In February 1991, David and Elizabeth Trull, and their two sons, Nathaniel and Benjamin, were traveling to Maine after a day of cross-country skiing in New Hampshire when their Volkswagen Vanagon slid on black ice and collided with an oncoming car. Benjamin, age 9, died in the accident, and Nathaniel, age 13, and Elizabeth suffered severe brain injuries. In this diversity products liability action, the Trulls seek damages from Volkswagen on the ground that defects in the design of the Vanagon made their injuries more severe than they otherwise would have been.1 The district court granted summary judgment for Volkswagen on the Trulls’ breach of warranty claims, and a jury found for the automobile manufacturer on the remaining liability claims.
*92The Trulls contend that the district court made a number of evidentiary errors that tainted the jury’s deliberations and also improperly imposed on plaintiffs the burden of proving the nature and extent of the enhanced injuries attributable to the Vanagon’s design. We find no reversible error in the district court’s evidentiary rulings. The remaining matter of who, under New Hampshire law, should bear the burden in a so-called “crashworthiness” case, poses sophisticated questions of burden allocation involving not only a choice of appropriate precedent but also an important policy choice. It is a question of fairly recent origin that has divided courts across the country. Although we do not criticize the district court for abuse of discretion in its decision declining to certify, we choose to exercise our own discretion and think it preferable to allow the state’s highest court to choose its own path. We therefore grant the Trulls’ motion to certify the burden of proof question to the New Hampshire Supreme Court.
I. Background
Although certain particulars of the Trulls’ 1991 accident are germane to issues raised on appeal, we think it more helpful to relate them in the context of our discussions of those claims. At this juncture, therefore, we provide only a brief procedural history of the case.
The Trulls filed their lawsuit in early 1994, alleging claims of negligence, breach of warranty and strict liability on behalf of all four family members. Plaintiffs had two primary theories of recovery: (1) the Vanagon was defective because it was a forward control vehicle constructed in such a way that it lacked sufficient protection against a frontal impact, and (2) the Vana-gon was defective because the rear bench seats, on which Nathaniel and Benjamin were seated, did not have shoulder safety belts as well as lap belts.
In May 1997, defendants filed an in limine motion seeking to exclude from trial certain evidence relating to the post-accident condition of the Vanagon because the vehicle had been disposed of (“spoliat-ed,” in the relevant jargon) by the Trulls’ insurance company and therefore was unavailable to be inspected by Volkswagen’s experts. In particular, David Trull, who had gone with a friend shortly after the accident to retrieve personal items from the vehicle, was prepared to testify that Elizabeth’s seat belt was broken. Other evidence indicated that Benjamin had been wearing a seat belt before the accident but was not wearing it when he was found in the wreckage after the crash. Volkswagen claimed that it would be unduly prejudicial to allow testimony about the post-accident condition of the seat belts when the defendant had had no opportunity to inspect them itself. The district court granted the motion to the extent that plaintiffs were barred from using lay testimony or expert opinion about the seat belts’ post-accident condition.
In December 1997, the district - court granted summary judgment for the defendants on the breach of warranty claims, a disposition that is not challenged on appeal. In April 1998, just before opening statements in the trial, Elizabeth Trull withdrew her remaining claims, assertedly because of the court’s exclusion of the seat-belt evidence and its determination that plaintiffs bore the burden of proving the nature of any enhanced injuries caused by a defect in the vehicle. Because Elizabeth was ejected from the Vanagon, she maintained that evidence about her broken seat belt was crucial for her to show a link between a defect and her injuries. The defendants subsequently moved to dismiss the claims of both Elizabeth and David Trull with prejudice,2 and the district court granted the motion on May 4. The trial *93proceeded with Nathaniel’s and Benjamin’s claims.
The jury returned verdicts for the defendants on all remaining claims, and the district court denied plaintiffs’ motion for a new trial. This appeal followed.
II. Discussion
Five of the plaintiffs’ six assertions of error concern the admission or exclusion of evidence, and the district court’s decisions in each instance are reversible only for an abuse of discretion. See Acosta-Mestre v. Hilton Int'l of Puerto Rico, 156 F.3d 49, 56-57 (1st Cir.1998). The last issue, the legal determination of the appropriate burden of proof in New Hampshire in a crashworthiness case, is a question of law entitled to de novo scrutiny. We discuss first why we find no reversible error in the court’s evidentiary decisions, addressing each in turn, before explaining why we choose to consult the New Hampshire Supreme Court on the burden of proof issue.
One procedural matter needs attention before we reach these claims of error, however, and we therefore begin with that issue, namely, whether Elizabeth has preserved her claims for appellate review.
A. Elizabeth’s Appeal.
Appellees contend that Elizabeth is not properly a party on appeal because the district court dismissed her claims with prejudice, and there has been no appeal of that dismissal. Appellants counter that their explicit statements withdrawing Elizabeth’s claims based on the court’s pretrial rulings on the burden of proof and the exclusion of the broken seat-belt testimony, together with their offer of proof on those matters, sufficed to preserve her right to appeal.
