UNITED STATES of America, Plaintiff-Appellee, v. Richard I. BERGER, Defendant-Appellant, 587 F.3d 1038

MILAN D. SMITH, JR., Circuit Judge: Defendant-Appellant Richard I. Berger appeals the sentence imposed by the district court following our affirmance of his conviction for twelve counts of bank and securities fraud. Berger argues that, in sentencing him on remand, the district court erred by: (1) not adhering to the civil loss causation principle in finding shareholder loss, as described by the Supreme Court in Dura Pharmaceuticals, Inc. v.

Broudo, 544 U.S. 336, 342-48, 125 S.Ct. 1627, 161 L.Ed.2d 577 (2005); and (2) applying an erroneous standard of proof in determining total loss for sentencing enhancement purposes. While we decline to extend the Dura Pharmaceuticals principle to criminal securities fraud, we conclude that the district court’s loss calculation approach was nevertheless flawed. Thus, although we conclude that the district court used the correct standard of proof in determining the total loss, we vacate Berger’s sentence and remand to the district court for resentencing.

FACTS AND PROCEDURAL BACKGROUND Craig Consumer Electronics, Inc. (Craig) was a publicly traded consumer electronics business that primarily distributed its products to retail electronics stores. During the relevant time frame, Berger was Craig’s President, Chief Executive Officer, and Chairman of the Board. Two other corporate officers, Donna Richardson and Bonnie Metz, participated in the fraudulent scheme and were convicted along with Berger for their involvement. In August 1994, Craig entered into a $50 million revolving credit agreement with a consortium of banks.

Under the agreement, the amount Craig was permitted to borrow was based on the value of its current inventory and accounts receivable. To determine the fluctuating amount Craig was eligible to borrow, Berger and his co-defendants were required to provide the lending banks with a daily certification concerning those assets. Berger and his accomplices began the fraudulent scheme as early as 1995. Starting at that time and continuing through September 1997, Craig lacked sufficient qualifying accounts receivable and inventory to continue borrowing the funds needed for Craig’s ongoing operations.

To conceal Craig’s true financial condition from the lending banks, Berger and his cohorts employed various accounting schemes to falsify the information contained in the certifications. Relying on these false statements, the banks lent millions of dollars to Craig based on either nonexistent or substantially overstated collateral.- In May 1996, Craig made an initial public offering (IPO) of its stock.

In connection with the IPO, Berger publicly misrepresented the company’s fiscal viability, misstating Craig’s financial condition in several mandatory reports filed with the Securities and Exchange Commission (SEC). At the time of the IPO, Craig was actually operating in default of its credit agreement with the lending banks, and was substantially overdrawn on its credit line. None of this information was disclosed in Craig’s mandatory SEC filings, or to its lenders. In 1997, an audit of the company’s records by Craig’s accounting firm uncovered various accounting irregularities.

As a result of the audit, Craig was required to restate its earnings for 1995 and part of 1996, thereby revealing that its earnings were substantially lower than those shown in its previous financial statements. In the months following this restatement, Craig’s stock price fell from $4.99 to $0.99 per share. In July 1997, Craig’s stock was delisted from the Nasdaq because of its failure to meet Nasdaq’s minimum bid price. The securities fraud and accounting irregularities noted were not publicly revealed until after the delisting.

The lending banks did not discover the full extent of the fraud until August 1997, when Craig filed for bankruptcy. In March 2003, Berger was indicted for thirty-six counts of bank and securities fraud including: conspiracy, loan fraud, falsification of corporate books and records, making false statements to accountants of a publicly traded company, and making false statements in reports filed with the SEC. Berger went to trial and was convicted on twelve of those counts.

In September 2004, the district court, believing controlling authority prohibited it from applying any sentencing facts not found by the jury, calculated an applicable sentencing range of zero to six months and sentenced Berger to six months imprisonment. The district court also ordered Berger to pay restitution of $3.14 million and a $1.25 million fine. Berger appealed his conviction and restitution order, and the government cross-appealed the sentence. We affirmed the conviction and the restitution amount.

