This class suit against Plaza Finance Company seeks statutory (not compensatory) damages for violation of the Truth in Lending Act, 15 U.S.C. §§ 1601 et seq.; § 1640(a). The district court granted summary judgment for the defendant largely on the basis of a district court decision, since reversed by the Fifth Circuit in Edwards v. Your Credit, Inc., 148 F.3d 427 (5th Cir.1998), in a case virtually identical to this one. A number of similar cases are pending elsewhere. Alvin C. Harrell, "Consumer Credit," 52 Consumer Finance L.Q. Rep. 104, 106 (1998).

In view of the procedural posture of our case, we construe the facts as favorably to the plaintiff as the record will permit. The defendant specializes in making small, short-term loans to individuals who have bad credit (the default rate on its loans is 25 percent). It lent the named plaintiff $307 for ten months, of which $7 went to pay for a premium for "nonfiling insurance," that is, insurance against losses to the finance company resulting from its failure to file (record) its security interest. The Truth in Lending disclosure form that the plaintiff was given listed, besides $307 as the "amount financed" (see Federal Reserve Board Regulation Z, 12 C.F.R. § 226.18(b)), a "finance charge" (basically the amount of interest charged for the loan, but including certain service charges as well, see 15 U.S.C. § 1605(a)) of $128 and an annual interest rate of 83 percent. In computing the interest rate for Truth in Lending purposes, the finance company uses the finance charge as the interest and the amount financed as the principal. Hence, other things being equal, the disclosed interest rate is higher the higher the finance charge and lower the higher the amount financed. This creates an incentive for lenders, who want the interest rate to look as low as possible, to shift items from the finance charge (interest) to the amount financed (principal).

The loan to the plaintiff was secured by a wage assignment plus a security interest in various items of the borrower's personal property, such as a television set. Unless such an interest is recorded in the local UCC registry of security interests in personal property, the holder of the interest will not be able to enforce it against a subsequent secured creditor who records his own security interest. UCC § 9-312(5)(a), 810 ILCS 5/9-312(5)(a); Midwest Decks, Inc. v. Butler & Baretz Acquisitions, Inc., 272 Ill.App.3d 370, 208 Ill. Dec. 455, 649 N.E.2d 511, 516 (Ill.App.1995). (Filing generally is unnecessary to perfect a purchase-money security interest, UCC §§ 9-302(1)(d), 307(2), 810 ILCS 5/9-302(1)(d), 307(2), but that type of security interest is not involved in this case.) Because the rules in Article 9 of the Uniform Commercial Code regulating filing are complex and demanding, see, e.g., UCC §§ 9-103, 9-401, 9-403(2), it is easy for a lender to make a mistake, such as filing in the wrong place or failing to renew the filing statement after its expiration; and lenders sometimes insure themselves against the consequences of such a mistake by buying nonfiling insurance. The Truth in Lending Act permits a lender to include as a service fee a premium for such insurance. 15 U.S.C. § 1605(d)(2). The $7 that Plaza charged the plaintiff was the per-loan premium it pays to Voyager Property and Casualty Insurance Company. This is the same insurance company that was involved in the Fifth Circuit's Edwards decision. Although Voyager is licensed to sell nonfiling insurance in Illinois, where the loan to the plaintiff was made, there is no suggestion that Illinois has determined that Voyager's contract with Plaza and its performance of that contract are consistent with the terms of the license or the insurance law of Illinois. Indeed, we can find nothing in the record or in the law of Illinois to indicate what Illinois understands by "nonfiling insurance."

The plaintiff argues that the "insurance" which Voyager has sold Plaza either is not insurance at all, or is not nonfiling insurance but instead default insurance--and a premium for default insurance, unlike a premium for nonfiling insurance, must be included in the finance charge. 12 C.F.R. § 226.4(b)(5). Interest is compensation not only for the time value of money but also for the risk of default, and so a charge to the borrower that is intended to compensate the lender for that risk is functionally part of the borrower's interest expense. A premium for default insurance is such a charge, and so it belongs in the "finance charge" (interest) column of the disclosure form.

The usual purpose of insurance is to shift risk from an individual or other entity that is risk averse, and so would prefer to substitute a cost certain (a fixed insurance premium) for the risk of incurring a larger cost, to an entity that by pooling independent risks can minimize the overall risk to itself. Union Labor Life Ins. Co. v. Pireno, 458 U.S. 119, 127-31, 102 S. Ct. 3002, 73 L. Ed. 2d 647 (1982); Group Life & Health Ins. Co. v. Royal Drug Co., 440 U.S. 205, 211, 99 S. Ct. 1067, 59 L. Ed. 2d 261 (1979); Clark Equipment Co. v. Dial Corp., 25 F.3d 1384, 1388 (7th Cir.1994); Sears, Roebuck & Co. v. Commissioner, 972 F.2d 858, 863 (7th Cir.1992). The simplest example is life insurance. A person who does not want to bear the financial risk of dying young can buy life insurance, for which he pays a fixed premium. The financial risk of his early death is shifted to the insurance company, which by pooling that risk with the risk of its policy holders who die old can eliminate the risk of incurring an unexpectedly steep loss from a premature death. By pooling, the insurance company shields itself from that risk; by joining the pool, the insured eliminates the risk to himself.