Because the spoliation issue also is raised on behalf of Benjamin’s estate, and because we uphold the district court’s exclusion of the seat-belt evidence, Elizabeth’s status is irrelevant with respect to that claim. The court’s ruling on the burden of proof, however, which we hold in abeyance in order to certify the issue to the New Hampshire Supreme Court, see infra at 100-03, is a different matter. If the state court disagrees with the district court’s resolution of that issue, Elizabeth would be entitled to bring her claims in a new trial only if she has preserved her appellate rights.3
After careful consideration of the record and relevant legal principles, we conclude that she has not. As noted above, Elizabeth withdrew her claims on the morning of the first day of trial, April 7, 1998. Three weeks later, on April 30, as the eighteen-day trial was near completion, defendants moved for dismissal of Elizabeth and David’s claims with prejudice. In an eight-page memorandum in support of their motion, defendants argued that dismissal with prejudice of Elizabeth and David’s claims was justified given the procedural irregularity and timing of Mrs. Trull’s “withdrawal,” failure of David Trull to submit either claims or evidence, the extensive preparation that was undertaken by defendants to try Elizabeth and David Trull’s claims, the amount of time the court has invested in ruling on evidentiary and other matters, the substantial rights defendants have developed under the rulings already made by the court, and the *94extreme cost to defendants and the court in holding a second trial in this matter.
Although the motion informed the court that plaintiffs objected, no opposition was filed. The docket reports that on May 4, the court orally granted defendants’ m.otion. On May 6, in a document entitled “JUDGMENT,” the district court included its May 4 grant of the dismissal motion in a list of five dispositions, which also included the jury verdicts rendered on May 4.4 The same list was repeated in a June 11 document entitled “POST-JUDGMENT JUDGMENT,” which contained one additional item, the court’s June 10 order denying plaintiffs’ motion for new trial. That motion did not seek reversal of the dis- . missal.
Plaintiffs’ subsequent notice of appeal requested review of six separate orders, but it, too, made no reference to the May 4 dismissal order. So far as we can tell, therefore, the record contains neither an objection to the dismissal filed by the plaintiffs nor a request that the court substitute a less final disposition. Although counsel stated when he announced withdrawal of Elizabeth’s claims that he was making an offer of proof “for purposes of appellate practice,” we think this was not enough to preserve her claims in the face of defendants’ subsequent motion to dismiss with prejudice. That motion explicitly alerted plaintiffs to defendants’ view that Elizabeth was not entitled to pursue an appeal precisely because of the way in which she had withdrawn from the trial.
For whatever reason, plaintiffs repeatedly sidestepped the dismissal issue. It was not raised in their renewed motion for certification of the burden of proof question, their motion for new trial, or their brief on appeal. Although' ambiguous, each document seems to presume that, if the court were .to agree with plaintiffs’ arguments, Elizabeth’s claims would be resurrected. See, e.g., Plaintiffs’ Brief on Appeal at 16 (In argument challenging exclusion of seat-belt evidence, plaintiffs asserted that “[t]he Court committed reversible error when it prohibited Elizabeth Trull from pursuing her claims.”). Not until their reply brief, in response to defendants’ arguments that Elizabeth had waived her appellate rights, did plaintiffs address the dismissal order, and then referred to it only in passing. Their position apparently is that, notwithstanding the dismissal, their offer of proof was sufficient in the context of this case to preserve Elizabeth’s issues.
We, however, can see no basis upon which to relax the rule requiring parties to specifically designate the orders from which they wish to take an appeal. See United States v. Velez Carrero, 140 F.3d 327, 330 (1st Cir.1998); Fed. R.App. Pro. 3(c). Plaintiffs could not have been oblivious to the district court’s dismissal of Elizabeth’s claims — the court documented its order twice, once before and once after the new trial motion. The request for dismissal, three weeks after the offer of proof, and the court’s subsequent grant of that motion, represented a new stage in the proceedings on Elizabeth’s claims. If plaintiffs wished to preserve Elizabeth’s claims without putting on what they believed would be a meritless case in light of the court’s rulings on spoliation and burden of proof, they could have stipulated to dismissal pending appeal or asked the court to change the dismissal to one without prejudice.5 Instead, plaintiffs chose to sit *95idly by as the district court disposed of her claims with finality. Plaintiffs’ failure to appeal the decision taking Elizabeth out of the case — indeed, their failure to offer any substantive response to defendants’ request for dismissal — is a procedural default that bars further litigation of her claims.
B. Exclusion of seat-belt evidence based on the destruction of the Vanagon (spoliation).
Appellants contend that the district court erred in excluding evidence about the post-accident condition of Elizabeth’s and Benjamin’s seat belts because appellants bore no responsibility for the destruction of the Vanagon. They argue that such a serious evidentiary sanction is reserved for instances of misconduct, in which evidence is knowingly destroyed in the face of contemplated litigation, and there is serious prejudice to the other party. The short answer to this assertion is that our case law does not require bad faith or comparable bad motive.
[B]ad faith is not essential. If such evidence is mishandled through carelessness, and the other side is prejudiced, we think that the district court is entitled to consider imposing sanctions, including exclusion of the evidence.... Although deterrence may play a role, the primary aim is remedial, at least absent willful destruction.
Sacramona v. Bridgestone/Firestone, Inc., 106 F.3d 444, 447, 446 (1st Cir.1997). Fairness to the opposing party therefore plays a substantial role in determining the proper response to a spoliation motion, and punishment for egregious conduct is not the sole rationale for the most severe sanction of exclusion.