However, we vacated Berger’s sentence and remanded to the district court for resentencing in light of United States v. Booker, 543 U.S. 220, 125 S.Ct. 738, 160 L.Ed.2d 621 (2005). On remand, using a preponderance of the evidence standard, the district court found several facts that significantly increased Berger’s sentencing range. Among other things, the district court found that Berger’s fraud caused a loss of $3.14 million to the various banks with which Craig did business, thereby triggering a thirteen-level enhancement under U.S.S.G. § 2F1.1.

To determine the loss to shareholders, the court adopted one of the government’s suggested calculation methods, the so-called “modified market capitalization theory,” i.e., comparing the change in stock value of other, unaffiliated companies after accounting irregularities in those companies’ records were disclosed to the market. The court determined that the average depreciation of those selected companies’ stock was 26.5% and applied that figure to the value of Craig’s initial public offering (although in Craig’s case, the fraud was never disclosed to the market before trading was halted).

The court calculated the resulting shareholder loss at $2.1 million. Therefore, the total calculated loss was $5.2 million, which triggered a fourteen-level sentencing enhancement, from level sixteen to thirty. This enhancement increased the applicable sentencing range from 21-27 months to 97-121 months. The district court imposed a 97-month sentence. Berger appeals the sentence, arguing that the district court committed two significant legal errors in calculating the applicable sentencing range. JURISDICTION AND STANDARD OF REVIEW We have jurisdiction pursuant to 28 U.S.C. § 1291.

We review de novo the district court’s

interpretation of the Sentencing Guidelines, United States v. Kimbrew, 406 F.3d 1149, 1151 (9th Cir.2005), which are relevant to this case because the Guidelines address the permissible methods for loss calculation. We review for abuse of discretion the district court’s application of the Guidelines to the facts of this case. Id. “If the district court makes a material miscalculation in the advisory guidelines range, ... we must vacate the sentence and remand for resentencing.” United States v. Zolp, 479 F.3d 715, 721 (9th Cir.2007).

Whether the district court violated due process by using an improper standard of proof is a question of constitutional law that we review de novo. United States v. Johansson, 249 F.3d 848, 853 (9th Cir.2001). DISCUSSION Berger raises two issues on appeal. First, he argues that in calculating loss in securities fraud cases, district courts must employ the civil securities fraud “loss causation” approach as described in Dura Pharmaceuticals, 544 U.S. 336, 125 S.Ct. 1627, 161 L.Ed.2d 577, and that the district court erred by not doing so here.

He also contends that the district court erred in finding facts resulting in a significant sentencing enhancement by a preponderance of the evidence — rather than a clear and convincing evidence — standard of proof. I. Loss Causation Principles Applied to Criminal Securities Fraud Berger first argues that the district court erred by including losses in its $2.1 million shareholder loss figure that did not actually occur, or that were not caused by his fraudulent conduct. A.

Civil Securities Fraud Standard The Supreme Court has ruled that to sustain a damages claim for civil securities fraud under 15 U.S.C. §§ 78j(

b) and 78u-4, a plaintiff must show that the fraud was publicly revealed and that the disclosure caused the shareholders to suffer loss. Dura Pharms., 544 U.S. at 346-47, 125 S.Ct. 1627 (finding that plaintiff must show that “share price fell significantly after the truth [of the fraud] became known”). In so holding, the Dura Pharmaceuticals Court rejected the notion that stock overvaluation resulting from so-called “fraud-on-the-market” may form the basis for a plaintiffs damages award in a private securities action. Id. at 341^43, 125 S.Ct. 1627.

In other words, a shareholder’s allegation that he was led to buy stock at a price that was artificially inflated due to fraud does not state a claim for loss. Id. (noting that while “one might say that the inflated purchase price suggests that the misrepresentation ... ‘touches upon’ a later economic loss” that is insufficient because “[t]o ‘touch upon’ a loss is not to cause a loss” (internal citations omitted) (citing 15 U.S.C. § 78u-4(b)(4))).

