UNITED STATES of America, Plaintiff-Appellee, v. Jamie OLIS, Defendant-Appellant, 429 F.3d 540

EDITH H. JONES, Circuit Judge: Jamie Olis appeals from a judgment of conviction for which he was sentenced to 292 months in prison for securities fraud, mail and wire fraud, and conspiracy. The charges arose from Olis’s work as a tax lawyer and accountant at Dynegy Corporation (“Dynegy”) on a transaction called “Project Alpha.” Olis argues that the evidence was insufficient for conviction and that the district court improperly calculated his sentence. We hold that the conviction is factually supported, but Olis must be resentenced.

Olis sufficiently preserved a Booker challenge to the court’s application of the sentencing guidelines as a mandatory scheme, and the district court overstated the loss caused by Olis’s crimes. We therefore AFFIRM the conviction, and VACATE and REMAND for resentencing. I. BACKGROUND The conviction arises from Olis’s position as Senior Director of Tax Planning and International (and later, Vice President of Finance) at Dynegy on a transaction called “Project Alpha,” a complex five-year deal involving natural gas transactions.

Project Alpha was a plan to borrow $300 million and make it appear to the outside world (and in particular to Dynegy’s auditor Arthur Andersen) as if the money was generated by Dynegy’s business operations. Project Alpha was designed to generate positive cash flow to Dynegy “from operations” during 2001 and negative cash-flow in 2002-05. Specifically, a special purpose entity (“SPE”) called ABG Gas Supply was created and owned by Deutsche Bank and Credit Suisse. During 2001, ABG Gas bought natural gas at market prices and sold it to Dynegy at a discount.

Dynegy then sold the gas at market prices, netting $300 million. During 2002-05, Project Alpha arranged that ABG Gas would buy gas at market prices and resell it to Dynegy at above-market prices. That money would flow to the banks, which would recoup the $300 million, plus interest. To support the accounting characterization of the deal as cash from operations, ABG Gas and the lenders could not be guaranteed full repayment on their investment. Further, ABG Gas had to be sufficiently “independent” from Dynegy, and the owners of ABG Gas had to bear risk.

But contrary to these requirements, Olis, his boss Gene Foster and his colleague Helen Sharkey, secretly put into place the “parent level” hedge and the “tear-up” agreements among Dynegy, ABG’s owner banks, and Citibank to ensure that the banks would not lose any money. The Government’s proof indicated that Olis, Foster, and Sharkey intentionally concealed the parent level hedge and tear-ups from Jim Hecker, the Arthur Andersen partner responsible for signing off on Dy-negy’s SEC statements, in order to obtain the desired accounting treatment of the transaction.

On April 25, 2002, following its review of Project Alpha, the SEC required Dyne-gy to restate the cash flow as derived from a “financing” rather than “operations.” Because Dynegy was now seen to be borrowing rather than earning money from Project Alpha, Dynegy’s stock price was adversely affected. Foster, Sharkey, and Olis were indicted for conspiracy to commit mail fraud, wire fraud, and securities fraud (count 1), securities fraud (count 2), mail fraud (count 3) and wire fraud (counts 4-6). Foster and Sharkey pled guilty to one count each in exchange for maximum sentences of five years.

Foster testified against Olis at trial. The jury convicted Olis on all counts. The district court sentenced Olis, applying the Sentencing Guidelines as mandatory, to 292 months in prison, three years supervised release, and a $25,000 fine. The offense level was extraordinarily high based on the court’s findings that the fraudulent scheme caused a loss of $105 million to one shareholder, the University of California Retirement System (“UCRS”); that Olis employed “sophisticated means” and a “special skill” to carry out the fraud; and that there were more than fifty victims of the fraud. Olis has appealed. II.

SUFFICIENCY OF EVIDENCE Olis contends, almost perfunctorily, that the evidence does not support his conviction. In particular, he disputes the proof that he conspired to conceal two critical features of Project Alpha from Dynegy’s outside auditor Arthur Andersen— the “parent level” hedge and the “tear-up” agreements. This court will not disturb a jury’s verdict unless the record demonstrates that a rational jury could not have found each of the elements of the offense beyond a reasonable doubt. United States v. Dahlstrom, 180 F.3d 677, 684 (5th Cir. 1999).

