960 F.2d 657 (1992)
Charles E. COVEY, as Trustee of V. Jobst & Sons, Inc., Plaintiff-Appellee, Cross-Appellant,
COMMERCIAL NATIONAL BANK OF PEORIA, et al., Defendants-Appellants, Cross-Appellees.

Nos. 91-1843, 91-1896.
United States Court of Appeals, Seventh Circuit.

Argued February 12, 1992.
Decided April 3, 1992.
As Amended April 10, 1992.
Gregg N. Grimsley (argued), Vonachen, Lawless, Trager & Slevin, Peoria, Ill., for appellants Federal Deposit Ins. Corp., Continental Illinois Nat. Bank and Trust Co. of Chicago and Commercial Nat. Bank of Peoria.

James S. Brannon (argued), Peoria, Ill., for appellee Charles E. Covey.

Before EASTERBROOK and KANNE, Circuit Judges, and SHARP, District Judge.[*]

EASTERBROOK, Circuit Judge.

A preference-recovery action requires us to revisit one aspect of In re Xonics Photochemical, Inc., 841 F.2d 198 (7th Cir.1988): how to value contingent liabilities. In 1982 a construction enterprise reorganized. Jobst Corporation was created as a holding company, with stock in four subsidiaries its only assets. With the consent of the lenders, the holding company assumed all of the debt that the enterprise owed to Commercial National Bank of Peoria and Continental Illinois National Bank. The subsidiaries are V. Jobst & Sons, a commercial and industrial construction contractor with headquarters in Peoria, Illinois; New Order, a commercial contractor with headquarters in Tulsa, Oklahoma; Homeway of Illinois, a manufacturer and marketer of single-family homes (and its subsidiary, Homeway of Texas); and Strehlow Corporation, a real estate venture. For clarity we call Jobst Corporation “the parent” and V. Jobst & Sons “Jobst.”

Early in 1984 the family’s senior managers discovered large cost overruns at the Homeways. A partnership (JDJ) between Jobst and Jarvis & DeLoach, Inc., which served as general contractor for residential construction projects (and the principal customer of the Homeways’ products), also was in financial trouble. The partnership abandoned work on several projects, turning them over to the bonding company. Jobst’s managers, and the two principal lenders, believed that New Order and the operations of Jobst outside the JDJ partnership would continue to be profitable if they were refinanced. On May 3 the banks loaned $250,000 to Jobst, $250,000 to New Order, and issued letters of credit from which the beneficiaries eventually drew $775,000. In exchange Jobst guaranteed the parent’s debt to the banks. To back up the guarantee, Jobst gave the banks a security interest in all of its assets — equipment, receivables, everything.

Ten months later Jobst was being liquidated under Chapter 7 of the Bankruptcy Code of 1978. The Homeways, the parent corporation, and the group’s principal manager and equity investor also were bankrupt. See In re Strehlow, 84 B.R. 241 (Bankr.S.D.Fla.1988). The banks sold Jobst’s assets and collected its receivables, obtaining a total of $1,472,937.40. The trustee filed this adversary proceeding, seeking to recover this sum as a preference under 11 U.S.C. § 548. Before the transaction of May 3, Jobst owed nothing to the banks. Because the transaction took place within the year before the bankruptcy, the Code authorizes recapture if the transaction was designed to hinder or delay other creditors (§ 548(a)(1)) or if the debtor received “less than a reasonably equivalent value” and the transaction occurred while the debtor was insolvent or rendered it insolvent (§ 548(a)(2)). The trustee contended that the guarantee and security interest not only were designed to defeat other creditors’ interests (§ 548(a)(1)) but also rendered Jobst insolvent yet did not supply “reasonably equivalent value” (§ 548(a)(2)). Recovery of the $1.47 million the banks obtained from their security interest would inure to the benefit of Jobst’s other creditors, of which there are two groups: suppliers, who had about $2 million in unsecured credit outstanding, and the bonding company, which eventually expended more than $24 million to complete JDJ’s contractual commitments. The trustee renews in this court its argument based on § 548(a)(1), but we need not consider that subject in light of § 548(a)(2).

The bankruptcy judge concluded that the size of Jobst’s contingent liability to the bonding company was unknowable on May 3, 1984. The contingent liability to the banks on the guarantee could be estimated with greater confidence: the judge concluded that there was approximately a 40% chance that the group would pull through its financial distress, and correspondingly a 60% chance that the banks would collect from Jobst under the guarantee. Holding that the value of a contingent liability is the amount the firm must pay if the contingency comes to pass (the parent’s $7.4 million debt), times the probability that this will occur, the judge concluded that Jobst’s guarantee had a value exceeding $4 million. ($7,400,000 × 0.6 = $4,440,000.) Its principal manager put the value of the firm immediately before May 3 at $3.5 million. A contingent liability exceeding $3.5 million then made the firm insolvent, allowing recovery under § 548(a)(2). The district judge affirmed.

