delivered the opinion of the Court.

This case focuses on the 1984 version of a Federal Motor Vehicle Safety Standard promulgated by the Department of Transportation under the authority of the National Traffic and Motor Vehicle Safety Act of 1966,80 Stat. 718,15 U. S. C. §1381 et seq. (1988 ed.). The standard, FMVSS 208, required auto manufacturers to equip some but not all of their *8651987 vehicles with passive restraints. We ask whether the Act pre-empts a state common-law tort action in which the plaintiff claims that the defendant auto manufacturer, who was in compliance with the standard, should nonetheless have equipped a 1987 automobile with airbags. We conclude that the Act, taken together with FMVSS 208, pre-empts the lawsuit.

I

In 1992, petitioner Alexis Geier, driving a 1987 Honda Accord, collided with a tree and was seriously injured. The car was equipped with manual shoulder and lap belts which Geier had buckled up at the time. The car was not equipped with airbags or other passive restraint devices.

Geier and her parents, also petitioners, sued the car’s manufacturer, American Honda Motor Company, Inc., and its affiliates (hereinafter American Honda), under District of Columbia tort law. They claimed, among other things, that American Honda had designed its car negligently and defectively because it lacked a driver’s side airbag. App. 3. The District Court dismissed the lawsuit. The court noted that FMVSS 208 gave ear manufacturers a choice as to whether to install airbags. And the court concluded that petitioners’ lawsuit, because it sought to establish a different safety standard — i. e., an airbag requirement — was expressly preempted by a provision of the Act which pre-empts “any safety standard” that is not identical to a federal safety standard applicable to the same aspect of performance, 15 U. S. C. § 1392(d) (1988 ed.); Civ. No. 95-CV-0064 (D. D. C., Dec. 9, 1997), App. 17. (We, like the courts below and the parties, refer to the pre-1994 version of the statute throughout the opinion; it has been recodified at 49 U. S. C. § 30101 et seq.)

The Court of Appeals agreed with the District Court’s conclusion but on somewhat different reasoning. It had doubts, given the existence of the Act’s “saving” clause, 15 U. S. C. § 1397(k) (1988 ed.), that petitioners’ lawsuit involved the po*866tential creation of the kind of “safety standard” to which the Safety Act’s express pre-emption provision refers. But it declined to resolve that question because it found that petitioners’ state-law tort claims posed an obstacle to the accomplishment of FMVSS 208’s objectives. For that reason, it found that those claims conflicted with FMVSS 208, and that, under ordinary pre-emption principles, the Act consequently pre-empted the lawsuit. The Court of Appeals thus affirmed the District Court’s dismissal. 166 F. 3d 1236,1238-1243 (CADC 1999).

Several state courts have held to the contrary, namely, that neither the Act’s express pre-emption nor FMVSS 208 preempts a “no airbag” tort suit. See, e. g., Drattel v. Toyota Motor Corp., 92 N. Y. 2d 35, 43-53, 699 N. E. 2d 376, 379-386 (1998); Minton v. Honda of America Mfg., Inc., 80 Ohio St. 3d 62, 70-79, 684 N. E. 2d 648, 655-661 (1997); Munroe v. Galati, 189 Ariz. 113, 115-119, 938 P. 2d 1114, 1116-1120 (1997); Wilson v. Pleasant, 660 N. E. 2d 327, 330-339 (Ind. 1995); Tebbetts v. Ford Motor Co., 140 N. H. 203, 206-207, 665 A. 2d 345, 347-348 (1995). All of the Federal Circuit Courts that have considered the question, however, have found pre-emption. One rested its conclusion on the Act’s express pre-emption provision. See, e.g., Harris v. Ford Motor Co., 110 F. 3d 1410, 1413-1415 (CA9 1997). Others, such as the Court of Appeals below, have instead found preemption under ordinary pre-emption principles by virtue of the conflict such suits pose to FMVSS 208’s objectives, and thus to the Act itself. See, e. g., Montag v. Honda Motor Co., 75 F. 3d 1414, 1417 (CA10 1996); Pokorny v. Ford Motor Co., 902 F. 2d 1116, 1121-1125 (CA3 1990); Taylor v. General Motors Corp., 875 F. 2d 816, 825-827 (CA11 1989); Wood v. General Motors Corp., 865 F. 2d 395, 412-414 (CA1 1988). We granted certiorari to resolve these differences. We now hold that this kind of “no airbag” lawsuit conflicts with the objectives of FMVSS 208, a standard authorized by the Act, and is therefore pre-empted by the Act.

*867In reaching our conclusion, we consider three subsidiary questions. First, does the Act’s express pre-emption provision pre-empt this lawsuit? We think not. Second, do ordinary pre-emption principles nonetheless apply? We hold that they do. Third, does this lawsuit actually conflict with FMVSS 208, hence with the Act itself? We hold that it does.

II

We first ask whether the Safety Act’s express pre-emption provision pre-empts this tort action. The provision reads as follows:

“Whenever a Federal motor vehicle safety standard established under this subchapter is in effect, no State or political subdivision of a State shall have any authority either to establish, or to continue in effect, with respect to any motor vehicle or item of motor vehicle equipment[,] any safety standard applicable to the same aspect of performance of such vehicle or item of equipment which is not identical to the Federal standard.” 15 U. S. C. § 1892(d) (1988 ed.).

American Honda points out that a majority of this Court has said that a somewhat similar statutory provision in a different federal statute — a provision that uses the word “requirements” — may well expressly pre-empt similar tort actions. See, e. g., Medtronic, Inc. v. Lohr, 518 U. S. 470, 502-504 (1996) (plurality opinion); id., at 503-505 (Breyer, J., concurring in part and concurring in judgment); id., at 509-512 (O’Connor, J., concurring in part and dissenting in part). Petitioners reply that this statute speaks of pre-empting a state-law “safety standard,” not a “requirement,” and that a tort action does not involve a safety standard. Hence, they conclude, the express pre-emption provision does not apply.

We need not determine the precise significance of the use of the word “standard,” rather than “requirement,” however, for the Act contains another provision, which resolves the *868disagreement. That provision, a “saving” clause, says that “[c]ompliance with” a federal safety standard “does not exempt any person from any liability under common law.” 15 U. S. C. § 1397(k) (1988 ed.). The saving clause assumes that there are some significant number of common-law liability eases to save. And a reading of the express pre-emption provision that excludes common-law tort actions gives actual meaning to the saving clause’s literal language, while leaving adequate room for state tort law to operate — for example, where federal law creates only a floor, i. e., a minimum safety standard. See, e. g., Brief for United States as Amicus Curiae 21 (explaining that common-law claim that a vehicle is defectively designed because it lacks antilock brakes would not be pre-empted by 49 CFR §571.105 (1999), a safety standard establishing minimum requirements for brake performance). Without the saving clause, a broad reading of the express pre-emption provision arguably might pre-empt those actions, for, as we have just mentioned, it is possible to read the pre-emption provision, standing alone, as applying to standards imposed in common-law tort actions, as well as standards contained in state legislation or regulations. And if so, it would pre-empt all nonidentical state standards established in tort actions covering the same aspect of pei'-■formance as an applicable federal standard, even if the federal standard merely established a minimum standard. On that broad reading of the pre-emption clause little, if any, potential “liability at common law” would remain. And few, if any, state tort actions would remain for the saving clause to save. We have found no convincing indication that Congress wanted to pre-empt, not only state statutes and regulations, but also common-law tort actions, in such circumstances. Hence the broad reading cannot be correct. The language of the pre-emption provision permits a narrow reading that excludes common-law actions. Given the presence of the saving clause, we conclude that the pre-emption clause must be so read.

*869HH f — 1 1 — 1

We have just said that the saving clause at least removes tort actions from the scope of the express pre-emption clause. Does it do more? In particular, does it foreclose or limit the operation of ordinary pre-emption principles insofar as those principles instruct us to read statutes as preempting state laws (including common-law rules) that “actually conflict” with the statute or federal standards promulgated thereunder? Fidelity Fed. Sav. & Loan Assn. v. De la Cuesta, 458 U. S. 141, 153 (1982). Petitioners concede, as they must in light of Freightliner Corp. v. Myrick, 514 U. S. 280 (1995), that the pre-emption provision, by itself, does not foreclose (through negative implication) “any possibility of implied [conflict] pre-emption,” id., at 288 (discussing Cipollone v. Liggett Group, Inc., 505 U. S. 504, 517-518 (1992)). But they argue that the saving clause has that very effect.

We recognize that, when this Court previously considered the pre-emptive effect of the statute’s ianguage, it appeared to leave open the question of how, or the extent to which, the saving clause saves state-law tort actions that conflict with federal regulations promulgated under the Act. See Freightliner, supra, at 287, n. 3 (declining to address whether the saving clause prevents a manufacturer from “us[ing] a federal safety standard to immunize itself from state common-law liability”). We now conclude that the saving clause (like the express pre-emption provision) does not bar the ordinary working of conflict pre-emption principles.