The evidence, and all inferences reasonably drawn from it, must be viewed in the light most favorable to the verdict, regardless whether the conviction is based on direct or circumstantial evidence. Id. Olis asserts that the evidence demonstrated that everyone working on Project Alpha, including Arthur Andersen accountants, knew that the bank owners of ABG Gas were fully hedged against the risk of loss from variable gas prices.

Olis’s boss Foster, testified, however, as a star prosecution witness and co-indictee, that he and Olis wrongly agreed to the tear-ups and the parent hedge and hid them from Arthur Andersen. Jim Hecker, an audit partner at Arthur Andersen, testified that he advised Dynegy against tear-ups, and Dynegy subsequently did not reveal this aspect of Project Alpha to him.

A reasonable jury, basing its conclusion on the testimony of Foster and Hecker, together with the incriminating emails among Olis and his co-indictees and a wealth of other evidence, could easily have found Olis guilty beyond a reasonable doubt of all the charged crimes. III. SENTENCING Far more problematic are some of the issues Olis raises concerning his Booker objection, the district court’s use of the 2001 version of the Sentencing Guidelines, and the reasonableness of the district court’s loss calculation, all of which contributed to Olis’s sentence of imprisonment. We address each in turn.

A Booker Objection Olis first argues that under Booker, his Sixth Amendment right to a jury trial was violated because the district court enhanced his sentence under the mandatory guidelines regime based on facts not proved to the jury beyond a reasonable doubt. See United States v. Booker, 543 U.S. 220, 125 S.Ct. 738, 756, 160 L.Ed.2d 621 (2005)(“any fact (other than a prior conviction) which is necessary to support a sentence exceeding the maximum authorized by the facts established by a plea of guilty or a jury verdict must be admitted by the defendant or proved to the jury beyond a reasonable doubt.”).

During sentencing, the district court determined the following facts:

(1) Olis was responsible for an approximately $105 million loss to UCRS, which enhanced his base offense by twenty-six levels under the Sentencing Guidelines;

(2) Olis’s offense involved sophisticated means, requiring a two-level enhancement;

(3) Olis used a special skill, in a manner that significantly facilitated the commission or concealment of the offense, resulting in another two-level enhancement; and

(4) Olis’s scheme included fifty or more victims, requiring a four-level sentencing enhancement. None of these findings was proven beyond a reasonable doubt to the jury or admitted by Olis. Relying on these judge-found facts, and as mandated by the Guidelines, the court calculated Olis’s total offense level to be 40. Olis had no criminal history for guidelines purposes. These two determinations yielded a sentencing range of 292 to 365 months in prison. See U.S.S.G. Ch. 5 Pt. A.

The district court, noting that it was “required to follow ... the Federal Sentencing Guidelines,” stated that it took “no pleasure in sentencing [Olis] to 292 months,” but that it was the court’s job “to follow the law.” The district court’s findings on the enhancements dramatically increased Olis’s sentencing range beyond the minimum span permitted by the jury’s verdict. The Government asserts, however, that Olis did not properly preserve his Booker objection and that we should review Olis’s sentencing points for plain error. We disagree.

Olis repeatedly objected before and during his sentencing hearing to both the district court’s loss calculation and the burden of proof utilized by the court. His objections regarding the loss calculation alerted the court to cases that acknowledged the potential for a constitutional violation when sentencing facts are not found by at least clear and convincing evidence. Olis’s objections were overruled and there is nothing to indicate that the district court made its findings on any basis other than a preponderance of the evidence. In United States v.

Akpan, 407 F.3d 360, 375-76 (5th Cir. 2005), this court held that although one defendant “never explicitly mentioned the Sixth Amendment, Apprendi, or Blakely until his Rule 28(

j) letter,” his objections during sentencing that the court’s loss calculation had not been proven at trial adequately apprised the district court of a Sixth Amendment objection. Although Olis, like the defendant in Akpan, never explicitly mentioned the Sixth Amendment, Apprendi, or Blakely, his repeated objections also adequately apprised the court that he was raising a constitutional error with respect to the loss calculation. Because Olis preserved his error by objecting in the district court, we must “ ‘vacate the sentence and remand, unless we can say the error is harmless under Rule 52(

a) of the Federal Rules of Criminal Procedure.’ ” Akpan, 407 F.3d at 376 (citation omitted). We review the record de novo to determine whether the district court’s error was harmless. United States v. Ahmed, 324 F.3d 368, 374 (5th Cir.2003). Further, under Fed. Rule Crim. Proc. 52(a), the Government bears the burden of showing harmless error by “demonstrating beyond a reasonable doubt that the federal constitutional error of which a defendant complains did not contribute to the sentence that he received.” Akpan, 407 F.3d at 377.

