Against this backdrop, the other allegations in the Notices of Removal are similarly unavailing. Even if the Notices of Removal were sufficient to establish that Congress and the EPA may have expected that the defendants would use MTBE, they leave unstated the reason for that expectation. Of course, constraints on the supply of other oxygenates constitute one plausible explanation, but it is also possible that Congress and the EPA based their *130expectation on price differences among the oxygenates or the fact that the major oil companies also controlled the natural gas from which MTBE is derived. That it may have been more convenient or less expensive for the defendants to use MTBE does not mean it would have been impossible for them to use other, less-polluting additives. Moreover, most of the facts alleged by the defendants in their Notices of Removal only indicate that MTBE would be one of several oxygenates used, and many of the floor statements cited by the defendants put ethanol and ethyl tertiary butyl ether, another oxygentate, on equal footing with MTBE with respect to the means by which gasoline refiners would comply with the terms of the CAA. In sum, the defendants have not met their burden of providing “candid, specific and positive” allegations, Willingham, 395 U.S. at 408, 89 S.Ct. 1813 (internal quotation marks omitted), that they were acting under federal officers when they added MTBE, and not some approved alternative, to their reformulated gasoline.
Moreover, the defendants’ reliance on Watson v. Philip Moms Cos., 420 F.3d 852 (8th Cir.2005), cert. granted, — U.S. -, 127 S.Ct. 1055, 166 L.Ed.2d 797 (2007) (No. 05-1284), is misplaced. In that case, users of so-called “light” cigarettes sued a cigarette manufacturer in state court under a state law barring deceptive trade practices, alleging that it had designed its cigarettes to deliver more tar and nicotine to smokers than its use of the terms “light” and “lowered tar and nicotine” would suggest. Specifically, pursuant to a voluntary agreement with the Federal Trade Commission (“FTC”) entered into in 1970, Philip Morris and other cigarette manufacturers had used a standard testing mechanism designed by the FTC to measure the amount of tar and nicotine in its products and had displayed the results in all of its advertisements. This filter test was known to be imperfect, as it did not accurately measure the tar and nicotine inhaled by a person under real-world conditions. Rather, it was intended to provide a standard methodology for comparing the relative amounts of these substances present in different varieties of cigarettes. In addition, the FTC permitted manufacturers to advertise a cigarette as “light” or “low tar” if the tar collected using the FTC’s testing apparatus fell below a certain level. The plaintiffs contended that Philip Morris manipulated the design and content of its “light” cigarettes so that they would score well on the FTC test, even though they were as or more harmful than regular cigarettes.
Philip Morris removed the case from state to federal court on the ground that it was acting under a federal officer for the purposes of cigarette testing and advertising. The district court denied the plaintiffs’ motion to remand, and the Eighth Circuit affirmed, holding that Philip Morris was acting at the direction of the FTC. Specifically, it noted that the FTC specified the testing procedures to be used with the same particularity that the Defense Department specified the formula for Agent Orange, determined the language to be used to disclose the tar and nicotine content of each type of cigarette, monitored the industry’s testing labs, independently verified the manufacturers’ results, monitored the content of cigarette advertisements, and threatened enforcement actions against manufacturers that did not comply with the terms of the “voluntary” agreement in any way. Watson, 420 F.3d at 858-60. The Eighth Circuit further observed that “[t]he FTC involved itself in the tobacco industry to an unprecedented extent” and that it engaged in “an unusually high level of governmental participation and control.” Id. at 860.