No risk-shifting purpose is discernible in the arrangement between Plaza and Voyager. Their original contract expressly capped Voyager's potential liability to Plaza at 90 percent of the premiums paid by Plaza. This meant that no risk was shifted to Voyager. If Plaza paid Voyager premiums of $50,000, and had insured losses of $100,000, for a total loss-related cost of $150,000, it would receive $45,000 in insurance proceeds from Voyager, leaving it with a net cost of $105,000 ($150,000 - $45,000)--a net cost greater than the premiums, greater even than the incurred loss, that is, the loss against which Plaza nominally insured. Plaza was in effect a self-insurer, and the cost of self-insurance is not within the dispensation to exclude premiums for nonfiling insurance from the finance charge. 12 C.F.R. Part 226, Supp. I, § 226.4(e)(4) (Official Staff Commentary on Regulation Z).

Of course insurance policies always have limits; but when the limit is at or below the premium, the insured is not shifting risk by buying the policy. This cannot be the end of the analysis, however. Insurance has other functions besides risk-shifting, including smoothing costs over time, providing assistance in defending against claims (liability insurers typically provide a defense if their insured is sued for conduct covered or even just arguably covered by the policy), and exploiting various tax opportunities. Wisconsin Power & Light Co. v. Century Indemnity Co., 130 F.3d 787, 791 (7th Cir.1997); Sears, Roebuck & Co., supra. But at argument Plaza's lawyer was able to identify only one possible function of Voyager's insurance, and that is to avoid having to include $7 in the finance charge. By its lawyer's own acknowledgment, Plaza is getting nothing in the way of a service from its so-called insurer, whether for bearing risk or for anything else, in exchange for the $7 that it pays Voyager--or rather for the 70cents it pays Voyager, for it gets the rest of the $7 back. If the only function of the insurance policy is to monkey with the disclosed interest rate, it is not a bona fide policy; it is a fraud.

And that is, so far as the record indicates, the only function of the "nonfiling insurance" that Plaza bought from Voyager: to enable the lender (Plaza) to shift $7 of what otherwise would be an interest charge from the finance-charge column on the Truth in Lending disclosure form to the amount-financed column. See 12 C.F.R. § 226.18(b)(2). The expense to the borrower is the same, but the disclosed annual interest rate is lower. Remember that items added to the finance charge count as interest in computing the disclosed annual interest rate, and so increase that rate, while items added to the amount financed increase the denominator in the interest-rate calculation and so reduce the disclosed rate. In this case, the disclosed interest rate would have risen to 89 percent if $7 had been subtracted from the amount financed and added to the finance charge. The smaller the loan, the bigger the difference that the shift of $7 makes. Consider a very simple example: a one-year loan (amount financed) of $50 (and Plaza makes loans that small), and a finance charge of $40 payable at the end of the year, and hence an annual interest rate of 80 percent. (Simple rather than compound interest is assumed, to keep the example simple.) If $7 is shifted from the amount financed to the finance charge, the annual interest rate skyrockets to 109 percent ($47/$43). That's a scary figure; it might frighten off even the necessitous borrowers who are Plaza's principal customers.

During the period covered by the complaint, the insurance contract with Voyager was altered to drop the 90-percent-of-premiums cap. This change eliminated the most illusory feature of the contract. But the plaintiff, who received her loan from Plaza after the change, contends that Voyager and Plaza have an informal understanding that Plaza will not submit claims in excess of 90 percent of the premiums that it has paid during a given period; hence the change in the policy was merely cosmetic. If so, the inference that the arrangement between these two companies is not an insurance policy of any sort, but a sham that has no purpose other than to facilitate an evasion of the Truth in Lending Act's requirements, would be compelling in the absence of contrary evidence not yet presented by Plaza. It is true that the Fifth Circuit in Edwards granted summary judgment for the finance company on this, the "complete sham" claim of a Truth in Lending violation, 148 F.3d at 442-43, but it did so on a record that, unlike the record in this case, did not contain the 90 percent of premiums cap. But we note that if the plaintiff decides to pursue the "complete sham" claim on remand, a division of the class into subclasses (see Fed.R.Civ.P. 23(c)(4)(B); Marisol A. by Forbes v. Giuliani, 126 F.3d 372, 378-79 (2d Cir.1997); 7B Charles A. Wright, Arthur R. Miller & Mary K. Kane, Federal Practice and Procedure § 1790, pp. 276-84 (2d ed.1986)) may be indicated because persons who borrowed from Plaza before it changed the terms of the insurance policy have an even stronger claim than those who, like the named plaintiff, borrowed after the change.