In its nineteen-page ruling on spoliation, the district court thoughtfully examined various factors, including prejudice and motive, in deciding whether and how to sanction plaintiffs for the disposal of the Vanagon. Its analysis, which we do not recount in detail, by no means reflects an abuse of discretion. The district court found no suggestion of bad faith, but did opine that “there are clear indications that the plaintiffs acted negligently in allowing the destruction of the Vanagon.” Order at 12. Presumably, the court was referring to the fact that David Trull failed to prevent disposal of the car or to preserve evidence of the seat-belt condition after he had examined the Vanagon while it was in the insurance company’s possession. Trull had inspected the vehicle in an effort “to figure out what had happened,” and the thought “cross[ed] his mind” at the time that Volkswagen might be at fault for something.6 Although Trull and his friend later discussed returning to photograph the seat belts, they never did.
In assessing prejudice, the court rejected the plaintiffs’ arguments that defendants were not unfairly disadvantaged because plaintiffs’ experts, too, were unable to examine the car and because thé defendants were able to mount a sophisticated defense to thé claims using other evidence. The claim that the parties were equally disadvantaged by loss of the vehicle ignores the very evidence that plaintiffs *96sought to include. Had the district court allowed David Trull’s lay testimony concerning the condition of the seat belts, Volkswagen would have had no basis for challenging his recollection or any expert opinion derived from his testimony. And while Volkswagen was able to vigorously defend without an expert evaluation of the Vanagon, the court recognized that defendants would be forced to resort to a “battle of the experts” armed with a small amount of information of questionable reliability. Order at 14. In this situation, the district court’s judgment is unassailable.7
C. Exclusion of evidence of the Eurovan
The Trulls contend that the district court erred in excluding evidence concerning the Eurovan, a Volkswagen vehicle manufactured for sale in Europe before the time of the accident. The Trulls sought to show that the Eurovan had been designed to solve the problems inherent in the Vanagon, thereby demonstrating that Volkswagen was aware that the Vanagon was unreasonably dangerous and that a feasible, affordable alternative was available. The district court rejected the proffered evidence as of doubtful relevance and because of “a danger of unfair prejudice.”
The Eurovan’s significance relates to the body design of the two vehicles. The 1986 Vanagon owned by the Trulls was a so-called “forward control vehicle,” which had the engine in the rear and consequently did not have the usual extended front end. The lack of a front “crumple zone” to protect occupants from a front-end collision was among the design defects that the Trulls alleged contributed to their injuries. Volkswagen, however, countered that the Vanagon’s front structure was specially designed to absorb collision forces and protect occupants.
The Eurovan, which was introduced in the late 1980s, is a front-engine vehicle and has a more traditional, longer front end. In offering the Eurovan evidence, the plaintiffs sought to show that it was, in essence, a “subsequent remedial measure” taken by Volkswagen to fix the Vanagon’s defective design. Although evidence concerning subsequent remedial measures generally is not admissible at trial, see Fed.R.Evid. 407, measures that take place before the accident at issue do not fall within the prohibition. See Bogosian v. Mercedes-Benz of North America, 104 F.3d 472, 481 (1st Cir.1997); Raymond v. Raymond Corp., 938 F.2d 1518, 1523 (1st Cir.1991). The district court does have its usual discretion, however, to exclude the evidence under Fed.R.Evid. 403 if its probative value is substantially outweighed by the danger of unfair prejudice. See Bogosian, 104 F.3d at 481; Raymond, 938 F.2d at 1523-24.
The district court ultimately rejected the Eurovan evidence as unduly prejudicial in light of its doubtful probative value, and'we again cannot fault its reasoning. There was no question that an alternative design was feasible, as it was undisputed that Volkswagen knew that a more conventional, extended front-end design was an available option at the time of the Trulls’ accident.8 Indeed, the jury heard *97ample evidence that other manufacturers’ vans had the front-end, or “snub-nose” design, and plaintiffs elicited testimony from their expert indicating that such vehicles were safer. Moreover, plaintiffs proffered no evidence that Volkswagen specifically designed the Eurovan’s front end to remedy known safety problems in the Vanagon. In addition, the proffered Eurovan evidence appears irrelevant to the remaining claims in the case; the claims of the front-seat passengers, David and Elizabeth, had been withdrawn, and the boys’ claims centered on the lack of shoulder belts, not on the lack of a front crumple zone.
In these circumstances, we doubt that admission of the evidence would have been proper. Without question, it was not an abuse of discretion to exclude it.
D. Expert opinion based on Fatal Accident Reporting System (FARS) data
Plaintiffs claim the district court erred in allowing defendants’ experts to testify about an article entitled “Usage and Effectiveness of Rear-Seat Belt Restraints in Severe Frontal Crashes” because the article contained unreliable and irrelevant data and constituted hearsay. We note, initially, that the court did not allow the article itself into evidence, but simply permitted defendants’ experts — two of the article’s authors — to testify about its contents and the basis for their conclusion that lap/shoulder belts are no more effective than lap belts alone in preventing deaths. The study described in the article was based on data compiled by the National Highway Traffic Safety Administration (NHTSA) through the Fatal Accident Reporting System (FARS), and the statistics cover every fatal frontal collision reported for the period 1975 to 1995. The FARS system was implemented by the federal government to collect data on fatal accidents in the United States, and the information it contains is drawn from police reports, motor vehicle registration files, vital statistics, and state highway department records. See Fatal Accident Reporting System 1978 Annual Report Foreward (1978).
