POSNER, Circuit Judge.

Tina Barron, age 18, was rendered paraplegic and forced onto the welfare rolls as the result of an accident in 1984 in which the car she was riding in (driven by her sister) skidded on a rain-slick highway in North Carolina and turned over. Because Barron was not wearing a seatbelt (although the car was equipped with seatbelts), she was flung out of the car, either through the closed sunroof, as she claims, or through the window on the passenger's side of the front seat, which was also closed. Although a citizen of Illinois, Barron brought suit against Ford Motor Company of Canada, the manufacturer of the car, in a Florida state court. The parties being of diverse citizenship and Ford not a citizen of Florida, Ford was able to remove the case to federal district court in Florida, from which it sought to transfer the case, on grounds of convenience, 28 U.S.C. § 1404(a), to the Eastern District of North Carolina. The plaintiff countered with a request to transfer the case to the Central District of Illinois, and her request was granted and the case transferred. Applying Florida's rules on conflict of laws (and thus anticipating Ferens v. John Deere Co., 494 U.S. 516, 110 S.Ct. 1274, 108 L.Ed.2d 443 (1990), which holds that the transferee court must apply the conflict of laws rules of the transferor jurisdiction whether the defendant or the plaintiff requested the transfer), Judge Mihm ruled that the law of North Carolina — as it happens, the only state in the United States not to recognize strict liability in products cases, Smith v. Fiber Controls Corp., 300 N.C. 669, 678, 268 S.E.2d 504, 509-10 (1980); Warren v. Colombo, 93 N.C. App. 92, 102, 377 S.E.2d 249, 255 (1989) — governed the substantive issues in the case. 716 F. Supp. 377 (C.D.Ill. 1989). A two-week jury trial, in which the plaintiff tried to prove that she had been ejected through the sunroof and that Ford had been negligent in making the sunroof out of tempered rather than laminated glass, ended in a verdict for Ford.

Barron argues to begin with that the judge should have applied Illinois law, which imposes strict liability in products cases, rather than North Carolina law, which requires the plaintiff to prove negligence. It is not easy to see what difference this would have made to the outcome of the case; and before entangling itself in messy issues of conflict of laws a court ought to satisfy itself that there actually is a difference between the relevant laws of the different states. International Administrators, Inc. v. Life Ins. Co., 753 F.2d 1373, 1376 n. 4 (7th Cir. 1985). There might seem to be all the difference in the world between negligence and strict liability, but in a products liability case this often is not true. The plaintiff must prove either that the product was defective or that it was unreasonably dangerous, and in determining whether a product is "defective" or "unreasonably" dangerous the court weighs costs and benefits just as it would do in a negligence case. See Flaminio v. Honda Motor Co., 733 F.2d 463, 467 (7th Cir. 1984), and references cited there. The difference between the two grounds of liability bites when the defendant is being sued for a product defect created by a component that he had bought, and he could not be sued for negligence because he used all due care in buying, inspecting, and installing it, so that he was without blame yet is liable nonetheless if liability is strict. Id. That is not an issue in this case. Ford did not buy the sunroof; it designed, manufactured, and installed it.

So this appears to be a case of a false conflict; but if he had to choose, Judge Mihm was right to apply North Carolina law rather than (as the plaintiff urges) Illinois law. It is true that Barron was a citizen of Illinois before she made a visit of several months' duration to her sister in North Carolina, in the course of which the accident occurred; and after a brief period of hospitalization in North Carolina she returned to Illinois, where she remains today, a public charge. And it is true that Ford of Canada (or for that matter its American parent) is not a citizen of North Carolina. But in Florida as in most states (for remember that it is Florida conflict of laws principles that govern this case), despite the inroads that "interest analysis" and "most significant relationship" inquiry have made on the simplicities of the old common law conflicts principles, the presumption remains that the law of the state in which the accident occurred governs tort claims arising from the accident, Bishop v. Florida Specialty Paint Co., 389 So.2d 999, 1001 (Fla. 1980); State Farm Mutual Automobile Ins. Co. v. Olsen, 406 So.2d 1109, 1111 (Fla. 1981), although the presumption is rebuttable. Wal-Mart Stores, Inc. v. Budget Rent-A-Car Systems, 567 So.2d 918, 920 (Fla.App. 1990). That state is concerned in an accident even if neither party is a resident — clearly so in a case such as this, where one of the parties to the accident, albeit not to the lawsuit, was a resident of the state (Tina Barron's sister). And its tort principles may be informed by familiarity with local conditions, which is another reason to apply those principles rather than the tort law of the plaintiff's or the defendant's domicile. William F. Baxter, "Choice of Law and the Federal System," 16 Stan. L.Rev. 1 (1963).