The plaintiff's alternative ground for contending that Plaza has violated the Act is, however, both of uniform strength across the entire class and solid. It is that if there is real insurance here, and not a complete sham, it is insurance against default rather than insurance against the consequences of not filing a security interest. A lender files a security interest only if it thinks it might want to seize the collateral in the event of default. The plaintiff has presented evidence from which it can be inferred, with enough confidence to defeat summary judgment, that Plaza does not want to seize the collateral for its loans, because they are tiny loans to people who own no valuable property. The taking of the security interest is intended merely to frighten the borrower. FTC, Trade Regulation Rule on Credit Practices, 49 Fed.Reg. 7740, 7762-63 (March 1, 1984); American Financial Services Ass'n v. FTC, 767 F.2d 957, 973 (D.C.Cir.1985); Kathleen E. Keest & Gary Klein, Truth in Lending § 3.9.6.3, p. 142 n. 729 (3d ed.1995). Had the named plaintiff defaulted on her $307 loan, it would not have been worth Plaza's while to institute legal process to obtain and sell her personal property. Rather than wanting to file and occasionally through neglect failing to do so and wanting insurance against that eventuality, Plaza doesn't want to bother to file, has no desire for insurance against the consequences of not filing (there are no consequences, so there is nothing to insure against), and never has submitted a nonfiling claim to Voyager.

Most of the claims that it submits to Voyager are for losses caused by a default by a borrower who cannot be located (a "skip"). In these cases, the borrower's property can't be located either. Failure to have recorded a security interest in the property can have no consequence for Plaza in such a case; Plaza can't enforce a security interest in property that it can't find. Some of the claims involve borrowers who have defaulted because they died, and a few involve borrowers who have entered Chapter 7 bankruptcy. But in neither of these classes of case is there any indication that property in which Plaza had a security interest was ever snatched away from it by a creditor who had a superior interest by reason of having recorded it.

Insurance companies sometimes pay claims that don't actually come within the scope of the insurance policy. They do so because of their own inadvertence, fraud by the insured, ambiguity in the policy, or a desire to maintain good relation with valued customers. We want to emphasize, lest lenders find themselves obliged by innocent mistake to pay potentially very large statutory damages in class actions, that nonfiling insurance doesn't cease to be so merely because the insurance company sometimes pays on claims that don't fit within the strict terms of the policy. But so far as the record discloses, Voyager has never paid a claim by Plaza for a loss due to nonfiling. Not rarely; never. So far as the record discloses, the insurance contract is in fact a contract for default insurance, not nonfiling insurance, and the law treats the two types of insurance differently.

If an insurance company issues an automobile liability policy to a person who does not own or drive an automobile, and then pays the claims that the person submits for accidents caused by his bicycle, the policy would not be an automobile liability policy but a bicycle liability policy. Cf. Nolker v. Wallace, 317 So. 2d 4 (La.App.1975). And so it is here: penetrating form to substance is the right approach in a Truth in Lending Act case in which an insurance policy says "nonfiling" but all the claims submitted and paid under it are for defaults the costs of which to the insured have nothing to do with the insured's having failed to record a security interest. This is not just our idea, or that of the Supreme Court, Ford Motor Credit Co. v. Cenance, 452 U.S. 155, 158, 101 S. Ct. 2239, 68 L. Ed. 2d 744 (1981) (per curiam); the Federal Reserve Board, in its regulation interpreting and applying the Act, has made clear that it is the actual character of a policy of insurance--what it really insures--rather than the name, that controls its classification for purposes of the Act. See 12 C.F.R. Pt. 226, Supp. I, § 226.4(d)(10), p. 314 (1998) (Official Staff Commentary).

We emphasize that in speaking of penetrating the outer form to find the inner substance, we (and the Supreme Court, and the Federal Reserve Board) are not referring to the "substance" of the credit transaction. The Act is not a usury law; it does not limit interest rates; all it requires is truthful and (it is hoped) informative disclosure of the interest rate and the other terms of credit. The relevant substance is truth; and its protection is inconsistent with allowing a lender to call something "nonfiling insurance" that is not insurance against nonfiling losses and indeed may not be insurance at all. Since the Truth in Lending Act permits premiums for nonfiling insurance to be included in the amount financed but requires default insurance premiums to be included in the interest charge, a lender cannot be permitted to designate a premium as being for nonfiling insurance if it is really, clearly, and always for default insurance. Otherwise the Act would be easily evaded, cf. Mourning v. Family Publications Service, Inc., 411 U.S. 356, 366-68 and n. 26, 93 S. Ct. 1652, 36 L. Ed. 2d 318 (1973); Gibson v. Bob Watson Chevrolet-Geo, Inc., 112 F.3d 283, 287 (7th Cir.1997), as appears to be happening here.