We see no abuse of discretion in the district court’s decision to allow the experts’ testimony. Under Fed.R.Evid. 803(8), data compilations of public agencies setting forth “matters observed pursuant to duty imposed by law as to which matters there was a duty to report” are not considered hearsay. Volkswagen reports in its brief, and appellants do not contest, that the FARS data on which the article was based are collected and reported pursuant to a legislative mandate by the NHTSA, a public agency. See 49 U.S.C. § 30168(a)(1)(A). Additionally, the evidence appears to satisfy Fed.R.Evid. 703, which allows admission of expert testimony based on “facts or data ... of a type reasonably relied upon by experts in the particular field.” See International Adhesive Coating Co. v. Bolton Emerson Int’l, 851 F.2d 540, 544-45 (1st Cir.1988); see generally Seese v. Volkswagenwerk, A.G., 648 F.2d 833, 845-46 (3d Cir.1981). Indeed, appellants’ expert, John Stilson, acknowledged that the FARS data are among the appropriate sources to consult when evaluating the crashworthiness of a vehicle. Consequently, the testimony was not excludible as hearsay.
Appellants’ complaint that the study was irrelevant because it included data gathered after the date of their accident similarly misfires. A threshold question for the jury was whether the Vanagon was defective because its rear seats were equipped only with lap belts. Regardless of the state of knowledge at the time Volkswagen designed and manufactured the Trulls’ vehicle, the company could be neither negligent nor strictly liable if more comprehensive, .up-to-date information demonstrated that a seat belt system using only lap belts was reasonably safe. Although statistics focusing solely on rear lap belts in forward control vehicles would allow a more precise comparison, the una*98vailability of such evidence does not render improper the evidence that was available. The dissimilarities went to the weight of the evidence and were a proper subject for cross-examination. See International Adhesive Coating Co., 851 F.2d at 544. In sum, we find no reversible error in the admission of this expert testimony.9
E. Evidence from prior lawsuit
Plaintiffs contend that the district court erred in allowing into evidence a diagram depicting the paths of the Vanagon and the other car involved in the crash, an AMC Concord. The Trulls had relied on the diagram, which showed the vehicles colliding at a 90-degree angle, in an earlier state court lawsuit they filed against the State of New Hampshire for failure to remedy the icy road conditions. The Trulls had introduced the diagram as an exhibit through the testimony of Sgt. Leon Shackley, the officer in charge of the accident investigation for the Conway Police Department.
In the present lawsuit, the Trulls maintained that their accident involved a nearly straight, head-on collision, while Volkswagen attempted to show that the accident was a side impact, with the front end of the Concord contacting the passenger side of the Vanagon at an angle of about 55 degrees. The angle of impact was significant in the Volkswagen case because the Trulls’ claims focused on the Vanagon’s lack of a “front crumple zone” and the absence of shoulder belts to prevent the boys from surging forward in the back seat of the car.
When Shackley was called as a witness by the Trulls in this case, Volkswagen sought on cross-examination to show that the Trulls had changed their position about the angle of the collision and that — the point of significance — their original view matched Volkswagen’s. The Trulls objected on various grounds to admission of the diagram, including that the document did not qualify as a prior inconsistent statement and that its admission would improperly bring evidence of another lawsuit into the case. The district court permitted Volkswagen to introduce the diagram during cross-examination of Shackley, but barred Volkswagen from identifying the defendant or referring to the claims at issue. The exchange was brief, and went as follows:
Q. Sergeant, in another lawsuit brought by the Trulls, you were called as a witness in the trial of that case; right?
A. Yes.
Q. Do you recall that in that case the Trulls, through their lawyer, introduced this diagram ... as a description of the paths of the vehicles in the accident?
A. I don’t specifically recall it. It was over a year ago or almost a year ago when that one went.
Volkswagen’s attorney then showed Shack-ley the transcript from the earlier proceeding.
Q. Do you see the question from the Trulls’ counsel to you?
A. Yes.
Q. And the question — the sense of the question was whether that diagram was *99an accurate portrayal of how the vehicles traveled in this accident; correct?
A. Yes.
Q. And then the Trulls lawyer asked that that be admitted as evidence; correct?
A. Yes.
At that point, the district court gave the following limiting instruction:
Members of the jury, I’m going to instruct you at this time that the mere fact that another lawsuit may have been brought by the Trulls is not a matter which you should in any way consider in deciding this case. The mere fact of that lawsuit is not material and relevant to the decisions that you have to make in this case. However, to the extent that you find in any evidence presented to you any positions taken consistent or inconsistent with claims made in this case, then you may weigh and consider that evidence along with all the other evidence in this case. But the mere fact of that lawsuit is something that you should disregard in your deliberations.
Subsequent to this testimony, Volkswagen sought to re-call David Trull for questioning about the diagram and about a related statement made by the Trulls’ attorney in the earlier trial. The statement was made during a bench conference, as that attorney, Randall Cooper, advocated admission of a photograph of the accident scene. He stated that the photograph “certainly demonstrates that the reconstruction by the Police Department is accurate in showing how the vehicle slid, and there was a side-to-front impact.” Before finally deciding whether to allow Trull to be re-called, the district court conducted voir dire questioning of Trull and Cooper. They both emphasized that the angle of the accident was not at issue in the earlier trial and that the diagram and statement by Cooper consequently were of little weight and should not be deemed an inconsistent position.