In this case, the Government points to no evidence proving beyond a reasonable doubt that the district court would have sentenced Olis to nearly twenty-five years in prison had it acted under an advisory Sentencing Guidelines scheme as required by Booker. Therefore, we vacate Olis’s sentence and remand for resentencing. As Booker requires the district court to “consider” the guidelines before issuing a “reasonable” sentence, 125 S.Ct. at 757, 767, we must review the specific sentencing issues that have arisen in this case and provide an analytical framework to aid the district court in resentencing. B.

The 2001 Sentencing Guidelines Olis contends that the district court erred by using the 2001 version of the Sentencing Guidelines, rather than the 2000 version, to calculate his sentence. Courts are required to “use the Guidelines Manual in effect on the date that the offense of conviction was committed.” U.S.S.G. § 1B1.11(b)(1). The guidelines add, “If a defendant is convicted of two offenses, one before and one after the effective date of the revised edition of the guidelines, the revised edition applies to both offenses.” U.S.S.G. § 1B1.11(b)(3).

The jury convicted Olis for crimes that were committed during the years 2000 through early 2002. The jury found in Count 2 that the securities fraud in which Olis participated was coextensive with the conspiracy, and in Count 6 that Olis committed or aided and abetted wire fraud on March 13, 2002, the date that Dynegy’s 2001 Form 10-K was electronically filed with the SEC. The dates of these offenses plainly fall after the effective date of the November 2001 revisions. The jury additionally convicted Olis of the Count 1 conspiracy that lasted from on or about August 2000 through April 2002.

This court has held that conspiracy “is a continuing offense” and that “[s]o long as there is evidence that the conspiracy continued after the effective date of the [amendments to the] guidelines, the Ex Post Facto Clause is not violated.” United States v. Buckhalter, 986 F.2d 875, 880 (5th Cir.1993). Moreover, unless a conspirator effectively withdraws from the conspiracy, he is to be sentenced under the amendments to the guidelines, even if he did not commit

an act in furtherance of the conspiracy after the date of the new guidelines, or did not know of acts committed by other co-conspirators after the date of the new guidelines, where it was foreseeable that the conspiracy would continue past the effective date of the amendments. United States v. Devine, 934 F.2d 1325, 1332 (5th Cir.1991). Devine applied to a defendant whose criminal conspiracy straddled the period before and after the effective date of the Guidelines, but its reasoning also applies where, as here, the conspiracy continued into the period covered by revised Guidelines.

Olis never withdrew from the conspiracy. Because three of the counts are governed by the 2001 amendments to the guidelines, the other three counts (mail fraud committed in April 2001 and wire fraud committed in August and October 2001) are also controlled by the November 2001 amendments. C. Loss Calculation under the Sentencing Guidelines The most significant determinant of Olis’s sentence is the guidelines loss calculation. By the district court’s reasoning, this added twenty-six levels to his base offense level and alone placed Olis in a punishment range exceeding fifteen years’ imprisonment.

This court reviews a district court’s factual findings at sentencing for clear error and its legal analysis de novo. Nixon v. Epps, 405 F.3d 318, 322 (5th Cir.2005). While the district court need only make a “reasonable estimate of loss,” U.S.S.G. § 2B1.1 app. note to (c)(2002), this court first determines whether the trial court’s method of calculating the amount of loss was legally acceptable. United States v. Saacks, 131 F.3d 540, 542-43 (5th Cir. 1997); United States v. Krenning, 93 F.3d 1257, 1269 (5th Cir.1996).

Although otherwise amended in 2001, the guideline covering securities fraud has continuously provided that a sentencing court should use the greater of actual or intended loss. § 2B1.1, cmt. n 2.2(a) (2001). The guidelines measure criminal culpability in theft and economic crimes according to their pecuniary impact on victims. Actual loss, which is at issue here, “means the reasonably foreseeable pecuniary harm that resulted from the offense.” § 2B1.1, cmt. n. 2(a)(i).