*131There are several reasons why we are not persuaded that the decision in Watson is useful to our resolution of this appeal. First, even if we assume Watson was correctly decided, its conclusion was by no means inevitable. The plaintiffs did not challenge any conduct undertaken by or at the direction of the FTC, including the testing of the cigarettes using the government-specified methodology or the public disclosure of the test results. Instead, they alleged that Philip Morris manipulated its cigarettes to achieve particular test results, even though the “light” cigarettes were as or more harmful than regular cigarettes, and as a result, the marketing of Philip Morris’s cigarettes as “light” was misleading. While the FTC permitted Philip Morris to use the word “light” in its advertisements on the basis of the test results, it did not require the company to do so. Moreover, at least three federal courts facing similarly stated allegations regarding “light” cigarettes did not find the federal officer removal statute satisfied and remanded those cases back to state court. See Paldrmic v. Attria Corp. Servs., Inc., 327 F.Supp.2d 959 (E.D.Wis.2004); Virden v. Altria Group, Inc., 304 F.Supp.2d 832 (N.D.W.Va.2004); Tremblay v. Philip Morris, Inc., 231 F.Supp.2d 411 (D.N.H.2002); cf. Brown v. Philip Morris Inc., 250 F.3d 789, 801 (3d Cir.2001) (in Bivens action, “[t]he mere fact that a tobacco company has complied with the requirements of a federal law cannot suffice to transform it into a federal actor any more than the compliance of a myriad of private enterprises with federal law and administrative regulations could of itself work such a transformation”).
More significantly, the grounds for removal in Watson were more compelling than those in the present cases. For one thing, the FTC’s level of control and monitoring over the cigarette industry, which the court described as “unprecedented,” 420 F.3d at 860, was stronger than that of the EPA over the gasoline industry. The FTC required the regulated companies to test its products using an apparatus and a methodology created and specified by the federal government; it required manufacturers to disclose the results of those tests in their advertisements; and it held a sword over the manufacturers’ heads by threatening to take action against any company that deviated from the requirements set forth in the “voluntary” agreement. Id. at 858-61. By contrast, while federal regulations at issue here have much to say about gasoline content, they allow refiners to use any of several additives to meet federal oxygenation requirements and say nothing regarding the marketing of gasoline containing MTBE, a highly dangerous compound that, like tar and nicotine, poses a threat to human health if ingested.
Lastly, we are mindful of Judge Gruen-der’s concurring opinion in Watson, which cautioned that “our decision today should not be construed as an invitation to every participant in a heavily regulated industry to claim that it, like Philip Morris, acts at the direction of a federal officer.” Id. at 863. Judge Gruender observed, and we agree, that “in most instances, a contract, principal-agent relationship, or near-employee relationship with the government will be necessary to show the degree of direction by a federal officer necessary to invoke removal under 28 U.S.C. § 1442(a)(1).” Id. Judge Gruender concurred with the court’s opinion because he found that “the FTC’s direction and control of the testing and marketing practices at issue is extraordinary,” id. at 863-64, but there is no similar direction and control in the present cases.
The federal officer removal statute is not to be construed grudgingly, Arizona v. Manypenny, 451 U.S. 232, 242, 101 S.Ct. *1321657, 68 L.Ed.2d 58 (1981), but we do not believe it was intended to be construed so broadly that it would federalize a broad spectrum of state-law tort claims against entities regulated by—though not acting under—officers or agencies of the United States. As we have found that removal was inappropriate under the federal officer removal statute because the defendants did not act under an officer of the United States, we need not address the last requirement for removal under the federal officer removal statute, i.e., whether defendants have offered a “colorable” federal defense.
2. Bankruptcy Removal
Pursuant to the bankruptcy removal statute, “[a] party may remove any claim or cause of action in a civil action other than ... a civil action by a governmental unit to enforce such governmental unit’s police or regulatory power, to the district court for the district where such civil action is pending, if such district court has jurisdiction of such claim or cause of action under section 1334 of this title.” 28 U.S.C. § 1452. Section 1334, in turn, grants the federal district courts original jurisdiction over all cases under Title 11 of the United States Code. 28 U.S.C. § 1334(a). The state plaintiffs clearly are governmental units. See 11 U.S.C. § 101(27). The question we must answer, then, is whether these actions are intended to enforce California’s and New Hampshire’s “police or regulatory powers,” bringing them within the exception to removal in section 1452.
We have never had occasion to define the parameters of a governmental unit’s police or regulatory power in the context of section 1452. We have, however, addressed that question in dealing with the automatic stay provision of the bankruptcy code. That clause, in relevant part, exempts from the automatic stay “the commencement or continuation of an action or proceeding by a governmental unit ... to enforce such governmental unit’s ... police and regulatory power, including the enforcement of a judgment other than a money judgment....” 11 U.S.C. § 362(b)(4). As “[t]he language of the police and regulatory power exceptions in the automatic stay context and in the removal context is virtually identical, and the purpose behind each exception is the same,” PG & E Corp., 433 F.3d at 1123, we think it proper to look to judicial interpretations of section 362 for guidance.
