TIMBERS, Circuit Judge.

General Motors Corporation (GM) appeals from a judgment entered in the Northern District of Ohio, John M. Manos, District Judge, on October 6, 1981 upon a jury verdict awarding Dennis C. Sours and his father Olen Sours $2,282,402.00 and $49,-336.41, respectively, for personal injuries sustained by Dennis, on September 30,1976, when his 1968 Chevrolet Camaro slid off the road and rolled over in a one-car accident, and for hospital and medical expenses incurred by Olen. Dennis’ neck was broken in this crash as the result of the partial collapse of the roof of the car when it rolled over. Dennis was left a quadriplegic.

The jury, in its answers to special interrogatories, found that GM was liable for the injuries and damages sustained, first, because it negligently had designed the car’s roof so that it could not withstand the low-speed roll-over, and, second, because the roof structure constituted a defective product under the principles of strict liability. The jury also found that Dennis’ injuries were proximately caused by the collapse of the car’s roof and that the car had not been substantially modified so as to relieve GM of liability.

On appeal GM claims that there was insufficient evidence to support the jury’s findings of negligence, product defect, and lack of substantial modification. GM also claims that the district court erred in excluding evidence that Dennis was not wearing his seat belt at the time of the accident.

After careful consideration, we find each of GM’s claims of error to be without merit. We affirm.

On the day of the accident, plaintiff’s 1 automobile slid, the passenger or right side *1513 first, off Ohio Route 303 near its intersection with Ohio Route 88 in Portage County, Ohio. The car skidded into a ditch bank, still sideways, and then rolled over a full 360 degrees, coming finally to rest again on its wheels, facing in the direction from which it came. During the course of the roll, the roof of the car crashed into the ground and was deformed, collapsing partially into the passenger compartment over the head of the plaintiff driver. Although none of the three passengers was injured seriously, plaintiff sustained serious injuries as stated above. Neither plaintiff nor the three passengers were wearing seat belts when the accident occurred.

Plaintiff commenced the instant diversity action in September 1978 against GM, whose Chevrolet division had manufactured the car. He sought compensatory and punitive damages resulting from Dennis’ injuries and Olen’s out-of-pocket expenses and loss of his son’s companionship. The amended complaint filed in November 1979 set forth three grounds for recovery: (1) negligence in the design of the car’s roof; (2) strict liability for the defective design of the car’s roof; and (3) breach of express and implied warranties. The breach of warranty claims were dismissed prior to trial upon GM’s motion; they are not before us on this appeal.

GM filed its answer to the amended complaint on February 6, 1981, generally denying the allegations of the complaint. GM alleged, as it urged at trial, that plaintiff’s injuries were caused not by the collapse of the roof but by his dive into the right roof rail as the car rolled over. GM also pleaded the affirmative defenses of contributory negligence and assumption of risk, based on its claim that plaintiff’s injuries occurred because he was not wearing his seat belt.

On February 17, 1981, plaintiff moved to strike the latter defenses on the ground that evidence of failure to use an automobile seat belt is inadmissible under Ohio law to prove contributory negligence, much less the deliberate participation in a known peril necessary to sustain the defense of assumption of risk. Shortly before trial plaintiff renewed his claims in the form of a motion in limine to prohibit any reference at trial to Dennis’ failure to use a seat belt. The district court, based on its interpretation of controlling Ohio law, granted both of plaintiff’s motions in a memorandum opinion filed September 17, 1981. To preserve its position, GM made an offer of proof at trial with respect to the seat-belt evidence.

The sixteen day trial began in Cleveland on September 14, 1981. At the close of plaintiff’s case, GM moved for a directed verdict, which the court granted as to the punitive damage claim but denied in all other respects. At the close of the entire case, GM renewed its motion for a directed verdict, which again was denied.

On October 6, the jury returned its verdict in the form of answers to special interrogatories as set forth above. On October 9, judgment against GM was entered upon the jury verdict. Thereafter, GM’s motions for a new trial and for judgment n.o.v. were denied. GM thereupon filed the instant appeal.

We hold that there was sufficient evidence to support the jury’s verdict, including its specific findings that negligence in the design of the car’s roof proximately caused plaintiff’s injuries, that product defect proximately caused his injuries, and that there was no substantial modification. We also hold that the seat belt evidence, even if admissible, would have had no bearing on the case as it was tried by the parties and therefore its exclusion, if error at all, was harmless. We affirm in all respects.

II.

In view of the importance of applying the correct law in resolving the issues with which we deal in the remainder of this opinion, it is critical that we determine at the outset what is the applicable law. We take it to be common ground that, in this diversity action arising from an accident that occurred in Ohio, the law of that state

*1514 as enunciated by the Ohio Supreme Court governs. Erie Railroad Co. v. Tompkins, 304 U.S. 64 (1938).