Nothing in the language of the saving clause suggests an intent to save state-law tort actions that conflict with federal regulations. The words “Compliance” and “does not exempt,” 15 U. S. C. § 1397(k) (1988 ed.), sound as if they simply bar a special kind of defense, namely, a defense that compliance with a federal standard automatically exempts a defendant from state law, whether the Federal Government meant that standard to be an absolute requirement or only a minimum one. See Restatement (Third) of Torts: Products *870Liability § 4(b), Comment ~e (1997) (distinguishing between state-law compliance defense and a federal claim of preemption). It is difficult to understand why Congress would have insisted on a compliance-with-federal-regulation precondition to the provision’s applicability had it wished the Act to “save” all state-law tort actions, regardless of their potential threat to the objectives of federal safety standards promulgated under that Act. Nor does our interpretation conflict with the purpose of the saving provision, say, by rendering it ineffectual. As we have previously explained, the saving provision still makes clear that the express pre-emption provision does not of its own force pre-empt common-law tort actions. And it thereby preserves those actions that seek to establish greater safety than the minimum safety achieved by a federal regulation intended to provide a floor. See supra, at 867-868.

Moreover, this Court has repeatedly “decline[d] to give broad effect to saving clauses where doing so would upset the careful regulatory scheme established by federal law.” United States v. Locke, ante, at 106-107; see American Telephone & Telegraph Co. v. Central Office Telephone, Inc., 524 U. S. 214, 227-228 (1998) (AT&T); Texas & Pacific R. Co. v. Abilene Cotton Oil Co., 204 U. S. 426, 446 (1907). We find this concern applicable in the present ease. And we conclude that the saving clause foresees — it does not foreclose— the possibility that a federal safety standard will pre-empt a state common-law tort action with which it conflicts. We do not imderstand the dissent to disagree, for it acknowledges that ordinary pre-emption principles apply, at least sometimes. Post, at 899-900 (opinion of Stevens, J.).

Neither do we believe that the pre-emption provision, the saving provision, or both together, create some kind of “special burden” beyond that inherent in ordinary pre-emption principles — which “special burden” would specially disfavor pre-emption here. Cf. post, at 898-899. The two provisions, read together, reflect a neutral policy, not a specially *871favorable or unfavorable policy, toward the application of ordinary conflict pre-emption principles. On the one hand, the pre-emption provision itself reflects a desire to subject the industry to a single, uniform set of federal safety standards. Its pre-emption of all state standards, even those that might stand in harmony with federal law, suggests an intent to avoid the conflict, uncertainty, cost, and occasional risk to safety itself that too many different safety-standard cooks might otherwise create. See H. R. Rep. No. 1776, 89th Cong., 2d Sess., 17 (1966) (“Basically, this preemption subsection is intended to result in uniformity of standards so that the public as well as industry will be guided by one set of criteria rather than by a multiplicity of diverse standards”); S. Rep. No. 1301,89th Cong., 2d Sess., 12 (1966). This policy by itself favors pre-emption of state tort suits, for the rules of law that judges and juries create or apply in such suits may themselves similarly create uncertainty and even conflict, say, when different juries in different States reach different decisions on similar facts.

On the other hand, the saving clause reflects a congressional determination that occasional nonuniformity is a small price to pay for a system in which juries not only create, but also enforce, safety standards, while simultaneously providing necessary compensation to victims. That policy by itself disfavors pre-emption, at least some of the time. But we can find nothing in any natural reading of the two provisions that would favor one set of policies over the other where a jury-imposed safety standard actually conflicts with a federal safety standard.

Why, in any event, would Congress not have wanted ordinary pre-emption principles to apply where an actual conflict with a federal objective is at stake? Some such principle is needed. In its absence, state law could impose legal duties that would conflict directly with federal regulatory mandates, say, by premising liability upon the presence of the very windshield retention requirements that federal law re*872quires. See, e. g., 49 CFR § 571.212 (1999). Insofar as petitioners’ argument would permit common-law actions that "actually conflict” with federal regulations, it would take from those who would enforce a federal law the very ability to achieve the law’s congressionally mandated objectives that the Constitution, through the operation of ordinary preemption principles, seeks to protect. To the extent that such an interpretation of the saving provision reads into a particular federal law toleration of a conflict that those principles would otherwise forbid, it permits that law to defeat its own objectives, or potentially, as the Court has put it before, to “‘destroy itself.’” AT&T, supra, at 228 (quoting Abilene Cotton, supra, at 446). We do not claim that Congress lacks the constitutional power to write a statute that mandates such a complex type of state/federal relationship. Cf. post, at 900, n. 16. But there is no reason to believe Congress has done so here.

The dissent, as we have said, contends nonetheless that the express pre-emption and saving provisions here, taken together, create a “special burden,” which a court must impose “on a party” who claims conflict pre-emption under those principles. Post, at 898. But nothing in the Safety Act’s language refers to any “special burden.” Nor can one find the basis for a “special burden” in this Court’s precedents. It is true that, in Freightliner Corp. v. Myrick, 514 U. S. 280 (1995), the Court said, in the context of interpreting the Safety Act, that “[a]t bestS” there is an “inference that an express pre-emption clause forecloses implied pre-emption.” Id., at 289 (emphasis added). But the Court made this statement in the course of rejecting the more absolute, argument that the presence of the express pre-emption provision entirely foreclosed the possibility of conflict pre-emption. Id., at 288. The statement, headed with the qualifier “[a]t best,” and made in a case where, without any need for inferences or “special burdens,” state law obviously would survive, see id., at 289-290, simply preserves a legal possibility. This *873Court did not hold that the Safety Act does create a “special burden,” or still less that such a burden necessarily arises from the limits of an express pre-emption provision. And considerations of language, purpose, and administrative workability, together with the principles underlying this Court’s pre-emption doctrine discussed above, make clear that the express pre-emption provision imposes no unusual, “special burden” against pre-emption. For similar reasons, we do not see the basis for interpreting the saving clause to impose any such burden.

A “special burden” would also promise practical difficulty by further complicating well-established pre-emption principles that already are difficult to apply. The dissent does not contend that this “special burden” would apply in a case in which state law penalizes what federal law requires — i. e., a case of impossibility. See post, at 892-893, n. 6, 900, n. 16. But if it would not apply in such a ease, then how, or when, would it apply? This Court, when describing conflict preemption, has spoken of pre-empting state law that “under the circumstances of th[e] particular ease . . . stands as an obstacle to the accomplishment and execution of the full purposes and objectives of Congress” — whether that “obstacle” goes by the name of “conflicting; contrary to;... repugnance; difference; irreconcilability; inconsistency; violation; curtailment; . . . interference,” or the like. Hines v. Davidowitz, 312 U. S. 52, 67 (1941); see Jones v. Rath Packing Co., 430 U. S. 519, 526 (1977). The Court has not previously driven a legal wedge — only a terminological one — between “conflicts” that prevent or frustrate the accomplishment of a federal objective and “conflicts” that make it “impossible” for private parties to comply with both state and federal law. Rather, it has said that both forms of conflicting state law are “nullified” by the Supremacy Clause, De la Cuesta, 458 U. S., at 152-153; see Locke, ante, at 109; English v. General Elec. Co., 496 U. S. 72, 78-79 (1990), and it has assumed that Congress would not want either kind of conflict. The Court *874has thus refused to read general “saving” provisions to tolerate actual conflict both in cases involving impossibility, see, e. g., AT&T, 524 U. S., at 228, and in “frustration-of-purpose” eases, see, e. g., Locke, ante, at 103-112; International Paper Co. v. Ouellette, 479 U. S. 481, 493-494 (1987); see also Chicago & North Western Transp. Co. v. Kalo Brick & Tile Co., 450 U. S. 311, 328-331 (1981). We see no grounds, then, for attempting to distinguish among types of federal-state conflict for purposes of analyzing whether such a conflict warrants pre-emption in a particular case. That kind of analysis, moreover, would engender legal uncertainty with its inevitable systemwide costs (e.g., conflicts, delay, and expense) as courts tried sensibly to distinguish among varieties of “conflict” (which often shade, one into the other) when applying this complicated rule to the many federal statutes that contain some form of an express pre-emption provision, a saving provision, or as here, both. Nothing in the statute suggests Congress wanted to complicate ordinary experience-proved principles of conflict pre-emption with an added “special burden.” Indeed, the dissent’s willingness to impose a “special burden” here stems ultimately from its view that “frustration-of-purpos[e]’’ conflict pre-emption is a freewheeling, “inadequately considered” doctrine that might well be “eliminate^].” Post, at 907-908, and n. 22. In a word, ordinary pre-emption principles, grounded in longstanding precedent, Hines, supra, at 67, apply. We would not further complicate the law with complex new doctrine.

IV

The basic question, then, is whether a common-law “no airbag” action like the one before us actually conflicts with FMVSS 208. We hold that it does.

In petitioners’ and the dissent’s view, FMVSS 208 sets a minimum airbag standard. As far as FMVSS 208 is concerned, the more airbags, and the sooner, the better. But that was not the Secretary’s view. The Department of *875Transportation’s (DOT’s) comments, which accompanied the promulgation of FMVSS 208, make clear that the standard deliberately provided the manufacturer with a range of choices among different passive restraint devices. Those choices would bring about a mix of different devices introduced gradually over time; and FMVSS 208 would thereby lower costs, overcome technical safety problems, encourage technological development, and win widespread consumer acceptance — all of which would promote FMVSS 208’s safety objectives. See generally 49 Fed. Reg. 28962 (1984).