A further reason to doubt that Barron was placed at a disadvantage by the application of North Carolina law to her case is that it is North Carolina law that arms her with the ground of appeal she presses most strongly — that the judge should not have let Ford introduce any evidence about seatbelt use because North Carolina has a strong common law rule, now codified by statute, that evidence that a plaintiff didn't fasten his seatbelt is inadmissible in any civil action. Hagwood v. Odom, 88 N.C. App. 513, 516-17, 364 S.E.2d 190, 191-92 (1988). This may be, as we shall see, an overbroad statement of the common law rule; and yet the statute (which is not applicable to this case, however, because the case arose before its effective date) goes even further than the rule as we have stated it. For it provides that "evidence of failure to wear a seat belt shall not be admissible in any criminal or civil trial, action, or proceeding except in an action based on a violation of this section." N.C.Gen.Stat. § 20-135.2A(d) (emphasis added). The section in question is a mandatory seatbelt law, and evidence of nonuse can of course be introduced in a proceeding to impose a penalty for violation of the law. But if the statute is read literally, that is the only type of proceeding in which such evidence can be introduced. So if an irate passenger ripped off his seat belt, tore it from its moorings, and used it to strangle the driver, in the ensuing murder trial the prosecution would be forbidden to identify the murder weapon because to do so would be to show that the defendant had not been wearing his seatbelt.

The literal reading of the statute — more to the point, of the common law rule that preceded it, at least as the plaintiff understands that rule — would thus rule out Ford's presenting evidence, as it did, that Tina Barron was not wearing her seatbelt when the accident occurred. It is true that this was not the focus, at least not the nominal focus, of the testimony about seat belts. Ford's point was merely that the provision of seatbelts was a part of the automobile's overall restraint system, so that the reasonableness of making the sunroof of tempered rather than of laminated glass was a function in part of the other steps Ford had taken to prevent occupants from being flung out. But it could not make the argument without indicating that Barron had not been wearing her seatbelt, since if she had been and it had not kept her from flying through the sunroof Ford's argument about its total restraint system would fall flat on its face.

The literal interpretation of North Carolina's rule, though, is almost certainly incorrect. In State v. Brewer, 328 N.C. 515, 522, 402 S.E.2d 380, 385 (1991), a prosecution of a woman for murdering her disabled daughter by abandoning her car with the daughter in it on a railroad crossing, the Supreme Court of North Carolina remarked, without criticism, the introduction of evidence that the daughter knew how to release her seatbelt; it never occurred to anyone that such evidence might be inadmissible.

If it were inadmissible under North Carolina law, moreover, this might not help the plaintiff in this case. Even in diversity cases the rules of evidence applied in federal courts are the federal rules of evidence rather than state rules, Lovejoy Electronics, Inc. v. O'Berto, 873 F.2d 1001, 1005 (7th Cir. 1989); In re Air Crash Disaster Near Chicago, 701 F.2d 1189, 1193 (7th Cir. 1983); Romine v. Parman, 831 F.2d 944 (10th Cir. 1987), save with respect to matters of presumptions, privilege, and competency of witnesses, Fed.R.Evid. 302, 501, 601, none of which is involved here. If North Carolina's rule against the admission of testimony about a failure to wear one's seatbelt is a rule of evidence, it is inapplicable to this case; and there is no counterpart rule in the federal law of evidence.