At argument Plaza's lawyer reminded us that his client did not conceal the $7 charge from the plaintiff; she knew she was paying it. What she didn't know was what interest rate she was paying. Maybe she didn't care. But the provision of the Truth in Lending Act that requires disclosure of the annual interest rate assumes that borrowers, or some appreciable number of them at any rate, do care; and we are not at liberty to question the premises of a valid statute. Id. at 287.

We do worry about opening a can of worms in which class action lawyers can rummage for discrepancies between the terms and actual performance of contracts relating to the various service charges that the Act permits to be included in the amount financed. But to forbid inquiry into performance would open a big loophole, as a host of cases (besides Edwards, a virtual clone of this case) attest. Ford Motor Credit Co. v. Cenance, supra, 452 U.S. at 158, 101 S. Ct. 2239 (expressly refusing to place form over substance); Mourning v. Family Publications Service, Inc., supra, 411 U.S. at 366-68, 93 S. Ct. 1652; Walker v. Wallace Auto Sales, Inc., 155 F.3d 927, 932-33 (7th Cir.1998); Adiel v. Chase Federal Savings & Loan Ass'n, 810 F.2d 1051, 1053 (11th Cir.1987). Nothing in the language or history of the Act, or the cases interpreting it, authorize courts to forbid plaintiffs to prove if they can that a finance charge has been mislabeled as part of the amount financed in circumstances that negate an inference that the mislabeling is due to a mistake by a third party, such as a nonfiling insurer who occasionally pays what is really a loss due to default rather than nonfiling. We need not consider the outer bounds of "occasionally," since the plaintiff's submission is that the parties to the so-called insurance contract never intended for any nonfiling claim to be submitted and paid and that no such claim ever was submitted or paid.

We emphasize that the facts recited in this opinion are merely those adduced in the summary judgment proceedings and viewed in a light favorable to the plaintiff. The trial may cast the facts in a different light. But the grant of summary judgment was error.

REVERSED.

EASTERBROOK, Circuit Judge, dissenting.

The Truth in Lending Act, 15 U.S.C. §§ 1601-67f, separates the elements of a credit transaction into two categories: the "amount financed" and the "finance charge." Once these have been identified, an annual percentage rate of interest (APR) may be computed mechanically, and it must be disclosed to the consumer. The APR is a useful way to compare different offers, but only if every lender puts the same elements into the "amount financed" and "finance charge." Under 15 U.S.C. § 1605(d)(1) the filing fee for perfecting a security interest belongs in the "amount financed" rather than the "finance charge," and to produce parallel treatment § 1605(d)(2) puts the "premium payable for any insurance in lieu of perfecting any security interest otherwise required by the creditor in connection with the transaction" into the same category. See also 12 C.F.R. § 226.4(e)(2). But other insurance goes into the "finance charge," see § 1605(a)(5), (b), (c), because it is an alternative to swallowing the loss if the creditor dies or defaults--and credit risks are a big part of what the finance charge covers.

Plaza Finance Company took security interests in borrowers' property but did not perfect them by filing. It purchased a nonfiling insurance policy from Voyager Property and Casualty Insurance Company for $7 per borrower, less than the fee Plaza would have paid to record a financing statement under the Uniform Commercial Code. Insurance was superior to filing from Plaza's perspective, because the insurance sometimes would enable it to collect the value of the security, just as a perfected security interest would, without the cost of filing (the state never refunds the fee) and without the expense and hassle of retrieving and selling the chattel. Plaza disclosed the cost of this insurance to Adams and its other customers as part of the amount financed. Adams borrowed $300, but Plaza told her that the "amount financed" was $307. There were no hidden charges, unlike Walker v. Wallace Auto Sales, Inc., 155 F.3d 927 (7th Cir.1998), and the cost of credit (the periodic and total payments) was fully revealed. Adams knew how much she was borrowing and how much she would have to pay back.

