The Department of Labor appeals from the dismissal of a suit that it brought against an insurance company for monetary relief under ERISA. The trustees of a union pension fund had obtained from Continental Casualty Company a one-year extension of the fiduciary liability insurance policy that Continental had previously issued to them. According to the Department of Labor, the extension provided only $1 million in additional coverage yet Continental charged a premium of $970,000 for the extension and the fund paid the premium. The Department sued both the trustees and Continental — the trustees for having breached their fiduciary duty to the fund by obtaining insurance for their own protection at a price disproportionate to any possible benefit to the fund, and Continental for having knowingly participated in the trustees’ breach. Relief sought included an order that “the defendants, jointly and severally, ... restore to the [pension fund] all losses sustained as a result of the breaches of fiduciary duties or participation therein” and that Continental “disgorge all payments it received ... as a result of its participation in the breaches of fiduciary duties.”

The claim against the trustees was settled, but for less than the amount sought by the Department, so the case proceeded to trial against Continental. Midway in the trial the Supreme Court decided Mertens v. Hewitt Associates, — U.S.-, 113 S.Ct. 2063, 124 L.Ed.2d 161 (1993), holding that the remedial statute under which the plaintiff had proceeded in that ease — and the Department of Labor in this one — did not, in authorizing injunctions or “other appropriate relief,” authorize “money damages ..., the classic form of legal relief.” Id. — U.S. at-, 113 S.Ct. at 2068 (emphasis in original), interpreting 29 U.S.C. § 1132(a)(3). The Court in passing described restitution, in contrast to damages, as a form of relief traditionally available in equity, — U.S. at -, 113 S.Ct. at 2069, so the Department in our case immediately reduced its demand for relief against Continental to the net amount that the insurance company had received for the challenged extension, some $818,000 after deduction from the premium of broker and other fees. For restitution as normally understood in civil cases seeks to deprive the defendant of money or any other thing of value that he gained from tortious or otherwise wrongful activity or that it would be unconscionable for him to retain because received from the plaintiff in circumstances under which he knew or should have known that the plaintiff expected compensation. In either case the defendant would be unjustly enriched if allowed to keep the gain; and $818,000 is all that Continental gained from the extension of coverage.

But en route to its interpretation of “other equitable relief,” the Supreme Court in Mer-tens had said that it was far from clear that a suit against a party that was (in Mertens as in this case) not a fiduciary but merely a knowing participant in a fiduciary’s breach of duty was within the scope of the statute, regardless of the nature of the relief sought. The Court pointed out that no provision of ERISA, which it described as a carefully drafted statute, makes a nonfiduciary liable for knowing participation in a fiduciary’s breach of duty, even though such liability was well established under the common law of trusts. But since the parties had assumed the applicability of the statute and quarreled only over the remedy, the Court decided to place decision on the narrow ground that the statute did not authorize the remedy of damages. Id. — U.S. at -, 113 S.Ct. at 2067-68. The doubts that the Court expressed about the existence of nonfiduciary liability under ERISA thus were dictum, which we are not bound by and which the Department of Labor urges us not to follow.