Well, but is it a rule of evidence for purposes of the Erie doctrine, or is it a substantive rule and therefore binding in a diversity case (or any other case in which state law supplies the rule of decision)? The difference is this. A pure rule of evidence, like a pure rule of procedure, is concerned solely with accuracy and economy in litigation and should therefore be tailored to the capacities and circumstances of the particular judicial system, here the federal one; while a substantive rule is concerned with the channeling of behavior outside the courtroom, and where as in this case the behavior in question is regulated by state law rather than by federal law, state law should govern even if the case happens to be in federal court. Massachusetts Mutual Life Ins. Co. v. Brei, 311 F.2d 463, 465-66 (2d Cir. 1962); Olin Guy Wellborn III, "The Federal Rules of Evidence and the Application of State Law in the Federal Courts," 55 Texas L.Rev. 371, 404, 406 (1977). The North Carolina rule could be either. It is a rule of evidence if it is motivated by concern that jurors attach too much weight to a plaintiff's failure to wear his seatbelt. It is a substantive rule if it is designed not to penalize persons who fail to fasten their seatbelts. Many rules mix procedural or evidentiary with substantive policy concerns, examples being the parol evidence rule, the "mend the hold" doctrine that we discussed in Harbor Ins. Co. v. Continental Bank Corp., N.A., 922 F.2d 357, 364 (7th Cir. 1990), the rule against admission of evidence of subsequent repairs in a tort case, Flaminio v. Honda Motor Co., supra, 733 F.2d at 471, and the doctrine of election of remedies, Olympia Hotels Corp. v. Johnson Wax Development Corp., 908 F.2d 1363, 1371 (7th Cir. 1990). The more broadly the North Carolina rule is interpreted, the stronger the inference that its predominant character is that of a rule of evidence.

In choosing between a broad and a narrow interpretation we must consider why a defendant would want to introduce evidence of a plaintiff's failure to wear his seatbelt. The purpose would not be to show that the plaintiff was contributorily negligent. Wearing a seatbelt does not make it less likely that you will have an accident — it could make it more likely, at least if you're the driver, by making you feel safer. The purpose of the evidence would be to show that the plaintiff failed to take reasonable measures to reduce the severity of the accident if one occurred. The duty to take such measures (which is closely related to the contract duty of mitigation of damages) goes by the name "avoidable consequences." Brooks v. Allison Division, 874 F.2d 489, 490 (7th Cir. 1989); Tacket v. General Motors Corp., 836 F.2d 1042, 1047 (7th Cir. 1987). It is to that doctrine that evidence of not fastening one's seatbelt is primarily relevant. Miller v. Miller, 273 N.C. 228, 239-40, 160 S.E.2d 65, 74 (1968); Evra Corp. v. Swiss Bank Corp., 673 F.2d 951, 958 (7th Cir. 1982); Mays v. Dealers Transit, Inc., 441 F.2d 1344, 1352 (7th Cir. 1971). North Carolina courts accept the doctrine of avoidable consequences, but hold that it does not apply to the use of seatbelts. Miller v. Miller, supra, 273 N.C. at 239-40, 160 S.E.2d at 74. The nonuse of seatbelts is so widespread that the North Carolina courts, bearing in mind that the law rarely requires a person to exercise more than average care, refuse to pronounce that nonuse unreasonable. Id. 273 N.C. at 232-34, 160 S.E.2d at 69-70. Therefore, as a matter of state substantive law, evidence that the plaintiff had failed to fasten his seatbelt would be irrelevant to show that his damages should be cut down as a penalty for unreasonable behavior, and irrelevant evidence is not admissible under the federal rules of evidence. See Fed.R.Evid. 402. Miller dates from the 1960s, and seatbelt usage is more common today, in North Carolina as elsewhere, we presume; but the North Carolina courts have not seen fit to reexamine their rule.

Hagwood v. Odom, supra, 88 N.C. App. at 516-17, 364 S.E.2d at 191-92, the clearest articulation of North Carolina's common law rule against seatbelt evidence, holds not that such evidence is inadmissible on all issues but only that it is inadmissible to establish the plaintiff's failure to exercise due care to minimize the consequences of an accident should one occur. Hagwood shows that Judge Adams was right, and his brethren wrong, in Seese v. Volkswagenwerk A.G., 648 F.2d 833, 841-42 (3d Cir. 1981); id. at 852-54 (dissenting opinion). That case held that in North Carolina evidence of the nonuse of a seatbelt is inadmissible not only to show that the plaintiff had been careless in failing to fasten his seatbelt but also to show that the provision of the seatbelt option was a reasonable component of an overall plan for protecting occupants from the consequences of a crash. The court in Seese did not have the benefit of Hagwood, which was decided years later, and which shows that the majority in Seese guessed wrong about the North Carolina rule. We add that the statute itself, in speaking of "evidence of failure to wear a seat belt," could well be thought not to be referring to evidence that the vehicle was equipped with seatbelts.