The district court declined to allow the questioning of Trull. On appeal, plaintiffs repeat their objections to the introduction of the diagram and assert that its admission constituted prejudicial, reversible error.
This issue strikes us as a classic example of the value of leaving matters of evidence to the discretion of the district court. Unquestionably, the Trulls in the earlier trial put forward as accurate the diagram from which they now wish to distance themselves, and it is a matter of record that their attorney described the accident as a “side-to-front impact.” With only that much before us, we have simply a prior factual claim contradictory to a factual position taken in this case by the same party, and thus a presumptively admissible statement against interest. See Kassel v. Gannett Co., 875 F.2d 935, 952-53 (1st Cir.1989); Estate of Spinosa v. International Harvester Co., 621 F.2d 1154, 1157 (1st Cir.1980). With the limited scope of questioning permitted by the court and the curative instruction, we cannot say the district judge abused his discretion in admitting the diagram during the Shackley cross-examination. Through their own questioning, the Trulls had the opportunity to dimmish the force of the diagram by eliciting that it was not drawn with technical precision and that Shackley no longer considered it to be accurate.
In addition, the court limited the impact of the diagram by rejecting Volkswagen’s request to question David Trull about it. The court’s decisions reflect the give-and-take of a trial, and the careful consideration that a judge is obliged to give to the parties’ competing positions. Although the voir dire questioning of Trull and his attorney, which occurred after admission of the diagram, suggested that the inconsistency may have resulted more from inattentiveness than deliberate strategy, this softer view of the contradiction did not retroactively make erroneous or prejudicial the diagram’s admission. The district court consistently dealt with this issue *100thoughtfully and with respect for each party’s position. We find no abuse of discretion.
F. Exclusion of correspondence.
Plaintiffs sought admission into evidence of more than a dozen letters written between Volkswagen and two government agencies, the National Transporta^ tion Safety Board (NTSB) and the NHTSA. The correspondence related primarily to Volkswagen’s knowledge of the added protective benefits provided to rear seat occupants by shoulder belts. We decline to dwell on this issue. The district court allowed plaintiffs to introduce a sampling of the letters, concluding that the remainder were cumulative and that their admission and the resulting testimony would be unduly confusing for the jury. Plaintiffs’ brief fails to identify crucial, non-cumulative information that was omitted as a result of the court’s selective admissions. We find no abuse of discretion.
G. Certification of Burden of Proof
The crashworthiness/enhanced injury theory of liability under which the Trulls brought their claims arises in automobile accident eases when plaintiffs claim that a defect in their vehicle caused them to suffer greater injuries than would have occurred had the vehicle been properly designed. See Larsen v. General Motors Corp., 391 F.2d 495, 503 (8th Cir.1968). In such cases, the plaintiffs seek to recover damages from the car manufacturer for at least a portion of their injuries.
Whether the plaintiff or the defendant bears the burden of segregating the injuries attributable to the vehicle’s defect is a question that has divided the courts that have considered it. See Polston v. Boomershine Pontiac-GMC Truck, Inc., 952 F.2d 1304, 1310 (11th Cir.1992) (citing cases); McLeod v. American Motors Corp., 723 F.2d 830, 833-34 (11th Cir.1984) (same); Czarnecki v. Volkswagen of America, 172 Ariz. 408, 837 P.2d 1143, 1147-48 (App.1991) (same); see also Restatement (Third) of Torts: Products Liability § 16, at 243-53 (1998); H. Vickies & E. Oldham, “Enhanced Injury Should Not Equal Enhanced Liability,” 36 S. Tex. L.Rev. 417, 426-38, 440 (1995); G. Tietz, M. Bushman & J. Podraza, Jr., “Crashworthiness and Erie: Determining State Law Regarding the Burden of Proving and Apportioning Damages,” 62 Temp. L.Rev. 587, 595-611 (1989). Both positions appear to have ample justification, and a given state’s stance is not easily predicted.10 The New Hampshire Supreme Court has not yet faced the issue, and the Trulls therefore moved for certification in the district court. That court declined to certify, finding that New Hampshire law was adequately revealing and that it required the burden to be placed on the plaintiffs. The Trulls have renewed their request for certification on appeal.11
*101We are not persuaded that New Hampshire’s approach to this issue can be discerned from review of the state’s existing tort law cases. Because of the sharp division in the courts that have addressed the issue, and the policy concerns that underlay the decision, we have chosen to consult the New Hampshire Supreme Court through certification. See Kansallis Finance Ltd. v. Fern, 40 F.3d 476, 481 (1st Cir.1994) (absent controlling state law precedent, federal appeals court sitting in diversity has discretion to certify state law questions to highest state court); Polston, 952 F.2d at 1310 (certifying the same question to the Georgia Supreme Court). Before framing our query, we summarize the competing perspectives.
The minority view, which is the one endorsed by Volkswagen and used by the district court in this case, places the burden on the plaintiff to prove the nature and extent of his enhanced injuries. This so-called Huddell/Caiazzo approach, derived from Huddell v. Levin, 537 F.2d 726, 737-38 (3d Cir.1976) (New Jersey law), and Caiazzo v. Volkswagenwerk, A.G., 647 F.2d 241, 250 (2d Cir.1981) (New York law), requires a plaintiff to prove three things: (1) that an alternative, safer design would have been practicable under the circumstances; (2) the injuries, if any, that would have resulted if the alternative design had been used; (3) the extent of enhanced injuries attributable to the defective design. See, e.g., Huddell, 537 F.2d at 737-38. The latter two elements are, of course, corollaries, both in essence requiring the plaintiff to sort out which injuries would not have occurred, and which injuries (including additional ones) would have occurred, had there been no defect.
