United States v. Sidney J. Dickler, Richard R. Petrucci. Sidney J. Dickler, in No. 94-3517. Richard R. Petrucci, in No. 94-3518

OPINION OF THE COURT STAPLETON, Circuit Judge: These are appeals from the judgments of sentence imposed on Sidney Dickler and Richard Petrucci after each entered a plea of guilty to impeding the functions of the Resolution Trust Corporation (“RTC”) in violation of 18 U.S.C. § 1032(2). The defendants attack their sentences on two grounds: that the court prohibited them from submitting evidence at their sentencing hearings relevant to the calculation of the “loss” caused by their criminal conduct for purposes of U.S.S.G. § 2F1.1 and that the district court erred in calculating that loss. We will reverse the judgments and remand for resen-tencing. I. Sidney Dickler and Richard Petrucci were charged in a three-count indictment with offenses relating to their operation of two companies: Action Repossession Services, Inc. (“Action Repossession”) and Action Motors, Inc. (“Action Motors”). Action Repossession was in the business of repossessing cars on behalf of financial institutions. Action Motors was a used car dealer. Dickler and Petrucci were principals in both companies. Counts One and Two of the indictment charged Dickler and Petrucci with participating in a scheme to defraud two federally insured financial institutions, Horizon Financial Savings (“Horizon”) and Atlantic Financial (“Atlantic”), in violation of 18 U.S.C. § 1344. According to the indictment, the criminal activity began in 1985. The RTC became the conservator of Horizon on June 8, 1989, and of Atlantic on January 11, 1990. Count Three charged defendants with participating in a scheme to impede the functions of the RTC in violation of 18 U.S.C. § 1032(2). Action Repossession was under contract with the victim banks to repossess cars when a car owner defaulted on his or her car loan, or when a lease terminated and the car was not voluntarily returned. Under its agree *821 ment with the banks, Action Repossession was to repossess and store the vehicle, prepare a condition report on the repossessed vehicle, and solicit three bona fide bids for the vehicle from prospective buyers. 1 The condition report and bids were then to be sent to the banks, who would either accept one of the bids or reject them all. If the bids were rejected, the banks might sell the car at auction. If one of the bids was accepted, the bank would send the vehicle title and bill of sale to Action Repossession, who was then expected to transfer title to the vehicle to the successful bidder on the bank’s behalf. The defendants’ fraudulent scheme involved submitting false bids to the banks. Instead of soliciting bona fide bids from used car dealers or individuals, Dickler and Pe-trucci submitted bids with the names of fictitious bidders with the intent that Action Motors would purchase the car for resale whenever a false bid was accepted. Thus, under the scheme, when the bank accepted one of the bids and sent the title and bill of sale to Action Repossession, Action Motors would acquire the vehicle instead of the fictitious bidder and, after repairing and “detailing” it, 2 would then resell the vehicle for a profit. Prior to trial, Dickler and Petrucci each entered a plea of guilty to Count Three of the indictment pursuant to a plea agreement. The respective plea agreements, which were the same in all aspects relevant to this appeal, provided that: (1) that conduct charged in Counts One and Two could be considered “relevant conduct” for purposes of the pre-sentence investigative report (“PSR”), (2) the relevant loss for purposes of applying U.S.S.G. § 2F1.1 was more than $120,000 and less than $500,000, (3) each defendant would assist the government in the investigation of other bank fraud violations, (4) a special assessment of $50 would be paid, (5) the defendants’ offense levels under the Sentencing Guidelines should be reduced for acceptance of responsibility, (6) the offense levels should be increased two levels for aggravating roles, and (7) at sentencing, the government would move to dismiss the remaining counts and recommend that the defendants be given sentences in the middle range of the applicable guideline range. In compliance with Federal Rule Criminal Procedure 11(f), the court, before accepting the pleas of Dickler and Petrucci, asked the government to summarize its evidence as to Count Three. 3 The government summarized the defendants’ false bid scheme. It submitted documentation to explain how the scheme worked, including a bid sheet, a condition report, a bill of sale, odometer disclosure statement, and an internal record of Action Possession indicating the purchase, repair, and retail sale of the vehicle. In the course of explaining the illustrative documentation to the court, the government indicated that the scheme included not only the preparation and submission of bids from fictitious bidders but also the preparation and submission of false condition reports. Specifically, the government represented that it had spoken with the former lessee of the vehicle described in the sample condition report, who had explained that the vehicle had been in better condition when it was repossessed than represented on the condition report (e.g., he refuted that the tires were poor, that the car had no hubcaps, and that the tail light was broken), and that the signature purporting to *822 be his on the condition report was not. Each defendant, upon questioning by the court, agreed with the government’s factual summary and entered his plea. A PSR was subsequently prepared for each defendant. In relevant part, the PSRs contained the following factual allegations and legal conclusions: the defendants would order employees to falsify the condition reports being sent to the financial institutions, the defendants’ fraud caused the banks to lose monies on the vehicles because the submitted bids were “low balled,” the defendants had obtained approximately $386,223 from the banks and the RTC through the submission of false bids (calculated by deducting the bid price and cost for detailing and repairing the vehicles from the price at which Action Motors sold the repossessed vehicles), and the amount they obtained represented the amount of loss for purposes of calculating the offense level under U.S.S.G. § 2Fl.l(b). Sentencing hearings were held on September 2,1994. A separate hearing was held for each defendant although the court held that any relevant information in the first hearing (Dickler’s) could be incorporated into the second. The focus of the hearings was the calculation of loss under U.S.S.G. § 2Fl.l(b). Although they had stipulated to a loss of at least $120,000 in their respective plea agreements, the defendants objected to the PSR’s loss calculation because it focused on the gain they realized from reselling the repossessed vehicles rather than the actual loss to the victims. They maintained that