If our understanding of the scope of the rule is correct, then the rule is substantive, all right; it is founded on the desire of the North Carolina courts not to penalize the failure to fasten one's seatbelt, because nonuse is so rampant in the state that the average person could not be thought careless for failing to fasten his seatbelt. Miller v. Miller, supra, 273 N.C. at 232-34, 160 S.E.2d at 69-70. But, so understood, the rule, while applicable in diversity cases, does not cover Ford's use of seatbelt evidence in this case. The purpose was not to show that Tina Barron failed to mitigate the consequences of the accident but to show that Ford had been reasonable in deciding not to make the sunroof out of laminated glass — that it had provided as it were a backup system in the form of seatbelts. One could put the point more strongly. It would be careless for an automobile manufacturer to rely upon an openable window, such as a sunroof, as part of the occupant-restraint system; the rolled-down glass of an open window is no restraint. It might therefore be sensible to make the windshield out of laminated glass (if laminated glass is in fact, as the plaintiff argues, more likely to contain the occupants in an accident than tempered glass), because the windshield cannot be opened anyway, but not to make the sunroof out of laminated glass. Even the choice of laminated glass for the windshield could be questioned, since an occupant could be severely injured by being hurled against a rigid pane of glass — which is another way of saying that a seatbelt is a superior method of restraint. The very idea that a sunroof — for which people pay extra precisely so that they can open it in good weather — is part of a car's restraint system has elements of fantasy, and the plaintiff's counsel readily conceded at argument that people don't think in these terms. At all events, the presence of a seatbelt system is at least relevant to the manufacturer's effort to show that he used due care (or, in the language of strict liability, did not design the car defectively) with regard to the danger of an occupant's being ejected in the event of an accident; and this is regardless of the doctrine of avoidable consequences.

It is true that many people don't use seatbelts, no matter how easy the manufacturers make their use. That is, indeed, the basic rationale of the North Carolina rule. The seemingly irrational distaste of automobile occupants for "buckling up" is one of the major forces behind the call for "passive restraints," such as the airbag. We may assume that the duty of care is shaped by customers' known propensities, whatever one might think of them from an ethical or prudential standpoint; that in some settings even "idiot proof" could be a doctrine of tort law, cf. Estrada v. Schmutz Mfg. Co., 734 F.2d 1218, 1220 (7th Cir. 1984), by extrapolation from the rule that requires a potential injurer to take special precautions when children are known to be in the vicinity. Prosser and Keeton on the Law of Torts, § 33, at pp. 199-200 (W. Page Keeton et al. eds. 5th ed. 1984). But this is just to say that Ford's evidence that it had provided a seatbelt system as its main defense against the type of injury that occurred in this case was not conclusive on the issue of due care. The question is whether it was barred by the North Carolina rule. It was not. Either the rule is a substantive rule, and inapplicable because limited to the (partial) defense of avoidable consequences, or an evidentiary rule and therefore inapplicable to proceedings in federal courts. Either way, the seatbelt evidence was admissible.

The plaintiff complains separately that Ford harped on the seatbelt issue unmercifully, for example by asking the plaintiff whether she had taken a driver's education course in which the use of seatbelts had been explained (her answer was yes). But the plaintiff's lawyer set her up for this damaging line of inquiry. An auxiliary theory of her case was that Ford should have warned occupants of the car that the sunroof wouldn't hold them in and that they should therefore fasten their seatbelts. Stegall v. Catawba Oil Co., 260 N.C. 459, 464, 133 S.E.2d 138, 142 (1963); Ziglar v. E.I. du Pont de Nemours Co., 53 N.C. App. 147, 151, 280 S.E.2d 510, 513 (1981). It was relevant to this issue for Ford to show that Barron didn't need any warnings about the dangers of not fastening her seatbelt because she had learned all about that in the driver's ed course.