A

The history of FMVSS 208 helps explain why and how DOT sought these objectives. See generally Motor Vehicle Mfrs. Assn. of United States, Inc. v. State Farm Mut. Automobile Ins. Co., 463 U. S. 29, 34-38 (1983). In 1967, DOT, understanding that seatbelts would save many lives, required manufacturers to install manual seatbelts in all automobiles. 32 Fed. Reg. 2408, 2415. It became apparent, however, that most occupants simply would not buckle up their belts. See 34 Fed. Reg. 11148 (1969). DOT then began to investigate the feasibility of requiring “passive restraints,” such as airbags and automatic seatbelts. Ibid. In 1970, it amended FMVSS 208 to include some passive protection requirements, 35 Fed. Reg. 16927, while making clear that airbags were one of several “equally acceptable” devices and that it neither “ 'favored’ [n]or expected the introduction of airbag systems.” Ibid. In 1971, it added an express provision permitting compliance through the use of nondetachable passive belts, 36 Fed. Reg. 12858, 12859, and in 1972, it mandated full passive protection for all front seat occupants for vehicles manufactured after August 15, 1975, 37 Fed. Reg. 3911. Although the agency’s focus was originally on airbags, 34 Fed. Reg. 11148 (1969) (notice of proposed rule-making); State Farm, 463 U. S., at 35, n. 4; see also id., at 46, n. 11 (noting view of commentators that, as of 1970, FMVSS *876208 was “ ‘a de facto airbag mandate’ ” because of the state of passive restraint technology), at no point did FMVSS 208 formally require the use of airbags. From the start, as in 1984, it permitted passive restraint options.

DOT gave manufacturers a further choice for new vehicles manufactured between 1972 and August 1975. Manufacturers could either install a passive restraint device such as automatic seatbelts or airbags or retain manual belts and add an “ignition interlock” device that in effect forced occupants to buckle up by preventing the ignition otherwise from turning on. 37 Fed. Reg. 3911 (1972). The interlock soon became popular with manufacturers. And in 1974, when the agency approved the use of detachable automatic seatbelts, it conditioned that approval by providing that such systems must include an interlock system and a continuous warning buzzer to encourage reattachment of the belt. 39 Fed. Reg. 14593. But the interlock and buzzer devices were most unpopular with the public. And Congress, responding to public pressure, passed a law that forbade DOT from requiring, or permitting compliance by means of, such devices. Motor Vehicle and Schoolbus Safety Amendments of 1974, § 109,88 Stat. 1482 (previously codified at 15 U. S. C. § 1410b(b) (1988 ed.)).

That experience influenced DOT’S subsequent passive restraint initiatives. In 1976, DOT Secretary William T. Coleman, Jr., fearing continued public resistance, suspended the passive restraint requirements. He sought to win public acceptance for a variety of passive restraint devices through a demonstration project that would involve about half a million new automobiles. State Farm, supra, at 37. But his successor, Brock Adams, canceled the project, instead amending FMVSS 208 to require passive restraints, principally either airbags or passive seatbelts. 42 Fed. Reg. 34289 (1977).

Andrew Lewis, a new DOT Secretary in a new administration, rescinded the Adams requirements, primarily because DOT learned that the industry planned to satisfy those *877requirements almost exclusively through the installation of detachable automatic seatbelts. 46 Fed. Reg. 53419-53420 (1981). This Court held the rescission unlawful. State Farm, supra, at 34, 46. And the stage was set for then-DOT Secretary, Elizabeth Dole, to amend FMVSS 208 once again, promulgating the version that is now before us. 49 Fed. Reg. 28962 (1984).

B

Read in light of this history, DOT’s own contemporaneous explanation of FMVSS 208 makes clear that the 1984 version of FMVSS 208 reflected the following significant considerations. First, buckled up seatbelts are a vital ingredient of automobile safety. Id., at 29003; State Farm, supra, at 52 (“We start with the accepted ground that if used, seatbelts unquestionably would save many thousands of lives and would prevent tens of thousands of crippling injuries”). Second, despite the enormous and unnecessary risks that a passenger runs by not buckling up manual lap and shoulder belts, more than 80% of front seat passengers would leave their manual seatbelts unbuckled. 49 Fed. Reg. 28983 (1984) (estimating that only 12.5% of front seat passengers buckled up manual belts). Third, airbags could make up for the dangers caused by unbuckled manual belts, but they could not make up for them entirely. Id., at 28986 (concluding that, although an airbag plus a lap and shoulder belt was the most “effective” system, airbags alone were less effective than buckled up manual lap and shoulder belts).

Fourth, passive restraint systems had their own disadvantages, for example, the dangers associated with, intrusiveness of, and corresponding public dislike for, nondetachable automatic belts. Id., at 28992-28993. Fifth, airbags brought with them their own special risks to safety, such as the risk of danger to out-of-position occupants (usually children) in small cars. Id., at 28992, 29001; see also 65 Fed. Reg. 30680, 30681-30682 (2000) (finding 158 confirmed airbag-induced fatalities as of April 2000, and amending rule *878to add new requirements, test,procedures, and injury criteria to ensure that “future air bags be designed to create less risk of serious airbag-induced injuries than current air bags, particularly for small women and young children”); U. S. Dept, of Transportation, National Highway Traffic Safety Administration, National Accident Sampling System Crash-worthiness Data System 1991-1993, p. viii (Aug. 1995) (finding that airbags caused approximately 54,000 injuries between 1991 and 1993).

Sixth, airbags were expected to be significantly more expensive than other passive restraint devices, raising the average cost of a vehicle price $320 for full frontal airbags over the cost of a car with manual lap and shoulder seatbelts (and potentially much more if production volumes were low). 49 Fed. Reg. 28990 (1984). And the agency worried that the high replacement cost — estimated to be $800 — could lead ear owners to refuse to replace them after deployment. Id., at 28990, 29000-29001; see also id., at 28990 (estimating total investment costs for mandatory airbag requirement at $1.3 billion compared to $500 million for automatic seatbelts). Seventh, the public, for reasons of cost, fear, or physical intrusiveness, might resist installation or use of any of the then-available passive restraint devices, id., at 28987-28989 — a particular concern with respect to airbags, id., at 29001 (noting that “[a]irbags engendered the largest quantity of, and most vociferously worded, comments”).

FMVSS 208 reflected these considerations in several ways. Most importantly, that standard deliberately sought variety — a mix of several different passive restraint systems. It did so by setting a performance requirement for passive restraint devices and allowing manufacturers to choose among different passive restraint mechanisms, such as airbags, automatic belts, or other passive restraint technologies to satisfy that requirement. Id., at 28996. And DOT explained why FMVSS 208 sought the mix of devices that it expected its performance standard to produce. Id., at *87928997. DOT wrote that it had rejected a proposed FMVSS 208 “all airbag” standard because of safety concerns (perceived or real) associated with airbags, which concerns threatened a “backlash” more easily overcome “if airbags” were “not the only way of complying.” Id., at 29001. It added that a mix of devices would help develop data on comparative effectiveness, would allow the industry time to overcome the safety problems and the high production costs associated with airbags, and would facilitate the development of alternative, cheaper, and safer passive restraint systems. Id., at 29001-29002. And it would thereby build public confidence, id., at 29001-29002, necessary to avoid another interlock-type fiasco.

The 1984 FMVSS 208 standard also deliberately sought a gradual phase-in of passive restraints. Id., at 28999-29000. It required the manufacturers to equip only 10% of their car fleet manufactured after September 1,1986, with passive restraints. Id., at 28999. It then increased the percentage in three annual stages, up to 100% of the new car fleet for cars manufactured after September 1, 1989. Ibid. And it explained that the phased-in requirement would allow more time for manufacturers to develop airbags or other, better, safer passive restraint systems. It would help develop information about the comparative effectiveness of different systems, would lead to a mix in which airbags and other non-seatbelt passive restraint systems played a more prominent role than would otherwise result, and would promote public acceptance. Id., at 29000-29001.

Of course, as the dissent points out, post, at 908, FMVSS 208 did not guarantee the mix by setting a ceiling for each different passive restraint device. In fact, it provided a form of extra credit for airbag installation (and other nonbelt passive restraint devices) under which each airbag-installed vehicle counted as 1.5 vehicles for purposes of meeting FMVSS 208’s passive restraint requirement. 49 CFR §571.208, S4.1.3.4(a)(1) (1999); 49 Fed. Reg. 29000 (1984). *880But why should DOT have bothered to impose an airbag ceiling when the practical threat to the mix it desired arose from the likelihood that manufacturers would install, not too many airbags too quickly, but too few or none at all? After all, only a few years earlier, Secretary Dole’s predecessor had discovered that manufacturers intended to meet the then-current passive restraint requirement almost entirely (more than 99%) through the installation of more affordable automatic belt systems. 46 Fed. Reg. 53421 (1981); State Farm, 463 U. S., at 38. The extra credit, as DOT explained, was designed to “encourage manufacturers to equip at least some of their cars with airbags.” 49 Fed. Reg. 29001 (1984) (emphasis added) (responding to comment that failure to mandate airbags might mean the “end of... airbag technology”); see also id., at 29000 (explaining that the extra credit for airbags “should promote the development of what may be better alternatives to automatic belts than would otherwise be developed” (emphasis added)). The credit provision reinforces the point that FMVSS 208 sought a gradually developing mix of passive restraint devices; it does not show the contrary.