We recently have described the precise contours of our inquiry as follows:

“A federal court sitting in diversity must apply the law of the state’s highest court. If the highest court has not spoken, the federal court must ascertain from all available data what the state law is and apply it. If the state appellate court announces a principle and relies on it, that is a datum not to be disregarded by the federal court unless it is convinced by other persuasive data that the highest court of the state would decide otherwise.” (citations omitted).

Clutter v. Johns-Manville Sales Corp., 646 F.2d 1151, 1153 (6th Cir.1981). See also Coleman v. Western Elec. Co., 671 F.2d 980, 983-84 (6th Cir.1982); 1A Moore’s Federal Practice ¶ 0.307-0.309[3], at 3077-3134 (2d ed. 1982).

III.

GM challenges the jury verdict in four respects. First, it argues that, under the “multiple factor” risk-benefit analysis announced in Knitz v. Minster Machine Co., 69 Ohio St.2d 460, 432 N.E.2d 814, cert. denied, 459 U.S. — (1982), there was insufficient evidence to support the finding of product defect. Second, it argues that its compliance with federal standards was a conclusive defense to the claims of product defect and negligence. Third, it argues that there was insufficient evidence of negligent design to sustain the jury verdict. Finally, it argues that the finding of no substantial modification is insupportable. We shall discuss each of these arguments seriatum.

(A)

The crux of strict products liability law, in Ohio as in other states, is that a product must perform in accordance with the expectations of the ordinary user. 2 A product that does not measure up to such expectations is “unreasonably dangerous” and hence defective within the meaning of Restatement (Second) of Torts § 402A (1965), which Ohio has expressly adopted. Temple v. Wean United, Inc., 50 Ohio St.2d 317, 364 N.E.2d 267 (1977). “A product will be found unreasonably dangerous if it is dangerous to an extent beyond the expectations of an ordinary consumer when used in an intended or reasonably foreseeable manner.” Leichtamer v. American Motors Corp., 67 Ohio St.2d 456, 467, 424 N.E.2d 568, 577 (1981). The Ohio Supreme Court recently has reaffirmed its adherence to this basic principle. Knitz v. Minster Machine Co., supra, 69 Ohio St.2d at 465-66, 432 N.E.2d at 817-18.

The jury in the instant case was instructed in accordance with this consumer expectation test. We cannot say — nor does GM seriously contend — that under this test the evidence was insufficient to support the jury finding of product defect. On the face of it, a consumer might well expect that the roof of his car would withstand a low-speed roll-over without significant collapse. As GM’s own expert succinctly stated, the roof design was intended “to provide structural strength and integrity — it does more than just keep the rain out.” GM as the manufacturer therefore recognized the need to design a crashworthy roof. Moreover plaintiffs introduced substantial persuasive evidence on this issue. For example, they introduced testimony bearing on the manufacturer’s conceded responsibility to design a “crashworthy” vehicle, as well as the hazards presented by roll-over accidents. They also introduced credible evidence of the “encroachment” problem — that being the tendency of the roof on the 1968 Camaro body type (“F-37”) to deform and collapse into the passenger area under stress such as that to which the roof was subjected in the *1515 forseeable low-speed roll-over accident here involved.

GM challenges this evidence, contending that plaintiffs’ experts, standards and tests gave a misleading presentation of the F-37 body and its “crashworthiness” in roll-over accidents. GM focuses on the evidence that emerged from its own experts, standards, and test results. These depicted the F-37 roof as a safer design than did plaintiffs evidence. It is the classic function of the jury, however, to resolve just this sort of conflict in the evidence.

The essential point here is that the jury had before it evidence that under certain foreseeable accident conditions the F-37 roof would collapse “or encroach” into the passenger area. In this regard the instant case is strikingly similar to the seminal Ohio case of Leichtamer v. American Motors Corp., supra, in which the Ohio Supreme Court held, in affirming the court of appeals, that there was sufficient evidence of product defect to sustain a jury verdict against American Motors where the “roll-bar” on its CJ-7 jeep failed to protect passengers in a roll-over accident. There the Ohio Supreme Court stated that the jeep had been advertised as an off-road vehicle, meant to be used in rugged terrain, and that jeep owners consequently could expect protection from the roll-bar in a foreseeable accident. The adequacy of the protection actually provided — judged in terms of the ordinary consumer’s expectations — was held to be a jury question. “[A]n instruction may be given on the issue of strict liability in tort if the plaintiff adduces sufficient evidence that an unreasonably dangerous product design caused or enhanced plaintiff’s injuries in the course of a foreseeable use.” 67 Ohio St.2d at 465, 424 N.E.2d at 576 (citation omitted).

