dissenting:
/
This case is less about bullets than about federal/state relations. It raises important questions of when it is appropriate for this court to certify issues of New York law to the New York Court of Appeals. I believe that federal courts in general, and this circuit in particular, have tended to be far too reluctant to certify questions to the state courts. See, e.g., Martin Flumenbaum & Brad S. Karp, Certification of Unsettled Law Issues, N.Y.L.J., Jan. 29, 1992, at 3 (noting that the procedure for certification to the New York Court of Appeals “has been used only sparingly by the Second Circuit,” which in 1992 had certified only five issues over the preceding six-year period). Specifically, federal courts have all too often refused to certify when they can rely on state lower court opinions to define state law. I view this reluctance as both wrong and unjust.
Reluctance to certify is wrong because it leads to precisely the kind of forum shopping that Erie R.R. Co. v. Tompkins, 304 U.S. 64, 73-77, 58 S.Ct. 817, 819-22, 82 L.Ed. 1188 (1938), was intended to prevent. See Hanna v. Plumer, 380 U.S. 460, 468, 85 S.Ct. 1136, 1142, 14 L.Ed.2d 8 (1965) (noting that one of the aims of the Eñe decision was “discouragement of forum-shopping”). This is especially so in situations where there is some law in the intermediate state courts, but no definitive holding by the state’s highest tribunal. In such cases, and in the absence of certification, the party that is favored by the *158lower court decisions will almost invariably seek federal jurisdiction.1 It will do this in order to prevent the state’s highest court from reaching the issue, in the expectation that the federal court — unlike the state’s highest court — will feel virtually bound to follow the decisions of the intermediate state courts. See In re Brooklyn Navy Yard Asbestos Litig., 971 F.2d 831, 850 (2d Cir.1992) (suggesting that “lower state court decisions ... have great weight in informing” the decisions of the federal courts) (citation and internal quotation marks omitted). The result is to “funnel all similar litigation [in]to federal court.” Todd v. Societe BIC, S.A., 9 F.3d 1216, 1222 (7th Cir.1993) (en banc).
If the federal court treats the plaintiff more favorably than the state tribunal would, then the plaintiff always files in federal court; similarly any departure in the manufacturer’s favor leads the defendant to remove any suit filed in state court. In either case, the state loses the ability to develop or restate the principles that it believes should govern the category of cases. Certification then ensures that the law we apply is genuinely state law.
Id.2
When federal courts, in effect, prevent state courts from deciding unsettled issues of state law, they violate fundamental principles of federalism and comity. As the Supreme Court has noted, while certification may cause delay in any given case, “[i]t does, of course, in the long run save time, energy, and resources and helps build a cooperative judicial federalism.” Lehman Bros. v. Schein, 416 U.S. 386, 391, 94 S.Ct. 1741, 1744, 40 L.Ed.2d 215 (1974); see also Arizonans For Official English v. Arizona, — U.S. -, -, 117 S.Ct. 1055, 1073, 137 L.Ed.2d 170 (1997). Federal courts that refuse to certify end up “mak[ing] important state policy, in contravention of basic federalism principles.” Hakimoglu v. Trump Taj Mahal Assocs., 70 F.3d 291, 302 (3d Cir.1995) (Becker, J., dissenting); see also Richard Alan Chase, Note, A State Court’s Refusal to Answer Certified Questions: Are Inferences Permitted?, 66 St. John’s L. Rev. 407, 415 (1992) (“A federal court demonstrates respect for state sovereignty when it certifies a question to the state’s highest court and defers to its judgment on unresolved issues of state law.”).
Reluctance to certify is unjust because, as has happened with some frequency,3 the federal court, having refused to certify, may decide an issue of state law one way, only to *159discover that the state’s highest court, when presented with the issue in a later case, reaches the opposite result. In such a situation, the party who lost in federal court has been unjustly denied her state-law rights, and often has been left with no means of effective redress.
The case of DeWeerth v. Baldinger provides a striking example. In that case, the plaintiff, DeWeerth, had originally prevailed in the federal district court, which ruled that as a matter of New York law she had established ownership of a valuable impressionist painting that had been stolen from her forty years earlier. See DeWeerth v. Baldinger, 658 F.Supp. 688 (S.D.N.Y.1987). On appeal, this court reversed the district court’s judgment, finding that New York limitations law requires a showing of reasonable diligence in locating stolen property, and that DeWeerth had failed to make such a showing. See DeWeerth v. Baldinger, 836 F.2d 103, 108-12 (2d Cir.1987). We elected not to certify this question of New York law to the New York Court of Appeals, endeavoring instead to resolve the issue on our own. See id. at 108 n. 5.
Three years later, the New York Court of Appeals was presented with precisely the same issue, and held that the statute of limitations does not require a showing of reasonable diligence. See Solomon R. Guggenheim Found. v. Lubell, 77 N.Y.2d 311, 567 N.Y.S.2d 623, 626-27, 569 N.E.2d 426, 429-30 (1991). The Court of Appeals remarked, somewhat acidly:
Although the [Second Circuit in DeWeerth ] acknowledged that the question posed by the case was an open one, it declined to certify it to this Court, stating that it did not think that it “[would] recur with sufficient frequency to warrant use of the certification procedure.” Actually, the issue has recurred several times in the three years since DeWeerth was decided, including the case now before us. We have reexamined the relevant New York case law and we conclude that the Second Circuit should not have imposed a duty of reasonable diligence on the owners of stolen art work for purposes of the Statute of Limitations.
Id. (citations omitted).
Following the New York Court of Appeals’ decision in Guggenheim, DeWeerth attempted to reopen her case in order to have her dispute decided in accordance with the actual law of New York. See DeWeerth v. Balding-er, 804 F.Supp. 539 (S.D.N.Y.1992). Understandably, however, this court ruled that the subsequent clarification of state law did not justify disturbing a final judgment. See DeWeerth v. Baldinger, 38 F.3d 1266, 1272-73 (2d Cir.1994). Thus, despite the fact that “the DeWeerth panel’s prediction was wrong,” id. at 1273, and that DeWeerth was, in fact, entitled to the painting under New York law, she was left without a remedy. This happened not because of any decision by the highest court of New York, but rather because of the will of the federal courts.4
Before the advent of the certification procedure, this court had little choice but to “do [its] best to predict what the highest state *160court w[ould] do.” Cooper v. American Airlines, Inc., 149 F.2d 355, 359 (2d Cir.1945).5 But now, “[c]ertification is an alternative to prognostication,” Todd, 9 F.3d at 12226 — an alternative that respects the right of the state courts to define their own substantive law, deters forum shopping, and ensures that litigants like Gerta DeWeerth (and Carolyn McCarthy) will have their state-law created rights determined by “the only judges who can give definitive answers” as to those rights. Todd, 9 F.3d at 1222. It is, moreover, an alternative that does this without in any way requiring state courts to decide anything they do not wish to take on.
This last is why the statement quoted by the majority, ante, Op. at 153, that the certification “procedure must not be a device for shifting the burdens of this Court to those whose burdens are at least as great,” Dorman v. Satti, 862 F.2d 432, 435 (2d Cir.1988) (citation and internal quotation marks omitted), while true enough, can rarely be determinative. In the first place, it is well known that state court judges have expressed both publicly and privately their desire for certification 7 and their irritation with the fact that federal courts often decide interesting and important questions rather than certifying them to the courts that should be deciding them. More importantly, a state court that feels overburdened, or that for any other reason does not wish to decide the certified question, is always free to refuse to answer it. See, e.g., N.Y. Comp.Codes R. & Regs, tit 22, § 500.17(d) (noting that the New York Court of Appeals “will examine the merits presented by the certified question ... to determine whether to accept the certification”); Ira P. Robbins, The Uniform Certification of Questions of Law Act: A Proposal for Reform, 18 J. Legis. 127, 137 (1992) (“[T]he ultimate power to accept or reject a certified question rests exclusively in the discretion of the answering court. Th[is] procedural safeguard[] more than proteet[s] the answering court from a surfeit of certification cases because as a practical matter that court completely controls its docket and may reject certified-question cases if the number becomes overwhelming. The answering court need not even offer a reason for declining to answer____”) (footnotes omitted).
Just as the Supreme Court of the United States can decide for itself whether it wishes to grant certiorari, so too can state courts make analogous decisions as to certification. If anything, it is the failure to certify when certification is plausible that burdens state courts, because such a failure puts pressure *161on the state courts to accept direct appeals in subsequent cases in order to rectify a federal court error and to do this even at times when it might be inconvenient for them to do so.