Finally, FMVSS 208’s passive restraint requirement was conditional. DOT believed that ordinary manual lap and shoulder belts would produce about the same amount of safety as passive restraints, and at significantly lower costs — if only auto occupants would buckle up. See id., at 28997-28998. Thus, FMVSS 208 provided for rescission of its passive restraint requirement if, by September 1, 1989, two-thirds of the States had laws in place that, like those of many other nations, required auto occupants to buckle up (and which met other requirements specified in the standard). Id., at 28963, 28993-28994, 28997-28999. The Secretary wrote that “coverage of a large percentage of the American people by seatbelt laws that are enforced would largely negate the incremental increase in safety to be expected from an automatic protection requirement.” Id., at 28997. *881In the end, two-thirds of the States did not enact mandatory buckle-up laws, and the passive restraint requirement remained in effect.

In sum, as DOT now tells us through the Solicitor General, the 1984 version of FMVSS 208 “embodies the Secretary’s policy judgment that safety would best be promoted if manufacturers installed alternative protection systems in their fleets rather than one particular system in every car.” Brief for United States as Amicus Curiae 25; see 49 Fed. Reg. 28997 (1984). Petitioners’ tort suit claims that the manufacturers of the 1987 Honda Accord “had a duty to design, manufacture, distribute and sell a motor vehicle with an effective and safe passive restraint system, including, but not limited to, airbags.” App. 3 (Complaint, ¶ 11).

In effect, petitioners’ tort action depends upon its claim that manufacturers had a duty to install an airbag when they manufactured the 1987 Honda Accord. Such a state law— i e., a rule of state tort law imposing such a duty — by its terms would have required manufacturers of all similar cars to install airbags rather than other passive restraint systems, such as automatic belts or passive interiors. It thereby would have presented an obstacle to the variety and mix of devices that the federal regulation sought. It would have required all manufacturers to have installed airbags in respect to the entire Distriet-of-Columbia-related portion of their 1987 new car fleet, even though FMVSS 208 at that time required only that 10% of a manufacturer’s nationwide fleet be equipped with any passive restraint device at all. It thereby also would have stood as an obstacle to the gradual passive restraint phase-in that the federal regulation deliberately imposed. In addition, it could have made less likely the adoption of a state mandatory buckle-up law. Because the rule of law for which petitioners contend would have stood “as an obstacle to the accomplishment and execution of” the important means-related federal objectives that we have just discussed, it is pre-empted. Hines, 312 U. S., at *88267; see also Ouellette, 479 U. S., at 493; De la Cuesta, 458 U. S., at 156 (finding conflict and pre-emption where state law limited the availability of an option that the federal agency considered essential to ensure its ultimate objectives).

Petitioners ask this Court to calculate the precise size of the “obstacle,” with the aim of minimizing it, by considering the risk of tort liability and a successful tort action’s incentive-related or timing-related compliance effects. See Brief for Petitioners 45-50. The dissent agrees. Post, at 900-905. But this Court’s pre-emption cases do not ordinarily turn on such compliance-related considerations as whether a private party in practice would ignore state legal obligations — paying, say, a fine instead — or how likely it is that state law actually would be enforced. Rather, this Court’s pre-emption cases ordinarily assume compliance with the state-law duty in question. The Court has on occasion suggested that tort law may be somewhat different, and that related considerations — for example, the ability to pay damages instead of modifying one’s behavior — may be relevant for pre-emption purposes. See Goodyear Atomic Corp. v. Miller, 486 U. S. 174, 185 (1988); Cipollone, 505 U. S., at 536-539 (Blackmun, J., concurring in part, concurring in judgment in part, and dissenting in part); see also English, 496 U. S., at 86; Silkwood v. Kerr-McGee Corp., 464 U. S. 238, 256 (1984). In other cases, the Court has found tort law to conflict with federal law without engaging in that kind of an analysis. See, e. g., Ouellette, supra, at 494-497; Kalo Brick, 450 U. S., at 324-332. We need not try to resolve these differences here, however, for the incentive or compliance considerations upon which the dissent relies cannot, by themselves, change the legal result. Some of those considerations rest on speculation, see, e. g., post, at 901 (predicting risk of “no airbag” liability and manufacturers’ likely response to such liability); some rest in critical part upon the dissenters’ own view of FMVSS 208’s basic purposes — a view *883which we reject, see, e. g., post, at 901-904' (suggesting that pre-existing risk of “no airbag” liability would have made FMVSS 208 unnecessary); and others, if we understand them correctly, seem less than persuasive, see, e. g., post, at 902 (suggesting that manufacturers could have complied with a mandatory state airbag duty by installing a different kind of passive restraint device). And in so concluding, we do not “put the burden” of proving pre-emption on petitioners. Post, at 907. We simply find unpersuasive their arguments attempting to undermine the Government’s demonstration of actual conflict.

One final point: We place some weight upon DOT’s interpretation of FMVSS 208’s objectives and its conclusion, as set forth in the Government’s brief, that a tort suit such as this one would “ ‘stan[d] as an obstacle to the accomplishment and execution’ ” of those objectives. Brief for United States as Amicus Curiae 25-26 (quoting Hines, supra, at 67). Congress has delegated to DOT authority to implement the statute; the subject matter is technical; and the relevant history and background are complex and extensive. The agency is likely to have a thorough understanding of its own regulation and its objectives and is “uniquely qualified” to comprehend the likely impact of state requirements. Med-tronic, 518 U. S., at 496; see id., at 506 (Breyer, J., concurring in part and concurring in judgment). And 'DOT has explained FMVSS 208’s objectives, and the interference that “no airbag” suits pose thereto, consistently over time. Brief for United States as Amicus Curiae in Freightliner Corp. v. Myrick, O. T. 1994, No. 94-286, pp. 28-29; Brief for United States as Amicus Curias in Wood v. General Motors Corp., O. T. 1989, No. 89-46, pp. 7, 11-16. In these circumstances, the agency’s own views should make a difference. See City of New York v. FCC, 486 U. S. 57, 64 (1988); Hillsborough County v. Automated Medical Laboratories, Inc., 471 U. S. 707, 714, 721 (1985); De la Cuesta, supra, at 158; Blum v. Bacon, 457 U. S. 132, 141 (1982); Kalo Brick, supra, at 321.

*884We have no reason to suspect that the Solicitor General’s representation of DOT’s views reflects anything other than ‘•'the agency’s fair and considered judgment on the matter.” Auer v. Robbins, 519 U. S. 452, 461-462 (1997); cf. Hillsborough County, supra, at 721 (expressing reluctance, in the absence of strong evidence, to find an actual conflict between state law and federal regulation where agency that promulgated the regulation had not, at the time the regulation was promulgated or subsequently, concluded that such a conflict existed). The failure of the Federal Register to address pre-emption explicitly is thus not determinative.

The dissent would require a formal agency statement of pre-emptive intent as a prerequisite to concluding that a conflict exists. It relies on cases, or portions thereof, that did not involve conflict pre-emption. See post, at 908-909; California Coastal Comm’n v. Granite Rock Co., 480 U. S. 572, 583 (1987); Hillsborough, supra, at 718. And conflict preemption is different in that it turns on the identification of “actual conflict,” and not on an express statement of preemptive intent. English, supra, at 78-79; see Hillsborough, supra, at 720-721; Jones, 430 U. S., at 540-543. While “[p]re-emption fundamentally is a question of congressional intent,” English, supra, at 78, this Court traditionally distinguishes between “express” and “implied” pre-emptive intent, and treats “conflict” pre-emption as an instance of the latter. See, e. g., Freightliner, 514 U. S., at 287; English, supra, at 78-79; see also Cipollone, supra, at 545, 547-548 (Scalia, J., concurring in judgment in part and dissenting in part). And though the Court has looked for a specific statement of preemptive intent where it is claimed that the mere “volume and complexity” of agency regulations demonstrate an implicit intent to displace all state law in a particular area, Hillsborough, supra, at 717; see post, at 908-909, n. 23— so-called “field pre-emption” — the Court has never before required a specific, formal agency statement identifying conflict in order to conclude that such a conflict in fact exists. *885Indeed, one can assume that Congress or an agency ordinarily would not intend to permit a significant conflict. While we certainly accept the dissent’s basie position that a court should not find pre-emption too readily in the absence of clear evidence of a conflict, English, supra, at 90, for the reasons set out above we find such evidence here. To insist on a specific expression of agency intent to pre-empt, made after notice-and-comment rulemaking, would be in certain cases to tolerate conflicts that an agency, and therefore Congress, is most unlikely to have intended. The dissent, as we have said, apparently welcomes that result, at least where “frustration-of-purpos[e]” pre-emption by agency regulation is at issue. Post, at 907-908, and n. 22. We do not.