Another issue is whether the judge should have allowed the plaintiff's lawyer to rehabilitate one of the plaintiff's witnesses, a former employee of Ford who testified that for safety's sake the sunroof should have been made of laminated glass. On cross-examination Ford brought out the fact that this employee had been fired, which — the plaintiff argues — allowed the impression to linger that his testifying against Ford was retaliatory in motive. The plaintiff wanted to show that in fact the employee had been fired for blowing the whistle on Ford's unsafe practices. The judge refused to allow the attempted rehabilitation. He feared that it would turn the trial into a trial of whether this man's discharge had been wrongful.

This sort of judgment, balancing the probative value of testimony against the costs in protracting the trial and muddying the issues, involves imponderable best weighed by the judge on the spot, not the appellate panel. United States v. Shapiro, 879 F.2d 468, 472 (9th Cir. 1989); United States v. Martinez, 775 F.2d 31, 37 (2d Cir. 1985). So we ask, not whether we think we would have made the same ruling, but whether the judge acted reasonably. He did. Even without the attempted rehabilitation, the jury was (as it seems to us) as likely to infer that the witness had been fired for disagreeing with Ford's judgment about what type of glass to use in the sunroof as that he had turned against Ford because it had fired him for being a lousy employee. And the judge allowed the plaintiff to place in evidence any memos the man had written while at Ford on the subject of laminated versus tempered glass, and she put in some letters he had written concerning the retention powers of the windshield in the Pinto.

The appeal raises other issues but only one has sufficient merit to warrant discussion. It relates to the question whether Barron was ejected through the sunroof or through the passenger window. Why this should have been thought to make a potential difference to the outcome of the trial is a little unclear, since presumably the passenger window was made of the despised tempered glass as well. Apparently, though, Ford's theory was that the passenger window had been shattered as the car was turning over and Barron had been thrown through the empty space, while her theory was that she had struck the sunroof while it was still intact, and her head had shattered it, which might not have happened had it been made of laminated glass. She presented testimony that the rescue workers had smashed the passenger window — after the accident occurred — testimony that if believed scotched Ford's theory, and Ford presented evidence contra. One of the witnesses the plaintiff's lawyer wanted to call was a state trooper who is an expert in determining the causes of highway accidents. He would have testified that because the sunroof was shattered, Tina Barron was found on the side of the road, and the rescue workers had broken the front passenger window, she must have exited the car through the sunroof. The judge refused to allow this testimony on the ground that there was nothing "expert" about it; all the trooper was doing was making an obvious logical deduction from premises furnished by other witnesses.

It is commonplace for an expert witness, who is not after all an eyewitness, to testify on the basis of premises supplied by such witnesses. He adds expertise to their evidence and comes up with an opinion bearing on some issue in the case. An expert is a witness allowed to offer an opinion. Fed.R.Evid. 702. This trooper used no expertise, the judge ruled, and therefore could not be allowed to testify on the route of the plaintiff's ejection from the car.

That is cutting things too fine. It is true in a logical but not a realistic sense that if the rescue workers were believed, the conclusion that Barron had gone through the sunroof followed ineluctably. But conceivably a jury might have thought it highly improbable that someone could sail through the sunroof, and this would lead them to doubt the rescue workers' testimony. The trooper's testimony, simplistic as it might seem to the sophisticated, would have alleviated that concern. A veteran of literally thousands of accident investigations, his testimony would have bolstered the plausibility of the plaintiff's theory of how this accident occurred. He should have been allowed to testify, and Ford's lawyer could then have tried to undermine his testimony on cross-examination by showing its crucial dependence on the rescue workers' evidence, which Ford had controverted.

So there was error, but, as with many evidentiary rulings, it was harmless. Even it, given the trooper's testimony, the jury would have been sure that Barron had gone through the sunroof, it is highly unlikely that they would have brought in a verdict for her. The evidence of Ford's negligence was very weak, even if one believed that laminated glass was safer in the relevant respect than tempered glass. Common sense says that sunroofs meant to be opened need not be built to restrain an occupant who declines to fasten her seatbelt.

AFFIRMED.