Volkswagen argues that this approach is conceptually clear and consistent with basic tort principles. It asserts in its brief that, “because in a crashworthiness case plaintiffs essential claim is that the design defect caused the plaintiff to suffer injuries he would not have received but for a defect in the vehicle’s design, the existence and extent of these more severe or ‘enhanced’ injuries are an essential element of the plaintiffs own theory of recovery for which the plaintiff bears the burden of proof.” See Caiazzo, 647 F.2d at 251.12
The company further argues that this approach reflects New Hampshire’s approach to tort law, which is premised on “the fundamental tenet ... that the plaintiff retains the ultimate burden of persuasion in negligence actions,” Pillsbury-Flood v. Portsmouth Hospital, 128 N.H. 299, 512 A.2d 1126, 1129 (1986), and which, in the context of a suit for aggravation of pre-existing injuries, places the burden of proving the extent of aggravation on the plaintiff, see Valliere v. Filfalt, 110 N.H. 331, 266 A.2d 843, 845 (1970).
The more widely used Fox-Mitchell approach, derived from Fox v. Ford Motor Co., 575 F.2d 774, 787-88 (10th Cir.1978) (Wyoming law), and Mitchell v. Volkswagenwerk, A.G., 669 F.2d 1199, 1206-1208 (8th Cir.1982) (Minnesota law), requires a plaintiff to prove only that the design defect was a “substantial factor in producing damages over and above those which were probably caused as a result of the original impact or collision,” see Mitchell, 669 F.2d at 1206. Once the plaintiff makes that showing, the burden shifts to *102the defendant to show which injuries were attributable to the initial collision and which to the defect. Under this line of cases, a manufacturer who fails to make such an allocation is held liable for all of the plaintiffs injury. Thus, in eases where the injury is “indivisible” — such as, for example, certain instances of paraplegia or death- — manufacturers are treated as joint tortfeasors for all injuries. See id.13
The Mitchell and Fox courts also claim a traditional tort law heritage for their approach. In explicitly rejecting the Hud-dell-Caiazzo allocation of the burden, the court in Fox noted “the orthodox doctrines of joint liability of concurrent tort-feasors for injuries which flow from their concurring in one impact,” 575 F.2d at 787. Under this doctrine, the Mitchell court pointed out, wrongdoers who each play a substantial role in creating an indivisible harm are treated as joint and several tortfeasors. See 669 F.2d at 1207. According to the Mitchell court, use of the Huddell-Caiazzo model in enhanced injury cases involving an indivisible injury departs from this well established approach and “will generally result in complete exoneration of the negligent manufacturer.” Id. at 1207-08. The court observed:
By placing the burden of proof on a plaintiff to prove that the designer was the sole cause of not only an enhanced indivisible- injury, but, in addition, that he would not otherwise have received injuries absent a defect, the injured victim is relegated to an almost hopeless state of never being able to succeed against a defective designer. The public interest is little served.... A rule of law which requires a plaintiff to prove what portion of indivisible harm was caused by each party and what might have happened in lieu of what did happen requires obvious speculation and proof of the impossible. This approach converts the common law rules governing principles of legal causation into a morass of confusion and uncertainty.
Id. at 1204-05.
The Fox-Mitchell construct thus was designed to ensure that plaintiffs whose injuries were in part traceable to a design defect would be able to recover from the manufacturer, even when they could not prove precisely how the defect made their injuries more severe. Using the joint tort-feasor setting as a model, this approach “is a result of a choice made as to where a loss due to failure of proof shall fall — on an innocent plaintiff or on defendants who are clearly proved to have been at fault,” see Mitchell, 669 F.2d at 1208 (quoting Mathews v. Mills, 288 Minn. 16, 178 N.W.2d 841, 845 (1970)).
Although Volkswagen emphasizes that the New Hampshire Supreme Court explicitly has rejected burden-shifting for the purpose of assisting plaintiffs, the primary case it cites addresses a plaintiffs argument that the burden of negating causation should have been shifted to the defendants. See Pillsbury-Flood, 512 A.2d at 1128. The Fox-Mitchell approach does not relieve a plaintiff of the threshold obligation of proving causation, and thus liability; it is only after a plaintiff has demonstrated that the design defect was a “substantial factor” in producing damages over and above those that otherwise would have occurred that the burden shifts to the'defendant to apportion damages. Indeed, in rejecting the burden-shifting proposed by the plaintiff in Pillsbury-Flood, the New Hampshire court noted that its rule ensures that defendants will not have to defend against improbable claims, and that defendants will not have to disprove *103the element of causation. Proving a negative poses practical and theoretical problems that should not be placed on defendants.