Nor do we agree with the dissent that the agency’s views, as presented here, lack coherence. Post, at 904-905. The dissent points, ibid., to language in the Government’s brief stating that

“a claim that a manufacturer should have chosen to install airbags rather than another type of passive restraint in a certain model of car because of other design features particular to that car... would not necessarily frustrate Standard 208’s purposes.” Brief for United States as Amicus Curiae 26, n. 23 (emphasis added).

And the dissent says that these words amount to a concession that there is no conflict in this very case. Post, at 905. But that is not what the words say. Rather, as the italicized phrase emphasizes, they simply leave open the question whether FMVSS 208 would pre-empt a different kind of tort case — one not at issue here. It is possible that some special design-related circumstance concerning a particular kind of car might require airbags, rather than automatic belts, and that a suit seeking to impose that requirement could escape pre-emption — say, because it would affect so few cars that its rule of law would not create a legal “obstacle” to 208’s mixed-fleet, gradual objective. But that is not what peti*886tioners claimed. They have argued generally that, to be safe, a car must have an airbag. See App. 4.

Regardless, the language of FMVSS 208 and the contemporaneous 1984 DOT explanation is clear enough — even without giving DOT’s own view special weight. FMVSS 208 sought a gradually developing mix of alternative passive restraint devices for safety-related reasons. The rule of state tort law for which petitioners argue would stand as an “obstacle” to the accomplishment of that objective. And the statute foresees the application of ordinary principles of pre-emption in cases of actual conflict. Hence, the tort action is pre-empted. •

The judgment of the Court of Appeals is affirmed.

It is so ordered.

with whom Justice Souter, Justice Thomas, and Justice Ginsburg join, dissenting.

Airbag technology has been available to automobile manufacturers for over 30 years. There is now general agreement on the proposition “that, to be safe, a car must have an airbag.” Ante this page. Indeed, current federal law imposes that requirement on all automobile manufacturers. See 49 U. S. C. §30127; 49 CFR. §571.208, S4.1.5.3 (1998). The question raised by petitioners’ common-law tort action is whether that proposition was sufficiently obvious when Honda’s 1987 Accord was manufactured to make the failure to install such a safety feature actionable under theories of negligence or defective design. The Court holds that an interim regulation motivated by the Secretary of Transportation’s desire to foster gradual development of a variety of passive restraint devices deprives state courts of jurisdiction to answer that question. I respectfully dissent from that holding, and especially from the Court’s unprecedented extension of the doctrine of pre-emption. As a preface to an explanation of my understanding of the statute and the regulation, these preliminary observations seem appropriate.

*887“This is a case about federalism,” Coleman v. Thompson, 501 U. S. 722, 726 (1991), that is, about respect for “the constitutional role of the States as sovereign entities.” Alden v. Maine, 527 U. S. 706, 713 (1999). It raises important questions concerning the way in which the Federal Government may exercise its undoubted power to oust state courts of their traditional jurisdiction over common-law tort actions. The rule the Court enforces today was not enacted by Congress and is not to be found in the text of any Executive Order or regulation. It has a unique origin: It is the product of the Court’s interpretation of the Anal commentary accompanying an interim administrative regulation and the history of airbag regulation generally. Like many other judge-made rules, its contours are not precisely defined. I believe, however, that it is fair to state that if it had been expressly adopted by the Secretary of Transportation, it would have read as follows:

“No state court shall entertain a common-law tort action based on a claim that an automobile was negligently or defectively designed because it was not equipped with an airbag;

“Provided, however, that this rule shall not apply to cars manufactured before September 1, 1986, or after such time as the Secretary may require the installation of airbags in all new cars; and

“Provided further, that this rule shall not preclude a claim by a driver who was not wearing her seatbelt that an automobile was negligently or defectively designed because it was not equipped with any passive restraint whatsoever, or a claim that an automobile with particular design features was negligently or defectively designed because it was equipped with one type of passive restraint instead of another.”

Perhaps such a rule would be a wise component of a legislative reform of our tort system. I express no opinion about *888that possibility. It is, however, quite clear to me that Congress neither enacted any such rule itself nor authorized the Secretary of Transportation to do so. It is equally clear to me that the objectives that the Secretary intended to achieve through the adoption of Federal Motor Vehicle Safety Standard 208 would not be frustrated one whit by allowing state courts to determine whether in 1987 the lifesaving advantages of airbags had become sufficiently obvious that their omission might constitute a design defect in some new cars. Finally, I submit that the Court is quite wrong to characterize its rejection of the presumption against pre-emption, and its reliance on history and regulatory commentary rather than either statutory or regulatory text, as “ordinary experience-proved principles of conflict pre-emption.” Ante, at 874.

I

The question presented is whether either the National Traffic and Motor Vehicle Safety Act of 1966 (Safety Act or Act), 80 Stat. 718, 15 U. S. C. §1381 et seq. (1988 ed.),1 or the version of Standard 208 promulgated by the Secretary of Transportation in 1984, 49 CFR §571.208, S4.1.3-S4.1.4 (1998), pre-empts common-law tort claims that an automobile manufactured in 1987 was negligently and defectively designed because it lacked “an effective and safe passive restraint system, including, but not limited to, airbags.” App. 3. In Motor Vehicle Mfrs. Assn. of United States, Inc. v. State Farm Mut. Automobile Ins. Co., 463 U. S. 29, 34-38 (1983), we reviewed the first chapters of the “complex and convoluted history” of Standard 208. It was the “unacceptably high” rate of deaths and injuries caused by automobile accidents that led to the enactment of the Safety Act in 1966. Id., at 33. The purpose of the Act, as stated by Congress, *889was “to reduce traffic accidents and deaths and injuries to persons resulting from traffic accidents.” 15 U. S. C. § 1381. The Act directed the Secretary of Transportation or his delegate to issue motor vehicle safety standards that “shall be practicable, shall meet the need for motor vehicle safety, and shall be stated in objective terms.” § 1392(a). The Act defines the term “safety standard” as a “minimum standard for motor vehicle performance, or motor vehicle equipment performance.” § 1391(2).

Standard 208 covers “[o]ccupant crash protection.” Its purpose “is to reduce the number of deaths, of vehicle occupants, and the severity of injuries, by specifying vehicle crashworthiness requirements . . . [and] equipment requirements for active and passive restraint systems.” 49 CFR § 571.208, S2 (1998). The first version of that standard, issued in 1967, simply required the installation of manual seat-belts in all automobiles. Two years later the Secretary formally proposed a revision that would require the installation of “passive occupant restraint systems,” that is to say, devices that do not depend for their effectiveness on any action by the vehicle occupant. The airbag is one such system.2 The Secretary’s proposal led to a series of amendments to Standard 208 that imposed various passive restraint requirements, culminating in a 1977 regulation that mandated such restraints in all cars by the model year 1984. The two commercially available restraints that could satisfy this mandate *890were airbags and automatic seatbelts; the regulation allowed each vehicle manufacturer to choose which restraint to install. In 1981, however, following a change of administration, the new Secretary first extended the deadline for compliance and then rescinded the passive restraint requirement altogether. In Motor Vehicle Mfrs. Assn., we affirmed a decision by the Court of Appeals holding that this rescission was arbitrary. On remand, Secretary Elizabeth Dole promulgated the version of Standard 208 that is at issue in this case.

The 1984 standard provided for a phase-in of passive restraint requirements beginning with the 1987 model year. In that year, vehicle manufacturers were required to equip a minimum of 10% of their new passenger ears with such restraints. While the 1987 Honda Accord driven by Ms. Geier was not so equipped, it is undisputed that Honda complied with the 10% minimum by installing passive restraints in certain other 1987 models. This minimum passive restraint requirement increased to 25% of 1988 models and 40% of 1989 models; the standard also mandated that "after September 1, 1989, all new cars must have automatic occupant crash protection.” 49 Fed. Reg. 28999 (1984); see 49 CFR § 571.208, S4.1.3-S4.1.4 (1998). In response to a 1991 amendment to the Safety Act, the Secretary amended the standard to require that, beginning in the 1998 model year, all new cars have an airbag at both the driver’s and right front passenger’s positions.3

Given that Secretary Dole promulgated the 1984 standard in response to our opinion invalidating her predecessor’s rescission of the 1977 passive restraint requirement, she provided a full explanation for her decision not to require air*891bags in all cars and to phase in the new requirements. The initial 3-year delay was designed to give vehicle manufacturers adequate time for compliance. The decision to give manufacturers a choice between airbags and a different form of passive restraint, such as an automatic seatbelt, was motivated in part by safety concerns and in part by a desire not to retard the development of more effective systems. 49 Fed. Reg. 29000-29001 (1984). An important safety concern was the fear of a “public backlash” to an airbag mandate that consumers might not fully understand. The Secretary believed, however, that the use of airbags would avoid possible public objections to automatic seatbelts and that many of the public concerns regarding airbags were unfounded. Id., at 28991.