This is not to say that certification should be used indiscriminately. As Judge Cabranes has explained, federal courts must be careful to certify only in appropriate cases. See L. Cohen & Co., Inc. v. Dun & Bradstreet, Inc., 629 F.Supp. 1419, 1422-24 (D.Conn.1986); see also Barnes v. Atlantic and Pac. Life Ins. Co., 514 F.2d 704, 705 n. 4 (5th Cir.1975) (“[W]e use much judgment, restraint and discretion in certifying. We do not abdicate.”). But “appropriate” must mean virtually any case in which 1) a significant and dispositive issue of state law is in genuine doubt (despite the existence of nonbinding lower court decisions, especially where those decisions — like the ones relied upon by the majority — were never appealed to the state’s highest court), and 2) certification is specifically requested by the party that did not invoke federal court jurisdiction. By such a standard, I believe that certification is not used enough, and that cases like the one before us are especially suited to its use.8
II
In cases that are dramatic and involve “hot” issues, there is a tendency for the parties to describe themselves as raising new issues that are remarkable in their legal context.9 But in fact, such cases are usually best looked at in the most traditional of ways. Courts must see how these cases fit into old categories before considering whether it is either necessary or proper to expand those old categories or to create new ones. And so it is with the ease before us. For this reason, I begin with the most traditional of the causes of action that the plaintiffs have raised — negligence—and address it in its most “black letter” terms. Cf. Andrew Jay McClurg, The Tortious Marketing of Handguns: Strict Liability is Dead, Long Live Negligence, 19 Seton Hall Legis. J. 777, 778 (1995) (“[Vjictims of gun violence and their lawyers should refocus their sights on the more prosaic liability theory of common law negligence. In other words, it is time to go back to basics.”). In doing this I do not, of course, seek to determine whether liability for negligence lies in a case like this one in New York. I examine the issue only to discern whether the question is sufficiently open to warrant certification.
A. Liability for negligence
To hold a defendant hable in negligence in New York, a plaintiff must show: 1) a duty on the part of the defendant; 2) a breach of that duty by conduct involving an “unreasonable risk of harm,” W. Page Keeton, Dan B. Dobbs, Robert E. Keeton & David G. Owen, Prosser and Keeton on the Law of Torts, § 53 at 358 (5th ed.1984) [hereinafter Prosser & Keeton]; 3) damages suffered by the plaintiff; and 4) causation, both in fact and proximate, between the breach and the plaintiffs harm. See, e.g., Febesh v. Elcejay Inn Corp., 157 A.D.2d 102, 555 N.Y.S.2d 46, 47 (1990); Stagl v. Delta Airlines, Inc., 52 F.3d 463, 467 (2d Cir.1995). So viewed, three of the four elements of a cause of action for negligence — damages, causation, and conduct involving an unreasonable risk of harm — are either readily present or sufficiently cognizable under New York law on the facts of this case that a federal court would err mightily to hold on its own to the contrary. The fourth, the existence of a duty, is a more difficult question. But it is one that, I submit, only the New York Court of Appeals (or the New York Legislature) can answer.
1. Conduct involving an unreasonable risk of harm
The plaintiffs alleged in their complaint that the defendant created an unreasonable *162risk of harm both by manufacturing Black Talons10 and by marketing them to the general public. On appeal, the plaintiffs have pursued only their negligent marketing claim.
Could a New York jury find that there was an undue risk of harm, if not in producing Black Talons, then in advertising them for use by (and selling them to) the general public? Put differently, could a jury find that the benefit gained by making Black Talons available to the public was outweighed by their potential harm.11 I believe that a reasonable jury could so find.
The distinctive feature of the Black Talon bullet is that it “is designed to expand upon impact exposing razor-sharp edges at a 90-degree angle to the bullet. This expansion dramatically increases the wounding power of the bullets.” McCarthy v. Sturm, Ruger and Co., 916 F.Supp. 366, 368 (S.D.N.Y. 1996); see also Winchester Halts Sale of Disintegrating Bullet to Public, Salt Lake Trib., Nov. 23, 1993, at A4 [hereinafter Winchester Halts Sale ] (“The 9mm bullets disintegrate upon impact into tiny razor-like claws that tear up tissues and organs.”); Wendell Jamieson, Winchester Disarms a Killer Bullet, Newsday (New York), Nov. 23,1993, at 5 (noting that “the devastating bullets ... bloom into claw-like prongs on impact and canse hideous, gaping wounds”). The defendant marketed the product to civilian consumers in a slick black box bearing the words “BLACK TALON” surrounded by the sharp, gleaming claws of a bird of prey. It is certainly possible that a fact-finder would conclude that any benefit that comes from marketing and selling bullets with this additional destructive feature to private citizens is outweighed by their potential for significant harm.
There is nothing novel in finding negligence on these grounds. New York law recognizes that a defendant can be held liable for negligently marketing a product. See, e.g., Kaufman v. Eli Lilly and Co., 65 N.Y.2d 449, 492 N.Y.S.2d 584, 589, 482 N.E.2d 63, 68 (1985); Bikowicz v. Sterling Drug, Inc., 161 A.D.2d 982, 557 N.Y.S.2d 551, 552 (1990). And there is no reason why that principle should not allow recovery against a manufacturer who introduces a harmful product into general circulation, where the social utility of marketing that product to the public is outweighed by its risk of harm. See MeClurg, supra; Joseph A. Page, Liability for Unreasonably and Unavoidably Unsafe Products: Does Negligence Doctrine Have a Role to Play?, 72 Chl-Kent. L. Rev. 87 (1996);12 see also Ezagui v. Dow Chem. Corp., 598 F.2d *163727, 736 (2d Cir.1979) (applying New York law) (finding a prima facie case of negligence because “Parke-Davis, as a matter of law, owed plaintiff a duty to exercise reasonable care in the design and in the decision to market Quadrigen”) (emphasis added).13
The fact that the New York legislature has not chosen to forbid the distribution of Black Talons in no way alters the conclusion that the defendant may have been negligent in marketing them to the general public.14 There are all sorts of situations in which the general distribution of an object is legal, but the decision to market and sell it to certain persons is nonetheless negligent because it poses an undue risk of harm. When fireworks were legal, it was still negligent to market and sell them to children. See Erwin v. Dunn, 201 S.W.2d 240, 242 (Tex.Civ.App. 1947). Similarly, even in the absence of a statute, serving alcohol to intoxicated adults is negligent. And this is so despite the fact that serving alcohol to sober ones is not. See Morrissey v. Sheedy, 26 A.D.2d 683, 272 N.Y.S.2d 430, 431-32 (1966); Tyrrell v. Quigley, 186 Misc. 972, 60 N.Y.S.2d 821, 822 (N.Y.Sup.Ct.1946). Selling tanks to the armed forces is fine; selling them to the general public is, I would think, clearly negligent.
I therefore conclude that there is little doubt that if the New York Court of Appeals *164chooses not to allow this type of negligent marketing claim to go forward, it will do so not because it holds, as a matter of law, that the defendant’s actions did not create an unreasonable risk of harm, but rather because it finds no duty, that is, because it believes that sound public policy does not dictate the imposition of liability, despite the fact that the defendant may have acted negligently. See infra Part II.A.4; McClurg, supra, at 795-99; Page, supra, at 98-99.
2. Damages
No one questions that the plaintiffs suffered harm of the most serious sort on the evening of December 7, 1993. Carolyn McCarthy lost her husband, who was shot in the head with a Black Talon bullet. Kevin McCarthy was shot in the head and hand, and now suffers paralysis. Maryanne Phillips was shot in her left arm, which was nearly destroyed by a Black Talon bullet. Whether these damages can be said to have resulted from the defendant’s decision to market Black Talons is a question of causation, and is considered immediately below.
S. Causation
Under New York law, a plaintiff alleging negligence must prove that the defendant’s breach of duty was a cause of the plaintiffs injury. Generally, there must be cause in fact, a causal tendency (that is, a “causal link”), and proximate cause. See Guido Calabresi, Concerning Cause and the Law of Torts: An Essay for Harry Kalven, Jr., 43 U. Chi. L. Rev. 69, 71-73 (1975) [hereinafter Calabresi, Concerning Cause ].
The first of these, the requirement that the defendant’s negligent act was a but for cause of the injury, see, e.g., Saulpaugh v. State, 132 A.D.2d 781, 517 N.Y.S.2d 328, 329 (1987), is readily met. To establish cause in fact, plaintiffs must generally prove that the specific damages they suffered resulted from the defendant’s unduly risky conduct, and would not have occurred in the absence of the defendant’s negligence.15 In this case, cause in fact necessitates a showing that the damages complained of resulted from the uniquely destructive nature of the Black Talons. This is so because it is the marketing of bullets of this sort, with these uniquely destructive characteristics, that the plaintiffs allege to be unduly risky.
The plaintiffs have clearly made such a showing here. Their allegations, if true, will support a finding that, but for the defendant’s marketing of its product to persons like Colin Ferguson, the plaintiffs would not have suffered such extensive injuries. Indeed, the majority concedes that “[t]he injuries to Dennis and Kevin McCarthy and Maryanne Phillips were enhanced by the ripping and tearing action of the Black Talons.” Ante, Op. at 151; cf. Bolm v. Triumph Corp., 33 N.Y.2d 151, 350 N.Y.S.2d 644, 647-48, 305 N.E.2d 769, 771-72 (1973) (holding that a manufacturer can be held liable in negligence “for defects in design which do not cause accidents but do enhance or aggravate injuries”).
Similarly there can be no doubt that a causal tendency was shown.16 Knowing what we know now, we can surely say that the negligent marketing of bullets like the Black Talon increases the chances of such injuries occurring. Cf. Martin v. Herzog, 228 N.Y. 164, 126 N.E. 814, 816 (1920) (“[Ejvidence of a collision occurring more than an hour after sundown between a car and an unseen buggy, proceeding without lights, is evidence from which a causal connection may be inferred between the collision and the lack of signals.”).
*165The existence of proximate cause is more of a question, but under New York law it too could properly be found by a jury. Despite the fact that the question of proximate cause is normally one for the fact finder, see Derdiarian v. Felix Contracting Corp., 51 N.Y.2d 308, 434 N.Y.S.2d 166, 167, 414 N.E.2d 666, 668 (1980), the district court in the instant case concluded that, as a matter of law, the defendant’s alleged negligence was not a proximate cause of the plaintiffs’ injuries. See McCarthy, 916 F.Supp. at 372. This holding was erroneous.