the figure did not accurately represent the victims’ loss because it did not account for the effect market forces and other external factors had on their resale price. Although the defendants conceded that the court was entitled to set the loss at $120,000 based on their stipulation, they contended that no higher loss figure was permissible because the fraudulent bids they had submitted, although falsified as to identity of the purchaser, represented the fair market value of the vehicles and, accordingly, there was no loss to the victims. According to Petrucci’s testimony, the defendants submitted bids in the name of fictitious purchasers because it was difficult to obtain bona fide bids from outside bidders, a contention supported by the testimony of another used car dealer who testified that most dealers were reluctant to become involved in the repossession bid market. The defendants were concerned, according to Pe-trucci, that the failure to submit three bids in accordance with the standard practice of the banks would jeopardize their repossession business, which generated approximately three times the income produced by their used car dealership. Testimony as to the banks’ method for evaluating bids on repossessed vehicles was presented by James Sweeney, a former collection manager for Atlantic Financial. Sweeney explained that at the time an automobile was repossessed, the bank would determine a high and low bid for the vehicle. The high bid would be eighty-five percent of the National Automobile Dealers Association (“NADA”) book’s average value for that car. To obtain the low bid figure, the bank would reduce that figure based on the vehicle’s condition and mileage, as represented in the vehicle condition reports and accompanying photographs. Sweeney testified that it was generally known in the repossession bid industry that banks valued their repossessed vehicles in this manner and thus that banks did not generally receive bids for more than eighty-five percent of “book” value. Sweeney further explained that once the statutory no-bid period passed, it was important to obtain and accept a reasonable bid as quickly as possible in order to avoid mounting storage charges and further erosion of the bid price. 4 The parties stipulated to testimony that Atlantic’s bid evaluation methods were standard industry practice. Sweeney also testified that, at Dickler’s suggestion, defendants would occasionally fix up the repossessed vehicle before selling it on the bank’s behalf but, although this generated a higher return *823 value for the bank, they elected not to sell most of the vehicles in this manner because they were more interested in simply disposing of the vehicles quickly. According to Petrucei, Horizon had likewise declined defendants’ suggestion to improve the condition of their vehicles before soliciting buyers. The defendant also presented evidence to explain why their resale price was generally significantly higher than their purchase bid price. Petrucei testified that all of the repossessed vehicles purchased from the banks and the RTC were detailed, and that many were also repaired before they were resold. Various witnesses testified that cosmetic and repair work increases the price of a used vehicle disproportionately to the cost of the work. Moreover, cars sold by used car dealers sell at higher prices than repossessed vehicles because repossessed vehicles are purchased on an “as is” basis and cannot be test-driven. 5 The defendants also attempted to present evidence that the vehicle condition reports had not been falsified, but the court would not permit this testimony, stating that defendants could not now present evidence that contradicted the facts to which they had agreed during the change of plea hearings. The district court did not find the defendants’ arguments and evidence persuasive. The court rejected all of the defendants’ substantive objections to the PSR’s loss calculation and adopted the PSR’s focus on the defendants’ resale prices less the amount they paid for the vehicles and their costs for detailing work and repairs. The court allowed only a deduction for computational errors of $34,765.50, resulting in a final loss calculation of $351,457.50. Thus, under U.S.S.G. § 2F1.1, the court added nine levels to the base offense level of six because the loss involved in the offense was greater than $350,000 and less than $500,000. That figure was increased another two levels for more than minimal planning, and increased an additional two levels for aggravating role, resulting in a total offense level of nineteen. That figure was then reduced three levels for acceptance of responsibility, for a total adjusted offense level of sixteen. This offense level, with a criminal history category of I, corresponded to a sentencing range for each defendant of twenty-one to twenty-seven months. Within that range, the court sentenced Diekler to twenty-four months of imprisonment and Petrucei to twenty-one months. These timely appeals followed. 6 II. The threshold issue is whether the district court erred in refusing to admit evidence at the sentencing hearing tending to show that the condition reports submitted to the banks were not falsified. The district court so ruled based on its view that when a defendant under oath expressly admits facts at a plea hearing in the course of persuading the court to accept his plea, he may not thereafter deny those facts any more than he may thereafter deny the facts alleged in the indictment and admitted by his plea. The defendants do not challenge this legal conclusion as a general proposition. 7 Rather, *824 they insist that there were special circumstances here that should relieve them of the consequences that would ordinarily flow from an admission at a plea hearing. First, according to the defendants, they did not unambiguously admit that their scheme involved falsified condition reports. Second, they claim surprise, pointing out that the indictment alleged only the submission of bids from fictitious bidders and insisting that they were “blindsided” by the government at the plea hearings. We cannot accept either contention. The transcript of the change of plea hearings simply does not support the defendants’ first contention. The government Represented that it had evidence tending to show that false condition reports were a part of the defendants’ scheme. Each defendant, in response to questioning from the court while under oath, then acknowledged that he agreed with what the government said he had done. There was no ambiguity; there were clear admissions in each instance. While we agree that the indictment did not put the defendants on notice of the government’s false condition report contention, that fact provides no adequate explanation for the defendants’ failure to take exception to that contention at the plea hearings. Both defendants were sophisticated businessmen who were represented by counsel. They and their counsel sat in court and listened to the government’s representation regarding what it would prove if the case went to trial. The government’s description of its case was neither lengthy nor complex, and each defendant was asked point blank by the court whether he agreed or did not agree with the government’s version of the facts. The defendants were thus in no position to contend at the sentencing hearing that falsified condition reports were not a part of their scheme. III. A. The defendants pleaded guilty to a violation of 18 U.S.C. § 1032(2). 