Plaintiff’s evidence here that the F-37 roof was defective parallels closely the evidence found sufficient in Leichtamer to submit the question of product defect to the jury. In both cases a protective structure was shown to lack the strength to resist the stress of a foreseeable roll-over accident. It cannot be seriously claimed that there was insufficient evidence here, under Ohio’s consumer expectation test, to submit the question of product defect to the jury.

GM’s chief argument is that a different test should have been applied: the three factor risk-benefit analysis formulated by the Ohio Supreme Court in Knitz v. Minster Machine Co., supra 3 — one month after the notice of appeal in the instant case was filed. GM’s argument in this respect turns on a misreading of Knitz in which the Ohio Supreme Court reaffirmed its adherence to the consumer expectation standard as the touchstone of strict products liability. It then went on to delineate an alternative standard for cases in which consumer expectations were not likely to be helpful, either because impossible to gauge or somehow distorted. Only for those peculiar circumstances was the Knitz variant designed. Indeed the alternative standard aids plaintiffs in cases where a product embodies “excessive preventable danger”, 69 Ohio St.2d at 465, 432 N.E.2d at 818, even though it satisfies ordinary consumer expectations. GM now seeks to beat what was intended as a plaintiffs’ sword into a shield for manufacturers asserting that the risk-benefit analysis must be undertaken even where consumer expectations do provide useful guidance. We disagree.

We believe it is clear that the alternative standard set forth in Knitz was not meant to eclipse the consumer expectation standard; rather, it was intended to serve as a refinement of that general principle in those situations where expectations were likely to be distorted:

“Unlike the factual setting in Leichtamer, there are situations in which ‘the consumer would not know what to expect, because he would have no idea how safe the product could be made’ .... Difficulty could arise, for example, where the injured party is an innocent bystander *1516 who is ignorant of the product and has no expectation of its safety, or where a new product is involved and no expectation of safety has developed. Conversely, liability would be barred hypothetically where industrial workmen gradually learn of the dangers involved in the machinery they must use to make a living and come to ‘expect’ the dangers. In such cases, the policy underlying such tort requires that ‘a product may be found defective in design, even if it satisfies ordinary consumer expectations’ .... ”

69 Ohio St.2d at 465, 432 N.E.2d at 818 (footnotes and citations omitted) (emphasis added).

As the court in Knitz pointed out, reasonable expectations of auto safety could be gauged in Leiehtamer, where the jeep had been advertised as a safe off-road vehicle. But just as jeep owners could expect protection on their excursions into perilous off-road terrain, so automobile owners can expect protection from their highway-bound vehicles in foreseeable accidents. The automobile is hardly a new product, nor was plaintiff herein an innocent bystander or an industrial workman using a machine under conditions of perceived duress. The evidence in the instant case established that roll-over accidents are a common occurrence. In both Leiehtamer apd the instant case, the same kind of consumer expectations were implicated. The expectations of ordinary automobile owners with respect to foreseeable accidents in the course of everyday on-road vehicle operation probably are easier to define than the adventurers’ expectations concerning inherently hazardous off-road performance in open jeeps, advertising notwithstanding.

Since the jury was properly instructed regarding the consumer expectation standard and not the Knitz variant, we find it neither necessary nor appropriate to consider whether plaintiff would have satisfied the alternative standard (even if the trial judge had had the omniscience to forecast the yet-to-be decided Ohio Supreme Court decision in Knitz). 4

Furthermore, the battle of the experts and their proofs, which GM invites us to resolve, is not ours to decide de novo. That is the prerogative of the jury. The question before us is whether the jury’s verdict that the F-37 roof was a defective product because it failed to meet the expectations of ordinary consumers as to strength and crashworthiness must be set aside as “manifestly opposed to the clear weight of the evidence.” Coffey v. Shenk, 39 Ohio App.2d 156, 163, 316 N.E.2d 917, 922 (1974). On the record before us it cannot seriously be maintained that the jury’s verdict falls within that category. “It is only when the record shows a serious mistake or misapprehension so violent as to shock the senses that a judgment will be reversed as being against the manifest weight of the evidence.” Nye v. Schuler, 110 Ohio App. 443, 445, 165 N.E.2d 16, 18 (1959). That most certainly is not the case before us.

We hold that the jury was properly instructed regarding the proper standard for determining product defect and that there was ample evidence to support the jury’s verdict on that issue.

(B)

GM also challenges the jury verdict with what strikes us as a notably ill-conceived argument that the F-37 roof complied with certain federal safety standards and thus as a matter of law could not have been defective or negligently designed. We disagree.

First, there was no federal standard for roof strength in effect at the time the F-37 body was designed. The subsequently promulgated Federal Motor Vehicle Safety *1517 Standard 216 (“FMVSS 216”), 49 C.F.R. § 571.216 (1982), did no more than establish a minimum standard; it did not presume to lay down a template for proper roof design. Furthermore, there was substantial dispute below as to whether the F-37 roof in fact met the FMVSS 216 standard.