In City of New York v. Exxon Corp., 932 F.2d 1020 (2d Cir.1991), New York City sued fifteen corporations to recover the costs of removing hazardous substances contained in industrial wastes produced by the defendants from several city-owned landfills. One of those defendants later filed for bankruptcy protection and sought to stay the action. We declined to grant a stay, holding that “Congress meant to except damage actions for completed violations of environmental laws from the action of the stay.” Id. at 1024. In support, we cited the legislative history of the automatic stay provision, which stated that “where a government unit is suing a debt- or to prevent or stop violation of [laws relating to] fraud, environmental protection, consumer protection, [or] safety, or similar police or regulatory laws, or attempting to fix damages for violation of such a law, the action or proceeding is not stayed under the automatic stay.” Id. (quoting H.R.Rep. No. 595, at 343 (1978), reprinted in 1978 U.S.C.C.A.N. 5963, 6299). This approach is consistent with that of other cases, which have held that “the term ‘police or regulatory power’ refers to the enforcement of state laws affecting health, welfare, morals, and safety, but not regulatory laws that directly conflict with the control of the res or property *133by the bankruptcy court.” In re Missouri, 647 F.2d 768, 776 (8th Cir.1981); accord In re Javens, 107 F.3d 359, 369-70 (6th Cir.1997); Hillis Motors, Inc. v. Hawaii Auto. Dealers’ Ass’n, 997 F.2d 581, 591 (9th Cir.1993).
While the similarity between the present cases and Exxon strongly suggests that the present proceedings fall within the “police or regulatory power” exception to removal under section 1452, the defendants suggest that the exception will not apply where “the government action directed against the debtor relates mainly to the protection of a pecuniary interest rather than the enforcement of regulatory police powers for the protection of the general public.” Defs.’ Br. at 67 (quoting In re Greenwald, 34 B.R. 954, 956-57 (Bankr. S.D.N.Y.1983)). In so arguing, the defendants allude to the “pecuniary purpose” test adopted in other circuits, see, e.g., Lockyer v. Mirant Corp., 398 F.3d 1098, 1108 (9th Cir.2005); Chao v. Hosp. Staffing Servs., Inc., 270 F.3d 374, 385 (6th Cir.2001), under which it is necessary to determine whether the governmental action relates primarily to the government’s pecuniary interest in the debtor’s property, in which case the automatic stay applies, or to matters of safety and welfare, in which case the suit may proceed. Cases that employ the pecuniary purpose test also use the “public purpose” test, under which the reviewing court determines whether the government seeks to “effectuate public policy” or “adjudicate private rights.” Mirant, 398 F.3d at 1109 (internal quotation marks omitted). While these tests were developed in connection with section 362, the automatic stay provision, the Ninth Circuit has applied them in the context of section 1452. PG & E Corp., 433 F.3d at 1124.
Although we do not find it necessary to pass on the validity of these tests at this time, we find them satisfied in this case. The California and New Hampshire actions relate primarily to matters of public health and welfare, and the money damages sought will not inure, strictly speaking, to the economic benefit of the states. Instead, the clear goal of these proceedings is to remedy and prevent environmental damage with potentially serious consequences for public health, a significant area of state policy. Thus, even under the tests advocated by the defendant companies, these proceedings represent efforts by California and New Hampshire to enforce their “police or regulatory power” and are not subject to removal under section 1452.
Apart from the question of whether these lawsuits are actions by governmental units to enforce their police or regulatory powers, the defendants contend that the police power exception in section 1452 does not apply to actions seeking money damages for past conduct. While the “pecuniary purpose” test, discussed above, concerns whether or not a suit is for a police or regulatory purpose, this argument presupposes that any action seeking money damages is barred. The defendants’ argument is without merit for two reasons. First, the automatic stay provision of section 362 bars only a governmental unit’s attempt to enforce a money judgment, not, as here, a suit seeking entry of a money judgment. See SEC v. Brennan, 230 F.3d 65, 71 (2d Cir.2000) (“It is well established that the governmental unit exception of § 362(b)(4) permits the entry of a money judgment against a debtor so long as the proceeding in which such a judgment is entered is one to enforce the governmental unit’s police or regulatory power.”); see also Penn Terra Ltd. v. Dep’t of Envtl. Res., 733 F.2d 267, 272 (3d Cir.1984). Second, section 1452 contains no provision distinguishing between lawsuits seeking to enforce a money judgment and those seek*134ing merely the entry of a money judgment. Instead, unlike the exception to the automatic stay provision of section 362, it applies to any suit by a governmental unit to enforce its police or regulatory powers. In sum, as the present cases are such suits, they may not be removed pursuant to section 1452.