8 Because their conduct involved a fraudulent bidding scheme, the court sentenced defendants under U.S.S.G. § 2F1.1, the guideline provision applicable to crimes of fraud and deceit. This guideline section provides for a base offense level of six with graduated enhancements of the offense level according to the size of “the loss” to the victim attributable to the fraudulent conduct. The court determined their loss to be $351,457.50 which added nine levels. While the general definition section of the Sentencing Guidelines does not define “the loss,” the commentary to § 2F1.1 discusses this concept. 9 Application note 7 states in relevant part: Valuation of loss is discussed in the Commentary to § 2B1.1 (Larceny, Embezzlement, and Other Forms of Theft). 10 As in theft cases, loss is the value of the money, property, or services unlawfully taken; it does not, for example, include interest the victim could have earned on such funds had the offense not occurred. *825 Consistent with the provisions of 2X1.1 (Attempt, Solicitation or Conspiracy), if an intended loss that the defendant was attempting to inflict can be determined, this figure will be used if it is greater than the actual loss. Frequently, loss in a fraud case will be the same as in a theft case. For example, if the fraud consisted of selling or attempting to sell $40,000 in worthless securities, or representing that a forged check for $40,000 was genuine, the loss would be $40,000. There are, however, instances where additional factors are to be considered in determining the loss or intended loss, [for example]: In a case involving a misrepresentation concerning the quality of a consumer product, the loss is the difference between the amount paid by the victim for the product and the amount for which the victim could resell the product received. In fraudulent loan application cases and contract procurement cases, the loss is the actual loss to the victim (or if the loss has not yet come about, the expected loss).... However, where the intended loss is greater than the actual loss, the intended loss is to be used. In a case involving diversion of government program benefits, loss is the value of the benefits diverted from intended recipients or uses. U.S.S.G. § 2F1.1 cmt. (n.7). We have previously interpreted § 2F1.1 and this commentary as requiring that the method for calculating the victim’s loss correspond to the nature of the defendant’s conduct. Thus, where the defendant’s fraud is similar to theft in that the defendant has “taken” something from the victim without giving the victim something of value in return, the value of the thing or service taken will reflect the victim’s loss. However, where the defendant intended to and did give the victim something of value in exchange for what was fraudulently taken, the value of the object or service taken will not reflect the victim’s actual loss and another method of calculating actual loss must be used. United States v. Kopp, 951 F.2d 521, 528-31 (3d Cir.1991). Our decision in United States v. Kopp provides a good illustration of this rule. The defendant in Kopp had submitted falsified loan documents to a bank and had thereby obtained a loan of $13.75 million. Although the amount fraudulently taken was $13.75 million, the court rejected this figure as the loss under § 2F1.1 because the defendant had secured the loan with a mortgage on other property which enabled the bank to recover the loan proceeds when the defendant defaulted. Thus, we explained: [The defendant] did not “take” $13.75 million for nothing, as a thief would. Furthermore, all thefts involve an intent to deprive the victim of the value of the property taken.... [T]he same is not always true for fraud: some fraud involves an intent to walk away with the full amount fraudulently obtained, while other fraud is committed to obtain a contract the fraud perpetrator intends to perform. Mechanical application of the theft guideline in fraud cases would frustrate the legislative purpose of the guidelines and contravene the specific language of the Commission. 951 F.2d at 529 (citing United States v. Schneider, 930 F.2d 555, 558 (7th Cir.1991)). In applying this flexible, fact-driven concept of loss, we have thus held that in situations where value passes in only one direction—from the victim to the perpetrator— the perpetrator’s gain will normally reflect the victim’s loss. On the other hand, where value passes in both directions, we have held that the victim’s loss will normally be the difference between the value he or she gave up and the value he or she received (or, if greater, the difference between what the perpetrator intended the victim to give up and to receive). Even where value flows in both directions, if it is not feasible to estimate with reasonable accuracy the victim’s loss or intended loss, we have indicated that a sentencing court may look to the perpetrator’s *826 gain as a surrogate for the victim’s loss. United States v. Badaracco, 954 F.2d 928, 937-38 and n. 10 (3d Cir.1992) (citing Kopp, 951 F.2d at 531). Where property is received by the perpetrator of a fraud and promptly resold without alteration, for example, the proceeds from the resale will normally approximate the market value of the property when the victim parted with it; in such a situation, the defendant’s gain can rationally serve as a surrogate for the victim’s loss. The guideline provision governing fraud offenses refers to the victim’s loss, however, and the defendant’s gain may be used only when it is not feasible to estimate the victim’s loss and where there is some logical relationship between the victim’s loss and the defendant’s gain so that the latter can reasonably serve as a surrogate for the former. Without this logical connection, the defendant’s gain cannot be said to be an “estimation” of actual loss, and as our precedent and the Sentencing Guidelines make clear, it is a reasonable estimation of loss, not an alternative, unrelated value, that the sentencing court must ascertain. U.S.S.G. § 2F1.1 cmt. (n.8) (“For the purposes of subsection (b)(1), the loss need not be estimated with precision. The court need only make a reasonable estimate of the loss, given the available information- The offender’s gain from committing the fraud is an alternative estimate that ordinarily will underestimate the loss.”) (emphasis added); cf. United States v. Holloman, 981 F.2d 690 (3d Cir.1992) (upholding use of defendant’s gain as surrogate for victim’s loss where value to defendant of stolen cancelled checks in counterfeiting scheme reflected bank’s potential loss), cert. denied, — U.S. —, 113 S.Ct. 3002, 125 L.Ed.2d 695 (1993). In Kopp, we specifically rejected the use of the defendant’s gain as “an alternative estimate, when ... the true loss is measurable.” 