Second, and perhaps more telling, the very federal safety statute upon which GM relies makes it abundantly clear that compliance with the regulations promulgated thereunder does not immunize a manufacturer from common law liability. 15 U.S.C. § 1397(c) (1976). This same effect is given to statutory regulation under Ohio law. Knitz v. Minster Machine Co., supra, 69 Ohio St.2d at 464, 432 N.E.2d at 817 (statutory regulation only a guide); Hageman v. Signal L.P. Gas, Inc., 486 F.2d 479, 483 (6th Cir.1973) (same).

Relying on Temple v. Wean United, Inc., 50 Ohio St.2d 317, 364 N.E.2d 267 (1977), a negligence case, GM argues that compliance with safety standards defeats the claim of negligent design as a matter of law. But Temple is circumscribed by its unique facts and was limited by its terms to the context of a negligence action. It does not support GM’s assertion that a product that so conforms cannot be defective under principles of strict liability. In Temple, the court construed an Ohio statute that required employers to comply with orders of the Ohio Industrial Commission. One of these orders set forth a list of “Acceptable Methods of Guarding” power presses. Since the manufacturer, Wean United, had incorporated one of these “acceptable” mechanisms into its presses, the court held that the design was not negligent as a matter of law. 50 Ohio St.2d at 325, 364 N.E.2d at 273. The utilization of officially approved design features, however, is far different from compliance with the threshold FMVSS 216 standard, which by its terms sets only a minimum level of roof strength and does not purport to relieve the complying manufacturer of common law liability; indeed, the authorizing statute specifically disclaims any such intent. Furthermore, the Ohio Supreme Court in the more recent case of Knitz has stated that such compliance is only a guide and not conclusive.

We hold that GM’s alleged compliance with FMVSS 216, along with its other evidence of adherence to industry customs and standards, was properly left for the jurors to factor into the calculus that comprises reasonable design in a case of strict products liability.

(C)

GM argues — somewhat obliquely in our view — that there was insufficient evidence to support the jury’s finding of negligence in the design of the F-37 roof. Under Ohio law, “[i]t is the duty of a manufacturer to use reasonable care under the circumstances to so design his product as to make it not accident or foolproof, but safe for the use for which it is intended”. Temple v. Wean United, Inc., supra, 50 Ohio St.2d at 326, 364 N.E.2d at 273 (quoting Gossett v. Chrysler Corp., 359 F.2d 84, 87 (6th Cir.1966)).

There was ample evidence from which the jury could find or infer that GM did not design the F-37 roof with reasonable care. Plaintiffs’ expert, Marcosky, testified that GM’s testing methods were deficient because they were static, so that GM could not have determined properly the crashworthiness of the F-37 roof in rollover accidents. Marcosky testified that this failure to conduct tests under dynamic conditions amounted to a breach of due care, because such dynamic tests would have alerted GM’s engineers to the encroachment problem. There also was evidence, based on proving-ground test films, that GM designers were aware that the roofs of other 2-door hardtop models, similar in design to the F-37, were crushing into the occupant zone in various roll-over and drop tests. On the basis of this evidence the jury was justified in finding that GM failed to exercise due care in the design of the F-37 roof.

GM’s sufficiency argument with respect to the jury’s finding of negligent design is tenuous at best in other respects. First, GM indirectly claims in its reply brief that *1518 the negligence issue should have been analyzed under the Knitz alternate test. This is a non sequitur. The Knitz risk-benefit analysis was expressly intended only for claims of product defect that defy the consumer expectation standard. GM suggests no reason why these two separate theories — strict products liability and negligence — should be collapsed into one.

Second, even if we were to accept GM’s questionable logic in this respect, we would adhere to our conclusion set forth above that the jury’s finding of product defect was amply supported by the evidence. Since we view the two theories as separate and independent — as did the jury, under proper instructions from the trial court — we hold that plaintiff’s evidence of GM’s lack of due care was sufficient to support the jury verdict on the issue of negligent design.

Third, GM also urges that its compliance,' albeit anticipatory, with a subsequently promulgated federal safety standard must immunize it from negligence as a matter of law. As we have indicated above, this is not so. Just as compliance with minimal standards does not resolve the claim of product defect, neither does it conclusively rebut the claim of negligent design. As the Ohio Supreme Court stated in Knitz, compliance is merely a factor to be considered by the jury. And that court’s prior decision in Temple, as we have explained above, is readily distinguishable from the instant case.

(D)

GM finally argues, on this aspect of the case, that the jury’s verdict on the issue of substantial modification of plaintiff’s car was not supported by the evidence.

In support of its claim that the car was in a substantially changed condition as a matter of law, GM argues to us — as it did to the jury — that the car was nine years old when purchased; that there was rust on the roof; and that the windshields and the rear window glass had been replaced, assertedly in an improper manner.