Xonics shows that to find the value of a contingent liability a court must determine the likelihood that the contingency will occur. Accord, In re Chase & Sanborn Corp., 904 F.2d 588, 594 (11th Cir. 1990); Douglas G. Baird, The Elements of Bankruptcy 150 (1992). To disregard the probability that the firm will not be called on to pay is to regard all firms as insolvent all of the time, for all firms face some (remote) contingencies exceeding the value of their assets. A firm’s product might prove dangerous, maiming hundreds of customers; all of an air carrier’s planes might fall out of the sky, or one of an electric utility’s nuclear stations melt down, creating stupendous liabilities; all of an insurer’s policyholders might die in the same year, generating obligations that exceed its assets. The probability of such occurrences is low, however, and it therefore makes sense to treat the firms as solvent.

According to the banks, the district and bankruptcy judges went awry in choosing the multiplicand. Instead of using the value of the debt guaranteed ($7.4 million), the courts should have used the value of Jobst’s assets ($3.5 million). A guarantee cannot be worth more than the assets standing behind it. Thus the right calculation, the banks submit, is $3,500,000 × 0.6 = $2,100,000, which leaves Jobst solvent after the transaction of May 3. They quote this language from Xonics, 841 F.2d at 200:

Suppose that on the date the obligations were assumed there was a 1 percent chance that Xonics Photochemical would ever be called on to yield up its assets [on its guarantees]…. Then the true measure of the liability created by these obligations on the date they were assumed would not be [their face amount of] $28 million; it would be a paltry $17,000. For at worst Xonics Photochemical would have to yield up all of its assets (net of other liabilities), that is, $1.7 million, and the probability of this outcome is by assumption only 1 percent…. Discounted, the obligations would not make Xonics insolvent….
The method used by the bankruptcy court here — multiplying the total debt guaranteed by the probability that the guarantor must make good — would have called for a different example in Xonics: $28,000,000 × 0.01 = $280,000. The banks ask us to adhere to the method illustrated in Xonics.

Discounting a contingent liability by the probability of its occurrence is good economics and therefore good law, for solvency, the key to § 548(a)(2), is an economic term. That Xonics departs from accounting conventions, see Note, Estimating Contingent Liabilities to Determine Insolvency in Bankruptcy Proceedings, 1989 B.Y.U.L.Rev. 1315, 1328-30, is true but not pertinent. Accountants may value assets at cost (“book value”), but if the market value of a firm’s assets exceeds its liabilities, it is solvent notwithstanding red ink in the balance sheet. The reverse is true as well: a firm whose assets are worth less than book value may be insolvent despite a financial statement showing positive net worth. Market value of both assets and liabilities determines solvency. The formula used in Xonics to find the value of contingent liabilities also is good economics — but only from the perspective of the creditor. The beneficiary of a guarantee never values that obligation at more than the issuer’s gross assets, and if other claims (say, secured debts) stand ahead of this one, the beneficiary does not value the guarantee at more than the issuer’s net assets. Valuing the guarantee for its own purposes, the creditor would proceed just as we did in Xonics. We did not hold, however, that this is the calculation the Code requires. Xonics used an illustration to demonstrate discounting; the parties did not debate, the case did not depend on, and we therefore did not decide, whether the creditor’s perspective is the right one. The subject is open to initial decision.

The Bankruptcy Code requires us to assess things from the debtor’s perspective. Consider a simple case. Debtor issues its note for $10 million. It has assets of $5 million, secured debt of $2 million, and no other debt. From the creditor’s perspective this note is worth somewhat less than $3 million. (Collection is costly and uncertain.) Together, Debtor’s creditors place a value of less than $5 million on its commitments. Nonetheless, Debtor is insolvent. Against assets of $5 million, there are claims of $12 million. Now turn the $10 million note into a guarantee of a parent corporation’s $50 million debt, coupled with a probability of 20% that Debtor will be called on to pay. The two commitments are economically equivalent: $10,000,000 = $50,000,000 × 0.2. The creditor will treat each instrument as worth a little less than $3 million. There is no reason not to treat the promises identically from Debtor’s perspective too. Each commitment renders the firm insolvent by creating aggregate claims exceeding its assets. If we use net assets as the maximum value of a contingent liability, it follows that no contingent liability ever renders any firm insolvent. A $5 million note issued by a firm with $4 million in assets propels the firm into insolvency, but a $5 billion guarantee by the same firm, on which the beneficiary is 99% certain to draw, would not: instead of multiplying $5 billion by 0.99, the court would multiply $4 million by 0.99. Yet all would concede that, from the debtor’s perspective, the guarantee is more costly than the unconditional note. (Consider: What happens if the firm unexpectedly receives an enormously valuable patent shortly after issuing the note or the guarantee? Which of these instruments will the equity owners more regret signing?)