951 F.2d at 530 (emphasis removed). Although we have subsequently refined the circumstances in which Kopp's specific actual loss calculation is applicable, we have not strayed from the concept that the loss calculation should represent the fraud victim’s actual loss. E.g., United States v. Shaffer, 35 F.3d 110, 114 (3d Cir.1994) (distinguishing Kopp’s focus on calculating victim’s actual loss at the time of sentencing and holding that in check kiting scheme, loss should be calculated at time crime is detected because this more accurately reflects the bank’s actual loss from the unsecured fraudulent “loans”); United States v. Mummert, 34 F.3d 201 (3d Cir.1994) (upholding face value of loan as reasonable calculation of bank’s actual loss where no assets had been pledged against the loan and no payments had been made thereon); see also United States v. Daddona, 34 F.3d 163, 170-71 (3d Cir.) (holding that actual loss may not include developer’s costs to complete construction project absent evidence linking those costs directly to defendant’s conduct in fraudulently securing construction bonds), cert. denied, — U.S. —, 115 S.Ct. 515, 130 L.Ed.2d 421 (1994). In this case, the district court found that the defendants’ conduct had deprived it of the ability to make a reasonable estimate of the loss. It reasoned that in each instance the bank expected to receive value equal to the best of three independently submitted bids and that because of the defendants’ conduct, no one could ever know what that value would have been. Accordingly, the district court looked to the defendants’ gain from their scheme as a surrogate for the victims’ loss. It took the gross amount received by the defendants from their sale of the cars and deducted the expense of repairing and detailing, as well as the amount paid to the banks for the vehicles. We find the district court’s analysis troublesome in a number of respects. B. We start with the proposition, based on the uncontradicted evidence, that there are two distinct markets involved here. First, there is an “as is” market in which a buyer purchases a repossessed vehicle in the condition it was in at the time of repossession and without an opportunity to test drive it. The second is a “reconditioned market” in which a buyer purchases a vehicle that has been repaired and detailed, and with an opportunity to test drive it. A vehicle normally has a substantially lower value in the “as is” market than in the “reconditioned” market. *827 In addition to the fact that the risk of paying more than a vehicle is worth is materially greater in the “as is” market, resulting in lower purchase prices in that market, the undisputed evidence indicated that when a vehicle is repaired and detailed, its value materially increases, often many times the amount invested in these activities. The banks knew that both markets existed and they chose to participate, with immaterial exceptions, only in the “as is” market. They did so primarily because they wanted to dispose of the vehicles and get paid quickly. Moreover, when the banks asked the defendants to secure three bids on an “as is” basis, they did so in an effort to secure the fair market value of the vehicles in the “as is” market, not because they expected the defendants to find a particular type of bidder who would bid on average above the fair value in that market. It follows, we believe, that the actual loss of the banks was the fair market value of the vehicles in the “as is” market, less what the defendants paid the banks for the vehicles. The record contains substantial evidence relevant to a determination of the fair market value of the vehicles in the “as is” market. According to the undisputed evidence, sellers in this market value vehicles by discounting the value of the vehicle reported in industry publications like the NADA “blue book” by fifteen percent and by further discounting the resulting value when there is above average mileage or below average condition. Based on this evidence, the defendants urge that the district court’s conclusion regarding the feasibility of estimating the loss cannot stand. We agree. It is true, as the district court found, that the defendants’ conduct with respect to the submission of false condition reports makes it difficult to now determine one factor in the evaluation formula—the condition of each particular vehicle at the time of repossession. Because the condition reports cannot be relied upon as evidence of the vehicles’ actual condition, we acknowledge that the defendants’ conduct has impaired the district court’s ability to estimate the banks’ losses in particular transactions. As the defendants stress, however, the relevant value here is not the value of a particular vehicle, but rather the value of a stream of many vehicles over a seven year period. Moreover, the uncontradicted evidence indicates that, because of the financial circumstances of their owners, repossessed vehicles, on average, have been less well maintained and are in poorer condition than vehicles of the same age generally. Finally, we know that, as a result, those selling in the “as is” market regard 85% of the “blue book” value as the high side of the value range and generally consider condition only for the purpose of determining whether the fair value is less than 85% of book value. One can argue persuasively on the basis of this record evidence that the fair market value of the stream in the “as is” market would be no more than 85% of the aggregate “blue book” value of all of the vehicles and that a comparison of this figure with the aggregate amount the banks received from the defendants for the vehicles would provide an estimate of the banks’ loss that should serve as a ceiling for purposes of loss calculation under § 2F1.1. We do not say that the evidence cited by the defendants compels a conclusion that it is feasible to arrive at a satisfactory estimate of the banks’ losses here. We hold, however, that it was sufficiently probative on that issue that the district court was not at liberty to accept the government’s evidence of the defendants’ gain as a surrogate without some explanation of why an estimate of loss based on this data was not feasible. Accordingly, we will remand to provide the district court with an opportunity to reevaluate the feasibility of making a satisfactory determination of the victim’s aggregate losses. C. The second arrow to the defendants’ bow is based on data gathered from Horizon’s records for the period from November of 1989 to November of 