Plaintiff, on the other hand, points to contrasting evidence that was before the jury: there was testimony that the roof was in “normal” condition prior to the accident; that the rust was merely surface rust primarily attributable to the years the car was left outside following the accident; that no irregularity was found in the roof other than that caused by the accident; and that the glass had been replaced in a proper manner, even though such glass did not contribute to roof strength in any way.

We hold that this conflicting evidence was properly left to the jury to resolve. Its verdict that there was no substantial modification was supported by the evidence.

In short, for the reasons stated above, we find no merit in any of GM’s arguments challenging the jury verdict.

IV.

This brings us to what we regard as a more substantial issue, namely, GM’s claim with respect to the trial court’s exclusion of evidence relating to plaintiff’s failure to use his seat belt.

GM claims that the exclusion of this evidence hampered its arguments on the issues of contributory negligence and mitigation of damages. It argues that plaintiff would have been protected, in whole or in part, had he worn the seat belt which was provided; that the jury was erroneously prevented from considering the ramifications of his failure to utilize the seat belt; and that the car was safe in roll-over accidents for occupants wearing seat belts which are an integral part of the car’s protective design.

This claim breaks down into two discrete issues: whether seat belt evidence is generally admissible under Ohio law where the failure to use a seat belt can be shown to have a logical bearing on the injuries sustained by a plaintiff; and, if so, whether the exclusion of the seat belt evidence in the instant case constituted prejudicial error.

*1519 (A) Admissibility of Seat Belt Evidence

Both sides recognize that there is a substantial question as to whether Ohio law permits the introduction of seat belt evidence, for any purpose, in actions to recover damages sustained by injured automobile drivers and passengers — even where the use of seat belts might have prevented or at least minimized the injuries sustained.

The district court below held that evidence of failure to use the seat belt was inadmissible under Ohio law. The court therefore granted plaintiff’s motions to strike GM’s affirmative defenses based on non-use and to prohibit any reference at trial to plaintiff’s non-use. In so ruling, the court relied on the only two Ohio cases which have considered this question. Roberts v. Bohn, 26 Ohio App.2d 50, 269 N.E.2d 53 (1971), rev’d on other grounds sub nom. Suchy v. Moore, 29 Ohio St.2d 99, 279 N.E.2d 878 (1972); Bertsch v. Spears, 20 Ohio App.2d 137, 252 N.E.2d 194 (1969).

In Roberts, an Ohio court of appeals stated:

“[W]e prefer the authority of the majority of the cases in other states which is that the determination as to whether an occupant of an automobile should or should not be required to wear a seat belt should be left to the Legislature. In the absence of a statute to the contrary, there is no common law duty imposed upon an occupant of an automobile to wear a seat belt in ordinary vehicular travel, and the failure to use an available seat belt is not contributory negligence as a matter of law. Therefore, evidence of the failure of an occupant of an automobile to use an available seat belt is generally not admissible in an action for personal injuries arising out of an automobile accident.”

26 Ohio App.2d at 56, 269 N.E.2d at 53 (citations omitted). GM argues that the subsequent reversal of Roberts, on other grounds, by the Ohio Supreme Court vitiates the court of appeals’ holding on the seat belt defense. This argument misses the point. Since the Ohio Supreme Court did not address the question of the seat belt evidence, the undisturbed holding by the court of appeals on this issue remains the most recent statement of Ohio law on the issue. Since we are called upon to construe Ohio law, the court of appeals’ decision on the seat belt issue in Roberts remains a beacon that illuminates what is an otherwise uncharted course. West v. American Telephone & Telegraph Co., 311 U.S. 223, 237 (1940) (federal court must follow intermediate state appellate court decisions unless it is convinced by “persuasive data” that highest state court would decide otherwise); Poyner v. Erma Werke GMBH, 618 F.2d 1106, 1190 (6th Cir.), cert. denied, 449 U.S. 841 (1980); Simpson v. Jefferson Standard Life Ins. Co., 465 F.2d 1320, 1323 (6th Cir.1972).

In Bertsch, the other pertinent Ohio case, the court of appeals held that failure to wear a seat belt did not constitute negligence per se. But that court left open the question of whether seat belt evidence might be admissible in other cases where non-use could be shown to have caused the accident or to have aggravated the plaintiff’s injuries. 20 Ohio App.2d at 139, 252 N.E.2d at 196.