To decide whether a firm is insolvent within the meaning of § 548(a)(2)(B)(i), a court should ask: What would a buyer be willing to pay for the debtor’s entire package of assets and liabilities? If the price is positive, the firm is solvent; if negative, insolvent. To illustrate, suppose Exxon Corporation were considering the acquisition of our hypothetical debtor with outstanding notes of $2 million and $10 million. It would acquire assets worth $5 million and debts of $12 million. Exxon would value this package at a negative $7 million. Now substitute the guarantee of $50 million in debt (with a 20% chance that the principal debtor will not pay). Again Exxon would treat this package as having a negative value of $7 million. That the guarantee exceeds the guarantor’s net assets would be of no moment to Exxon or an equivalently flush buyer. (The assumption of a buyer having assets exceeding the liabilities to be acquired is important. A stumblebum would pay 1¢ for the most hopelessly insolvent firm, as the deal puts none of the bum’s nonexistent assets at risk and could pay off if the debtor unexpectedly strikes it rich.)

After the transaction of May 3, Jobst had assets of $3.75 million, debt of $250,000 on the loan from the two banks, the benefit of the letters of credit, and contingent debt of $7.4 million plus any sums drawn against the letters of credit. (We disregard trade credit and the obligations to the bonding company.) The bankruptcy court found that the probability that the parent would fail to pay, and that the banks would collect from Jobst, was 60%. This means that Jobst was insolvent: a purchaser would not pay a positive price for its assortment of assets and claims.

To this the banks reply that the bankruptcy judge committed clear error in reckoning the probability at 60%. They assert that, according to the testimony of Jobst’s senior manager, “had the plan worked, 80% of the debt owed to the Banks would have been paid, leaving 20% for the Debtor to repay.” Any figure greater than 20% accordingly was error, the banks conclude. There are two things wrong with this approach. First, the testimony of a corporate insider is not conclusive; the bankruptcy judge noted that JDJ had turned some projects over to the bonding company before May 3, implying that Jobst was in serious straits. Second, the banks misinterpret the testimony. The manager did not say that there were four chances in five that the parent corporation would retire the debt without contributions from Jobst. Instead he testified that if everything went well, the parent would be able to pay 80% of its debt. If the parent would be unable to pay in full under the best circumstances, the banks were sure to turn to Jobst.

Next the banks contend that even if the transaction made Jobst insolvent, they are entitled to keep the $1.47 million they collected because Jobst received a “reasonably equivalent value”, § 548(a)(2)(A). The banks disbursed $1,206,773 (after adjustment for interest and a tax refund) and recovered $1,472,937.40. They submit that the ratio of outlay to recovery, 82%, is reasonable equivalence as a matter of law. In support they cite Durrett v. Washington National Insurance Co., 621 F.2d 201, 203 (5th Cir.1980), which suggested that value to the debtor within 70% of the creditor’s recovery is “reasonably equivalent”. We need not ask what percentage is “reasonably equivalent”, because the banks have the timing wrong. The question posed by § 548(a)(2)(A) is whether the debtor “received less than a reasonably equivalent value in exchange for such transfer or obligation”. Although a note or guarantee is not a “transfer” for purposes of 11 U.S.C. § 101(54), see Barnhill v. Johnson, ___ U.S. ___, 112 S.Ct. 1386, 118 L.Ed.2d 39 (1992) (check is not a transfer until paid), both note and guarantee are obligations. We have already determined that the guarantee was an obligation worth approximately $4.4 million from Jobst’s perspective, substantially more than the $1.2 million the entire corporate group received.

A moment’s thought reveals why the comparison between outlay and recovery cannot be the right one. Preference-recovery provisions help stave off a rush to separate assets from an ailing debtor. Competition among creditors to dismember the debtor may reduce the aggregate value of the assets, to the detriment of all. This may be what happened to Jobst, for the banks tell us that in May 1984 the firm was worth $3.5 million, but they realized less than half that when liquidating all of its assets. Collective proceedings permit a more orderly distribution, preserving a debtor’s going-concern value (if there is any). By providing for the recapture of last-minute payments to creditors, the preference-recovery provisions reduce the incentive to participate in asset-grabbing. See Levit v. Ingersoll Rand Financial Corp., 874 F.2d 1186, 1194-95 (7th Cir. 1989); Thomas H. Jackson, Avoiding Powers in Bankruptcy, 36 Stan.L.Rev. 725, 727-31, 756-68 (1984). Suppose a creditor nonetheless jumps the queue, seizing an asset and selling it for just enough to cover its loan (even if it would have been worth substantially more as part of an ongoing enterprise). By the banks’ reasoning, the value-reducing grab assures its right to keep the funds. For by depressing the asset’s value, the creditor has produced “reasonable equivalence” between its loan and its recovery. It cannot be that the very value-diminishing partitions of assets that the preference-recovery rules are designed to avert end up enabling the creditor to keep the booty.