While it is true “that an intervening intentional or criminal act will generally sever the liability of the original tort-feasor,” New York law is clear that “[t]hat doctrine has no application when the intentional or criminal intervention of a third party or parties is reasonably foreseeable.” Kush v. City of Buffalo, 59 N.Y.2d 26, 462 N.Y.S.2d 831, 835, 449 N.E.2d 725, 729 (1983). Moreover, “[a]n intervening act may not serve as a superseding cause, and relieve an actor of responsibility, where the risk of the intervening act occurring is the very same risk which renders the actor negligent.” Derdiarian, 434 N.Y.S.2d at 170, 414 N.E.2d at 671. It cannot be said that criminal acts like Colin Ferguson’s were, as a matter of law, not foreseeable. And it is precisely the unreasonable risk of this type of criminal activity that forms the basis of the plaintiffs’ negligence claim. Accordingly, whether the defendant’s alleged breach was a proximate cause of the plaintiffs’ injuries is a question for the jury in this case.
Jp. Duty
Three of the four elements of negligence liability have, without question, been sufficiently alleged under New York law. The only aspect of this case — viewed as a negligence action — that is problematic is the existence of a duty. “In order to establish a prima facie case of negligence under New York law, a claimant must show that ... the defendant owed the plaintiff a cognizable duty of care....” Stagl, 52 F.3d at 467; see also Strauss v. Belle Realty Co., 65 N.Y.2d 399, 492 N.Y.S.2d 555, 557, 482 N.E.2d 34, 36 (1985). In this case, the plaintiffs must show that the defendant was under a duty to protect them from the harm caused by foreseeable criminal intervenors, and that such a duty entailed refraining from aggressively marketing its product to the general public.
In many jurisdictions, the existence of a duty depends primarily on 1) the foreseeability of harm to the plaintiff that would flow from the defendant’s negligent acts,17 and 2) the absence of actions on the part of the plaintiff — like trespass or assumption of risk — that either destroy or lessen the defendant’s liability.18 In those jurisdictions, it is rarely worthwhile discussing duty separately from the existence of negligence, proximate cause, and special plaintiff attributes. In the absence of such plaintiff-based defenses, a general duty to refrain from causing foreseeable and unreasonable harm is assumed to exist.
That is not, however, the law in New York. Under New York law, the question of the existence of a “[djuty in negligence cases is [not] defined ... by foreseeability of injury.” Strauss, 492 N.Y.S.2d at 557, 482 N.E.2d at 36; see also Pulka v. Edelman, 40 N.Y.2d 781, 390 N.Y.S.2d 393, 396, 358 N.E.2d 1019, 1022 (1976) (“Foreseeability should not be confused with duty____ [Foreseeability] is applicable to determine the scope of duty— only after it has been determined that there *166is a duty.”). Rather, the determination of the existence of a duty is a decision of public policy that the New York Court of Appeals has expressly retained for itself (and the courts generally):
The question of the scope of an alleged tort-feasor’s duty is, in the first instance, a legal issue for the court to resolve. In this analysis, not only logic and science, but policy play an important role. The common law of torts is, at its foundation, a means of apportioning risks and allocating the burden of loss. While moral and logical judgments are significant components of the analysis, we are also bound to consider the larger social consequences of our decisions and to tailor our notion of duty so that the legal consequences of wrongs are limited to a controllable degree.
Waters v. New York City Housing Auth., 69 N.Y.2d 225, 513 N.Y.S.2d 356, 358, 505 N.E.2d 922, 923-24 (1987) (citations, brackets, and internal quotation marks omitted).
Unlike foreseeability and causation, which are issues generally and more suitably entrusted to fact finder adjudication, the definition of the existence and scope of an alleged tortfeasor’s duty is usually a legal, policy-laden declaration reserved for Judges to make prior to submitting anything to fact-finding or jury consideration. Common-law experience teaches that duty is not something derived or discerned from an algebraic formula. Rather, it coalesces from vectored forces including logic, science, weighty competing socioeconomic policies and sometimes contractual assumptions of responsibility. These sources contribute to pinpointing and apportioning of societal risks and to an allocation of burdens of loss and reparation on a fair, prudent basis.
... Courts traditionally and as part of the common-law process fix the duty point by balancing factors, including the reasonable expectations of parties and society generally, the proliferation of claims, the likelihood of unlimited or insurer-like liability, disproportionate risk and reparation allocation, and public policies affecting the expansion or limitation of new channels of liability.
Palka v. Servicemaster Management Servs. Corp., 83 N.Y.2d 579, 611 N.Y.S.2d 817, 820-21, 634 N.E.2d 189, 192-93 (1994).
This does not mean that the court is required- — or even permitted — to weigh such policy considerations to determine the existence of a duty in each individual New York negligence ease. Once the New York Court of Appeals has established that the relationship between plaintiffs and defendants in certain circumstances or categories of cases suffices to establish a duty of due care, all eases of like kind are covered by that finding, and there is no warrant to take a case from the jury for a separate judicial examination of duty.
It follows that, before we can be confident that there is a jury question as to negligence in this case, we must find precedents that establish a duty between the parties in cases akin to this one. I am not prepared to make such a finding. Nor, however, am I prepared to say that the New York Court of Appeals would not find that such a precedent exists or create one in this case. I am not, in other words, satisfied that the New York Court of Appeals has made a policy determination, one way or the other, in circumstances akin to those here.
The majority, instead, has no difficulty concluding that New York courts have determined that there is no duty here. It relies on the New York Court of Appeals’ statement that
[i]n the ordinary circumstance, common law in the State of New York does not impose a duty to control the conduct of third persons to prevent them from causing injury to others; liability for the negligent acts of third persons generally arises when the defendant has authority to control the actions of such third persons. This is so, we have said, even where “as a practical matter” defendant could have exercised such control.
Purdy v. Public Adm’r of Westchester, 72 N.Y.2d 1, 530 N.Y.S.2d 513, 516, 526 N.E.2d 4, 7 (1988) (citations omitted).
*167But the issue is nowhere near as simple as that. New York law is, in fact, quite clear that the defendant in many circumstances can be under a duty to the plaintiff that makes him liable for the harm caused by the intervening negligent acts of a third party. See, e.g., Modave v. Long Island Jewish Med. Ctr., 501 F.2d 1065, 1072 (2d Cir.1974) (Friendly, J.) (reciting “the unassailable proposition of New York law ‘that a wrongdoer is liable for the ultimate result, though the mistake or even negligence of the physician who treated the injury may have increased the damage which would otherwise have followed from the original wrong’”) (quoting Milks v. McIver, 264 N.Y. 267, 190 N.E. 487, 488 (1934)).
Moreover, in cases such as this one involving the introduction of goods into the stream of commerce, New York courts have had little difficulty in holding the original seller to have a duty not only to the purchaser and parties having a direct relationship with the purchaser, but also to third-party bystanders. Thus, in the celebrated case of Codling v. Paglia, 32 N.Y.2d 330, 345 N.Y.S.2d 461, 298 N.E.2d 622 (1973), the New York Court of Appeals found a duty to a bystander injured by the defendant’s product (an automobile) even though the defendant seller had been found to be free from negligence by a jury and even though the negligence of the user/purehaser of the car was in part the cause of the bystander’s injuries. See id., 345 N.Y.S.2d at 463, 298 N.E.2d at 624 (holding that, despite the negligent intervenor, “the manufacturer of a defective product may be held liable to an innocent bystander ... for damages sustained in consequence of the defect”).19
In fact, under appropriate conditions, a defendant can even be held liable for the intervening criminal acts of a third party. See, e.g., Stagl, 52 F.3d at 467; Nallan v. Helmsley-Spear, Inc., 50 N.Y.2d 507, 429 N.Y.S.2d 606, 612-18, 407 N.E.2d 451, 457-58 *168(1980) (holding that a commercial landlord has a duty to take reasonable precautionary measures to minimize the risk of foreseeable criminal activity and to make the premises safe for the visiting public); Stevens v. Kirby, 86 A.D.2d 391, 450 N.Y.S.2d 607, 610 (1982) (“A tavern owner owes a duty to his patrons to protect them from personal attack when he has reasonable cause to anticipate conduct on the part of third persons which is likely to endanger their safety.”).
What then does the New York Court of Appeals mean when it says that it has
imposed a duty to control the conduct of others where there is a special relationship: a relationship between defendant and [an intervenor] whose actions expose plaintiff to harm such as would require the defendant to attempt to control the third person’s conduct; or a relationship between the defendant and plaintiff requiring defendant to protect the plaintiff from the conduct of others[?]
Purdy, 530 N.Y.S.2d at 516, 526 N.E.2d at 7.