In this diversity action we are required to look first to the state’s highest court and then to its intermediate appellate courts on matters of substantive law. Here we have no controlling or even helpful precedent of the Ohio Supreme Court to aid our task. The two Ohio precedents that do exist on the admissibility of seat belt evidence are neither current nor uniform. While Roberts stated unequivocally that evidence of seat belt non-use is inadmissible, Bertsch did not so state, at least for those cases in which the injuries might be attributable to a failure to use seat belts. It also should be noted that both of these decisions involved defendants who negligently caused the accidents which resulted in injuries to the plaintiffs. In the case at bar, however, GM did not cause the accident; it is accused of failing to protect plaintiff adequately from injury caused by a foreseeable accident. In such a case, it may be appropriate to inquire into what steps, if any, plaintiff *1520 took to protect himself from an accident, such as using a seat belt. This difference may be significant, since the Ohio court in Roberts exhibited antipathy towards a defendant who would negligently cause an accident and then claim that plaintiff failed to protect himself against defendant’s negligence. Roberts, supra, 26 Ohio App.2d at 58, 269 N.E.2d at 59.

Roberts was decided in 1971. Since then the law on the admissibility of seat belt evidence has been in a state of flux. The policy implications of the various rules have been roundly debated, with a resulting patchwork-quilt of state law on the issue. In some states such evidence is wholly inadmissible, on the theory that automakers must design a vehicle that is safe for those who, foreseeably, will not wear seat belts. In others, the policy of encouraging seat belt use has led to approval of the seat belt defense under either a contributory negligence or mitigation of damages rationale. 5 Under the circumstances it is possible that the Ohio Supreme Court, if the question were properly presented, might adopt a rule different from that of Roberts. Bernhardt v. Polygraphic Company of America, 350 U.S. 198, 211 (1956) (Frankfurter, J., concurring). Thus we find ourselves in the somewhat unenviable position of attempting to anticipate a ruling of the Ohio Supreme Court in an important area of state law.

Our attempt to predict the Ohio Supreme Court’s position on the seat belt defense is complicated further by the peculiarly legislative nature of the issue. The penalties to be attached to seat belt non-use are uniquely amenable to resolution by the state legislature. Indeed a number of other state courts have left to their legislatures the determination of the evidentiary effect of a plaintiff’s failure to wear a seat belt. 6 In light of the nature of the undertaking and the contrasting developments in other jurisdictions, any decision we might reach concerning this delicate balance of policy interests might not pass muster by the Ohio Supreme Court. 7 We hesitate to announce *1521 a decision, unless absolutely required to do so, in an important area of state law where there is no sure guidance from the state courts. 8

Fortunately, we need not resort to tea leaves or to judicial tarot cards — or to a bold substitution of our own views for those of the Ohio Supreme Court — to resolve the issue of the admissibility of the seat belt evidence in the instant case. We hold that, regardless of whether the evidence should have been admitted, its exclusion, if error at all, was harmless. We turn next to the matter of harmless error.

(B) Harmless Error

The Ohio Rule on harmless error tracks Fed.R.Civ.P. 61:

“No error in either the admission or the exclusion of evidence and no error or defect in any ruling or order or in anything done or omitted by the court or by any of the parties is ground for granting a new trial, or for setting aside a verdict or for vacating, modifying or otherwise disturbing a judgment or order, unless refusal to take such action appears to the court inconsistent with substantial justice. The court at every stage of the proceedings must disregard any error or defect in the proceeding which does not affect the substantial rights of the parties.”

Ohio R.Civ.P. 61. Furthermore, the burden is on the appellant to demonstrate prejudice to its substantial rights. Westwood Chemical, Inc. v. Owens Corning Fiberglass, 445 F.2d 911, 918 (6th Cir.1971), cert. denied, 405 U.S. 917 (1972) (decided under Fed.R. Civ.P. 61, which is the same as the Ohio rule).

In the instant case the error, if any, in excluding the seat belt evidence must be considered harmless because, first, the jury was made aware, through exposition to the jury of GM’s theory of the injury, that plaintiff was not wearing his belt, and second, the jury rejected the theory of the injury in which failure to wear the seat belt was a proximate or contributing cause.

GM’s position, which was fully presented to the jury by GM’s questioning and argument, was that plaintiff’s injury resulted from his being thrown from his seat against the right roof rail during the roll, or, alternatively, from thereafter being thrown back against the roof on the left side of the car. This so-called “dive”, which seat belts might have prevented, and not the collapse of the roof onto plaintiff’s head, which seat belts in no way would have protected against, caused the injury, according to GM. In GM’s version, the roof, even if defectively weak, could not have been the cause of the injury and thus the jury could not have found for plaintiff. According to GM’s reasoning, the roof deformation was irrelevant; plaintiff was injured solely as the result of the unrestrained force of his dive, which presumably would have produced the same tragic result even if the roof had retained its shape.

GM argued this theory of the case to the jury at great length. The court properly instructed the jury that plaintiff could recover only if he proved that the injury was *1522 caused by the defective or negligently designed roof. 9 Such causation could not have been found if plaintiff simply dove, unrestrained, against the roof rail during the roll-over. Under the court’s instructions therefore the jury had to — and in fact did — find that the roof collapse and not the dive against the roof rail caused plaintiffs broken neck.