At length, we reach the trustee’s cross-appeal. Section 548(c) gives the creditor from which a preferential transfer has been recovered a lien against the assets of the estate in bankruptcy “to the extent that such transferee … gave value to the debtor”, if the creditor acted “in good faith”. The bankruptcy judge held that the banks acted in good faith and are entitled to a lien in the amount of $1,206,773. Rather than requiring the banks to turn over the entire $1.47 million, only to receive the $1.2 million back again, the bankruptcy judge ordered the banks to transfer only the difference to the trustee. The judge added the banks’ collections from third parties; the net recovery came to $408,000.

One of the trustee’s contentions — that the bankruptcy judge erred in concluding that the banks acted in good faith — is a long shot. Good faith is a finding of fact, and such findings must stand unless clearly erroneous. The bankruptcy judge concluded that the banks made the May 1984 loan in the expectation that Jobst and New Order could be restored to profitable operation. Testimony pro and con on the subject puts the judge’s finding beyond the scope of revision here. Anderson v. Bessemer City, 470 U.S. 564, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985).

The other argument fares better. Even after United Savings Ass’n v. Timbers of Inwood Forest Associates, Ltd., 484 U.S. 365, 108 S.Ct. 626, 98 L.Ed.2d 740 (1988), an undersecured creditor is entitled to interest that accrues before the filing of the petition in bankruptcy. The lien under § 548(c) made the banks secured to the extent of their advances. Thus they were entitled to an award of interest. But how much? They are secured only on their advances, not on the parent’s debt that Jobst guaranteed. Although there is some ambiguity in the record and the opinions of the district and bankruptcy judges, it appears that the banks applied the $1.47 million they collected from the liquidation first to the $500,000 in loans made in May 1984 and then to the parent’s debt. Interest continued accumulating on the $775,000 drawn on the letters of credit. The banks pursued other corporations also liable on this obligation, and their collections reduced the principal outstanding. Some $86,000 in interest accumulated on the letters of credit before the bankruptcy proceeding began in March 1985. The trustee submits that the banks are not entitled to a § 548(c) lien for this interest.

The trustee challenges the application of the proceeds. Had the $1.47 million been applied in sequence to the loan to Jobst, the draws against the letters of credit, and the guaranteed debt, none of the banks’ advances eligible for the § 548(c) lien would have been drawing interest, and 11 U.S.C. § 502(b)(2) would have disallowed their claim for interest on the guarantee. The banks’ brief in this court does not mention this problem, standing on the fact that the trustee conceded that $86,000 accumulated on the letters of credit. The concession is beside the point the trustee is making. Perhaps the banks’ disdain of the trustee’s actual argument reflects the absence of an answer to it.

A creditor usually may apply involuntary collections to an outstanding debt of its choice. First Wisconsin Financial Corp. v. Yamaguchi, 812 F.2d 370, 374 (7th Cir. 1987); Restatement of Security § 142 (1941). Cf. Restatement (2d) of Contracts §§ 258-60 (1981) (rules for application of voluntary payments). Rights of third parties may limit that privilege, cf. Jorgensen v. Aetna Casualty & Surety Co., 769 P.2d 809, 811-12 (Utah 1988), and at all events the black letter rule supposes that the application to a particular debt is lawful. We need not accept the trustee’s submission that an application reversed by § 548(a)(2) is “void” to see that third-party interests dominate here: the third parties are the competing creditors. The bank applied the funds in a way that maximized its return at the expense of Jobst’s other creditors. Treating the proceeds as if they had been properly applied to the parent’s debt not only is inconsistent with the premise of the preference recovery but also disregards the limit in § 548(c): none of the parent’s existing debt represented “value to the debtor”. The lien was allowable, therefore, only on the new advances plus interest on those advances. To protect other creditors the bankruptcy court should have treated the banks as if they had applied the $1.47 million in collections to those advances first.

Except to the extent it authorized the banks to retain $86,000 as interest on the letters of credit, the judgment is affirmed. The case is remanded for a recalculation of the recovery consistent with this opinion.

[*] Honorable Allen Sharp, Chief Judge, United States District Court for the Northern District of Indiana, sitting by designation.