It cannot mean that a duty will be imposed only where there is privity between the defendant and the victim. See, e.g., Codling, 345 N.Y.S.2d at 461-65, 298 N.E.2d at 622-25 (finding that a defendant can be held liable to a third-party bystander). Nor can it mean that the defendant must actually be able to control the intervenor. See, e.g., Modave, 501 F.2d at 1072 (finding that a negligent doctor can be held liable for the harm caused by another negligent doctor at a different hospital). Rather, the cases in which New York courts have refused to find the requisite relationship are ones in which the defendant did not have much of an opportunity to prevent the harm, see Pulka, 390 N.Y.S.2d at 395-96, 358 N.E.2d at 1021, and in which imposing a duty would expose the defendant to liability “beyond sound public policy,” Einhorn v. Seeley, 136 A.D.2d 122, 525 N.Y.S.2d 212, 216 (1988); see also Waters, 513 N.Y.S.2d at 358-59, 505 N.E.2d at 924. That is, they are the rather unusual cases where the allowance of liability would create “disproportionate risk and reparation allocation, and [would violate] public policies affecting the expansion or limitation of new channels of liability.” Polka, 611 N.Y.S.2d at 821, 634 N.E.2d at 193. Thus, “ ‘duty’ is not sacrosanct in itself, but is only an expression of the sum total of those considerations of policy which lead the law to say that the plaintiff is entitled to protection.” Prosser and Keeton, swpra, § 53 at 358.
What of this case? On the one hand, it seems that the defendant could have substantially reduced the harm caused by these unusually destructive bullets by not marketing them to the general public. And the danger of exposing the defendant to liability beyond sound public policy might not be present here, especially if the New York courts were to conclude that marketing Black Talons to the general public causes more harm than benefit.20 On the other hand, this case may well involve “the expansion ... of new channels of liability,” Palka, 611 N.Y.S.2d at 821, 634 N.E.2d at 193, since it involves a criminal intervenor in a case where no direct relationship exists between the injured plaintiff and the defendant. In this respect it is different from Modave (negligent intervenor and direct relationship between plaintiff and defendant), Codling (negligent intervenor and no direct relationship), and Nailon and Stevens (criminal intervenor and direct relationship).
Under the circumstances, it is hard to know whether the New York Court of Appeals would find a duty. The fact that the foreseeable intervenor behaved in a criminal, rather than a negligent, manner does not change matters for the purposes of proximate cause. See Kush, 462 4N.Y.S.2d at 835, 449 N.E.2d at 729 (noting that, although “an intervening intentional or criminal act will generally sever the liability of the original tort-feasor,” New York law is clear that “[t]hat doctrine has no application when the intentional or criminal intervention of a third party or parties is reasonably foreseeable”). Does it do so for purposes of duty? That is a question to which the New York Court of *169Appeals has given us no answer21 (although it should be noted that Codling makes not even a hint of a suggestion that the result would have been different if the intervenor had been a criminal). Since the Court of Appeals has neither countenanced nor foreclosed liability in cases involving a criminal intervenor and the absence of a direct relationship, I believe that we are bound to allow that Court to make a policy determination of duty in the instant case.
In this respect, the argument that, because it is legal to sell and advertise Black Talons, there can be no liability, is misplaced. As Judge Gabrielli has noted, “[t]he common-law duty of reasonable care to those within the ambit of foreseeable danger requires no buttressing by legislative enactment; nor does the absence of such legislation in the present instance exclude the possibility of liability.” Pulka, 390 N.Y.S.2d at 398, 358 N.E.2d at 1023-24 (Gabrielli, J., dissenting). There is all of the difference in the world between making something illegal and making it tortious. Making an activity tortious forces the people who derive benefit from it to internalize the costs associated with it, thereby making sure that the activity will only be undertaken if it is desired by enough people to cover its costs.22 It does not proscribe it *170altogether. As a result, very different policy considerations go into the decision of whether to forbid something and the decision of whether to find a duty that permits liability for the harm it causes. The fact that the New York legislature has failed to prohibit Black Talons is certainly one factor that the New York Court of Appeals is likely to consider in deciding the “policy-laden declaration reserved for Judges” that it must make in resolving whether a duty exists. Palka, 611 N.Y.S.2d at 820, 634 N.E.2d at 192. But the weight to be given to this factor is just the sort of thing that only the New York Court of Appeals itself can determine.23
It is precisely because the public policy decision of whether to impose a legal duty is one that the New York Court of Appeals has retained for itself that we should be especially inclined to certify the question to that Court, rather than to endeavor to resolve it ourselves. Certification is particularly warranted when “the resolution of [the issue] requires a careful weighing of competing state policies and potentially could affect a large number of people.” Jackson v. JohnsManville Sales Corp., 757 F.2d 614, 615 (5th Cir.1985). We are simply not entitled to make policy for the State of New York. Such is a task for the State’s highest tribunal.24
B. Strict products liability
The requirements for strict products liability in New York are not markedly dissimilar from those for negligent products liability. Once again: a) a duty is needed; b) that duty must be breached by the defendant’s manufacture or sale of a defective product; c) the plaintiff must suffer an injury; and d) the defect must be the cause of the plaintiffs injury.25 The primary difference between the two causes of action is that a plaintiff may recover in strict products liability without showing that the defendant’s conduct was wrongful, so long as its product was defective.
As to duty, damages, and causation, the same arguments made with respect to negligence apply here as well. And for those *171reasons, I believe that, in this ease, the questions of damages and causation are for the jury, while the issue of duty should be certified to the New York Court of Appeals.
The additional problem with the claim of strict products liability is that the plaintiffs must establish the existence of a defect. The majority notes that the New York Court of Appeals has explained that “a defectively designed product is one which, at the time it leaves the seller’s hands, is in a condition not reasonably contemplated by the ultimate consumer.” Robinson v. Reed-Prentice Div. of Package Mach. Co., 49 N.Y.2d 471, 426 N.Y.S.2d 717, 720, 403 N.E.2d 440, 443 (1980). Since the Black Talons performed in precisely the manner contemplated by Colin Ferguson, the majority concludes that under New York’s “consumer contemplation” test, the Black Talon is not a defective product.
There is no doubt that, in order for strict products liability to apply, there must be a defect, i.e. “something wrong with the product, and if nothing is wrong there will be no liability.” DeRosa v. Remington Arms Co., 509 F.Supp. 762, 769 (E.D.N.Y.1981) (citation and internal quotation marks omitted). But a large number of jurisdictions have held that there is something wrong with any product that is unreasonably dangerous, “even though it comports in all respects to its intended [and obvious] design.” Azzarello v. Black Bros. Co., 480 Pa. 547, 391 A.2d 1020, 1025 (1978); see also Turner v. General Motors Corp., 584 S.W.2d 844, 850 (Tex.1979). In other words, these jurisdictions employ a risk/benefit calculus to establish a design defect. “Under this approach, a product is defective as designed if ... the magnitude of the danger outweighs the utility of the product,” that is, if “the harmful consequences in fact from intended and reasonably foreseeable uses resulting from the way the product was designed and marketed up to the time of plaintiffs injury outweighf] the benefits in terms of wants, desires, and human needs served by the product.” Prosser & Keeton, supra, § 99 at 699.26
Historically, in determining whether a product is defectively designed, some courts have applied only the “risk/benefit test” while others have used only the “consumer contemplation test” alluded to by the majority. See id., § 99 at 698-700. But many courts have allowed a plaintiff to establish a defect, and hence liability, if the defendant’s product fails either of these tests. See, e.g., Turner, 584 S.W.2d at 850; Barker v. Lull Eng’g Co., 20 Cal.3d 413, 143 Cal.Rptr. 225, 236, 573 P.2d 443, 454 (1978).27
In Robinson, the case upon which the majority relies, the New York Court of Appeals appeared to require that the plaintiff meet both tests:
Where a product presents an unreasonable risk of harm, notwithstanding that it was meticulously made according to detailed plans and specifications, it is said to be defectively designed. This rule, however, is tempered by the realization that some products, for example knives, must by their very nature be dangerous in order to be functional. Thus, a defectively designed product is one which, at the time it leaves the seller’s hands, is in a condition not reasonably contemplated by the ultimate consumer and is unreasonably dangerous for its intended use; that is one whose utility does not outweigh the danger inherent in its introduction into the stream of commerce.
Robinson, 426 N.Y.S.2d at 720, 403 N.E.2d at 443.
This dual requirement would be highly unusual. And more recently, the Court of Appeals, after quoting this passage from Robinson, went on to say that
[i]n order to establish a prima facie case in strict products liability for design defects, *172the plaintiff must show that the manufacturer breached its duty to market safe products when it marketed a product designed so that it was not reasonably safe and that the defective design was a substantial factor in causing plaintiffs injury.
Voss v. Black & Decker Mfg. Co., 59 N.Y.2d 102, 463 N.Y.S.2d 398, 402, 450 N.E.2d 204, 208 (1983). Significantly, this latter statement of what constitutes a prima facie case makes no mention at all of consumer expectations as a separate requirement additional to that imposed by the risk/benefit test.