C. Alternative Grounds for Removal
Having concluded that the district court lacked jurisdiction over these actions under the federal officer removal statute or the bankruptcy removal statute, we now turn to the other bases for removal urged by the defendants, though rejected by the district court in MTBE V: preemption and federal question jurisdiction. We may affirm the denial of the motion to remand on either of those grounds, as “[a]n appellate court is free to affirm a district court decision on any grounds for which there is a record sufficient to permit conclusions of law, even grounds not relied upon by the district court.” Gmurzynska v. Hutton, 355 F.3d 206, 210 (2d Cir.2004) (internal quotation marks omitted).
As a preliminary matter, we note that preemption and federal question jurisdiction may apply only to the California action, not the New Hampshire action. Defs.’ Br. at 58. In the New Hampshire action, unlike the California action, not every defendant consented to removal. Id. Because removal on the basis of preemption or a substantial federal question— unlike removal under the federal officer or bankruptcy removal statutes—requires the consent of all defendants, Chicago, Rock Island & Pac. R.R. Co. v. Martin, 178 U.S. 245, 247-48, 20 S.Ct. 854, 44 L.Ed. 1055 (1900); 14C Charles Alan Wright, Arthur R. Miller, & Edward H. Cooper, Federal Practice and Procedure § 3731, at 258 (3d ed.1998), only the California action is arguably capable of being removed on the grounds discussed below.
1. Preemption
We begin by observing that the parties have largely ignored the issue of preemption in their papers to this court for the obvious reason that the case for preemption is weaker with respect to the California plaintiffs than it is with respect to perhaps any other plaintiffs. Many of the plaintiffs in MTBE V alleged that MTBE entered their groundwater as a result of being emitted from car tailpipes and falling back to the earth as rain. This allegation supports at least a colorable argument that those lawsuits are for the purposes of emission control, or at least so closely related to the issue of emission control that perhaps the applicable EPA regulations are of preemptive effect. See 42 U.S.C. § 7545(c)(4)(A) (preempting state regulation of fuel additives when for the purpose of motor vehicle emission control). Nevertheless, in this case, California does not appear to allege liability based on the release of MTBE through normal vehicle operations. Rather, it premises its theory of liability on leaks and discharges of MTBE from gasoline delivery systems into the state’s groundwater, before it has seen the inside of a vehicle’s gas tank. The defendants’ own papers acknowledge the weakness of their argument. Specifically, in their opposition to the plaintiffs’ motions for remand in two related cases, County of Nassau v. Amerada Hess Corp., No. 03-cv-9543 (S.D.N.Y. filed Dec. 2, 2003), and Water Authority of Western Nassau County v. Amerada Hess Corp., No. 03-cv-9544 (S.D.N.Y. filed Dec. 2, 2003), the defendants write:
That states and private parties have myriad other remedies, under both state and federal law, to address MTBE in groundwater—and, in particular, remedies against those who spill or leak *135MTBE gasoline—means that this Court need not be concerned that federal preemption interferes with the States’ historical role in protecting health and safety.... Rather, reflecting that clean air and national supply of reasonably priced gasoline were overriding federal concerns, Congress and EPA preempted only in the narrow area of fuel design, while preserving participation in the federal administrative process and state remedies against those who spill gasoline.
Defs.’ Opp’n to Pis.’ Mot. to Remand at 29 (emphasis added). This statement recognizes that state-law remedies are available to address MTBE in groundwater—the very remedies pursued in this action.