There is an initial question whether restitution should be classified as an equitable remedy; if not, the Department cannot succeed, because the statute confines it to equitable relief. Restitution is a remedy historically and today dispensed in law and eq[*756] uity proceedings alike. First National Bank v. Warren, 796 F.2d 999, 1000 (7th Cir.1986); Medtronic, Inc. v. Intermedies, Inc., 725 F.2d 440, 443 (7th Cir.1984); 1 Dan R. Dobbs, Law of Remedies § 1.2, p. 11; § 4.1(1), p. 556; § 4.1(3), pp. 564-65; §§ 4.2-4.3 (2d ed. 1993). Whether it is equitable depends merely on whether it is being sought in an equity suit. If the beneficiary of a trust sought an accounting of the profits of a defalcating trustee — a form of restitutionary relief — the accounting if ordered would be ordered in a suit in equity, and the remedy thus would be equitable, while a suit seeking the identical relief against a nonfiduciary would normally be a suit at law and the relief sought therefore legal. 1 Dobbs, supra, § 4.3(5), pp. 608-14. A special wrinkle here, however, is that the concept of liability for (in effect) aiding and abetting a fiduciary’s misconduct comes out of the law of trusts (as noted in Mertens, see — U.S. at-, 113 S.Ct. at 2067; see also Seminole Nation v. United States, 316 U.S. 286, 296, 62 S.Ct. 1049, 86 L.Ed. 1480 (1942); 4 Austin W. Scott & William F. Fratcher, The Law of Trusts § 326, p. 291 (4th ed.1989)) which along with the closely related concept of fiduciary obligation was invented by equity judges. Mertens v. Hewitt Associates, supra, — U.S. at -, 113 S.Ct. at 2068; Chauffeurs, Teamsters & Helpers, Local No. 391 v. Terry, 494 U.S. 558, 571 n. 8, 110 S.Ct. 1339, 1348 n. 8, 108 L.Ed.2d 519 (1990); 3 Scott & Fratcher, supra, § 197, p. 188. It is thus an equitable concept, so it can be argued that the Department of Labor is seeking restitution as an equitable remedy in an equity suit.

The Court’s search in Mertens was, however, for distinctively equitable relief on the one hand and, on the other, distinctively legal relief, such as damages, which though sometimes awarded by a court of equity under the “cleanup” doctrine, Medtronic, Inc. v. Intermedies, Inc., supra, 725 F.2d at 442-43, is the classic remedy at law. Unfortunately restitution straddles this divide. The Court may have seemed to place it on the equitable side; other cases, too, have described restitution as an equitable remedy. E.g., Chauffeurs, Teamsters & Helpers, Local No. 391 v. Terry, supra, 494 U.S. at 570, 110 S.Ct. at 1347; Tull v. United States, 481 U.S. 412, 424, 107 S.Ct. 1831, 1838, 95 L.Ed.2d 365 (1987). But we think it more likely that all the Court meant in any of these cases was that restitution, in contrast to damages, is a remedy commonly ordered in equity cases and therefore an equitable remedy in a sense in which damages, though occasionally awarded in equity cases, are not. Restitution is merely not an exclusively equitable remedy like an injunction. Granfinanciera, S.A. v. Nordberg, 492 U.S. 33, 49 n. 7, 109 S.Ct. 2782, 2794 n. 7, 106 L.Ed.2d 26 (1989); United States v. Fountain, 768 F.2d 790, 801 (7th Cir.1985); Austin v. Shalala, 994 F.2d 1170, 1176 n. 6 (5th Cir.1993). On this interpretation, which seems the more plausible and conforms to a correct understanding of both historical and current practice, restitution is a legal remedy when ordered in a ease at law and an equitable remedy (rather than a legal remedy pressed into service to provide complete relief in an equity case — the rationale of the “clean up” doctrine) when ordered in an equity case. This interpretation helps the Department of Labor in this case, since, although it was seeking restitution against a nonfiduciary, the basis for its claim was a duty founded on equitable rather than legal principles. Had the Department’s claim gone to trial, therefore, neither party would have been entitled to a jury trial. Cf. Wardle v. Central States, Southeast & Southwest Areas Pension Fund, 627 F.2d 820, 829 (7th Cir.1980); In re Vorpahl, 695 F.2d 318 (8th Cir.1982).

But was the Department of Labor seeking restitution? It was seeking not a profit, but merely a receipt, an insurance premium, net of some expenses; to call this a “profit,” it could be argued, would convert every suit for the price of a contract into a suit for restitution, contrary to the law. 1 Dobbs, supra, § 4.1(2), p. 559. But what the Department was really seeking, and what the concept of unjust enrichment and its remedial corollary restitution do encompass, was to recover a “negative unjust enrichment,” consisting of the unjust avoidance of a loss. 1 id., § 509, pp. 802-03; Dan B. Dobbs, Handbook on the Law of Remedies: Damages-Equity-Restitution § 4.5, p. 278 (1973). The Depart-[*757] merit’s argument is that Continental Casualty was already liable for the claims against the trustees that the $1 million extension of coverage was ostensibly obtained to cover, so that the $818,000 net premium reduced its underwriting losses by that amount. It benefited to that extent, and at the expense of the pension fund. “Negative unjust enrichment” is a clumsy term, but the concept is straightforward. If A steals from B to pay a debt to C, A is unjustly enriched at B’s expense, even though the consequence of the theft was to reduce A’s obligation to G rather than to put money in A’s pocket.