Moreover, in Denny v. Ford Motor Co., 87 N.Y.2d 248, 639 N.Y.S.2d 250, 254-58, 662 N.E.2d 730, 734-38 (1995), the Court of Appeals explained that, while the risk/benefit test is an essential element of a cause of action for design defect, the consumer contemplation test sounds more in commercial law than in torts, and is therefore an element of a claim of breach of an implied warranty. As a result, it is anything but clear whether, read in light of more recent developments in the jurisprudence of the New York Court of Appeals, Robinson in fact stands for the proposition that a plaintiff may not establish that a product was defectively designed for the purposes of attaching strict liability without proving that the product was in a condition not reasonably contemplated by the consumer.28
Indeed, one recent commentator, after an extensive fifty-state survey, did not include New York on his list of the twenty-five “[sjtates which use consumer expectations as part of any test for strict liability design defects.” John F. Vargo, The Emperor’s New Clothes, 26 U. Mem. L. Rev. 493, 951 index 2 (1996). Instead, he noted that, while “the Robinson court mixed the consumer expectation test with a risk-utility test,” id. at 822, that court may not have intended that result, see id. at 823, and, after Voss and Denny, “New York law ... appears to relegate the consumer expectation test to the separate legal theory of warranty,” id. at 837.29
Significantly, the Restatement of Torts, upon which the Court of Appeals relied in Robinson, now takes the position that the existence of a design defect should be determined solely by the risk/benefit test. See Restatement (Third) of Torts.- Products Liability § 2(b) & cmts. c, f (Tentative Draft No. 2,1995) [hereinafter, Restatement 3rd], According to the most recent Restatement, the consumer contemplation test no longer applies.30 See id. at § 2 cmt. f (rejecting “conformance to consumer expectations as a defense” and declaring instead that “[t]he mere fact that a risk presented by a product design is open and obvious, or generally known, and that the product thus satisfies expectations, does not prevent a finding that the design is defective”).
Since it is more than possible that the New York Court of Appeals would apply the risk/benefit test to determine whether the Black Talon is a defective product, and since, as discussed above, it cannot be said as a matter of law (especially by a federal court) that the benefits of the Black Talon outweigh *173its risks, I believe that the majority has erred in relying on the consumer expectations test to affirm the dismissal of the plaintiffs’ strict liability claim.
Nonetheless, the fact that a product fails the risk/utility test may not be sufficient to give rise to strict liability. This is so because, as the majority notes, New York law has appeared to require that the plaintiff also show a reasonable alternative design before she will be permitted to recover in strict liability for a defectively designed product. See Voss, 463 N.Y.S.2d at 402, 450 N.E.2d at 208. But see Vargo, supra, at 832-37 (suggesting that, despite Foss, a showing of a reasonable alternative design is no longer required in New York after Denny); Jerry J. Phillips, The Unreasonably Unsafe Product and Strict Liability, 72 Chi.-Kent L. Rev. 129, 159 (1996) (explaining that the Court of Appeals in Denny “noted that the strict tort claim was tried on a risk-utility standard, apparently without reasonable alternative design being a necessary factor”). Moreover, it has been noted (in a controversial comment in the latest tentative draft of the third Restatement of Torts)31 that this requirement
applies even though the plaintiff alleges that the category of product sold by the defendant is so dangerous that it should not have been marketed at all. Thus common and widely distributed products such as alcoholic beverages, tobacco, firearms, and above-ground swimming pools may be found to be defective only upon proof [that] ... reasonable alternative designs could have been adopted. Absent [such] proof ... courts do not impose liability based on a conclusion that an entire product category should not have been distributed in the first instance. That is, courts have not imposed liability for categories of products that are generally available and widely used and consumed, solely on the ground that they are considered socially undesirable by some segments of society.
Restatement 3rd, supra, § 2 cmt. c.
In the instant ease, however, a possible alternative design does exist. It consists of the elimination of the extra-destructive “talons.” The proposed Restatement contains a remarkably relevant discussion:
Several courts have suggested that the designs of some products are so manifestly unreasonable, in that they have low social utility and high degree of danger, that liability should attach even absent proof of a reasonable alternative design. In large part the problem is one of how the range of relevant alternative designs is described. For example, a toy gun that shoots hard rubber pellets with sufficient velocity to cause injury to children could be found to be defectively designed within the rule of § 2(b). Toy guns that do not produce injury would constitute reasonable alternatives to the dangerous toy. Thus, toy guns that project ping pong balls, soft gelatin pellets, or water might be found to be reasonable alternative designs to a toy gun that shoots hard pellets. However, if consideration is limited to toy guns that are capable of causing injury, then no reasonable alternative will, by hypothesis, be available. In that instance, the design feature that defines which alternatives are relevant — the capacity to injure — is precisely the feature on which the user places value and of which the plaintiff complains. If a court were to adopt this characterization of the product, it could conclude that liability should attach without proof of a reasonable alternative design. The court would condemn the product design as defective and not reasonably safe because the extremely high degree of danger posed by its use or consumption so substantially outweighs its negligible utility that no rational adult, fully aware of the relevant facts, would choose to use or consume the product.
Id., § 2 cmt. d.
It is worth noting that courts and commentators have been wrestling with the questions of what is a relevant safer alternative design, and whether entire categories of products can be deemed defective in the absence of an *174alternative design, with increasing frequency in recent years. See, e.g., Smith v. Keller Ladder Co., 275 N.J.Super. 280, 645 A.2d 1269, 1271 (App.Div.1994) (noting that in some cases, a jury might justifiably conclude “that even though there is presently no alternative design which would make a product safer, the product is so dangerous and of such little use that under the risk-utility analysis the manufacturer should bear the cost of liability of harm to others”) (citation, brackets, and internal quotation marks omitted); Wilson v. Piper Aircraft Corp., 282 Or. 61, 577 P.2d 1322, 1328 n. 5 (1978) (in banc) (“There might be cases in which the jury would be permitted to hold the defendant liable on account of a dangerous design feature even though no safer design was feasible____ If, for example, the danger was relatively severe and the product had only limited utility, the court might properly conclude that the jury could find that a reasonable manufacturer would not have introduced such a product into the stream of commerce.”); Carl T. Bogus, The Third Revolution in Products Liability, 72 Chi.-Kent L. Rev. 3,11 (1996) (“[I]f strict liability attaches to products with unreasonably dangerous features how can it not reasonably attach to unreasonably dangerous products?”); Page, supra, at 101-04; id. at 122 (“If, by definition, a product can be deemed ‘defective’ in its design or labeling upon proof by plaintiff that the product as designed or labeled was unreasonably dangerous, conceptually it would seem logical to classify a decision to market a whole category of products as ‘defective’ because of the manufacturer’s failure to make an alternative marketing decision that would have withheld the product from consumers. The manufacturer’s option to market would be evaluated in the same way as her option to use one of several alternative designs, warnings, or instructions-for-use.”). See generally Symposium on Generic Products Liability, 72 Chi.-Kent L. Rev. 1 (1996). The New York Court of Appeals has yet to confront the issue, and should be given the opportunity to do so here.32
I therefore believe that the question of whether the defendant can be held strictly liable for the alleged defective design of the Black Talon, like the question of duty, is “a complex question of New York common law for which no [controlling] New York authority can be found.” Riordan v. Nationwide Mut. Fire Ins. Co., 977 F.2d 47, 51 (2d Cir.1992). And that question, again like the question of duty, will turn in substantial part on considerations of public policy. See Azzarello, 391 A.2d at 1026 (noting that the question of whether “the utility of a product outweigh[s] the unavoidable danger it may pose” is “dependent] upon social policy”). Accordingly, I would certify both questions.33
*175CONCLUSION
I do not know whether the New York Court of Appeals would allow a cause of action for negligence or strict liability to proceed in this case. Nor do I know whether that Court should allow such liability.34 What I do know is that these questions are not for this Court to decide. And for that reason, I respectfully dissent.
. An analogous situation exists where there is authority from the state's highest court, but that authority is very old and has been ignored in recent years, during which time other jurisdictions have abandoned similar rulings. Cf. Harry Shulman, The Demise of Swift v. Tyson, 47 Yale L.J. 1336, 1350 (1938) (inquiring whether, under Erie, federal courts have "the power to disregard a state decision on the ground that it will doubtless be overruled by the state court at the first opportunity”); Arthur L. Corbin, The Common Law of the United States, 47 Yale L.J. 1351, 1352 (1938) (suggesting, critically, that under one reading of Erie, federal courts forced to apply state common law are "limited in a way in which the [state] judges themselves are not limited,” insofar as they cannot look to intervening decisions of the courts of other states, or to any other sources that might lead them to disregard or limit prior cases).
. See also Ira P. Robbins, The Uniform Certification of Questions of Law Act: A Proposal for Reform, 18 J. Legis. 127, 128 (1992) (noting that certification "resolves many of the problems associated with the Erie doctrine”); Appellate Certification of State Law Questions to State Courts, 9 No. 2 Fed. Litigator 56, 56 (1994) ("One of several considerations that can enter into a plaintiff's choice of forum when her citizenship is diverse from that of defendants is whether she would rather have a state court or a federal court decide matters of state law. Although federal courts are to decide such questions as state courts would, it is no secret that the tendencies of federal judges and state judges with respect to open questions in certain areas of the law are not always identical. To make the equation more complex, when a state supreme court has a procedure for certification of questions of law for determination by that court, a federal court may, but is not required to take advantage of it.”).
.See, e.g., Hakimoglu, 70 F.3d at 302-03 & n. 11 (Becker, /., dissenting) (collecting instances in which the Third Circuit has erroneously predicted the law of Pennsylvania). As Chief Judge Sloviter has explained, "[i]t is not that Third Circuit judges are particularly poor prognosticators. All of the circuits have similar problems in predicting state law accurately.” Dolores K. Sloviter, A Federal Judge Views Diversity Jurisdiction Through the Lens of Federalism, 78 Va. L. Rev. 1671, 1680 (1992); see also id. at 1680-81 & nn. 52-53 (cataloguing federal court errors in prediction); United Servs. Life Ins. Co. v. Delaney, 328 *159F.2d 483, 486-87 & nn. 5-9 (5th Cir.1964) (Brown, J., concurring) (discussing the “staggering” number of instances in which the courts of Texas, Alabama, and Florida have proved the Fifth Circuit’s guesses to be erroneous).