1990. According to the defendants, these were the only records of the victims available to them. During this period Horizon sold a total of 1,103 repossessed vehicles. Action Motors bought 103 or 9% of these vehicles. The rest were sold to others. The largest single purchaser was *828 a dealer in eastern Pennsylvania, Yelland. Horizon received an average return on all I,103 vehicles equal to 96.51% of a value designated on defendants’ exhibits as the “appraised value.” On vehicles sold to Yel-land the average return was 95.26% of “appraised value.” By comparison, Horizon received a 97.01% return on the vehicles sold to Action Motors. The district court was unpersuaded by this data for the following reasons: Defendant was the agent of the institutions and had an obligation to solicit bids to obtain the highest possible price for the cars. Defendant did not fulfill this obligation and solicited no independent bids. (Defendant’s position at ¶ 9). Rather, defendant’s company purchased the cars, hiding its true identity. Thus, the fair market value of the cars is unascertainable because of defendant’s own conduct. The fact that the institutions may have received a higher percentage of the cars’ appraisal value from defendant than they received, on average, from other purchasers is of little relevance. The cars that defendant purchased may have been in better condition than other cars that the financial institutions sold. App. at 161-62. To the extent this conclusion rests on the district court’s view that the market value of the cars in the “as is” market “is of little relevance” because the banks bargained for three bids solicited by Atlantic, we have already noted our disagreement. To the extent it rests on the district court’s speculation that the 103 repossessed vehicles purchased by Action Motors may have been in better condition on average than the 1,103 repossessed vehicles purchased by others, we find that speculation inadequate to support the district court’s rejection of this approach to loss estimation. Given the volume of vehicles and the duration of the period involved, we do not believe the district court was entitled to conclude without some supporting record explanation that the 103 vehicles assigned to Action for repossession were in materially better condition than those assigned to others for repossession. We confess that our study has left us without confidence that we understand exactly what the “appraised value” refers to and how it was derived. At the same time, the record appears to indicate that the “appraised value” on Horizon’s books came from a source not dependent on Action’s condition reports, was regularly recorded on Horizon’s books, and was presumably relied upon by it for some business purpose. 11 While it is not a necessary inference, we believe a trier of fact could infer from this information that the “appraised value” of the various vehicles was determined in some reasonably consistent manner. If one draws this inference, this data concerning a substantial sample of the relevant universe of transactions appears to indicate that Horizon received more from the vehicles it sold to Action Motors than it received from its other sales in the “as is” market. Unless one is willing to assume that the sales to others were also tainted with fraud, this would suggest that the sales to defendants, on average, were not at prices below market value in the “as is” market. It is not for us to decide in the first instance whether any or all of these inferences should be drawn. On remand, the district court should take a fresh look at the data from Horizon’s books and determine whether or not, based on that data, it is possible to estimate the banks’ losses with reasonable accuracy. D. The third arrow to the defendants’ bow is an alternative argument that grants, arguendo, the validity of the proposition that a reasonable estimate of victim loss is not feasible. If the district court was entitled to look to the defendants’ gain as a surrogate for the victim’s loss, the defendants insist *829 that it erred in deducting only the defendants’ repair and detailing expenses and the purchase price paid the banks. Specifically, the defendants argue that the district court, if it looked to gain, was required to deduct the commissions paid to salespersons at Action Motors for reselling the cars (allegedly $40,820.79) and auction and transportation expenses related to the resales (allegedly $7,180.00). In addition, the defendants argue for a further $16,017.96 reduction in their gain, a figure that reflects the difference between the vehicles they resold at auction and the vehicles they resold to individuals in the retail market. Since the district court set the gain at $351,457.50, one or more of these adjustments could make a difference in the guideline range. The district court found the argument regarding the defendants’ additional expenses “without merit” because the “focus of the Court should be on the loss to the victim, not the costs of committing the crime to the defendant.” App. at 165. “Even if the defendant[s] incurred these costs,” the court reasoned, “they are not clearly connected to the actual losses sustained in this case, which is the lost value on the ears that the financial institutions sold to the defendant[s].” Id. Having determined to look to the defendants’ gain as a surrogate for the victim’s loss, we believe the district court was not entitled to give credit for certain expenses that reduced the defendants’ gain and ignore others that would have the same effect on the ground that the latter were not clearly connected to the bank’s loss. 12 In short, we find it impossible to logically distinguish between the defendants’ repair costs and the commissions and auction costs they paid in order to realize their gain. 13 The other adjustment requested by the defendants is also appropriate. The argument is that if the sentencing court follows a gain approach because it cannot estimate the victim’s loss with reasonable precision, and the defendant demonstrates that a component of the total gain is attributable solely to its own efforts after the victim’s loss was complete, that component must be deducted. More specifically, the record shows the price received by Action Motors on the resale of each vehicle, including the price on those sold at auction and those sold at retail. The value received on retail resales over and above what would have been received at auction is attributable, according to the defendants, solely to their own efforts and must be deducted before their gain can serve as a reasonable surrogate for the victim’s loss. The district court rejected this argument because “the fact that the -defendants’ profit percent *830 age ... on his retail car sales was greater than his wholesale profit margin does not inform [the] Court about the actual losses on the cars.” App. at 166. Again, we believe the district court misunderstood