As a result, it is now clear in civil securities fraud actions that “the complaint must allege that the practices that the plaintiff contends are fraudulent were revealed to the market and caused the resulting losses.” Metzler Inv. GMBH v. Corinthian Colls., Inc., 540 F.3d 1049, 1063 (9th Cir.2008). B. Application of Civil Rule to Criminal Securities Fraud The Supreme Court has not applied its Dura Pharmaceuticals loss causation principle to sentencing enhancements in criminal securities fraud cases, but two federal circuit courts have suggested that they are applicable in this context. In United States v.

Olis, the Fifth Circuit intimated that the civil loss causation principle described in Dura Pharmaceuticals should inform criminal securities fraud sentencing. See Olis, 429 F.3d 540, 546 (5th Cir.2005) (“The civil damage measure should be the backdrop for criminal responsibility both because it furnishes the standard of compensable injury for securities fraud victims and because it is attuned to stock market complexities.”) (citing Dura Pharms., 544 U.S. 336, 341-43, 125 S.Ct. 1627, 161 L.Ed.2d 577).

Oils cited several out-of-circuit cases, including various so-called “cook the books” scenarios, and noted with approval that “each case takes seriously the requirement to correlate the defendant’s sentence with the actual loss caused in the marketplace, exclusive of other sources of stock price decline.” Id. at 547. And in United States v.

Rutkoske, the Second Circuit endorsed the application of Dura Pharmaceuticals’s principle to criminal sentencing even more strongly, stating that: [t]he Government contends that the principles set forth in Dura Pharmaceuticals, a civil case, should not apply to loss calculation in a criminal case. The dicta in [our decision in United States v. Ebbers, 458 F.3d 110, 128 (2d Cir.2006) ] strongly undermines that position.

Moreover, we see no reason why considerations relevant to loss causation in a civil fraud case should not apply, at least as strongly, to a sentencing regime in which the amount of loss caused by a fraud is a critical determinant of the length of a defendant’s sentence. 506 F.3d 170, 179 (2d Cir.2007). This court has not applied Dura Pharmaceuticals’s strict loss causation standard to criminal fraud cases, but we have endorsed a more general loss causation principle, permitting a district court to impose sentencing enhancements only for losses that “resulted from” the defendant’s fraud. United States v.

Hicks, 217 F.3d 1038, 1048 (9th Cir.2000). In Hicks, we stated that “[t]he Guidelines’ ‘relevant conduct’ provision requires a defendant’s sentence to be based on ‘all harm that resulted from the acts or omissions’ of the defendant.” Id. (quoting U.S.S.G. § lB1.3(a)(3) (1995)); id. at 1048-49 (holding that government must show both “but-for” and “proximate” causation in establishing loss). Berger now urges us to take the next step and follow the Second Circuit in expressly applying Dura Pharmaceuticals’s civil principle to criminal securities fraud sentencing. We decline to do so for two reasons.

First, we believe that the primary policy rationale of Dura Pharmaceuticals for proscribing overvaluation as a valid measure of loss does not apply in a criminal sanctions context. Second, application of Dura Pharmaceuticals’s civil rule to criminal sentencing would clash with the parallel principles in the Sentencing Guidelines, which have persuasive value in federal courts. See United States v. Staten, 466 F.3d 708, 710 (9th Cir.2006) (holding that failure to consider Guidelines note in applying sentencing enhancement was reversible error).

As noted, Dura Pha'rmaceuticals rejected the notion that an allegation by a private plaintiff that he purchased securities that were overvalued because of fraud is sufficient to state a damages claim for civil securities fraud. 544 U.S. at 342, 125 S.Ct. 1627 (reversing Broudo v. Dura Pharms., Inc., 339 F.3d 933 (9th Cir.2003)). A key component of the Court’s holding was that “as a matter of pure logic, at the moment the transaction takes place, the plaintiff has suffered no loss; the inflated purchase payment is offset by ownership of a share that at that instant possesses equivalent value.” Id.