Voyager sold Plaza a policy that Illinois insurance regulators approved as "nonfiling insurance." According to its terms, the policy covers only losses caused by the ability of a later creditor to obtain priority over Plaza as a result of Plaza's failure to file the financing statement. Such a policy won't pay off very often. Creditors that take interests in consumer durables such as television sets and refrigerators rarely seize and sell the chattels; the costs of doing so may exceed the amount realized; and as a result Plaza is unlikely to fall behind in a race among creditors--though the deterrent effect of seizing personal property may lead a lender to do so occasionally even if it loses money in the process. But genuine nonfiling claims may be rare. In an effort to retrieve more of the premiums, Plaza submitted, and Voyager paid, claims that are not covered under the policy. Plaintiffs contend that this practice converted the policy from one of nonfiling insurance into one of nonpayment insurance, which must be treated as part of the "finance charge." If the $7 had been moved from the "amount financed" to the "finance charge," the APR would have been higher--though the periodic and total payments would have been unaffected. My colleagues hold that if plaintiffs can prove that Plaza routinely submitted, and Voyager regularly paid, claims that are not covered by the policy's terms, then plaintiffs are entitled to collect statutory penalties (and their lawyers handsome fees)--though not compensatory damages, for plaintiffs suffered none. The conclusion that a course of performance that the parties mutually acknowledge (or can be deemed to have contemplated) takes precedence over the terms of the policy has the support of the majority opinion in Edwards v. Your Credit, Inc., 148 F.3d 427 (5th Cir.1998), but I think that Judge Smith's dissent, id. at 443-46, has the better of the argument.

The idea that borrowers can collect penalties because an insurance company paid out too much to their lender--a practice that increases the lender's total collections and thus drives down the rate of interest to risky customers--makes sense only under a statute that elevates form over substance, as the TILA does. Lenders must follow prescribed forms of disclosure; it is essential not only to disclose how much the borrower pays for credit (as Plaza did) but also to disclose each element on the right line of the right form. 15 U.S.C. § 1638. Borrowers can collect statutory penalties if the forms are disregarded, even if they do not suffer injury, indeed even if the practice benefits them. 15 U.S.C. § 1640(a)(2). Plaza's practice was better for borrowers than paying the $7 filing fee, though perhaps worse than another option. Plaza paid $7 per customer to Voyager and recovered approximately $6.30 (less its administrative costs of making claims); it could have saved another 70cents by omitting both filing and insurance. But under the TILA we assess not the difference among $7 for filing, $7 for nonfiling insurance, and $6.30 added to interest, but the difference between $7 in the "amount financed" and $7 in the "finance charge." What my colleagues hold, and what the majority of the fifth circuit held in Edwards, is that plaintiffs may prevail by piercing the form of Voyager's policy to get at the substance of the parties' practice, and thus to conclude that the $7 belonged in the "finance charge."

A substance-over-form approach is fundamentally incompatible with the TILA's penalty provisions, which exalt form over substance. It just won't do to have a system in which the propriety of classification can be known only after the fact. If my colleagues are correct, then a lender that tried to play it safe by putting the cost of nonfiling insurance into the "finance charge" could be challenged, and required to pay penalties, if a court later decided that the transaction really was nonfiling insurance that had to go into the "amount financed." But the TILA assumes that the classifications are knowable ex ante and fixed at the time of the loan; an ex ante approach to classification and disclosure is discordant with an ex post inquiry to determine the "substance" of the transaction. No wonder the TILA includes a rule against using hindsight to assess the adequacy of disclosure. 15 U.S.C. § 1634.

One reason my colleagues give for adopting a substance-over-form approach is that, when forms control, evasion is facilitated. That's true enough; all rules (= forms) create loopholes. To say this is not to decide the contest in favor of substance (= standards developed in common-law fashion by courts), however. If the cost of rules is loopholes and evasion, then the cost of standards is--cost. It is more expensive to apply and litigate about standards than to apply rules. I'll say more later about these costs; for now it is enough to observe that there is no a priori reason to think the costs of standards lower than the costs of rules. See Louis Kaplow, Rules versus Standards: An Economic Analysis, 42 Duke L.J. 557 (1992); Colin S. Diver, The Optimal Precision of Administrative Rules, 93 Yale L.J. 65 (1983); Isaac Ehrlich & Richard A. Posner, An Economic Analysis of Legal Rulemaking, 3 J. Legal Studies 257 (1974). Congress may prefer standards over rules in some statutes, yet choose rules over standards in others. Disclosure statutes frequently prefer form (to promote certainty) at the expense of substance. Consider the securities laws, which require issuers to reveal balance sheets that reflect historical cost, but do not call for information about replacement cost or likely future profitability, which investors really want to know. Shares of stock are regulated on the basis of their formal attributes, even though (for closely held firms) control of the stock is functionally the same as ownership of the assets. See Landreth Timber Co. v. Landreth, 471 U.S. 681, 105 S. Ct. 2297, 85 L. Ed. 2d 692 (1985), disapproving Sutter v. Groen, 687 F.2d 197 (7th Cir.1982). Likewise the TILA is rule-based, and we disserve that legislative choice by deciding that standards really are the way to go.