the import of the argument being made. We do not suggest that the district court needs to view this argument with an uncritical eye. We do say, however, that the underlying legal premise is a sound one. When a gain approach is used as a surrogate for loss fraud and the defendant demonstrates that a component of the gain as calculated by the government could not be a component of the victim’s loss, an appropriate adjustment is required. E. The fourth and final arrow to the defendants’ bow in their attack on the district court’s calculation of loss is premised on the fact that it included the defendants’ gain on all purchases that occurred between 1985 and December of 1992. According to the defendants, the victim’s loss under § 2F1.1 can include only loss attributable to “relevant conduct” under § lB1.3(a)(2) and “relevant conduct” can include only conduct proscribed by a criminal statute. Because the bids they submitted to the RTC prior to November 29, 1990, the effective date of 18 U.S.C. § 1032(2), were not in violation of that statute, the defendants maintain that it was reversible error to include any loss attributable to those bids. The conduct underlying the indictment involved two periods: the period during which defendants submitted false bids to Horizon (1985 to June 1989) and Atlantic (1985 to January 1990), and the period during which they submitted false bids to the RTC who had been appointed custodians of the failed banks (June 1989 or January 1990 to December 1992). Counts One and Two of the indictment, which were dismissed at sentencing, were based on the conduct during the earlier period and Count Three, to which the defendants pleaded guilty, was based on the conduct during the latter period. There is no question that the defendants’ actions in defrauding the banks during the early period is “relevant conduct” within the meaning of the Sentencing Guidelines and thus that any loss attributable to that conduct may be used to calculate the defendants’ offense level under U.S.S.G. § 2F1.1. 14 The defendants argue, however, that because the statute underlying their guilty pleas, 18 U.S.C. § 1032(2), was not enacted until November 29,1990, their acts vis-a-vis the RTC prior to that date cannot be considered relevant conduct for purposes of determining loss. The district court regarded all sales as relevant conduct without finding that the bids during the challenged period were criminal conduct. If the defendants are right, this would require a $101,562.23 reduction in the district court’s loss calculation. Although this court has not yet addressed the question, other courts of appeals have concluded that “relevant conduct” within the meaning of § IB 1.3 must be criminal conduct. See United States v. Sheahan, 31 F.3d 595, 600 (8th Cir.1994) (“We agree that the relevant conduct the sentencing court should consider in the section 2F1.1 loss calculation is that which is attributable to the defendant’s ‘criminal conduct.’ ”); United States v. Wilson, 980 F.2d 259, 261 (4th Cir.1992) (same). 15 The government does not contend *831 otherwise, and we agree. 16 The relevant criminal conduct need not be conduct with which the defendant was charged, United States v. Santiago, 906 F.2d 867 (2d Cir.1990), nor conduct over which the federal court has jurisdiction, United States v. Pollard, 986 F.2d 44 (3d Cir.), cert. denied, — U.S. -, 113 S.Ct. 2457, 124 L.Ed.2d 671 (1993). Thus, the district court’s use of the loss attributable to the challenged period could be upheld if the defendants’ conduct during that period was shown to constitute any state or federal crime, since it is clear that the conduct was part of the same on-going scheme as the offense conduct. See U.S.S.G. § lB1.3(a)(2). For some inexplicable reason, however, the government did not present evidence at the sentencing hearing or argue in its brief on appeal that the conduct during that period was otherwise criminal. Nevertheless, at oral argument before this court the government suggested a number of criminal offenses which the defendants were said to have committed by submitting fictitious bids to the RTC 17 and asked that we affirm the district court’s inclusion of the loss that occurred during the challenged period on this basis. While we think it highly likely that the defendants’ conduct during the challenged period did violate some criminal statute, we decline to accept the government’s invitation. Due process requires that the defendants have fair notice of exactly why the government believes their conduct during this period was criminal and a fair opportunity to counter the government’s case on that score. In order to be fair, such an opportunity may have to include an opportunity to offer additional evidence. Accordingly, on remand, the district court should require the government to identify the statute or statutes it relies upon and to identify the record evidence that satisfies each element of the offense proscribed. The defendants should then be afforded the opportunity to develop an appropriate record and argue to the contrary. 18 IV. For the foregoing reasons, we will vacate the defendants’ sentences and remand for resentencing. On remand, the question may arise whether the district court is restricted to resentencing the defendants based on the current record, plus whatever the defendants may have to offer in response to the government’s designation of a criminal offense applicable to the defendants’ pre-November 29, 1990, conduct vis-a-vis the RTC. Stated conversely, the issue may arise whether the district court may reopen the record to permit further development of the relevant facts in light of this opinion. We do not preclude the district court from permitting further development of the record and leave that for resolution by an exercise of the district court’s informed discretion. *832 We agree with the Fourth and the D.C. Circuit Courts of Appeal that, where the government has the burden of production and persuasion as it does on issues like enhancement of the offense level under § 2F1.1 based on the victim’s loss, its case should ordinarily have to stand or fall on the record it makes the first time around. It should not normally be afforded “a second bite at the apple.” United States v. Leonzo, 50 F.3d 1086, 1088 (D.C.Cir.1995) (remanding for re-sentencing on the existing record where government failed to sustain its burden of proving loss under § 2F1.1); United States v. Parker, 30 F.3d 542, 553-54 (4th Cir.) (no new evidence permitted on resentencing where prosecution had failed to introduce sufficient evidence that offense took place within 1000 feet of a “playground” within meaning of statute), cert. denied, - — U.S. -, 115 S.Ct. 605, 130 L.Ed.2d 515 (1994). At the same time, we perceive no constitutional or statutory impediment to the district court’s providing the government with an additional opportunity to present evidence on remand if it has tendered a persuasive