Although the standard did not require airbags in all ears, it is clear that the Secretary did intend to encourage wider use of airbags. One of her basic conclusions was that “[a]u-tomatic occupant protection systems that do not totally rely upon belts, such as airbags . . . , offer significant additional potential for preventing fatalities and injuries, at least in part because the American public is likely to find them less intrusive; their development and availability should be encouraged through appropriate incentives.” Id., at 28963; see also id., at 28966, 28986 (noting conclusion of both Secretary and manufacturers that airbags used in conjunction with manual lap and shoulder belts would be “the most effective system of all” for preventing fatalities and injuries). The Secretary therefore included a phase-in period in order to encourage manufacturers to comply with the standard by installing airbags and other (perhaps more effective) nonbelt technologies that they might develop, rather than by installing less expensive automatic seatbelts.4 As a further incen*892tive for the use of such technologies, the standard provided that a vehicle equipped with an airbag or other nonbelt system would count as 1.5 vehicles for the purpose of determining compliance with the required 10, 25, or 40% minimum passive restraint requirement during the phase-in period. 49 CFR § 571.208, S4.1.3.4(a)(l) (1998). With one oblique exception,5 there is no mention, either in the text of the final standard or in the accompanying comments, of the possibility that the risk of potential tort liability would provide an incentive for manufacturers to install airbags. Nor is there any other specific evidence of an intent to preclude common-law tort actions.

II

Before discussing the pre-emption issue, it is appropriate to note that there is a vast difference between a rejection of Honda’s threshold arguments in favor of federal pre-emption and a conclusion that petitioners ultimately would prevail on their common-law tort claims. I express no opinion on the possible merit, or lack of merit, of those claims. I do observe, however, that even though good-faith compliance with the minimum requirements of Standard 208 would not provide Honda with a complete defense on the merits,61 as*893sume that such compliance would he admissible evidence tending to negate charges of negligent and defective design.7 In addition, if Honda were ultimately found liable, such compliance would presumably weigh against an award of punitive damages. Silkwood v. Kerr-McGee Corp., 485 F. Supp. 566, 583-584 (WD Okla. 1979) (concluding that substantial compliance with regulatory scheme did not bar award of punitive damages, but noting that “[g]ood faith belief in, and efforts to comply with, all government regulations would be evidence of conduct inconsistent with the mental state requisite for punitive damages” under state law).8

The parties have not called our attention to any appellate court opinions discussing the merits of similar no-airbag claims despite the fact that airbag technology was available for many years before the promulgation of the 1984 standard — a standard that is not applicable to any automobiles manufactured before September 1, 1986. Given that an arguable basis for a pre-emption defense did not exist until that standard was promulgated, it is reasonable to infer that the manufacturers’ assessment of their potential liability for compensatory and punitive damages on such claims — even *894without any pre-emption defense — did not provide them with a sufficient incentive to engage in widespread installation of airbags.

Turning to the subject of pre-emption, Honda contends that the Safety Act’s pre-emption provision, 15 U. S. C. § 1392(d), expressly pre-empts petitioners' common-law no-airbag claims. It also argues that the claims are in any event impliedly pre-empted because the imposition of liability in cases such as this would frustrate the purposes of Standard 208. I discuss these alternative arguments in turn.

III

When a state statute, administrative rule, or common-law cause of action conflicts with a federal statute, it is axiomatic that the state law is without effect. U. S. Const., Art. VI, el. 2; Cipollone v. Liggett Group, Inc., 505 U. S. 504, 516 (1992). On the other hand, it is equally clear that the Supremacy Clause does not give unelected federal judges carte blanche to use federal law as a means of imposing their own ideas of tort reform on the States.9 Because of the role of States as separate sovereigns in our federal system, we have long presumed that state laws — particularly those, such as the provision of tort remedies to compensate for personal injuries, that are within the scope of the States’ historic police powers — are not to be pre-empted by a federal statute unless it is the clear and manifest purpose of Congress to do so. Medtronic, Inc. v. Lohr, 518 U. S. 470, 485 (1996); Gade v. National Solid Wastes Management Assn., 505 U. S. 88, 116-117 (1992) (Souter, J., dissenting) ("If the [federal] statute’s terms can be read sensibly not to have a pre-emptive effect, the presumption controls and no pre-emption may be inferred”).

*895When a federal statute contains an express pre-emption provision, “the task of statutory construction must in the first instance focus on the plain wording of [that provision], which necessarily contains the best evidence of Congress’ pre-emptive intent.” CSX Transp., Inc. v. Easterwood, 507 U. S. 658, 664 (1993). The Safety Act contains both an express pre-emption provision, 15 U. S. C. § 1392(d), and a saving clause that expressly preserves common-law claims, § 1397(k). The relevant part of the former provides:

“Whenever a Federal motor vehicle safety standard established under this subchapter is in effect, no State or political subdivision of a State shall have any authority either to establish, or to continue in effect, with respect to any motor vehicle or item of motor vehicle equipment[,] any safety standard applicable to the same aspect of performance of such vehicle or item of equipment which is not identical to the Federal standard.”10

The latter states:

“Compliance with any Federal motor vehicle safety standard issued under this subchapter does not exempt any person from any liability under common law.”11

*896Relying on § 1392(d) and legislative history discussing Congress’ desire for uniform national safety standards,12 Honda argues that petitioners’ common-law no-airbag claims are expressly pre-empted because success on those claims would necessarily establish a state “safety standard” not identical to Standard 208. It is perfectly clear, however, that the term “safety standard” as used in these two sections refers to an objective rule prescribed by a legislature or an administrative agency and does not encompass ease-specific decisions by judges and juries that resolve common-law claims. That term is used three times in these sections; presumably it is used consistently. Gustafson v. Alloyd Co., 513 U. S. 561, 570 (1995). The two references to a federal safety standard are necessarily describing an objective administrative rule. 15 U. S. C. § 1392(a). When the pre-emption provision refers to a safety standard established by a “State or political subdivision of a State,” therefore, it is most naturally read to convey a similar meaning. In addition, when the two sections are read together, they provide compelling evidence of an intent to distinguish between legislative and administrative rulemaking, on the one hand, and common-law liability, on the other. This distinction was certainly a rational one for Congress to draw in the Safety Act given that common-law liability — unlike most legislative or administrative rulemaking — necessarily performs an important remedial role in compensating accident victims. Cf. Silkwood v. Kerr-McGee Corp., 464 U. S. 238, 251, 256 (1984).

It is true that in three recent cases we concluded that broadly phrased pre-emptive commands encompassed common-law claims. In Cipollone v. Liggett Group, Inc., while we thought it clear that the pre-emption provision in the 1965 Federal Cigarette Labeling and Advertising Act applied only to “rulemaking bodies,” 505 U. S., at 518, we concluded that the broad command in the subsequent 1969 *897amendment that “[n]o requirement or prohibition . . . shall be imposed under State law” did include certain common-law claims. Id., at 548-549 (Scalia, J., concurring in judgment in part and dissenting in part).13 In CSX Transp., Inc. v. Easterwood, where the pre-emption clause of the Federal Railroad Safety Act of 1970 expressly provided that federal railroad safety regulations would pre-empt any incompatible state “ ‘law, rule, regulation, order, or standard relating to railroad safety/ ”14 we held that a federal regulation governing maximum train speed pre-empted a negligence claim that a speed under the federal maximum was excessive. And in Medtronic, Inc. v. Lohr, we recognized that the statutory reference to “any requirement” imposed by a State or its political subdivisions may include common-law duties. 518 U. S., at 502-503 (plurality opinion); id., at 503-505 (Breyer, J., concurring in part and concurring in judgment); id., at 509-512 (O’Connor, J., concurring in part and dissenting in part).

The statutes construed in those cases differed from the Safety Act in two significant respects. First, the language in each of those pre-emption provisions was significantly broader than the text of § 1392(d). Unlike the broader language of those provisions, the ordinary meaning of the term “safety standard” includes positive enactments, but does not include judicial decisions in common-law tort cases.

Second, the statutes at issue in Cipollone, CSX, and Med-tronic did not contain a saving clause expressly preserving common-law remedies. The saving clause in the Safety Act *898unambiguously expresses a decision by Congress that compliance with a federal safety standard does not exempt a manufacturer from any common-law liability. In light of this reference to common-law liability in the saving clause, Congress surely would have included a similar reference in § 1392(d) if it had intended to pre-empt such liability. Chicago v. Environmental Defense Fund, 511 U. S. 328, 338 (1994) (noting presumption that Congress acts intentionally when it includes particular language in one section of a statute but omits it in another).

The Court does not disagree with this interpretation of the term “safety standard” in § 1392(d). Because the meaning of that term as used by Congress in this statute is clear, the text of § 1392(d) is itself sufficient to establish that the Safety Act does not expressly pre-empt common-law claims. In order to avoid the conclusion that the saving clause is superfluous, therefore, it must follow that it has a different purpose: to limit, or possibly to foreclose entirely, the possible pre-emptive effect of safety standards promulgated by the Secretary. The Court’s approach to the case has the practical effect of reading the saving clause out of the statute altogether.15

Given the cumulative force of the fact that § 1392(d) does not expressly pre-empt common-law claims and the fact that § 1397(k) was obviously intended to limit the pre-emptive effect of the Secretary’s safety standards, it is quite wrong for the Court to assume that a possible implicit conflict with the purposes to be achieved by such a standard should have the same pre-emptive effect “ ‘as an obstacle to the accomplishment and execution of the full purposes and objectives of Congress.’” Ante, at 873. Properly construed, the Safety Act imposes a special burden on a party relying on an arguable, implicit conflict with a temporary regulatory policy— *899rather than a conflict with congressional policy or with the text of any regulation — to demonstrate that a common-law claim has been pre-empted.

hH <!