. This is quite different from cases in which the state’s highest court declines to hear an appeal from an intermediate court and later, on hearing a similar case, decides it differently from the way the earlier case was decided. In all such cases, it is the highest court of the jurisdiction whose law applies that has decided, for its own reasons, to let the result in the first case stand. Likewise, when the state’s highest court declines to accept certification from a federal court, it is the highest court of the jurisdiction whose law applies that, in effect, has decided to allow the federal court to guess as to the law of the state, and to permit that guess to be based on the usual sources of law employed by federal courts in diversity cases, including non-binding lower state court opinions, the law of other jurisdictions, etc. And a judgment — even a judgment to abstain — by the state’s highest court is all that a litigant is entitled to receive. When instead a federal court fails to certify, the highest court of the state is denied the right to decide whether it wishes to resolve the issue. The loser loses because a court that has no ultimate say as to the applicable law has made itself into a court of last resort. For these reasons, I am unconvinced by Judge Selya’s thoughtful arguments against certification. See Bruce M. Selya, Certified Madness: Ask a Silly Question. . ., 29 Suffolk U.L. Rev. 677 (1995). I agree with him that certification is not a panacea (and that it creates some procedural problems), but that is a far different issue.
. This is still the case today when federal courts are asked to ascertain the law of a state that has not provided for certification. See, e.g., Sargent v. Columbia Forest Prods., Inc., 75 F.3d 86, 89-90 (2d Cir.1996) (noting that Vermont has no certification statute, and for that reason, this court is required to guess — often incorrectly — as to the law of Vermont). But since this situation is due to the failure of the state to allow certification, it is fair to attribute the litigant's loss (even when it results from the federal court's bad guess as to state law) to a decision of the state, that is, of the jurisdiction which has the authority to determine the litigant’s rights. See id. at 90 (explaining that the plaintiff was prejudiced by Vermont's failure to provide for certification because the defendant was able to remove the case to the federal courts, which were forced to make what turned out to be an erroneous guess as to state law).
. See also Peterson v. U-Haul Co., 409 F.2d 1174, 1177 (8th Cir.) (admitting that prediction "is a hazardous and unsatisfactory method of deciding litigation"), modified on other grounds, 421 F.2d 837 (8th Cir.1969); Martinez v. Rodriquez, 410 F.2d 729, 730 (5th Cir.1969) ("No matter how many Federal Judges, trial, appellate, three-judge panel, or the full panoply of the court en banc, any decision would have been an Brie-guess. Now the guesswork has been eliminated, and we are quickly presented with a definitive explication of Florida law.”).
.See, e.g., Harrison v. Insurance Co. of North America, 294 Ala. 387, 318 So.2d 253, 254 (1975) (declaring that the certification "process can serve as a demonstration of cooperative judicial federalism which would justify those of us who think that the federal form of government has a contribution to make toward the preservation of justice in this country”) (citation and internal quotation marks omitted); M. Biyan Schneider, "But Answer Came There None": The Michigan Supreme Court and the Certified Question of State Law, 41 Wayne L. Rev. 273, 302 (1995) (noting that in a recent survey "state court judges seemed remarkably receptive to certification from federal courts”) (citing John B. Corr & Ira P. Robbins, Interjurisdictional Certification and Choice of Law, 41 Vand. L. Rev. 411 (1988)). Indeed, more than one state high-court judge in this circuit has told me in informal conversation that they believe this court should certify more frequently.
. The majority suggests that I am proposing a "new standard” for certification, in violation of Riordan v. Nationwide Mut. Fire Ins. Co., 977 F.2d 47 (2d Cir.1992). See ante, Op. at 158 n.3. Since, as discussed below, I think that this case presents "complex question[s] of New York common law for which no New York authority can be found,” Riordan, 977 F.2d at 51, I do not believe that I am advocating a standard that is in any way inconsistent with our cases.
. The plaintiffs in the instant case are guilty of this mistake. See McCarthy v. Sturm, Ruger and Co., 916 F.Supp. 366, 372 (S.D.N.Y.1996) ("Plaintiffs candidly argue that I should expand existing tort doctrines to cover this case....”).
. The plaintiffs' negligent manufacturing claim was based on the fact "that a manufacturer is obligated to exercise that degree of care in his plan or design so as to avoid any unreasonable risk of harm to anyone who is likely to be exposed to the danger when the product is used in the manner for which the product was intended, as well as an unintended yet reasonably foreseeable use.” Micallef v. Miehle Co., 39 N.Y.2d 376, 384 N.Y.S.2d 115, 121, 348 N.E.2d 571, 577 (1976) (citations omitted),
. This is nothing more than the traditional "Hand formula.” See United States v. Carroll Towing Co., 159 F.2d 169, 173 (2d Cir.1947) (L.Hand, J.) (inquiring whether the burden of avoiding the harm is outweighed by the product of the probability of accident and the gravity of the resulting injury).
. One commentator has described this type of negligent marketing claim in these terms:
The imposition of negligence-based tort liability on the manufacturers of unreasonably and unavoidably unsafe products would require proof that a reasonable person in defendant's position, with the knowledge that defendant possessed' — or in the exercise of due care should have possessed — would not have marketed the product. Such a decision would be warranted if the foreseeable risks of harm from marketing the product would exceed the costs of avoiding that harm, which, in the product-liability context, is measured by the loss of utility that society incurs from the unavailability of the product.
Page, supra, at 97 (footnotes omitted). The decision not to limit the distribution of a dangerous product can constitute negligent marketing. See id. at 97 n. 40. Another commentator has articulated three potential negligent marketing claims: 1) a claim that the manufacturer acted unreasonably in marketing a product "that presents an unusually high risk of harm and negligible utility for legitimate purposes”; 2) a claim that the manufacturer implemented "a marketing strategy that deliberately, recklessly, or negligently targeted] criminal consumers”; and 3) a claim that the manufacturer failed "to take reasonable steps in the marketing process to minimize the risk that its products [would] be purchased by persons likely to misuse them.” MeClurg, supra, at 799.
. One lower federal court has attempted to cast doubt on this proposition, suggesting that a negligent marketing claim in a case like this one "really amount[s] to an alternate pleading of the [strict] product liability theory," and that the act of marketing a product cannot "give rise to liability absent a defect in the manufacture or design of the product itself.” Hamilton v. Accu-Tek, 935 F.Supp. 1307, 1323 (E.D.N.Y.1996). To the extent that case suggests that there is no difference between a cause of action for negligent marketing of a product and one for strict liability for defective design, it is, in my opinion, manifestly incorrect.
Traditionally, products liability actions have been allowed to proceed on a number of grounds, including negligence and strict liability. See, e.g., Voss v. Black & Decker Mfg. Co., 59 N.Y.2d 102, 463 N.Y.S.2d 398, 401, 450 N.E.2d 204, 207 (1983) ("In New York, a plaintiff injured by an allegedly defective product may seek recovery against the manufacturer on the basis of any one or more of four theories of liability. Depending on the factual context in which the claim arises, the injured plaintiff ... may state a cause of action in contract, express or implied, on the ground of negligence, or ... on the theory of strict products liability.”) (citation and internal quotation marks omitted); id. ("As the law of strict products liability has developed in New York, a plaintiff may assert that the product is defective because of a mistake in the manufacturing process or because of an improper design or because the manufacturer failed to provide adequate warnings regarding the use of the product.”) (citations omitted).
It is true that, in cases such as this one, the question of defect in strict products liability may closely resemble the question of breach in negligence, as the existence of a defect, like the existence of negligence, may depend (at least in part) on whether the benefit of the product outweighs the risk of harm. See infra Part II.B; see also Denny v. Ford Motor Co., 87 N.Y.2d 248, 639 N.Y.S.2d 250, 255, 662 N.E.2d 730, 735 (1995) (noting that "the inquiry in design defect cases [is] close[] to that used in traditional negligence cases”). But a claim of negligent marketing differs from a claim of design defect in at least one important way:
Strict products liability for design defect ... differs from a cause of action for a negligently designed [or marketed] product in that the plaintiff is not required to prove that the manufacturer acted unreasonably in designing [or marketing] the product. The focus shifts from the conduct of the manufacturer to whether the product, as designed, was not reasonably safe. A manufacturer is held liable regardless of his lack of actual knowledge of the condition of the product because he is in the superior position to discover any design defects and alter the design before making the product available to the public.
Voss, 463 N.Y.S.2d at 401-02, 450 N.E.2d at 207-08. Conversely,
[t]o impose liability for negligence, it would not be necessary for the plaintiff to prove there was "something wrong” with the product. The plaintiff would need to show only that there was “something wrong” with the manufacturer’s conduct. To find for the plaintiff under a negligent marketing theory, the court essentially would be saying: "Your product may be fine, but there was 'something wrong’ with your selling it (or with the manner in which you sold it).”
McClurg, supra, at 801.
. The general rule is, of course, that the legality of an act does not insulate it from possible tort liability. Illegality may be negligence per se, as it is in New York, see Van Gaasbeck v. Webatuck Cent. Sch. Dist. No. 1, 21 N.Y.2d 239, 287 N.Y.S.2d 77, 80, 234 N.E.2d 243, 245 (1967), but legality "does not preclude a finding of negligence," Feiner v. Calvin Klein, Ltd., 157 A.D.2d 501, 549 N.Y.S.2d 692, 693 (1990). Legality may, however, be relevant to the issue of duty. See infra Part II.A.4.