Moreover, actual loss “incorporates [a] causation standard that, at a minimum, requires factual causation (often called ‘but for’ causation) and provides a rule for legal causation (i.e., guidance to courts regarding how to draw the line as to what losses should be included and excluded from the loss determination).” U.S.S.G. Supp 2 App. C, Amendment 617 (November 1, 2001). This explanation does not, contrary to the Government’s argument in brief, lessen the preexisting standards that held a defendant responsible at sentencing only to the extent that losses are caused directly by the offense conduct.

See, e.g., United States v. Hicks, 217 F.3d 1038, 1048-49 (9th Cir.2000); United States v. Marlatt, 24 F.3d 1005, 1007 (7th Cir.1994) (“[there is a] difference between ‘but for’ causation and the causation — for which the presence of but-for causation is ordinarily a necessary condition but rarely a sufficient one— that imposes legal liability. The distinction runs throughout the law. Criminal law is no exception”). District courts must take a “realistic, economic approach to determine what losses the defendant truly caused or intended to cause.” United States v.

West Coast Aluminum Heat Treating Co., 265 F.3d 986, 991 (9th Cir. 2001). The loss guideline is skeletal because it covers dozens of federal property crimes. Some flesh can be added, however, where the gravamen of the offense conduct is securities fraud perpetrated on an established market. Useful guidance appears in the applicable principles for recovery of civil damages for securities fraud. The civil damage measure should be the backdrop for criminal responsibility both because it furnishes the standard of compen-sable injury for securities fraud victims and because it is attuned to stock market complexities.

In civil eases, the principle of loss causation is well established. See Dura Pharmaceuticals, Inc. v. Broudo, — U.S. -, 125 S.Ct. 1627, 1631-32, 161 L.Ed.2d 577 (2005); see generally Greenberg v. Crossroads Systems, Inc., 364 F.3d 657 (5th Cir.2004); Private Securities Litigation Reform Act, 15 U.S.C. § 78u-4(b). Thus, there is no loss attributable to a misrepresentation unless and until the truth is subsequently revealed and the price of the stock accordingly declines.

Where the value of a security declines for other reasons, however, such decline, or component of the decline, is not a “loss” attributable to the misrepresentation. See also United States v. Grabske, 260 F.Supp.2d 866, 869-71 (N.D.Cal.2002). Although cases applying the guidelines to securities fraud convictions at first blush yield no consistent rule analogizing criminal responsibility with civil “loss causation,” disparity is often more apparent than real.

In cases where defendants promoted worthless stock in worthless companies, measuring the loss as the entire amount raised by the schemes is neither surprising nor complex, and is fully consistent with civil loss causation. See, e.g., United States v. Hedges, 175 F.3d 1312, 1314-15 n. 6 (11th Cir.1999)(noting that the corporation’s “stock became worthless when the conspiracy was discovered”). A few cases appear to rely on a “market capitalization” approach, basing loss on a gross correlation between stock price decline and the revelation of a fraudulent transaction. See, e.g., United States v.

Eyman, 313 F.3d 741 (2d Cir.2002); United States v. Moskowitz, 215 F.3d 265 (2d Cir.2000). Because the courts’ underlying reasoning is sparse and supporting facts are few, their methodology, which might be at odds with the greater precision required in civil loss causation, is unenlightening. The final type of case, most analogous to the one before us, concerns fraudulent transactions that “cook the books” and prop up a company’s stock but do not, aside from the exceptional Enron or WorldCom situation, render the company worthless.

Sentencing decisions in these cases acknowledge that because a company’s stock price is affected before and after the fraud, by numerous extrinsic market influences as well as the soundness of other business decisions by the company; the calculation of loss attributable to securities fraud requires careful analysis. See, e.g., United States v. Snyder, 291 F.3d 1291 (11th Cir.2002); United States v. Bakhit, 218 F.Supp.2d 1232, 1238 (C.D.Cal.2002); Grabske, 260 F.Supp.2d at 869-71.

Each of these eases utilized or recommended somewhat different approaches to estimating reasonably the amount of loss inflicted by a defendant’s securities fraud committed within an extant company. Nevertheless, 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. Several features of the decisions are noteworthy.