reason why fairness so requires. See United States v. Ortiz, 25 F.3d 934, 935 (10th Cir.1994) (holding that an order vacating sentence and remanding for resentencing contemplates a de novo hearing at which court can receive any evidence it could have considered during first sentencing hearing); United States v. Cornelius, 968 F.2d 703, 705 (8th Cir.1992) (holding that district court erred in refusing to consider defendants’ evidence upon resentencing); United States v. Jacobs, 955 F.2d 7, 10 (2d Cir.1992) (per curiam) (where original sentence had been vacated because there was insufficient evidence connecting conspiracy’s income to drug sales, district court could on remand consider “reliable new evidence” on this issue); United States v. Stern, 13 F.3d 489, 498 (1st Cir.1994) (“where a sentence is vacated and remanded for redetermination under correct principles, the government is not automatically foreclosed from offering evidence pertinent to the newly established rule.”). Where, as here, the government believes that it is not feasible to estimate the victim’s loss and its evidence, in the absence of the defendant’s evidence, would support a finding to that effect, it will frequently not be fair to expect the government to be prepared with evidence concerning any theory of loss calculation the defendant may advance at the sentencing hearing. If the government, for want of notice or any other reason beyond its control, does not have a fair opportunity to fully counter the defendant’s evidence and the government’s theory does not carry the day, the district court is entitled to permit further record development on remand. By making these observations, we do not suggest that the government should or should not be permitted to offer further evidence in this case on remand. The district court is in a far better position than we to assess the situation in the light of the circumstances surrounding the original sentencing hearing. V. We hold that the district court did not err when it refused to entertain evidence tending to show that the condition reports were not falsified. We further hold, however, that the district court’s findings and conclusions do not support the sentences imposed. Accordingly, we will reverse the judgments of the district court and remand for resentencing. On remand, the district court will revisit its conclusion that it is not feasible to estimate the banks’ and RTC’s loss with a reasonable degree of accuracy. At a minimum, this will involve a reevaluation of the defendants’ evidence concerning the trade practice regarding the valuation of vehicles to be sold in the “as is” market and Horizon’s sales proceeds between November 1989 and November 1990. If the district court concludes that it is feasible to estimate the victims’ losses, it will resentenee the defendants based on the stipulated loss of $120,000 unless the government demonstrates that a greater loss was incurred. If the district court once again determines it appropriate to look to the defendants’ gain as a surrogate for the banks’ loss, it will deduct from the gross gain all expenses necessarily incurred in realizing that gain and any component of the gain necessarily attributable solely to the defendants’ investment of their own resources after their purchase of the vehicles from the banks. Finally, the *833 district court shall not include any gain attributable to bids received by the ETC prior to November 29, 1990, unless it determines that those bids were made in violation of a specific state or federal criminal statute. Footnotes [1] As was apparently the custom in this business, Action Repossession operated pursuant to informal, oral contracts and the parties dispute the exact terms of the agreements. For example, counsel for Action Repossession indicated at oral argument before this court that the company did not solicit bids for the banks pursuant to the repossession contract, but simply provided this as an enhancement to the contract. This distinction is immaterial for purposes of this appeal, however, since it is clear that in soliciting bids for the banks and the RTC, pursuant to the terms of the agreement or otherwise, the banks and the RTC expected Action Repossession to solicit bona fide bids. [2] Petrucci defined detailing a car as thoroughly cleaning the interior and exterior (buffing, waxing, degreasing the engine, painting the engine compartment, cleaning the trunk, shampooing the interior, trunk, and carpeting), performing cosmetic repairs such as repairing holes in the upholstery, and in some cases adding or removing exterior details such as pin striping and window tinting. [3] Rule 11(f) requires that a court inquire into the facts of the case to satisfy itself that there is a factual basis for the plea before accepting even a freely given plea. [4] According to Sweeney, an institution may not seek bids on repossessed vehicles under Pennsylvania law until 15 days after repossession. He also indicated that bids were generally valid for only a short time because the NADA book was reissued every two weeks with continually depreciating values. [5] When Action Motors resold the repossessed vehicles they were covered by Action Motors’ insurance and could be test driven with Action Motors' dealer plates. [6] The district court had jurisdiction pursuant to 18 U.S.C. § 3231 as the defendants were charged with violations of federal law. We exercise appellate jurisdiction pursuant to 18 U.S.C. § 3742 and 28 U.S.C. § 1291. [7] We have held that facts relevant to sentencing contained in the indictment and plea agreement are conclusively established by the entry of a guilty plea even if they are not elements of the offense charged. United States v. Parker, 874 F.2d 174, 178 (3d Cir.1989) (holding that where indictment and plea agreement specified value of packages taken, entry of guilty plea conclusively established value for purposes of sentencing even though value was not an element of the offense charged); see also Crawford v. United States, 519 F.2d 347, 350 (4th Cir.1975) ("[T]he accuracy and truth of an accused’s statements at a Rule 11 proceeding in which his guilty plea is established are ‘conclusively’ established by that proceeding unless and until he makes a reasonable allegation why this should not be so.”), cert. denied, 423 U.S. 1057, 96 S.Ct. 791, 46 L.Ed.2d 647 (1976), and overruled on other grounds by United States v. Whitley, 759 F.2d 327 (4th Cir.), cert. denied, 474 U.S. 873, 106 S.Ct. 196, 88 L.Ed.2d 164 (1985); Nesbitt v. United States, 773 F.Supp. 795, 799 (E.D.Va.1991) ("Sworn statements in a plea proceeding are conclusive unless the mov-ant can demonstrate compelling reasons for *824 questioning their truth, such as ineffective assistance of counsel."). [8] This statute provides: “Whoever ... corruptly impeded or endeavors to impede the functions of [the Resolution Trust] Corporation ... shall be fined under this title or imprisoned not more