Plaintiff .argued to the jury, based on the testimony of experts and the physical evidence, that his neck was broken by the downward impact of the roof as it collapsed. This evidence demonstrated that he was held in his seat by the centrifugal force generated in the roll and a seat belt would have made no difference. Of the four occupants of the car (none of whom was wearing a seat belt), only plaintiff, who was beneath the collapsed portion of the roof, was seriously injured. This undoubtedly had a substantial impact on the jury and contributed to its acceptance of plaintiff’s version of the accident. This was the version of the accident the jury had to believe in order to find for plaintiff. For unless the roof collapse — and not plaintiff’s own unrestrained second collision with the inside of the car — was the proximate cause of his injury, the allegedly defective and negligently designed roof would not have been the basis for the verdict returned, as in fact it was under the court’s proper instructions.

As the case was tried therefore the seat belt evidence was irrelevant. GM was not prevented from arguing to the jury its version of the accident — a version that made it patently obvious that plaintiff was not wearing a seat belt. See United States v. Ross Corp., 385 F.2d 564, 566 (6th Cir.1967) (harmless error in refusing to admit exhibits the contents of which were presented to the jury through testimony). But the jury rejected GM’s explanation, finding instead that plaintiff was injured as the result of the collapse of the roof onto his head. Although one may speculate as to which evidence the jury found most persuasive, the escape of plaintiff’s three companions without serious injury provided a powerful reason to discredit GM’s explanation and to determine that the roof collapse — under which only plaintiff was positioned at the time of the crash — was the cause of plaintiff’s injury. 10

For all of GM’s broad-gauged arguments about the policies behind punishment of drivers who fail to use seat belts, it has totally failed to demonstrate that the seat belt evidence it sought would have helped its case below. GM was permitted to argue its theory that plaintiff dove into the roof rail. It was allowed to introduce evidence that safety belts are an integral part of a crashworthy auto design. The jury was thus made aware of GM’s seat belt argument, even without detailed evidence about plaintiff’s non-use of his belt. The jury simply rejected GM’s explanation. GM has been unable to persuade us, as it could not persuade the jury, how the seat belt would have benefitted plaintiff when, in the version of the injury accepted by the jury, the roof crashed in on his head. Under such circumstances, there would have been small comfort in being held upright in his seat. Since the jury ascertained that the injury occurred as a result of the roof collapsing on plaintiff’s head, the proffered seat belt evidence — duplicative at best — would have been of no value to GM. Its exclusion, if error at all, was harmless error. The exclusion of such evidence was in no way inconsistent with substantial justice. This case is similar to others in which we have found the admission or exclusion of evidence to be harmless, even if error. TCP Industries, Inc. v. Uniroyal, Inc., 661 F.2d 542, 550 (6th Cir.1981); Baker v. Elcona Homes Corp., 588 F.2d 551, 559 (6th Cir.1978), cert. de *1523 nied, 441 U.S. 933 (1979); Kennedy v. Delaware Leasing & Rental Corp., 441 F.2d 562, 564 (6th Cir.1971); J.W. Bateson Co. v. Romano, 266 F.2d 360, 362 (6th Cir.1959); Hong v. City of Detroit, 185 F.2d 764, 766 (6th Cir.1950).

To summarize: we hold that there was substantial evidence to support the jury verdict on the issues of product defect, negligent design and absence of substantial modification; and any error in the exclusion of the seat belt evidence, if error at all, was harmless error. We therefore affirm the judgment of the district court.

Affirmed.

. Although there are two plaintiffs-appellees— the father, Olen Sours, and his son, Dennis C. Sours — throughout this opinion, unless otherwise stated, when we refer to “plaintiff” we mean Dennis. Olen’s claims are limited to the hospital and medical expenses incurred on behalf of his son.

. “The article sold must be dangerous to an extent beyond that which would be contemplated by the ordinary consumer who purchases it, with the ordinary knowledge common to the community as to its characteristics.” Restatement (Second) of Torts § 402A comment i (1965).

. Knitz set forth three factors to be considered for purposes of the risk-benefit analysis: (1) the likelihood that the product will cause injury; (2) the gravity of the danger posed; (3) the mechanical and economic feasibility of an improved design.

. As plaintiffs point out, evidence was adduced that might well have permitted findings on each of the three Knitz factors. As to factor one, there was credible evidence of the frequency of roll-over accidents; as to factor two, there was evidence that roof-collapse presents a grave danger to auto occupants in roll-over accidents; and, as to factor three, plaintiff introduced evidence of a roll-cage design that would prevent roof-collapse. For the reasons stated above, however, we need not decide whether this evidence would have been sufficient to support a verdict on the alternative standard.