Which is not to deny that evasions should be dealt with. My point is that the TILA does not charge courts with the duty of plugging loopholes; that task belongs to Congress, with the aid of the Federal Reserve, which acts by regulation, so that lenders may know the rules and apply them ex ante. My colleagues assume that if standards are needed, then judges will proclaim them. That must be why the majority cites, as establishing the primacy of substance over form, Ford Motor Credit Co. v. Cenance, 452 U.S. 155, 158, 101 S. Ct. 2239, 68 L. Ed. 2d 744 (1981), and Mourning v. Family Publications Service, Inc., 411 U.S. 356, 366 n. 29, 93 S. Ct. 1652, 36 L. Ed. 2d 318 (1973). Yet both Cenance and Mourning were cases about the validity and application of regulations that mapped particular elements of economic substance to the forms of the TILA, creating new rules in the process. These decisions, and others like them, do not establish that courts may do the same. If the conduct of Plaza and Voyager shows that lenders are undercutting the TILA, then the right response is a new law, a new regulation, or perhaps a new staff interpretation from the Federal Reserve, not judicial creativity that undermines lenders' ability to classify elements of the transaction at the time credit is extended. Regulations adopted after notice and comment can reconcile economic substance with the need for ex ante decisions in a way that judicial decisions after trials cannot.

According to the majority, this has been accomplished already. "[T]he Federal Reserve Board, in its regulation interpreting and applying the Act, has made it clear that it is the actual character of a policy of insurance--what it really insures--rather than the name, that controls the classification for purposes of the Act. See 12 C.F.R. Pt. 226, Supp. I, § 226.4(d)(10), p. 314 (Official Staff Commentary)." Adams, 168 F.3d at 936. This citation is mysterious; there is no § 226.4(d)(10). Paragraph 10 of the staff commentary at page 314 of the 1998 volume of C.F.R. containing Part 226 (Regulation Z) refers to all of § 226.4(d), a subsection that specifies the treatment of credit life insurance premiums and debt cancellation coverage, neither of which is involved in our case. A lender seeking to know how to classify nonfiling insurance would not turn to § 226.4(d), let alone to the staff interpretation of that subsection; it would turn to § 226.4(e)(2), which does address nonfiling insurance (but is uninformative on the form vs. substance debate). Nor does the staff interpretation say what the majority says it says. Here is its text:

Single-interest insurance defined. The term single-interest insurance as used in the regulation refers only to the types of coverage traditionally included in the term vendor's single-interest insurance (or VSI), that is, protection of tangible property against normal property damage, concealment, confiscation, conversion, embezzlement, and skip. Some comprehensive insurance policies may include a variety of additional coverages, such as repossession insurance and holder-in-due-course insurance. These types of coverage do not constitute single-interest insurance for purposes of the regulation, and premiums for them do not qualify for exclusion from the finance charge under § 226.4(d). If a policy that is primarily VSI also provides coverages that are not VSI or other property insurance, a portion of the premiums must be allocated to the nonexcludable coverages and included in the finance charge. However, such allocation is not required if the total premium in fact attributable to all of the non-VSI coverages included in the policy is $1.00 or less (or $5.00 or less in the case of a multi-year policy).

Adams does not contend that Plaza misunderstands the usage of "single-interest insurance," a term that has nothing to do with her claim. Indeed, Adams does not cite or discuss any part of Regulation Z or the staff commentary. Perhaps the majority reads this paragraph more broadly as requiring allocating of premiums whenever a policy covers multiple risks. What the staff commentary tells us is important is the contents of the policy, which is to say form, rather than the payment history deduced from the claims submitted by the insured. So if this commentary matters, it undercuts my colleagues' position. But as far as I can see nothing the Federal Reserve has written directly addresses the issue. The staff commentary on § 226.4(e)(2) reads:

4. Non-filing insurance. The exclusion in § 226.4(e)(2) is available only if non-filing insurance is purchased. If the creditor collects and simply retains a fee as a sort of self-insurance against non-filing it may not be excluded from the finance charge. If the non-filing insurance premium exceeds the amount of the fees excludable from the finance charge under § 226.4(e)(1), only the excess is a finance charge. For example:

The fee for perfecting a security interest is $5.00 and the fee for releasing the security interest is $3.00. The creditor charges $10.00 for non-filing insurance. Only $8.00 of the $10.00 is excludable from the finance charge.

Nothing here suggests that nonfiling insurance purchased and paid for must be included in the finance charge just because the insurer indemnifies a client for casualties that were not actually covered by the policy.