First, given the time and evidentiary constraints on the sentencing process, the methods adopted in these cases are necessarily less exact than the measure of damage applicable in civil securities litigation. Second, Snyder and Bakhit rejected an oversimplified market capitalization measure of damages proffered by the Government in favor of a more nuanced approach modeled upon loss causation principles. See Snyder, 291 F.3d at 1295-96; Bakhit, 218 F.Supp.2d at 1238.

As the Bakhit court noted, the Government’s use of stock prices the day before and the day after the revelation of the fraud did not account either for the actual price at which most holders purchased the company’s shares, or for the influence of outside factors on the change in price. Bakhit, 218 F.Supp.2d at 1239. The Government’s approach measured paper losses in the company’s value, which have no correlation with losses to actual shareholders who bought or sold based on fraudulent information. Third, the cases rejected defendants’ arguments that attempted to reason away all losses caused by the fraud.

Finally, the factual variations among these cases reflect the importance of thorough analyses grounded in economic reality. In this case, the district court, faced with a “cook the books” fraud, overemphasized his discretion as factfinder at the expense of economic analysis. Thus, the court elected to rely solely on the Heil testimony concerning the purchase and sale of UCRS stock as a measure of the loss caused by Olis’s offense. When Heil’s testimony was offered at trial to prove guilt, Olis’s counsel was not placed on notice that the same evidence might later pertain to the guidelines loss calculation.

For that reason, other significant extrinsic causes of the UCRS loss were not explored, much less quantified, at trial. UCRS bought most of its Dynegy holdings at the top of the market. As Olis pointed out at sentencing, however, two-thirds of the drop in Dynegy’s price occurred either before the revelation of Project Alpha’s problems or more than a week after the announcement of the restatement of earnings caused by Project Alpha.

Taken on the court’s own terms, a substantial portion of the entire loss on the UCRS investment in Dynegy, over $100 million, could not have been caused by Olis’s work on Project Alpha. During sentencing, moreover, Olis offered the expert report of a Rice University expert, Professor Bala Dharan, which explored numerous forces at work on the Dynegy stock price during the relevant periods. The court refused to consider the report, criticizing the expert’s analysis of whether Olis could have “reasonably foreseen” the impact of his conduct on the stock market.

As the court observed, the economist was arguably stretching his expertise into an improper legal conclusion, but his statements on this matter are separate from his economic analysis of price and market movements. Professor Dhar-an’s report demonstrates that Dynegy stock declined during the period covering Project Alpha in tandem with the stocks of other publicly traded companies in the energy marketing and trading business. Further, Dynegy’s stock was negatively affected, even before the restatement of Project Alpha’s cash flow impact, by the company’s failed bid to acquire the faltering Enron.

These factors and others cited in the report suggested that attributing to Olis the entire stock market decline suffered by one large or multiple small shareholders of Dynegy would greatly overstate his personal criminal culpability. Because the district court’s approach to the loss calculation did not take into account the impact of extrinsic factors on Dynegy’s stock price decline, Olis is entitled to resentencing on this factor, subject to the principles just discussed. D.

Other Sentencing Issues Olis challenges, as duplicative, additional sentencing enhancements for the use of “sophisticated means”, U.S.S.G. § 2B1.1(b)(8)(C), and “special skill”, U.S.S.G. § 3B1.3. This court has held that double-counting is not impermissible unless the guidelines so state. United States v. Calbat, 266 F.3d 358, 363 (5th Cir.2001). These guidelines do not proscribe their joint application. Further, the district court properly cited a similar case, United States v. Minneman, 143 F.3d 274, 283 (7th Cir.1998), which found no impermissible double counting.

Olis used his special skills in accounting and tax matters to advance an extremely sophisticated, but fraudulent, scheme. The final 4-level enhancement against Olis was based on there being more than 50 victims of his crime, U.S.S.G. § 2B1.1 (b)(2)(B). The probation office counted as victims nearly all 140,000 employees of UCRS. This is a questionable assessment, as pension plans attempt to balance gains and losses to their beneficiaries, rendering any impact upon UCRS plan members far more attenuated than if they individually owned Dynegy shares.