In addition, for the reasons articulated in MTBE V, 341 F.Supp.2d at 403-11, we agree with Judge Scheindlin that the plaintiffs’ claims are not completely preempted by federal law. Specifically, we agree with her conclusions that the EPA considered the preemptive effect of its regulations to be “limited,” id. at 406; that the agency intended its regulations to preempt nonidentical state controls in the context of fuel emissions, not in every conceivable area related to fuel and fuel additives, id. at 406-07; and that Congress did not intend to preempt state regulations unrelated to emissions control, id. at 408-09. We also agree with the conclusions of other courts that have found that similar claims were not preempted by the CAA. See, e.g., Oxygenated Fuels Ass’n v. Davis, 331 F.3d 665, 673 (9th Cir.2003); Oxygenated Fuels Ass’n v. Pataki, 158 F.Supp.2d 248, 257 & n. 4 (N.D.N.Y.2001).
We also note that, since oral argument was held in these cases, the district court has denied the defendants’ motions for summary judgment on the related affirmative defense of conflict preemption. In re Methyl Tertiary Butyl Ether (“MTBE’’) Prods. Liab. Litig., 457 F.Supp.2d 324 (S.D.N.Y.2006). Judge Scheindlin held that there was no conflict preemption, as “[i]t was not physically impossible for defendants to comply with both standards because, even if state tort law demands that defendants not use MTBE, the federal law did not require the use of MTBE.” Id. at 335. Likewise, she held that there was no obstacle preemption, because the CAA amendments created a fuel-neutral program that could succeed even if tort liability were available for the use or misuse of MTBE. Id. at 335-43. While conflict preemption is a defense, not a basis for jurisdiction, this holding only reinforces the plaintiffs’ argument that state tort law and the federal regulations at issue are compatible with one another.
2. Substantial Federal Question
For the reasons stated in Judge Scheindlin’s opinion, we agree that the plaintiffs’ claims do not raise a substantial federal question giving rise to federal subject matter jurisdiction. See MTBE V, 341 F.Supp.2d at 400-03. The plaintiffs’ claims arise under and will be decided under state law, and although the defendants may refer to federal legislation by way of a defense, the jury’s verdict will not necessarily turn on a construction of that federal law. As the Supreme Court noted in Merrell Dow Pharmaceuticals Inc. v. Thompson, 478 U.S. 804, 106 S.Ct. 3229, 92 L.Ed.2d 650 (1986), “the mere presence of a federal issue in a state cause of action does not automatically confer federal-question jurisdiction.” Id. at 813, 106 S.Ct. 3229. Indeed, words written by Justice Cardozo more than seventy years ago are equally applicable here:
The most one can say is that a question of federal law is lurking in the background, just as farther in the background there lurks a question of consti*136tutional law, the question of state power in our federal form of government. A dispute so doubtful and conjectural, so far removed from plain necessity, is unavailing to extinguish the jurisdiction of the states.
Gully v. First Nat’l Bank, 299 U.S. 109, 117, 57 S.Ct. 96, 81 L.Ed. 70 (1936).
The California defendants also argue that in delegating to the EPA the authority to enact regulations requiring oxygenated fuels, Congress required the agency to “give greater importance to clean air gains than to other potential environmental concerns.” Defs.’ Br. at 58 (citing 42 U.S.C. § 7545(k)(l)). They contend that plaintiffs’ lawsuits, in seeking to privilege clean water at the expense of clean air, “implicate these federal determinations and thus require the application of federal law.” Id. at 59. This argument, however, presents a false dichotomy between clean air and clean water. That the defendants might use MTBE to meet fuel oxygenation requirements does not necessarily conflict with the state’s goal of securing the safety of its water supply. Because the California complaint focuses on spills, leaks, and discharges of MTBE into groundwater, as well as alleged misrepresentations and omissions related to the safety of MTBE, they are entirely separate issues. A jury finding in favor of the plaintiffs would undermine neither Congress’s intent in enacting the CAA amendments, nor the EPA regulations implementing them.
Conclusion
We hold that, if the criteria of a valid removal statute are met, sovereign immunity does not bar the removal of a ease commenced by a state in its own courts. Because the requirements of the applicable removal statutes have not been met, we vacate the order of the district court and remand with directions to return these cases to the forums from which they were removed.