Even though the Safety Act does not expressly pre-empt common-law claims, Honda contends that Standard 208 — of its own force — implicitly pre-empts the claims in this case.

“We have recognized that a federal statute implicitly overrides state law either when the scope of a statute indicates that Congress intended federal law to occupy a field exclusively, English v. General Elec. Co., 496 U. S. 72, 78-79 (1990), or when state law is in actual conflict with federál law. We have found implied conflict preemption where it is ‘impossible for a private party to comply with both state and federal requirements,’ id., at 79, or where state law ‘stands as an obstacle to the accomplishment and execution of the full purposes and objectives of Congress.’ Hines v. Davidowitz, 312 U. S. 52, 67 (1941).” Freightliner Corp. v. Myrick, 514 U. S. 280, 287 (1995).

In addition, we have concluded that regulations “intended to pre-empt state law” that are promulgated by an agency acting nonarbitrarily and within its congressionally delegated authority may also have pre-emptive force. Fidelity Fed. Sav. & Loan Assn. v. De la Cuesta, 458 U. S. 141, 153-154 (1982). In this case, Honda relies on the last of the implied pre-emption principles stated in Freightliner, arguing that the imposition of common-law liability for failure to install an airbag would frustrate the purposes and objectives of Standard 208.

Both the text of the statute and the text of the standard provide persuasive reasons for rejecting this argument. The saving clause of the Safety Act arguably denies the Secretary the authority to promulgate standards that would *900pre-empt common-law remedies.16 Moreover, the text of Standard 208 says nothing about pre-emption, and I am not persuaded that Honda has overcome our traditional presumption that it lacks any implicit pre-emptive effect.

Honda argues, and the Court now agrees, that the risk of liability presented by common-law claims that vehicles without airbags are negligently and defectively designed would frustrate the policy decision that the Secretary made in promulgating Standard 208. This decision, in their view, was that safety — including a desire to encourage “public acceptance of the airbag technology and experimentation with better passive restraint systems”17 — would best be promoted *901through gradual implementation of a passive restraint requirement making airbags only one of a variety of systems that a manufacturer could install in order to comply, rather than through a requirement mandating the use of one particular system in every vehicle. In its brief supporting Honda, the United States agreed with this submission. It argued that if the manufacturers had known in 1984 that they might later be held liable for failure to install airbags, that risk “would likely have led them to install airbags in all cars,” thereby frustrating the Secretary’s safety goals and interfering with the methods designed to achieve them. Brief for United States as Amicus Curiae 25.

There are at least three flaws in this argument that provide sufficient grounds for rejecting it. First, the entire argument is based on an unrealistic factual predicate. Whatever the risk of liability on a no-airbag claim may have been prior to the promulgation of the 1984 version of Standard 208, that risk did not lead any manufacturer to install airbags in even a substantial portion of its cars. If there had been a realistic likelihood that the risk of tort liability would have that consequence, there would have been no need for Standard 208. The promulgation of that standard certainly did not increase the pre-existing risk of liability. Even if the standard did not create a previously unavailable pre-emption defense, it likely reduced the manufacturers’ risk of liability by enabling them to point to the regulation and their compliance therewith as evidence tending to negate charges of negligent and defective design. See Part II, supra. Given that the pre-1984 risk of liability did not lead to widespread airbag installation, this reduced risk of liability was hardly likely to compel manufacturers to install airbags in all cars— or even to compel them to comply with Standard 208 during the phase-in period by installing airbags exclusively.

Second, even if the manufacturers’ assessment of their risk of liability ultimately proved to be wrong, the purposes of Standard 208 would not be frustrated. In light of the inevi*902table time interval between the eventual filing of a tort action alleging that the failure to install an airbag is a design defect and the possible resolution of such a claim against a manufacturer, as well as the additional interval between such a resolution (if any) and manufacturers’ “compliance with the state-law duty in question,” ante, at 882, by modifying their designs to avoid such liability in the future, it is obvious that the phase-in period would have ended long before its purposes could have been frustrated by the specter of tort liability. Thus, even without pre-emption, the public would have been given the time that the Secretary deemed necessary to gradually adjust to the increasing use of airbag technology and allay their unfounded concerns about it. Moreover, even if any no-airbag suits were ultimately resolved against manufacturers, the resulting incentive to modify their designs would have been quite different from a decision by the Secretary to mandate the use of airbags in every vehicle. For example, if the extra credit provided for the use of nonbelt passive restraint technologies during the phase-in period had (as the Secretary hoped) ultimately encouraged manufacturers to develop a nonbelt system more effective than the airbag, manufacturers held liable for failing to install passive restraints would have been free to respond by modifying their designs to include such a system instead of an airbag.18 It seems clear, therefore, that any *903potential tort liability would not frustrate the Secretary’s desire to encourage both experimentation with better passive restraint systems and public acceptance of airbags.

Third, despite its acknowledgment that the saving clause “preserves those actions that seek to establish greater safety than the minimum safety achieved by a federal regulation intended to provide a floor,” ante, at 870, the Court completely ignores the important fact that by definition all of the standards established under the Safety Act — like the British regulations that governed the number and capacity of lifeboats aboard the Titanic19 — impose minimum, rather than fixed or maximum, requirements. 15 U. S. C. § 1891(2); see Norfolk Southern R. Co. v. Shanklin, ante, at 359 (Breyer, J., concurring) (“[Fjederal minimum safety standards should not pre-empt a state tort action”); Hillsborough County v. Automated Medical Laboratories, Inc., 471 U. S. 707, 721 (1985). The phase-in program authorized by Standard 208 thus set minimum percentage requirements for the installation of passive restraints, increasing in annual stages of 10, 25, 40, and 100%. Those requirements were not ceilings, and it is obvious that the Secretary favored a more rapid increase. The possibility that exposure to potential tort lia*904bility might accelerate the rate of increase would actually further the only goal explicitly mentioned in the standard itself: reducing the number of deaths and severity of injuries of vehicle occupants. Had gradualism been independently important as a method of achieving the Secretary’s safety goals, presumably the Secretary would have put a ceiling as well as a floor on each annual increase in the required percentage of new passive restraint installations. For similar reasons, it is evident that variety was not a matter of independent importance to the Secretary. Although the standard allowed manufacturers to comply with the minimum percentage requirements by installing passive restraint systems other than airbags (such as automatic seatbelts), it encouraged them to install airbags and other nonbelt systems that might be developed in the future. The Secretary did not act to ensure the use of a variety of passive restraints by placing ceilings on the number of airbags that could be used in complying with the minimum requirements.20 Moreover, even if variety and gradualism had been independently important to the Secretary, there is nothing in the standard, the accompanying commentary, or the history of airbag regulation to support the notion that the Secretary intended to advance those purposes at all costs, without regard to the detrimental consequences that pre-emption of tort liability could have for the achievement of her avowed purpose of reducing vehicular injuries. See Silkwood v. Kerr-McGee Corp., 464 U. S., at 257.

My disagreement with Honda and the Government runs deeper than these flaws, however. In its brief, the Government concedes that “[a] claim that a manufacturer should have chosen to install airbags rather than another type of *905passive restraint in a certain model of car because of other design features particular to that car... would not necessarily frustrate Standard 208’s purposes.” Brief for United States as Amicus Curiae 26, n. 23.21 Petitioners’ claims here are quite similar to the claim described by the Government: their complaint discusses other design features particular to the 1987 Accord (such as the driver’s seat) that allegedly rendered it unreasonably dangerous to operate without an airbag. App. 4-5. The only distinction is that in this ease, the particular 1987 Accord driven by Ms. Geier included no passive restraint of any kind because Honda chose to comply with Standard 208’s 10% minimum requirement by installing passive restraints in other 1987 models. I fail to see how this distinction makes a difference to the purposes of Standard 208, however. If anything, the type of claim favored by the Government — e. g., that a particular model of car should have contained an airbag instead of an automatic seatbelt — would seem to trench even more severely upon the purposes that the Government and Honda contend were behind the promulgation of Standard 208: that having a variety of passive restraints, rather than only airbags, was necessary to promote safety. Thus, I conclude that the Government, on the Secretary’s behalf, has failed to articulate a coherent view of the policies behind Standard 208 that would be frustrated by petitioners’ claims.