. There are exceptions to this requirement, see, e.g., Modave v. Long Island Jewish Med. Ctr., 501 F.2d 1065, 1072-74 (2d Cir.1974); Summers v. Tice, 33 Cal.2d 80, 199 P.2d 1, 2-5 (1948) (in banc), but they need not detain us here.
. "There is a causal link [that is, a causal tendency] between an act or activity and an injuty when we conclude on the basis of the available evidence that the recurrence of that act or activity will increase the chances that the injury will also occur." Calabresi, Concerning Cause, supra, at 71. This concept is distinct from cause in fact. For instance, "but for the fact that the trolley driver had been speeding, the trolley car would not have been under a particular rotten tree when it fell and hit the car. Yet, unless speeding increases vibrations and vibrations increase the likelihood of trees falling, the admittedly but for cause would not be causally linked to the injury.” Id. at 72.
. See, e.g., Doucette v. Town of Bristol, 138 N.H. 205, 635 A.2d 1387, 1391 (1993) (“The existence of a duty under particular circumstances depends on what risks, if any, are reasonably foreseeable.”) (citation and internal quotation marks omitted); Estates of Morgan v. Fairfield Family Counseling Ctr., 77 Ohio St.3d 284, 673 N.E.2d 1311, 1319 (1997) ("In Ohio, the existence of a duty depends on the foreseeability of the injury.") (citation, brackets, and internal quotation marks omitted).
. See, e.g., Nutbrown v. Mount Cranmore, Inc., 140 N.H. 675, 671 A.2d 548, 551 (1996) (explaining that the doctrine of primary assumption of risk "is simply an alternative expression for the proposition that the defendant was not negligent, that is, there was no duty owed”) (citation and internal quotation marks omitted); Gallagher v. Cleveland Browns Football Co., 74 Ohio St.3d 427, 659 N.E.2d 1232, 1236 (1996) (noting that "a defendant who asserts [a primary assumption of risk] defense asserts that no duty whatsoever is owed to the plaintiff”).
. The fact that the court in Codling found liability based on the nascent doctrine of strict products liability, rather than negligent products liability, is of no matter. In Codling, the trial court allowed the plaintiffs cause of action for negligence to proceed to the jury, which found no unreasonable conduct on the part of the defendant. See Codling, 345 N.Y.S.2d at 464-65, 298 N.E.2d at 625. The Court of Appeals made no suggestion in that case that the defendant was under no duty to the bystander plaintiff, such that, had the jury found negligence, the plaintiff would not have been permitted to recover. Indeed, the Court of Appeals quoted Codling's statement that a product manufacturer should not be immune from liability to any persons injured by the product " 'where the manufacturer could reasonably contemplate injury to such persons by reason of the defect''' in a negligence case decided the very same year as Codling. Bolm, 350 N.Y.S.2d at 649-50, 305 N.E.2d at 773 (quoting Codling, 345 N.Y.S.2d at 467, 298 N.E.2d at 627).
It has been noted that "[t]he strict liability rule for products liability cases ... does not eliminate the requirement that, even where there is a defect in the product, there must be some duty owed to the plaintiff.” Walton v. Chrysler Motor Corp., 229 So.2d 568, 573 (Miss.1970), overruled on other grounds by Toliver v. General Motors Corp., 482 So.2d 213 (Miss.1985); see also Allabach v. Santa Clara County Fair Ass'n Inc., 46 Cal.App.4th 1007, 54 Cal.Rptr.2d 330, 333 n. 1 (1996). Many courts have indicated that the duty requirement is identical in strict liability and negligence claims. See, e.g., Wagatsuma v. Patch, 10 Haw.App. 547, 879 P.2d 572, 585 (1994); Briggs v. Hartford Ins. Co., 532 So.2d 1154, 1156 (La.1988). While the Third Circuit has rejected a district court’s statement that "courts have consistently applied the identical duty requirement to strict liability and negligence claims,” Griggs v. BIC Corp., 786 F.Supp. 1203, 1209 (M.D.Pa.), rev’d, 981 F.2d 1429, 1434-35 (3d Cir.1992), that case dealt with situations in which, while there was no duty in strict liability, there might still be a duty in negligence. I have uncovered no authority for the converse proposition: that a duty can exist in strict liability, but not in negligence. Such a position would be unlikely, to say the least.
Instead, in addition to Codling and Bolm, numerous cases suggest that, under New York law, the existence of a duty in strict liability establishes the existence of a duty in negligence as well. See, e.g., Colonno v. Executive I Assocs., 228 A.D.2d 859, 644 N.Y.S.2d 105, 107 (1996) (holding that, in a products liability action premised on negligence, "the principle emanating from strict products liability law now holds contractors to a general standard of reasonable care for the protection of third persons who may be foreseeably endangered by the contractor's negligence”) (citations and internal quotation marks omitted). While the New York Court of Appeals has not spoken directly to this issue, there is absolutely no indication that New York law contemplates that there can be no duty in negligence despite the presence of a duty in strict liability. For a federal court to presume so in a diversity case without any statement to that effect from the New York Court of Appeals would be an extraordinary act.
. Whether the imposition of liability in these circumstances would go beyond the dictates of sound public policy is, of course, a difficult policy issue, but it is certainly not out of the question that the Court of Appeals might make just that determination.
. There is one situation in which that question has often come up in state courts. When a car owner negligently leaves the keys in the ignition of an unlocked car, and a thief steals the car and runs over a third-party pedestrian, does the car owner owe a duty to the pedestrian? In other words, could the defendant be held liable for the criminal acts of an intervenor absent any direct relationship with the plaintiff? Historically, a majority of jurisdictions answered this question in the negative, finding either no duty or no proximate cause. See Zinck v. Whelan, 120 N.J.Super. 432, 294 A.2d 727, 729-30 & n. 9 (App.Div.1972) (collecting cases). However, "[a] substantial and growing number of jurisdictions, though still a minority, have held, in the ordinary fact case of theft and accident within a reasonable time thereafter that there are at least jury questions as to duty, negligence and proximate cause.” Id.., 294 A.2d at 730 (collecting cases). A number of these jurisdictions have reached this conclusion in the absence of any statute forbidding the driver from leaving the keys in the ignition. See Roadway Express, Inc. v. Piekenbrock, 306 N.W.2d 784, 785-86 (Iowa 1981) (collecting cases).
In New York, the Court of Appeals initially indicated that it would allow liability in this situation. See Maloney v. Kaplan, 233 N.Y. 426, 135 N.E. 838, 839 (1922) (Pound, I.) ("The driver of a spirited horse, who leaves it unhitched in the street, may be liable for accidents immediately brought about by one who frightens the animal or touches it with the whip, so that it runs away; so the driver of an automobile who leaves it in the street without properly securing it may be liable for accidents immediately brought about by his negligence, although others may start the car.”); see also Illidge v. Goodwin, 5 Carr. & Payne’s Rep. 190 (C.P.1831) ("If a man chooses to leave a cart standing in the street, he must take the risk of any mischief that may be done.”). Subsequently, however, the Court of Appeals refused to allow recovery, see Walter v. Bond, 292 N.Y. 574, 54 N.E.2d 691 (1944), aff'g 267 A.D. 779, 45 N.Y.S.2d 378 (1943), concluding that the car owner’s negligence could not be a proximate cause of the plaintiff's injury, see Wilson v. Harrington, 295 N.Y. 667, 65 N.E.2d 101 (1946), aff'g 269 A.D. 891, 56 N.Y.S.2d 157 (1945); accord Lotito v. Kyriacus, 272 A.D. 635, 74 N.Y.S.2d 599, 600-01 (1947).
Not long after, the New York legislature enacted what is now N.Y. Veh & Traf. L. § 1210(a), which provides that ”[n]o person driving or in charge of a motor vehicle shall permit it to stand unattended without first stopping the engine, locking the ignition, [and] removing the key from the vehicle." That "statute changed the prior case law and it is now clear that the intervention of an unauthorized person no longer operates to break the chain of causation.” Guaspari v. Gorsky, 36 A.D.2d 225, 319 N.Y.S.2d 708, 711 (1971).
Because they found no proximate cause, the courts of New York never reached the question of duty. By enacting § 1210(a), the slate legislature (reflecting the will of the people) imposed liability before the New York Court of Appeals had occasion to reconsider its decision and perhaps join the "substantial and growing number of jurisdictions” that have found not only proximate cause, but also a duty, in cases of this sort. As such, we have no indication of what the Court of Appeals would say about duty in this situation in the absence of a statute. It is worth noting, however, that the Court of Appeals’ earliest and only pronouncement on the subject, Maloney v. Kaplan, held, at least by implication, that a duty did exist.
. Cf. Note, Absolute Liability for Ammunition Manufacturers, 108 Harv. L. Rev. 1679, 1691 (1995) ("The primary advantage of [imposing tort liability on ammunition manufacturers] is that it will force consumers of ammunition to internalize costs that have heretofore been borne by third parties and society in general. Such internalization will provide manufacturers and consumers with the proper incentives to choose care and activity levels that more closely equate costs and benefits.”) (footnote omitted); id. at 1690 (noting that, unlike banning ammunition, imposing liability "does not reflect any moral or *170ethical condemnation of ammunition manufacturers”).