As it happens, the Fed has prescribed a backup rule to be applied when the statute and regulations are insufficiently precise. Neither the TILA nor Regulation Z defines "insurance." What to do? Well, we could search for common-law solutions, but what we are supposed to do is apply state law. Undefined terms "have the meanings given to them by state law or contract." 12 C.F.R. § 226.2(b)(3). We know how the parties' contract deals with Voyager's policy; does Illinois law override that contract? My colleagues do not discuss how Illinois defines "nonfiling insurance"--though the fact that Voyager's policy has been approved by state regulatory officials, and is sold as, nonfiling insurance is good evidence that state law at least permits Plaza to treat it as "nonfiling insurance", cf. Schmitt v. American Family Mutual Insurance Co., 161 F.3d 1115 (7th Cir.1998), and no more is required to show that plaintiffs cannot collect statutory penalties. Plaintiffs do not contend that Illinois permits an insurance policy to be recharacterized in light of performance as non-insurance, or as insuring a risk other than the one named in the policy. That ought to end this case. But the majority does not mention § 226.2(b)(3) or Illinois law, and despite § 226.2(b)(3) they insist that contracts may be disregarded.

What drives my colleagues' approach is their conviction that elements substantively related to creditworthiness should be included in the "finance charge" and therefore increase the APR. If this is so, one wonders why nonfiling insurance (or for that matter, filing fees) are excluded from the "finance charge" in the first place. Congress decided, for good or ill, to put these charges, and more besides, in the "amount financed" rather than the "finance charge." It did not express a preference for one over the other; it prescribed the contents of each. That prescription is essential, as I have explained, in order to make the APR a useful means of comparing different offers. The APR of a given offer adds nothing to the information about the periodic and total payments. Only an irrational consumer who knew the payment schedule would be distracted by the APR, and the TILA, like other disclosure statutes, assumes that consumers are rational (else disclosure does no good). What the APR does is facilitate comparison of offers. Is a payment schedule of $40 a month for 12 months better or worse for the consumer than a schedule of $30 a month for 18 months, when the consumer receives $300? Comparing the APR offers a way to answer that question (the rate for the first offer is 97%, and for the second 85%)--but only if every other element of the transaction has been classified precisely and predictably into categories, so that APRs calculated by different lenders (or by the same lender for different payment schedules) reflect the same items. By destabilizing this classification process the majority interferes with the attainment of this statutory goal.

Let us assume, however, that putting "effective nonpayment insurance" into the "finance charge," and thus increasing the APR, is a good thing. It does not follow that plaintiffs prevail, for "no legislation pursues its purposes at all costs. Deciding what competing values will or will not be sacrificed to the achievement of a particular objective is the very essence of legislative choice--and it frustrates rather than effectuates legislative intent simplistically to assume that whatever furthers the statute's primary objective must be the law." Rodriguez v. United States, 480 U.S. 522, 525-26, 107 S. Ct. 1391, 94 L. Ed. 2d 533 (1987) (emphasis in original). See also Board of Governors v. Dimension Financial Corp., 474 U.S. 361, 373-74, 106 S. Ct. 681, 88 L. Ed. 2d 691 (1986); NAACP v. American Family Mutual Insurance Co., 978 F.2d 287, 294, 298 (7th Cir.1992); Contract Courier Services, Inc. v. Research and Special Programs Administration, 924 F.2d 112, 115 (7th Cir.1991); In re Erickson, 815 F.2d 1090, 1094 (7th Cir.1987). One vital question is who, if anyone, possesses the authority to determine whether to achieve more of the statute's goals--Congress, the Executive Branch, or the Judicial Branch. For the TILA the rulemaking power has been delegated to the Federal Reserve. Another question concerns costs of decision by litigation, which can be substantial for all standards, and for reasons I have given are especially large under a form-based system such as the TILA that does not require proof of injury.

Costs of pursuing the majority's search for substance are likely to be high--and not just because the process uses hindsight to impose statutory penalties in the absence of actual injury. All trials are costly, and trials conducted without the benefit of a legal rule especially so. Parties ladle up evidence hoping to demonstrate whatever the judge later thinks important. My colleagues remand this case for evidentiary proceedings, but they do not articulate the legal rule that the district judge will apply. How frequent must the insurance company's payments for "skips" be to mark the policy as nonpayment rather than nonfiling insurance? Suppose 5% of Plaza's claims fit the terms of the policy; is this enough to negate the inference that it is "really" a nonpayment-insurance policy? Or must 10% or 20% of the claims be legitimate? Must the insurance company know that claims outside the policy's formal terms are false, or is suspicion enough? Must the knowledge precede issuance or renewal of a policy, or is it enough if a pattern develops during a policy's term? A regulation issued by the Fed might resolve these matters; my colleagues do not (and this is not their fault; it is not the proper place of the judiciary to invent such mechanical rules; thus the difference between judicial and legislative rulemaking comes to the fore).