Because “[sjhares are normally purchased with an eye toward a later sale!”,] ... if, say, the purchaser sells the shares quickly before the relevant truth begins to leak out, the misrepresentation will not have led to any loss.” Id. Moreover, the Court reasoned, “the common law has long insisted that a plaintiff in such a case show not only that had he known the truth he would not have acted but also that he suffered actual economic loss.” Id. at 343-44, 125 S.Ct. 1627 (citing e.g., Pasley v. Freeman, 100 Eng. Rep. 450, 457 (1789)).

Thus, the Court was concerned principally with the plaintiffs ability to show that he suffered actual loss caused directly — and exclusively — by the defendant’s fraudulent misrepresentation. The Dura Pharmaceuticals Court’s concern is not implicated in the criminal sentencing arena. As demonstrated, in a private civil fraud action, a court gauges loss from the perspective of the plaintiff-victim, i.e., whether the plaintiff can show the amount and cause of loss he sustained. Id.

Because a civil plaintiff bears the burden to show loss, it is logical to require that the plaintiff show that any loss he sustained was attributable directly to devaluation caused by revelation of the defendant’s fraud. It likewise follows that a plaintiffs mere allegation that he purchased overvalued stock is insufficient to state a claim, because the allegation does not by itself establish that the plaintiff personally incurred loss commensurate with the overvaluation.

In criminal sentencing, however, a court gauges the amount of loss caused, i.e., the harm that society as a whole suffered from the defendant’s fraud. See, e.g., Zolp, 479 F.3d at 720. Whether and to what extent a particular individual suffered actual loss is not usually an important consideration in criminal fraud sentencing.

Therefore, where the value of securities have been inflated by a defendant’s fraud, the defendant may have caused aggregate loss to society in the amount of the fraud-induced overvaluation, even if various individual victims’ respective losses cannot be precisely determined or linked to the fraud.

As a result, the principle underlying the Dura Pharmaceuticals Court’s reluctance to allow mere overvaluation as a basis for establishing loss is generally not present in the criminal sentencing context, and we are not persuaded that it would be appropriate to expand the Dura Pharmaceuticals rule to the criminal sentencing context. The Sentencing Guidelines provide further support for limiting the scope of Dura Pharmaceuticals’s loss causation rule in a criminal sentencing context. In arguing for this

interpretation, the government cites the commentary to the 1995 Guidelines (the version applied at Berger’s sentencing), specifically its endorsement of a flexible approach to loss calculation in criminal sentencing. E.g., Zolp, 479 F.3d at 718-19. The government notes that § 2F1.1 commentary note 8 of the 1995 Guidelines states that: The court need only make a reasonable estimate of the loss, given the available information.

This estimate, for example, may be based on the approximate number of victims and an estimate of the average loss to each victim, or on more general factors, such as the nature and duration of the fraud and the revenues generated by similar operation. The offender’s gain from committing the fraud is an alternative estimate that ordinarily will underestimate the loss. In addition, the government contends that § 2F1.1 condones measuring loss by overvaluation. See U.S.S.G. § 2F1.1, cmt. n.7(a) (1995).

That note states: [a] fraud may involve the misrepresentation of the value of an item that does have some value (in contrast to an item that is worthless). Where, for example, a defendant fraudulently represents that stock is worth $40,000 and the stock is worth only $10,000, the loss is the amount by which the stock was overvalued (i.e., $30,000).

Thus, were Dura Pharmaceuticals’s loss causation rule applied to criminal sentencing enhancements, that principle’s plain rejection of the overvaluation loss measurement method, see 544 U.S. at 343, 125 S.Ct. 1627, would collide with Congress’s clear endorsement of that method, see U.S.S.G. § 2F1.1, cmt. n.7(a). For these reasons, we decline to require, in finding facts relevant to sentencing, a showing that “share price fell significantly after the truth became known.” Dura Pharms., 544 U.S. at 347, 125 S.Ct. 1627.