than 5 years, or both.” 18 U.S.C. § 1032(2). [9] All references to the Sentencing Guidelines, unless otherwise noted, are to the 1994 edition of the Guidelines Manual which was in effect at the time of the defendants' sentencing and incorporates amendments through November 1, 1993. [10] The commentary to section 2B1.1, which has a similar table for enhancement of the base offense level based on loss, defines “loss” as: the value of the property taken, damaged, or destroyed. Ordinarily, when property is taken or destroyed the loss is the fair market value of the particular property at issue. Where the market value is difficult to ascertain or inadequate to measure the harm to the victim, the court may measure loss in some other way, such as reasonable replacement cost to the victim. U.S.S.G. § 2B1.1 cmt. (n.2); see also id. § 2B1.1 cmt. (background) (“The value of property stolen plays an important role in determining sentences for theft and other offenses involving stolen property because it is an indicator of both the harm to the victim and the gain to the defendant.”). [11] Horizon used an independent damage appraiser, J.M. Taylor, to assess the condition of its repossessed vehicles, but there was conflicting evidence regarding the use that Horizon made of these damage reports. The defendants point to evidence indicating that Horizon used the damage reports, along with the defendants’ vehicle condition reports, to determine an appropriate target bid, while a government witness testified that Horizon used them only to determine the residual value of the vehicle to the lessee. [12] While the district court was, of course, entitled to pass upon the probative value of the evidence regarding the other expenses, it made no finding that this evidence was not worthy of belief. To the contrary, based on the district court’s findings and conclusions and the fact that this evidence of other expenses was uncontra-dicted and unchallenged by the government, it appears that the district court regarded it as credible, but not legally relevant. [13] In United States v. Badaracco, 954 F.2d 928 (3d Cir.1992) we approved the use of the defendant’s gain as a surrogate for the victim’s loss and declined to deduct the defendant's expenses from the amount of his gain. Our refusal there to give the defendant credit for his expenses is not contrary to the position we take here. Ba-daracco involved a bank officer whose fraud involved conditioning construction loans on developers’ subcontracting out the electrical work to companies -in which the defendant had an interest. We found this type of fraud to be like embezzlement and concluded that an analogy to the theft guidelines was therefore proper. On this basis, we upheld the district court's use of the defendant's gain, measured in terms of the full amount of the contracts awarded to the defendant’s electrical companies, as a reasonable estimate of the loss under § 2F1.1. We refused to deduct the expenses of the defendant's companies because a three-party transaction was involved and those expenses neither conferred a benefit on the bank nor were logically related to anything received by the bank in the transaction. As we have pointed out, the banks’ loss in this case was the difference between what they gave up (the value of the cars at the time they were sold to defendants) and what they received (the amount paid by the defendants for the cars). Gain can logically serve as a surrogate for loss here only to the extent it reflects the value of what the banks gave up. Based on the uncontra-dicted evidence in this case, it appears that the expenses incurred by the defendants between the time they purchased the cars and resold them contributed to their resale value. Accordingly, the resale value cannot be used as a surrogate for the value of the vehicles at the time of their purchase by the defendants without deducting their expenses. [14] "Relevant conduct" includes acts that were part of the same course of conduct or common scheme or plan as the offense conduct if those acts would be grouped as multiple counts under U.S.S.G. § 3D1.2(d) had the defendant been convicted of those counts. See U.S.S.G. § lB1.3(a)(2) & cmt.(n.3). The bank fraud and RTC fraud counts would be grouped as related counts under § 3D1.2(d) because the offense level for both offenses is determined on the basis of total loss. Moreover, as part of the plea agreement, the defendants acknowledged their responsibility for the conduct charged in Counts One and Two and stipulated that that conduct could be considered "relevant conduct” for purposes of sentencing. See also U.S.S.G. § 6B 1.2(a) (conduct underlying charges dismissed pursuant to a plea agreement can not be excluded from consideration in sentencing by the terms of the plea agreement). [15] Without directly addressing this issue, other courts have presumed that relevant conduct is criminal. See, e.g., United States v. Palomba, 31 F.3d 1456, 1464 n. 8 (9th Cir.1994) (For purposes of provision permitting grouping of closely related counts, relevant conduct is defined inter alia with reference to similarity between charges *831 of conviction and "other criminal conduct."); United States v. Shonubi, 998 F.2d 84 (2d Cir.1993) (The guideline provision defining relevant conduct with respect to a "course of conduct” refers to identifiable pattern of criminal conduct.); United States v. Bethley, 973 F.2d 396, 401 (5th Cir.1992) (Relevant conduct involves “repeated instances of criminal behavior [that] constitute a pattern of criminal conduct.”), cert. denied, - U.S. -, 113 S.Ct. 1323, 122 L.Ed.2d 709 (1993). [16] While § 1B1.3 does not expressly so state, it defines relevant conduct in terms that are more consistent with the hypothesis that relevant conduct is limited to criminal behavior. [17] The government cited the federal mail fraud statute, 18 U.S.C. § 1341, and the Pennsylvania criminal fraud statutes, e.g., 18 Pa.Cons.Stat.Ann. § 4107 (deceptive business practices); id. § 4101 (criminal forgery). As we understand a conser-vatorship under the RTC’s statute, the bank does not cease to exist when the RTC is appointed conservator. Accordingly, it may he that the submission of fictitious bids to the RTC violated Pennsylvania's bank fraud statute (e.g., 18 Pa. Cons.Stat.Ann. § 4103) or even the then current version of 18 U.S.C. § 1344, the statute underlying Counts One and Two and making it a crime to defraud a bank insured by the Federal Deposit Insurance Corporation. [18] United States v. Pollard, 986 F.2d 44 (3d Cir.1993), ce rt. denied, — U.S.-, 113 S.Ct. 2457, 124 L.Ed.2d 671 (1993), does not provide support for the proposition that we must search the record ourselves and determine whether the defendants' conduct was otherwise criminal. In Pollard, the defendant did not argue that the relevant conduct was non-criminal, but only that the federal court was without jurisdiction to charge him based on that conduct.