. A number of state courts, or courts applying the law of a particular state, have held that seat belt evidence is inadmissible for purposes of proving contributory negligence, at least where the failure to wear seat belts did not cause the accident but only contributed to the severity of the injuries suffered. This rule has been followed in Arizona, Colorado, Delaware, the District of Columbia, Illinois, Kansas, Louisiana, Michigan, Missouri, New York, North Carolina, Oklahoma, Oregon, Tennessee, Texas, and Washington. On the other hand, courts in the following states — or courts applying the law of these states — have suggested that a failure to wear an available seat belt may constitute contributory negligence where that failure plays a substantial role in plaintiffs injuries: California, Connecticut, Indiana, Florida, Pennsylvania, New Jersey, Maryland, and Wisconsin. See Annotation, “Automobile Occupant’s Failure to Use Seat Belt as Contributory Negligence,” 92 A.L.R.3d 9 (1979).

On the question of mitigation of damages, there appears to be the same sort of split among the states. The courts of — or courts applying the law of — Alabama, Arizona, Colorado, Delaware, the District of Columbia, Kansas, Michigan, Missouri, New Jersey, New Mexico, North Carolina, Oregon, Tennessee, and Washington flatly refuse to consider seat belt non-use as a failure to mitigate damages. But where causal connection can be shown between non-use and injury, courts of — or courts applying the laws of — California, Connecticut, Illinois, Indiana, New York, Oklahoma, Pennsylvania, South Carolina, Virginia, and Texas have allowed the question to be submitted to the jury. See Annotation, “Non-use of Seat Belt as Failure to Mitigate Damages,” 80 A.L.R.3d 1033 (1977).

On the admissibility of seat belt evidence in comparative negligence jurisdictions, see Annotation, “Nonuse of Automobile Seat Belts as Evidence of Comparative Negligence,” 95 A.L. R.3d 239 (1979). •

. E.g., Kopsichke v. First Continental Corp., 610 P.2d 668, 683 (Mont.1980) (“In light of the history and the numerous legislative problems that must be considered to effectively extend the seat belt rule of law we ... reach the conclusion that to adopt a seat belt defense when the legislature has failed to do so would be ill-advised.”). See also State v. Ingram, 427 N.E.2d 444, 448 (Ind.1981); Fields v. Volkswagen of America, Inc., 555 P.2d 48, 62 (Okla.1976); Fischer v. Moore, 183 Colo. 392, 396, 517 P.2d 458, 460 (1973); Britton v. Doehring, 286 Ala. 498, 508, 242 So.2d 666, 675 (1970); Miller v. Haynes, 454 S.W.2d 293, 301 (Mo.App.1970); Miller v. Miller, 273 N.C. 228, 238, 160 S.E.2d 65, 73 (1968).

. Indeed the United States Supreme Court has held that where a federal court cannot make a confident prediction as to how the state’s high *1521 est court would rule on an issue, certification to that court — where available — is perhaps the wiser course. Lehman Bros. v. Schien, 416 U.S. 386, 390-91 (1974) (approving certification but noting that “[i]ts use in a given case rests in the sound discretion of the trial court.”); Clay v. Sun Insurance Office Ltd., 363 U.S. 207 (1960). The Rules of Practice of the Ohio Supreme Court, which provide for certification from lower state courts in certain instances, do not provide for certification from the federal courts. See also 1A Moore’s Federal Practice ¶ 0.309[3] n. 7, at 3132 (2d ed. 1982).

. Under our prior decisions, where there is substantial uncertainty as to how the state courts would rule on an issue, our inquiry has been limited to determining whether the district judge reached a “permissible conclusion in his determination of the local law.” Lee Shops, Inc. v. Schatten-Cypress Co., 350 F.2d 12, 17 (6th Cir.1965), ce rt. denied, 382 U.S. 980 (1966). See also Clairol, Inc. v. Boston Discount Center of Berkley, 608 F.2d 1114, 1120 n. 8 (6th Cir.1979); Insurance Co. of North America v. Federated Mutual Ins. Co., 518 F.2d 101, 106 & n. 3 (6th Cir.1975); Rudd-Melikan, Inc. v. Merritt, 282 F.2d 924, 929 (6th Cir.1960) (“Court of Appeals should accept the considered view of the district judge”). But see Transamerica Ins. Group v. Beem, 652 F.2d 663, 665 & n. 3 (6th Cir.1981).

. The trial court gave this instruction with regard to both the negligence and the strict liability causes of action.

. Moreover, there was direct evidence that plaintiff’s neck was crushed by the downward force of the roof and not by a dive into the roof rail. Malion, who was sitting in the front passenger seat, testified that after the crash the roof “was on top of his [plaintiffs] head, and it was like, bent and deformed and stuff.” Mal-ion also testified that he remained in his seat throughout the roll, although he too was not wearing a seat belt.