512 A.2d at 1129-30. A very similar rationale motivated the Mitchell court, which concluded that plaintiffs who have proven causation should not be faced with the “impossible burden of proving a negative fact,” 669 F.2d at 1204-05. In our view, the difference between shifting the initial burden of causation and shifting the burden of allocating damages is sufficiently significant that the result in Pillsbury-Flood does not necessarily lead to the Huddell-Caiazzo approach.14
In sum, we think that the burden of proof that the New Hampshire Supreme Court would choose for a crashworthiness case is not readily apparent from existing case law and is quintessentially a policy judgment appropriately made for the state by its own courts.15 The applicable burden of proof unquestionably may have been determinative of the jury’s verdict. The court instructed the jury that, to establish Volkswagen’s liability, the plaintiffs had to prove, along with causation, “the nature and extent of the injuries that were enhanced.” If the Fox-Mitchell approach were the appropriate standard, the plaintiffs would have met their burden to establish liability simply by showing that a defect in the vehicle was a “substantial factor” in producing damages over and above those that otherwise would have occurred. The burden of apportioning damages would have fallen on Volkswagen. We therefore certify to the New Hampshire Supreme Court, pursuant to its Rule 34, the question set forth in the attached certificate.
III. Conclusion
For these reasons, we affirm the various evidentiary rulings by the district court and certify the question of the appropriate burden of proof to the New Hampshire Supreme Court. We retain jurisdiction pending that court’s determination.
CERTIFICATION
For the reasons discussed in our opinion, the following question is respectfully certified to the Supreme Court of New Hampshire, pursuant to its Rule 34:
Under New Hampshire law, in a crash-worthiness or enhanced injury case, does the plaintiff bear the burden of demonstrating the specific nature and extent of the injuries attributable to the manufacturer, or does the burden of apportionment fall on the defendant once the plaintiff has proved causation?
In asking only this question, we do not intend to limit the scope of the court’s response. We welcome discussion of any relevant New Hampshire law that the Supreme Court deems appropriate.
The clerk of this court will transmit our certification, along with copies of the briefs and the record appendix in this case, to the New Hampshire Supreme Court. We shall await its reply with interest and appreciation.
It is so ordered.
*104ORDER
Aug. 26, 1999
The petition for rehearing is denied. The argument that the jury would have reached the same result even if the Fox-Mitchell approach had been used was not raised in petitioner’s brief at any point. Indeed, the entire thrust — the statement of the issue, the argument, and the conclusion — centered on the question of location of burden of .proof. For this reason, alone, the petition is unavailing. See Ramsdell v. Bowles, 64 F.3d 5, 8 (1st Cir. 1995) ( “arguments not presented in initial brief on appeal are waived”) (citing Pignons S.A. de Mecanique v. Polaroid Corp., 701 F.2d 1, 3 (1st Cir. 1983)); Aunyx Corp. v. Canon U.S.A., Inc., 978 F.2d 3, 9 (1st Cir. 1992) (“potential issues not supported by argument in appellant’s brief are deemed abandoned” )(citing Brown v. Trustees of Boston Univ., 891 F2d 337, 352 (1st Cir. 1989)).
Moreover, the petition would fail on its merits as well. At trial, the parties disputed what injuries were caused by the lack of shoulder harnesses. The Trulls argued that there would have been only bumps and bruises if there had been shoulder belts, while Volkswagen maintained that the same injuries would have occurred with shoulder belts as had occurred without them. A verdict for Volkswagen, under the instruction that the Trulls carried the burden of proving allocation, could have meant only that the jury was not persuaded that the Trulls had proven that all of the serious injury was attributable to the defect. A finding for Volkswagen in those circumstances does not mean the jury would reach the same result if Fox-Mitchell were the applicable standard, giving the auto manufacturer the burden of apportioning the injuries.
. The defendants are Volkswagen of America, Inc., and Volswagenwerk, A.G. For the sake of simplicity, and because nothing turns on the distinction, we treat both together as "Volkswagen.”
. Although the defendants included David Trull in their motion, and the district court included him as well in its disposition, plaintiffs report that his claims had dropped from the case earlier.
. We note that plaintiffs at times have suggested that the exclusion of the seat belt evidence by itself prevented Elizabeth from proceeding with her claims and at other times have indicated that the exclusion of the seat belt evidence was fatal only in light of the burden of proof ruling. If the former were true, our ruling on the spoliation motion would render any further discussion of her appeal unnecessary. Although we are reluc.tant to speculate about trial strategy, it is possible that, had the burden of proof issue been resolved differently, Elizabeth would have brought her claims and substituted more inferential evidence about the condition of Elizabeth’s seat belt for the testimony excluded by the district court. See note 7 infra. Because of the ambiguity, we have chosen to consider the procedural viability of Elizabeth’s appeal.
. The document stated that, "[i]n accordance with the following, judgment is hereby entered:
4. Order dated May 4, 1998, granting Defendants' Motion for Dismissal With Prejudice of Claims Made by Elizabeth and David Trull; and,
5. The Special Verdicts returned by the jury on May 4, 1998....”
. Had plaintiffs pursued either of these paths, their situation would be more similar to that of the plaintiffs in BIW Deceived v. Local S6, 132 F.3d 824, 828 (1st Cir.1997), in which we ruled that a party could consent to a judgment against it and still preserve its right to *95appeal. There, the plaintiffs sought a final judgment in favor of defendants for the explicit purpose of bringing an appeal of the district court's adverse preemption ruling. We noted that, despite the usual rule that a party to a consent judgment waives any objections to matters -within the scope of the judgment, a party who reserves its rights "unequivocally” may retain the right to appeal. Here, plaintiffs did not reserve their appellate rights at the time of the dispositive order, the dismissal with prejudice, and also have not appealed the ruling.