We instead reiterate our broader rule that “[t]he Guidelines’ ‘relevant conduct’ provision requires a defendant’s sentence to be based on ‘all harm that resulted from the acts or omissions’ of the defendant.” Hicks, 217 F.3d at 1048 (quoting U.S.S.G. § lB1.3(a)(3) (1995)). C.

The District Court’s Loss Valuation Approach While the district court was not required to follow Dura Pharmaceuticals’s loss causation approach, the loss-calculation method it did employ troubles us; it leaves us with little confidence that the government demonstrated, by the applicable standard of proof, that shareholder loss occurred, let alone that approximately $2.1 million of loss occurred.

Though the Guidelines state that courts may employ various methodologies to determine loss and that loss need not be established with precision, the fact that “[t]he court need only make a reasonable estimate of the loss,” U.S.S.G. § 2B1.1, cmt. n. 3, § 2F1.1, cmt. n. 8, does not obviate the requirement to show that actual, defendant-caused loss occurred. Rather, the plain language of the Guidelines commentary merely indicates that, in arriving at the loss figure, some degree of uncertainty is tolerable.

First, the Guidelines’ statement that the “estimate [of loss] ... may be based on the approximate number of victims and an estimate of the average loss to each victim,” U.S.S.G. § 2F1.1, cmt. n. 8, presupposes that the court has already determined that some defendant-caused loss occurred. Indeed, without any loss to victims, there would be nothing on which to base an estimate.

In the same way, the fact that the loss estimate “may be based on ... general factors, such as the nature and duration of the fraud and the revenues generated by similar operation,” id., or on the “offender’s gain from committing the fraud,” id., does not suggest that a court is relieved of the duty to determine that some loss actually occurred. Even the overvaluation method example in § 2F1.1 commentary note 7(

a) does not suggest that a showing of actual loss is unnecessary. That illustration provides a model for calculating the amount of loss where fraud caused the value of stock to decrease, but where the stock retained residual value. The example assumes that the stockholders were left holding stock that depreciated because of the fraud. In sum, each of these possible methodologies assumes that some loss was proximately caused by the defendant, while recognizing that the amount of loss may not be easily measurable.

In determining that the shareholder loss was $2.1 million in this case, the district court employed a eounterfactual approach. The method examined the effect on the stock value of other, unrelated companies after accounting irregularities were disclosed to the market. Using that method, the court determined that the average depreciation in value was 26.5%. That figure was applied to the value of Craig’s initial public offering. The court’s method appears to have assumed that defendant-caused shareholder loss existed, and only then purported to measure that loss.

Moreover, that measure of loss was not based on Craig’s finances or on the actual effect of Berger’s fraud, but rather on data from other companies in previous years and different economic conditions. More importantly, it was based on cases in which there had been disclosure of accounting irregularities to the market, despite the fact that Craig’s accounting irregularities were never disclosed while its stock was still publicly traded.

As a result, because the method did not properly establish that Berger’s sentence was based only on “ ‘all harm that resulted from the acts or omissions’ of the defendant,” Hicks, 217 F.3d at 1048 (quoting U.S.S.G. § lB1.3(a)(3) (1995)), it was an abuse of discretion. We therefore remand to the district court to redetermine, based on the principles described herein, how much of the shareholders’ loss was actually caused by Berger’s fraud.

While we do not dictate the exact method the district court must use, we note that whatever method is chosen should attempt to gauge the difference between Craig’s share price — as inflated through fraudulent representation — and what that price would have been absent the misrepresentation. II. Standard of Proof in Finding Sentencing Facts Berger next argues that, in circumstances such as these, due process requires the district court to find the loss amount by clear and convincing evidence, rather than by a preponderance of the evidence. A.