Why, if Continental was already liable for the claims against the trustees, the trustees would pay for the extension of coverage is not explained by this theory; but we need not worry about this, for quite apart from any doubts about the character of the remedy sought the Department of Labor has no claim. A majority of the Supreme Court has made clear its view that Congress’s omission to impose on nonfiduciaries a duty not to participate knowingly in an ERISA fiduciary’s breach of fiduciary obligations was not inadvertent; that Congress knew that at common law (including in that term the judge-made law of equity) nonfidueiaries were subject to “knowing participation” liability in trust cases, and knowing this decided not to cast the net of ERISA liability that wide. This is a strict constructionist’s approach that we might not embrace as an original matter. ERISA is a detailed and carefully drafted statute, true. But Congress is not omniscient, and even its detailed and carefully drafted statutes contain inadvertent omissions that courts might properly undertake to rectify if satisfied that by doing so they would not be upsetting a compromise — satisfied, that is, that the omission was indeed inadvertent rather than deliberate. We are impressed by the Department’s argument that the scope of ERISA’s preemption provision is so broad that pension funds may well be remediless against knowing participants in fiduciary misconduct (its remedy against the misbehaving fiduciaries themselves may be inadequate because of the limitations of their solvency, and proved to be so in this case), though the force of this point is diminished by the fact that the holding in Mertens, which we are not free to question, strips the pension funds, or here their surrogate the Department of Labor, of their principal remedy — a suit for damages.

The Supreme Court considered all these points in Mertens and found them unimpressive. The majority opinion goes out of its way to throw cold water on the idea of an implied liability of nonfiduciaries for knowing participation in fiduciaries’ misconduct. The discussion is dictum, but it is considered dictum. If we thought the Court had overlooked some point that might have altered its view, we would be less reluctant to buck the dictum. But it appears not to have. The Department’s brief in our case devotes only a few pages to criticism of the Mertens dictum, and the points it makes were addressed and rejected by the Court.

Justice Holmes has been derided for claiming in “The Path of the Law,” 10 Harv.L.Rev. 457, 461 (1897), that the law is merely a prediction of what the courts will do. The theory has many weaknesses, but also a valid core. In areas of profound uncertainty, such as whether a statute that does not explicitly impose duties on nonfiduciaries should be interpreted as doing so implicitly because of the background of trust law against which it was enacted and the vagueness and breadth of “other appropriate relief’ with no specified limitation as to whom the relief can be sought from, federal law is for all practical purposes what the Supreme Court says it is. When the Court’s view is embodied in a holding, the Court’s reluctance to overrule its precedents enables a confident prediction that that holding is “the law.” When the view is embodied in a dictum, prediction cannot be made with the same confidence. But where it is a recent dictum that considers all the relevant considerations and adumbrates an unmistakable conclusion, it would be reckless to think the Court likely to adopt a contrary view in the near future. In such a case the dictum provides the best, though not an infallible, guide to what the law is, and it will ordinarily be the duty of a lower court to be guided by it.

Thornton v. Evans, 692 F.2d 1064, 1078 (7th Cir.1982), decided by this court many[*758] years before Mertens, held that an ERISA plaintiff could obtain relief against nonfiduci-aries who conspired with a fiduciary to violate the statute. The line between conspiracy and knowing participation is so fine that we have grave doubts, unnecessary however to resolve in this case, that Thornton survived Mertens. But we have no doubt that the district judge was right to dismiss the Department’s suit on the authority of Mer-tens.

AFFIRMED.