. On the issue of fault, Trull was asked in his deposition whether he thought Volkswagen "was at fault for something” at the time he inspected the vehicle and noticed that a portion of the seat-belt mechanism was missing. He replied: “I won’t say that it didn’t cross my mind at that point, but unfortunately I had much more pressing personal matters to deal with at the time.”
. In its order, some eight months before trial, the district court indicated that it would be open to a limiting instruction in lieu of exclusion of the evidence, but it appears that none was proposed.
In addition, the plaintiffs were not barred from introducing all evidence concerning Elizabeth Trail's seat belt. In April 1998, the district court denied Volkswagen's motion to clarify and extend the exclusion order to cover all seat-belt evidence, apparently clearing the way for plaintiffs to introduce evidence that Elizabeth was wearing a seat belt but nonetheless was ejected from the vehicle— evidence that would have permitted an inference of a seat-belt defect. Although such an inference would have been less compelling than direct testimony that the belt was broken, it nonetheless weakens plaintiffs' assertion that the spoliation order entirely deprived Elizabeth of her claims.
. Volkswagen’s position was that extended front-end and forward-control vans were different types of vehicles, not just different designs for the same category of vehicle.
. Appellants rely on case law holding that evidence of prior accidents is admissible only if the proponent shows that the earlier accidents occurred under " 'circumstances substantially similar to those at issue in the case at bar,’ ” Moulton v. Rival Co., 116 F.3d 22, 26-27 (1st Cir.1997) (quoting McKinnon v. Skil Corp., 638 F.2d 270, 277 (1st Cir.1981)). The cases they cite, however, all involve attempts by plaintiffs to bring evidence of other accidents before the jury to demonstrate dangerousness or notice of danger, a particularly ’ prejudicial form of evidence when used to counter a defendant’s assertion that its product is safe. At bottom, the "substantially similar" requirement is a more particularized approach to the requirement that evidence be probative. In the circumstances here, where defendants sought to introduce evidence of other accidents, we think it was adequate for the court to determine more generally whether the evidence was probative and, if so, whether it was unfairly prejudicial.
. The recently revised Restatement of Torts lists 23 states as favoring the view that the defendant should bear the burden of allocating damages and six in support of placing the burden on the plaintiff. See Restatement (Third) of Torts: Products Liability § 16, Comment (d), at 244 (1998). The Reporters, both on the basis that placing the burden on the defendant has been endorsed by a substantial majority of the jurisdictions that have considered the problem, and, more importantly, on the basis of the principles of causation articulated in § 433B of the Restatement (Second) of Torts, suggest that this view reflects the correct approach. See Restatement (Third) of Torts, § 16, Comment (d), at 251-53. A 1995 law review article noted, however, that in the two preceding years there appeared to be a trend in favor of the minority view. See H. Vickies & M. Oldham, "Enhanced Injury Should Not Equal Enhanced Liability," 36 S. Tex. L.Rev. 417, 440 (1995).
. Technically, the Trulls have both appealed the district court’s rejection of the certification motion they filed in that court and made an independent request for certification to this court. Our consideration of the renewed request makes it unnecessary to determine whether the district court abused its discretion in refusing the earlier one. As we have noted, however, the fact that the district court, in the exercise of its discretion, reached *101a different conclusion from ours does not, on this record, indicate any abuse of discretion.
. The Caiazzo court explained its rule as follows:
We realize that a plaintiffs burden of offering evidence of what injuries would have resulted absent the alleged defect will be heavy in some instances and perhaps impossible in others. Where it is impossible, however, the plaintiff has merely failed to establish his prima facie case, i.e., that it is more probable than not that the alleged defect aggravated or enhanced the injuries resulting from the initial collision. Moreover, in those instances in which the plaintiff cannot offer any evidence as to what would have occurred but for the alleged defect, the plaintiff has not established the fact of enhancement at all.
647 F.2d at 251.
. Under this approach, the court initially decides whether an injury is indivisible. If the court finds that it is, the jury is instructed to determine only whether the negligent design of the manufacturer was a substantial factor in producing that injury. If the jury makes such a finding, the manufacturer is held jointly and severally liable for the damages. See Mitchell v. Volkswagenwerk, A.G., 669 F.2d 1199, 1209-1210 (8th Cir.1982). If the court finds that the injury is divisible, the jury must apportion the damages appropriately.
. We acknowledge that one New Hampshire case cited by Volkswagen tends to support its position. In Valliere v. Filfalt, 110 N.H. 331, 333, 266 A.2d 843, 845 (1970), the court endorsed an instruction placing the burden on the plaintiff to show "whether and to what extent the plaintiff's pre-existing condition was aggravated by the accident.” But, as is evident, the court treated causation together with the extent of the injury, understandably giving no consideration to the rationales that have polarized the opinions in the crashwor-thiness context, since virtually all of these cases were decided in more recent years. We therefore cannot say how the court would resolve the issue before us.
. We note that a New Hampshire trial court adopted the Huddell-Caiazzo approach, but did so without analysis. See Mansfield v. CT Corp., No. C-1 084-84, slip op. at 4-5 (N.H.Super. Ct. April 20, 1987).