Nature of Dispute Maintaining that preponderance of the evidence is the proper standard, the government first argues that Berger challenged only the methodology of the district court’s determination, which is a purely legal dispute. See, e.g., United States v. Hardy, 289 F.3d 608, 613 (9th Cir.2002). The government asserts that where there is no factual dispute, but only a legal dispute, preponderance of the evidence is the appropriate standard of proof for the district court to use in determining facts relevant to sentencing. In support, the government cites United States v.

Romero-Rendon, 198 F.3d 745, 748 (9th Cir.1999), withdrawn by 220 F.3d 1159, 1165 (9th Cir.2000). This argument fails for two reasons. First, contrary to the government’s characterization, Romero-Rendon declined to address the appropriate standard of proof; instead, the court held that the unchallenged pre-sentence report constituted clear and convincing evidence of the critical predicate fact. See id. at 1165. Here, however, Berger vigorously challenged the loss calculations before the district court.

Second, although the parties disputed the validity of the district court’s loss calculation, making such a calculation necessarily involved the district court’s determination of how much loss was suffered, an issue we have held to be one of fact. See, e.g., United States v. Garro, 517 F.3d 1163, 1167 (9th Cir.2008) (citing United States v. Lawrence, 189 F.3d 838, 844 (9th Cir.1999)) (holding that “[a] calculation of the amount of loss is a factual finding”). Thus, by challenging the loss amount, Berger raised both a factual and legal dispute.

We proceed to consider what standard of proof the district court should have employed in resolving the factual dispute. We review the issue de novo because it, like the question regarding calculation method, is a question of law. B. Standard of Proof for Finding Sentence-Enhancing Facts A district court typically uses a preponderance of the evidence standard when finding facts pertinent to sentencing. United States v. Armstead, 552 F.3d 769, 776 (9th Cir.2008) (citing United States v. Moreland, 509 F.3d 1201, 1220 (9th Cir.2007)). In United States v.

Restrepo, how ever, we concluded that, “when a sentencing factor has an extremely disproportionate effect on the sentence relative to the offense of conviction,” due process may require a district court to apply a heightened standard. 946 F.2d 654, 659-60 (9th Cir.1991) (en banc) (emphasis added) (citing McMillan v. Pennsylvania, 477 U.S. 79, 87-91, 106 S.Ct. 2411, 91 L.Ed.2d 67 (1986), for the proposition that disproportionate effect may require the application of the heightened standard but concluding that such an effect was not relevant in that case).

Since Restrepo, we have not been a model of clarity in deciding what analytical framework to employ when determining whether a disproportionate effect on sentencing may require the application of a heightened standard of proof. Some of our cases have explicitly stated that where the sentencing enhancements are based on charged conduct, i.e., the “offense of conviction,” employing a preponderance of the evidence standard does not implicate Restrepo’s due process concerns. See, e.g., United States v. Harrison-Philpot, 978 F.2d 1520, 1524 (9th Cir.1992).

We have also held that there is no “bright-line rule for the disproportionate impact test,” and instead look to the “ ‘totality of the circumstances,’ without considering any one factor as dispositive.” See, e.g., United States v. Jordan, 256 F.3d 922, 928 (9th Cir.2001) (citing United States v. Valensia, 222 F.3d 1173, 1182 (9th Cir.2000), cert. granted, judgment vacated, and remanded by, 532 U.S. 901, 121 S.Ct. 1222, 149 L.Ed.2d 133 (2001) (reversing and remanding on other grounds)).

We have followed the “totality of the circumstances” approach even where the enhancement was based on the “offense of conviction.” Johansson, 249 F.3d at 853-58. Berger argues that our case law is irreconcilable, or premised on an indefensible distinction between an enhancement based on the “offense of conviction” and one based on “uncharged” or “acquitted” conduct.