.Indeed, the fact that the Black Talon has not been banned should not be taken to suggest that a majority of New Yorkers necessarily favors its continued availability. To the contrary, the public outcry against Black Talons was so great even before Colin Ferguson’s crime that their manufacturer chose to withdraw them from public distribution in the face of mounting political pressure. See Jamieson, supra, at 5. Bills had been introduced in Congress both to ban the bullets, see id., and, alternatively, to tax them out of existence. See Winchester Bows to Pressure, Halts Talon Bullet Sales, Balt. Sun, Nov. 23, 1993, at 5A. Police groups, including New York City's Police Commissioner, had advocated banning the bullets. See Jamieson, supra, at 5. Even emergency room physicians had condemned the bullets because of the danger of contracting blood-borne diseases when the razor-sharp talons pierce both their surgical gloves and their skin. See Winchester Halts Sale, supra, at A4; All Things Considered: Winchester Will no Longer Make the Black Talon Bullet (National Public Radio broadcast, Nov. 23, 1993).
In fact, one of the plaintiffs in this action, Carolyn McCarthy, was elected to Congress largely on a platform of opposition to assault weapons and extra-destructive ammunition. See NBC Nightly News (NBC television broadcast, Nov. 6, 1996) ("Carolyn McCarthy has been elected to Congress. The widow of one of the victims of the 1993 massacre on New York’s Long Island Railroad turned politician after her local congressman voted against a ban on assault weapons. Last night she defeated him.’’).
. Naturally, that task may also be undertaken by the state’s legislature. See supra note 21 (noting that the New York legislature imposed liability, where the Court of Appeals had not, for the injuries caused to a third person when a car owner leaves the keys in an unlocked car and an intervening thief steals the car and strikes the third person).
. These are not the elements of a cause of action in strict products liability. Those elements are much more specific, and depend on the theory of liability being asserted. See, e.g., Fane v. Zimmer, Inc., 927 F.2d 124, 128 (2d Cir.1991) (explaining, for example, that in design defect cases the typical plaintiff must show "that (1) the product is 'defective' because it is not reasonably safe as marketed; (2) the product was used for a normal purpose; (3) the defect was a substantial factor in causing the plaintiff's injuries; (4) the plaintiff by the exercise of reasonable care would not have both discovered the defect and apprehended its danger; (5) the plaintiff would not have otherwise avoided the injury by the exercise of ordinary care”) (citation and internal quotation marks omitted). Rather, these are the basic requirements for any cause of action in tort for products liability, regardless of the particular theory of recovery.
. The difference between this risk/ulility test and the Learned Hand test for negligence is that the latter requires that a reasonable person knew or ought to have known of the undue danger at the time that she or he acted, while the former imposes strict liability and is not limited by what the actor knew or should have known. See Voss, 463 N.Y.S.2d at 401-02, 450 N.E.2d at 207-08.
. For a detailed analysis of the various tests adopted by state courts in determining the existence of a design defect, see John F. Vargo, The Emperor's New Clothes, 26 U. Mem L. Rev. 493, 538-47 (1996).
. The plaintiffs also claimed below that the defendant's decision to manufacture Black Talons as designed was negligent. It should be noted that, regardless of the applicability of the consumer expectations test in strict liability, New York law does not allow a manufacturer to rely on the obvious nature of the defect (that is, on reasonable consumer expectations) to defend against a claim of negligent design and manufacture. See Micallef v. Miehle Co., 39 N.Y.2d 376, 384 N.Y.S.2d 115, 348 N.E.2d 571 (1976).
. This commentator suggests that the Court of Appeals’
prior statement concerning consumer expectations could be explained in at least two different ways. The court may be adopting a modified Barker [v. Lull Eng'g Co., supral rule in design cases — the design defect may be established by application of the consumer expectations test, but if this test proves inappropriate, the plaintiff may establish design defect through [a variation of the risk/benefit calculus]. An alternative interpretation is that the consumer expectation test is used for manufacturing defects, and the risk-utility test is used for design cases.
Id. at 832.
.I have elsewhere argued in favor to the consumer expectations test, and have expressed doubts about the appropriateness of the risk/benefit test. See Guido Calabresi & Jeffrey O. Cooper, New Directions in Tort Law, 30 Val U.L. Rev. 859, 864-65 (1996). But my personal preferences are, of course, irrelevant to the task before us.
. For a description of the controversy surrounding this section of the Restatement, see Vargo, supra, at 519-36; Carl T. Bogus, The Third Revolution in Products Liability, 72 Chi.-Kent L. Rev. 3, 15-16 (1996).
. Whatever the merits of the contention that entire categories of products can be found defective even in the absence of evidence of an alternative design, it is important to emphasize that this is not a case in which the plaintiffs seek to impose liability on the manufacturers of all handguns or ammunition, and in which by definition there could be no reasonable alternative design short of failing to manufacture that type of product at all. Rather, this is a case in which the plaintiffs seek to impose liability for the decision to create extra-destructive ammunition that shreds human flesh upon impact and to market that product to the general public. There exists an alternative design — no talons — and an alternative marketing strategy — sale only to police officers, see Bogus, supra, at 15 n. 83 ("Some products may be unreasonably dangerous when sold to the public-at-large but not when sold to a particular class of purchasers. For example, x-ray equipment may be reasonably dangerous when sold to licensed facilities, and handguns may be reasonably dangerous when sold to law enforcement officers, even if such products are unreasonably dangerous when sold to the public-at-large.”) — either of which stop well short of imposing liability on the entire ammunition product category.
. A third potential cause of action may also exist on these facts — ultrahazardous activity liability. Like negligence and strict liability, ultra-hazardous activity liability requires the existence of a duty relationship which is breached by the defendant’s ultrahazardous activity, thereby causing demonstrable injury to the plaintiff. The district court refused to consider whether there was a breach in this case (in other words, whether the defendant's activities were ultrahazardous) because it found that there was no duty relationship. It agreed with the defendant that "the doctrine of ultrahazardous activity does not apply to products; rather, it is limited to activities involving the use of land." McCarthy, 916 F.Supp. at 371 (emphasis added) (citing Forni v. Ferguson, No. 132994/94 (N.Y. Sup.Ct. New York County Aug. 2, 1995), aff'd, 648 N.Y.S.2d 73 (N.Y.App.Div.1996)).
It seems to me particularly ironic to say that New York courts would limit the applicability of *175ultrahazardous activity liability solely to cases involving land, given that, in the seminal case of ultrahazardous activity liability in the United States — Sullivan v. Dunham, 161 N.Y. 290, 55 N.E. 923, 924 (1900) — the Court of Appeals declared that "the safety of the person is more sacred than the safety of property,” and held that prior cases that had allowed liability only for trespass upon land, and not trespass upon persons, did not prevent the imposition of liability on a defendant whose ultrahazardous activity had injured a passerby.
In fact, New York law has at times allowed ultrahazardous liability for activities not involving the defendant's use of land. See, e.g., Margosian v. U.S. Airlines, Inc., 127 F.Supp. 464, 467 (E.D.N.Y.1955) (finding that an airplane pilot whose plane crashed on the plaintiff's property was liable under New York law despite the lack of negligence because aviation was, at the time, an ultrahazardous activity) (citing Rochester Gas & Elec. Corp. v. Dunlop, 148 Misc. 849, 266 N.Y.S. 469 (Monroe County Ct.1933), and Restatement (First) Torts § 520 cmt. b (1939) (classifying aviation as an ultrahazardous activity)). Subsequently, New York courts have refused to allow ultrahazardous activity liability for the use of airplanes, concluding "in light of the technical progress achieved in the design, construction, operation and maintenance of aircraft generally, that flying should no longer be deemed to be an ultrahazardous activity.” Wood v. United Air Lines, 32 Misc.2d 955, 223 N.Y.S.2d 692, 697 (N.Y.Sup.Ct.1961), aff'd, 16 A.D.2d 659, 226 N.Y.S.2d 1022 (1962); accord Crist v. Civil Air Patrol, 53 Misc.2d 289, 278 N.Y.S.2d 430, 433-34 (N.Y.Sup.Ct.1967). But these cases do not undermine the principle that ultrahazardous activity liability may be imposed on activities other than the use of land. See, e.g., Siegler v. Kuhlman, 81 Wash.2d 448, 502 P.2d 1181, 1184-87 (1972) (en banc) (transporting gasoline by truck constitutes an ultrahazardous activity). Indeed, one court has found that the manufacturing and marketing of assault weapons to the general public constitutes an ultrahazardous activity. See In re 101 California Street, No. 959316 (Cal.Super.Ct. 2d Dep’t Apr. 10, 1995) (cited in Hamilton v. Accu-Tek, 935 F.Supp. 1307, 1324 (E.D.N.Y.1996)). But see Delahanty v. Hinckley, 564 A.2d 758, 760-61 (D.C.1989) (collecting cases refusing to allow ultrahazardous activity liability for the manufacture and sale of firearms). See generally McClurg, supra, at 788-91 (collecting cases and commentary).
Nonetheless, the plaintiffs chose not to pursue this argument on appeal, and I will go no further in discussing its possible viability or non-viability-
. In this respect, I have, in my dissent, emphasized the arguments in favor of liability, not because I am necessarily convinced by them, but because that is what I must do to determine whether there is sufficient uncertainty to warrant certification.