V

For these reasons, it is evident that Honda has not crossed the high threshold established by our decisions regarding *906pre-emption of state laws that állegedly frustrate federal purposes: it has not demonstrated that allowing a common-law no-airbag claim to go forward would impose an obligation on manufacturers that directly and irreconcilably contradicts any primary objective that the Secretary set forth with clarity in Standard 208. Gade v. National Solid Wastes Management Assn., 505 U. S., at 110 (Kennedy, J., concurring in part and concurring in judgment); id., at 111 (“A freewheeling judicial inquiry into whether [state law] is in tension with federal objectives would undercut the principle that it is Congress [and federal agencies,] rather than the courts[,] that pre-emp[t] state law”). Furthermore, it is important to note that the text of Standard 208 (which the Court does not even bother to quote in its opinion), unlike the regulation we reviewed in Fidelity Fed. Sav. & Loan Assn. v. De la Cuesta, 458 U. S., at 158, does not contain any expression of an intent to displace state law. Given our repeated emphasis on the importance of the presumption against pre-emption, see, e. g., CSX Transp., Inc. v. Easterwood, 507 U. S., at 663-664; Rice v. Santa Fe Elevator Corp., 331 U. S. 218, 230 (1947), this silence lends additional support to the conclusion that the continuation of whatever common-law liability may exist in a case like this poses no danger of frustrating any of the Secretary’s primary purposes in promulgating Standard 208. See Hillsborough County v. Automated Medical Laboratories, Inc., 471 U. S., at 721; Silkwood v. Kerr-McGee Corp., 464 U. S., at 251 (“It is difficult to believe that [the Secretary] would, without comment, remove all means of judicial recourse for those injured by illegal conduct”).

The Court apparently views the question of pre-emption in this case as a close one. Ante, at 883 (relying on Secretary’s interpretation of Standard 208’s objectives to bolster its finding of pre-emption). Under “ordinary experience-proved principles of conflict pre-emption,” ante, at 874, therefore, the presumption against pre-emption should control. Instead, the Court simply ignores the presumption, *907preferring instead to put the burden on petitioners to show that their tort claim would not frustrate the Secretary’s purposes. Ante, at 882 (noting that petitioners’ arguments “cannot, by themselves, change the legal result”). In view of the important principles upon which the presumption is founded, however, rejecting it in this manner is profoundly unwise.

Our presumption against pre-emption is rooted in the concept of federalism. It recognizes that when Congress legislates “in a field which the States have traditionally occupied...[,] we start with the assumption that the historic police powers of the States were not to be superseded by the Federal Act unless that was the clear and manifest purpose of Congress.” Rice v. Santa Fe Elevator Corp., 331 U. S., at 230; see Jones v. Rath Packing Co., 430 U. S. 519, 525 (1977). The signal virtues of this presumption are its placement of the power of pre-emption squarely in the hands of Congress, which is far more suited than the Judiciary to strike the appropriate state/federal balance (particularly in areas of traditional state regulation), and its requirement that Congress speak clearly when exercising that power. In this way, the structural safeguards inherent in the normal operation of the legislative process operate to defend state interests from undue infringement. Garcia v. San Antonio Metropolitan Transit Authority, 469 U. S. 528, 552 (1985); see United States v. Morrison, ante, at 660-663 (Bkeyer, J., dissenting); Kimel v. Florida Bd. of Regents, 528 U. S. 62, 93-94 (2000) (Stevens, J., dissenting); Allied-Bruce Terminix Cos. v. Dobson, 513 U. S. 265, 292-293 (1995) (THOMAS, J., dissenting); Gregory v. Ashcroft, 501 U. S. 452, 460-464 (1991). In addition, the presumption serves as a limiting principle that prevents federal judges from running amok with our potentially boundless (and perhaps inadequately considered) doctrine of implied conflict pre-emption based on frustration of purposes — i. e., that state law is pre-empted if it “stands as an obstacle to the accomplishment and execu*908tion of the full purposes and objectives of Congress.” Hines v. Davidowitz, 312 U. S. 52, 67 (1941).22

While the presumption is important in assessing the preemptive reach of federal statutes, it becomes crucial when the pre-emptive effect of an administrative regulation is at issue. Unlike Congress, administrative agencies are clearly not designed to represent the interests of States, yet with relative ease they can promulgate comprehensive and detailed regulations that have broad pre-emption ramifications for state law. We have addressed the heightened federalism and nondelegation concerns that agency pre-emption raises by using the presumption to build a procedural bridge across the political accountability gap between States and administrative agencies. Thus, even in cases where implied regulatory pre-emption is at issue, we generally “expect an administrative regulation to declare any intention to pre-empt state law with some specificity.”23 California Coastal *909Common v. Granite Rock Co., 480 U. S. 572, 583 (1987); see Hillsborough County v. Automated Medical Laboratories, Inc., 471 U. S., at 717-718 (noting that too easily implying pre-emption “would be inconsistent with the federal-state balance embodied in our Supremacy Clause jurisprudence,” and stating that “because agencies normally address problems in a detailed manner and can speak through a variety of means, including regulations, preambles, interpretive statements, and responses to comments, we can expect that they will make their intentions clear if they intend for their regulations to be exclusive”); Fidelity Fed. Sav. & Loan Assn. v. De la Cuesta, 458 U. S., at 154 (noting that preemption inquiry is initiated “[w]hen the administrator promulgates regulations intended to pre-empt state law”). This expectation, which is shared by the Executive Branch,24 serves to ensure that States will be able to have a dialog *910with agencies regarding pre-emption decisions ex ante through the normal notice-and-comment procedures of the Administrative Procedure Act (APA), 5 U. S. C. § 553.

When the presumption and its underpinnings are properly understood, it is plain that Honda has not overcome the presumption in this case. Neither Standard 208 nor its accompanying commentary includes the slightest specific indication of an intent to pre-empt common-law no-airbag suits. Indeed, the only mention of such suits in the commentary tends to suggest that they would not be pre-empted. See n. 5, supra. In the Court’s view, however, “[t]he failure of the Federal Register to address pre-emption explicitly is ... not determinative,” ante, at 884, because the Secretary’s consistent litigating position since 1989, the history of airbag regulation, and the commentary accompanying the final version of Standard 208 reveal purposes and objectives of the Secretary that would be frustrated by no-airbag suits. Preempting on these three bases blatantly contradicts the presumption against pre-emption. When the 1984 version of Standard 208 was under consideration, the States obviously were not afforded any notice that purposes might someday be discerned in the history of airbag regulation that would support pre-emption. Nor does the Court claim that the notice of proposed rulemaking that led to Standard 208 provided the States with notice either that the final version of the standard might contain an express pre-emption provision or that the commentary accompanying it might contain a statement of purposes with arguable pre-emptive effect. Finally, the States plainly had no opportunity to comment upon either the commentary accompanying the final version of the standard or the Secretary’s ex post litigating position that the standard had implicit pre-emptive effect.

Furthermore, the Court identifies no case in which we have upheld a regulatory claim of frustration-of-purposes implied conflict pre-emption based on. nothing more than an ex post administrative litigating position and inferences from *911regulatory history and final commentary. The latter two sources are even more malleable than legislative history. Thus, when snippets from them are combined with the Court’s broad conception of a doctrine of frustration-of-purposes pre-emption untempered by the presumption, a vast, undefined area of state law becomes vulnerable to preemption by any related federal law or regulation. In my view, however, “preemption analysis is, or at least should be, a matter of precise statutory [or regulatory] construction rather than an exercise in free-form judicial policymaking.” 1 L. Tribe, American Constitutional Law §6-28, p. 1177 (3d ed. 2000).

As to the Secretary’s litigating position, it is clear that “an interpretation contained in a [legal brief], not one arrived at after, for example, a formal adjudication or notice-and-comment rulemaking[,]... do[es] not warrant Chevron-style deference.” Christensen v. Harris County, ante, at 587. Moreover, our pre-emption precedents and the APA establish that even if the Secretary’s litigating position were coherent, the lesser deference paid to it by the Court today would be inappropriate. Given the Secretary’s contention that he has the authority to promulgate safety standards that pre-empt state law and the fact that he could promulgate a standard such as the one quoted supra, at 887, with relative ease, we should be quite reluctant to find preemption based only on the Secretary’s informal effort to recast the 1984 version of Standard 208 into a pre-emptive mold.25 See Hillsborough County v. Automated Medical *912Laboratories, Inc., 471 U. S., at 721; cf. Medtronic, Inc. v. Lohr, 518 U. S., at 512 (O’Connor, J., concurring in part and dissenting in part) (“It is not certain that an agency regulation determining the pre-emptive effect of any federal statute is entitled to deference”); Smiley v. Citibank (South Dakota), N. A., 517 U. S. 735, 743-744 (1996). Requiring the Secretary to put his pre-emptive position through formal notiee-and-eomment rulemaking — whether contemporaneously with the promulgation of the allegedly pre-emptive regulation or at any later time that the need for pre-emption "becomes apparent26 — respects both the federalism and non-delegation principles that underlie the presumption against pre-emption in the regulatory context and the APA’s requirement of new rulemaking when an agency substantially modifies its interpretation of a regulation. 5 U. S. C. §551(5); Paralyzed Veterans of America v. D. C. Arena L. P., 117 F. 3d 579, 586 (CADC 1997); National Family Planning & Reproductive Health Assn. v. Sullivan, 979 F. 2d 227, 240 (CADC 1992).

* * *

Because neither the text of the statute nor the text of the regulation contains any indication of an intent to pre-empt *913petitioners’ cause of action, and because I cannot agree with the Court’s unprecedented use of inferences from regulatory history and commentary as a basis for implied pre-emption, I am convinced that Honda has not overcome the presumption against pre-emption in this case. I therefore respectfully dissent. - .