He asserts that we have not always been diligent in maintaining this distinction, or even if we have, there is no basis for holding that facts relating to uncharged or acquitted conduct are subject to a heightened standard of proof, while facts relating to the offense of conviction are not. We decline Berger’s invitation to decide this case on such broad grounds, because a number of our cases squarely address the factual situation presented here. Those cases involve a defendant’s fraudulent conduct where sentencing enhancements for financial loss are based on the extent of the fraud conspiracy.

They hold that facts underlying the disputed enhancements need only be found by a preponderance of the evidence. United States v. Riley, 335 F.3d 919, 926-27 (9th Cir.2003); Armstead, 552 F.3d at 777-78; Garro, 517 F.3d at 1168-69. In Riley, the defendant pled guilty to one count of conspiracy to produce fictitious obligations, one count of possession of fictitious obligations, and one count of identification fraud. 335 F.3d at 923.

At sentencing, the district court applied a nine-level enhancement to the defendant’s sentence under § 2Fl.l(b)(l)(J), for a financial loss greater than $350,000 but less than or equal to $500,000. Id. We held that because this enhancement was “based entirely on the extent of the conspiracy to which Riley pled guilty,” it was properly determined based on a preponderance of the evidence. Id. at 926-27. In Armstead, a jury convicted the defendant of nine counts of bank fraud and one count of conspiracy to commit bank fraud. 552 F.3d at 774.

The district court applied a fourteen-level enhancement for the amount of loss under § 2B1.1(b)(1). Id. at 775-76. The defendant disputed $173,576.88 of the total loss, which represented the difference between a twelve-level and a fourteen-level increase. Id. at 777 n. 6. We held that the district court did not need to employ a clear and convincing standard of proof to the two-level enhancement because the loss amount was attributable to the extent of the conspiracy. Id. at 777-78.

Finally, in Garro, the defendant was convicted of eight counts of wire fraud, eleven counts of money laundering, and one count of tax evasion. 517 F.3d at 1165. The district court found that the amount of loss in the defendant’s fraudulent scheme exceeded $20 million, resulting in a sixteen-level sentencing enhancement. Id. at 1167. Again, citing Riley, we held that the district court’s enhancement was based on conduct for which the defendant had been charged and convicted, and therefore the district court properly used a preponderance of the evidence standard of proof. Id. at 1169.

Here, like in Riley, Armstead, and Garro, the enhancement under U.S.S.G. § 2F1.1 is based entirely on the extent of the fraud conspiracy for which Berger was convicted. See Riley, 335 F.3d at 926. Applying the holdings of Riley, Armstead, and Garro to this case, we conclude that the district court did not err in using a preponderance of the evidence standard to determine the amount of loss for purposes of § 2F1.1.

CONCLUSION For the reasons described, we vacate Berger’s sentence, affirm the district court’s ruling on the applicable standard of proof in finding sentence-enhancing facts in this context, and remand to the district court for resentencing before a new district judge consistent with the standards articulated in this opinion. AFFIRMED in part, REVERSED in part, VACATED and REMANDED. . Richardson, Craig’s Chief Financial Officer until May 31, 1997, pled guilty to three counts of the indictment prior to trial. Metz was at various times a Vice President in Craig’s Hong Kong and Cerritos, California locations. .

Our prior decision in this case provides a more detailed description of the scheme. See United States v. Berger, 473 F.3d 1080, 1083-85 (9th Cir.2007). . It is unclear the extent to which this decline resulted from the restated earnings, as opposed to unrelated external market forces or other factors. . The district court used the 1995 version of the Sentencing Guidelines to avoid creating a potential ex post facto problem. . But see United States v. Williamson, 439 F.3d 1125, 1137 n. 12 (9th Cir.2006) ("We review ... application of the Guidelines de novo.” (emphasis added)).

Because we conclude that the district court’s application of the Guidelines in calculating loss was erroneous under either de novo or abuse of discretion review, we do not attempt to resolve this conflict in our case law. . The government in this case concedes that, "in order to be a basis for an increase in base offense level under the guidelines, the losses from defendant’s securities fraud offenses must have resulted from those offenses.” .