Then there is the question whether Voyager's policy is "insurance" in the first place. My colleagues suggest that "insurance" usually means risk-shifting via pooling of hazards among persons or firms, but they acknowledge that insurance may spread a single firm's losses over time--and they might add that sometimes bona fide insurance policies have nothing to do with either form of risk adjustment. Many firms purchase medical insurance for their employees as part of ERISA welfare plans and agree to reimburse insurers the full cost, plus a loading charge. Realistically the employers are buying an administration service, not a risk-alleviation service. Workers' compensation policies have a similar function from the employers' perspective, and from the workers' perspective they serve a bonding (but not an insurance) function. Retrospectively-rated policies are similar; they may spread risk if the insured becomes insolvent before the final bill arrives, but if the insured is able to pay no risk is shifted. Do my colleagues mean to say that none of these devices can be "insurance" even if state law (which, recall, controls under 12 C.F.R. § 226.2(b)(3)) calls them insurance? My colleagues do not treat risk-spreading as the exclusive function of insurance, and thus imply that the definition of "insurance" in the McCarran-Ferguson Act, see Department of the Treasury v. Fabe, 508 U.S. 491, 113 S. Ct. 2202, 124 L. Ed. 2d 449 (1993); Group Life & Health Insurance Co. v. Royal Drug Co., 440 U.S. 205, 99 S. Ct. 1067, 59 L. Ed. 2d 261 (1979), will not be used under the TILA, but then what rule must the district court apply on remand? Defining "insurance" for McCarran-Ferguson purposes has been an enduring problem; judicial creation of a new federal definition under the TILA will be worse, because the TILA does not contain even a clue about the meaning of that word (yet another reason to turn to state law under § 226.2(b)(3), one would suppose).

Illustrating the costs and uncertainties involved, my colleagues disagree with the fifth circuit's treatment of the insurance question in Edwards (compare 148 F.3d at 442-43 with Adams, 168 F.3d at 935)--though they add that Edwards dealt with a record that "did not contain the 90 percent of premiums cap." This is a questionable distinction. Edwards argued that Voyager was unwilling to pay out more than 89.25% of premiums collected. 148 F.3d at 431. Although this rule did not appear in the policy, my colleagues' principal point is that how the parties behave prevails over what the contract says. To be consistent they should add that a tacit cap and an explicit cap amount to the same thing. This would make clear their disagreement with the fifth circuit, which on this issue at least deemed form--whether the 89.25% figure was communicated to the policyholder--controlling. 148 F.3d at 443.

Plaintiffs call the payout limit (in either express or tacit form) proof that the contract does not insure anything. Voyager, by contrast, calls it a stop-loss provision and observes that the Supreme Court has held maximum-payment provisions compatible with "insurance" for purposes of the McCarran-Ferguson Act. Union Labor Life Insurance Co. v. Pireno, 458 U.S. 119, 129-30, 102 S. Ct. 3002, 73 L. Ed. 2d 647 (1982). The actual language of the provision sounds more like a warning that excessive claims will lead to nonrenewal (or a premium increase)--for no insurer wants to lose money. This is how the district court in Edwards treated it after receiving evidence. 148 F.3d at 431. Drivers who have accidents must expect to pay more for insurance the next year, and lenders who have a bad loss experience likewise. Why is this incompatible with "insurance"? It is easy to imagine a complex and expensive fight on remand about the significance of the 90% provision but hard to see why the effort is worth the candle. My colleagues imply that it matters whether the "cap" is in the policy or is just an understanding, but as I have already explained this is incompatible with their quest for substance and creates still another complexity for the district court (and lenders) to grapple with.

Lenders underwrite both sides of this litigation--they pay their own lawyers, and through statutory penalties and attorneys' fees they ultimately cover the expenses of the plaintiffs' side too. Where does this money come from? Not from the lenders' shareholders. Financial markets are marked by fierce competition; investors can put their money in a low-risk basket of securities and will not accept lower returns from a sector of the economy beset by expensive litigation. In the long run, then, the funds to finance litigation of this kind come from borrowers--and the borrowers involved in this case are among those least able to afford the outlay. Today's case illustrates how easy it is to make the supposed beneficiaries of a legal rule worse off. Adams and Plaza were partners in contract--a contract in which fraud played no role (recall that the $7 charge was disclosed). As we have remarked elsewhere, borrowers "will pay for judicial liberality and everyone will pay for the loss of legal certainty that ensues when legal principles are bent however futilely to redistributive ends. The idea that favoring one side or the other in a class of contract disputes can redistribute wealth is one of the most persistent illusions of judicial power. It comes from failing to consider the full consequences of legal decisions. Courts deciding contract cases cannot durably shift the balance of advantages to the weaker side of the market; they can only make contracts more costly to that side in the future, because [the stronger side] will demand compensation for bearing onerous terms." Original Great American Chocolate Chip Cookie Co. v. River Valley Cookies, Ltd., 970 F.2d 273, 282 (7th Cir.1992). If Congress had commanded the result my colleagues reach, judges would have to go along no matter who bears the cost; but Congress did not, and I do not.