We note that, based on this reasoning, Dura Pharmaceuticals may be more relevant in the context of criminal restitution under, for instance, the Mandatory Victims Restitution Act of 1996 (MVRA), 18 U.S.C. § 3663A, which, unlike the sentencing enhancement scheme, focuses on harm to the victims as opposed to loss caused by the defendant. See, e.g., Berger, 473 F.3d at 1104 ("The MVRA requires a defendant to pay restitution to a victim who is 'directly and proximately harmed as a result of die fraud.”) (quoting 18 U.S.C. § 3663A(a)(2)). .

The Guidelines, including those for enhancements purposes, "are ordinarily applied in light of available commentary, including application notes.” Staten, 466 F.3d at 715 (citing United States v. Allen, 434 F.3d 1166, 1173 (9th Cir.2006) ("The application notes to the Guidelines are exactly that — notes about when a particular Guideline applies and when it does not.”)). .

Section 2F1.1 was repealed in 2001. . Similarly, the government points out that § 2B1.1 commentary note 3 provides that: The court need only make a reasonable estimate of the loss, given the available information. This estimate, for example, may be based upon the approximate number of victims and the average loss to each victim, or on more general factors such as the scope and duration of the offense. . As we discuss below, the standard in this case should be preponderance of the evidence. .

In concluding that the district court’s method was erroneous, we do not suggest that Berger’s fraud caused no loss to investors. The district court found that Craig’s spring 1997 stock value decline was "unrelated to Berger's criminal conduct" because it resulted not from disclosure of fraud, but from disclosure of the company's poor financial status. But that conclusion is valid only in the narrowest sense.

While revelation of Berger’s fraud to the public did not depreciate Craig’s stock value (as Craig was no longer publicly traded by that point), it appears that the stock value was overvalued, at least in part, because of the fraud. As a result, many investors were likely induced to buy Craig stock at its IPO price under the false pretenses created by Berger and his cohorts.

Had Craig’s financial troubles not been masked by fraud during the IPO, then surely (assuming the IPO had happened at all) Craig’s stock price would already have been significantly lower in spring 1997, Craig's earnings would not have required restatement, and the stock value would not have plummeted. . The government cites only a withdrawn version of Romero-Rendon, 198 F.3d at 748, withdrawn by 220 F.3d 1159. The government maintains that Romero-Rendon states that preponderance of the evidence is an adequate standard where the defendant did not challenge the accuracy of the sentencing report.

The superceding version of our decision in that case, however, did not so hold. . Berger's reliance on Staten and Zolp is misplaced. Neither Staten nor Zolp focused on the standard of proof, and therefore neither calls into question our holdings in Riley, Armstead, or Garro. See United States v. Johnson, 256 F.3d 895, 915 (9th Cir.2001) (en banc) (holding that where a statement is "merely a prelude to another legal issue that commands the panel's full attention,” it is not binding on later panels).

Rather, Staten considered the viability of the "disproportionate impact” rule following the Supreme Court's decision in Booker, Staten, 466 F.3d at 717-20, and Zolp addressed a district court's factual finding that the stock at issue was worthless after the defendant's fraud came to light, Zolp, 479 F.3d at 718-21.

While each noted that the clear and convincing standard of proof applies " 'where an extremely disproportionate sentence results from the application of an enhancement,' ” Id. at 718 (quoting Staten, 466 F.3d at 717), that statement was "merely a prelude to another legal issue that commandfed] the panel’s full attention,” Johnson, 256 F.3d at 915. To the contrary, Riley specifically held that the preponderance standard applied to a sentencing enhancement for the amount of loss under U.S.S.G. § 2F1.1. Riley, 335 F.3d at 926.

We note also that Staten and Zolp were both decided by three-judge panels, and "a three-judge panel may not overrule a prior decision of the court.” Miller v. Gammie, 335 F.3d 889, 899 (9th Cir.2003) (en banc). . The district judge in this case, the Honorable Robert M. Takasugi, passed away on August 4, 2009.