It is, however, inconceivable that fewer than fifty shareholders of Dynegy suffered a market loss from purchases or sales of stock caused by Olis’s fraud. This enhancement may unduly skew the guidelines range in “cook the books” securities frauds, but it clearly applies. See Coffee, supra at 246-47. Conclusion For these reasons, Olis’s conviction is affirmed, but he must be resentenced in accordance with Booker’s overall standard of reasonableness after the court “considers” the guidelines including a recalculation of the amount of loss for which Olis should be held responsible.

CONVICTION AFFIRMED; SENTENCE VACATED AND REMANDED. . We recite the facts in the light most favorable to the verdict. The description of the transaction at issue is extremely simplified for the sake of brevity. . Foster and Sharkey are set to be sentenced thirty days after the decision in this appeal. . We do not dwell on the elements of each and every count, because Olis's brief neither argues nor supports such a level of detail. .

In addition to arguing insufficiency of the evidence, Olis requests this court to order a new trial because the evidence preponderates heavily against the verdict such that a mistake was made. Olis concedes that no motion for a new trial was filed in the district court. A district court "is powerless to order a new trial except on the motion of the defendant.” United States v. Brown, 587 F.2d 187, 189 (5th Cir.1979)(citing Fed. R.Crim. P. 33). Olis cannot demonstrate that the district court erred in failing to grant him a new trial when he never sought such relief in the district court. .

Olis argued to the district court that McMillan v. Pennsylvania, 477 U.S. 79, 87-88, 106 S.Ct. 2411, 2417, 91 L.Ed.2d 67 (1986), and United States v. Kikumura, 918 F.2d 1084 (3d Cir. 1990) supported his contention that a standard higher than preponderance of the evidence should be used for enhancements that dramatically increase a sentence under the Sentencing Guidelines. . See generally, Wayne R.

LAfave, Substantive Criminal Law, § 6.4(c) (2d ed.2003) (noting that "even though A's conduct may actually cause B’s [injury], his conduct is not necessarily the 'legal' (or 'proximate') cause of B's [injury],'' and that "the requirement of [legal] causation in criminal law, more often than not, serves not to free defendants from all liability, but rather to limit their punishment consistent with accepted theories of punishment”). . 15 U.S.C.

Section 78u-4(b)(4) states: Loss causation: In any private action arising under this chapter, the plaintiff shall have the burden of providing that the act or omission of the defendant alleged to violate this

chapter caused the loss for which the plaintiff seeks to recover damages. . Under the PSLRA, the general method for calculating loss is the difference between the price paid by the plaintiff and the price after ameliorative information is released to the market, minus any amount of loss that may have been caused by other market factors present during the period of loss. Bloomenthal & Wolff, Securities & Federal Corporate Law § 13:46 (2d ed.2005) . John C. Coffee, "Are We Really Getting Tough on White Collar Crime," 15 Fed. Sent. Rptr. 245, 246 (2003). . See Benjamin E. Rosenberg, Commentary. Damages vs.

Loss — Two Answers to the Same Question, 9 No. 18 Andrews Derivatives Litig. Rep. 12 (2003). . Bakhit also refused to employ a Government theory of loss that had never before been utilized in litigation. Bakhit, 218 F.Supp.2d at 1238-39.

The Government does not further the goals of sentencing uniformity or fairness when, as seems to be happening in these cases, the Government persistently adopts aggressive, inconsistent, and unsupportable theories of loss. .In Snyder, the court noted that "because the price of BioCiyst stock was higher after disclosure of the fraud than its average price during the life of the fraud," using a method of determining loss, such as "subtracting the stock price after the fraud from the average stock price over the life of the fraud,” was not appropriate. Snyder, 291 F.3d at 1296.

Therefore, the court stated that the district court needed to "employ a slightly different, but nonetheless proper, methodology." Id. The court determined that the proper methodology would be to take the number of shares traded during the period the fraud was extant, and multiply by the difference between the average price of the shares during that period and the average price of the shares during the three days after the fraud was revealed. Id. at 1295-96.

In Grabske, the share price of the corporation increased when the fraud was committed, then dropped down once the fraud was disclosed, but still remained above the pre-fraud share price. The court rejected both the Government's and the defendant's loss calculation methods, and instead adopted a rescissionary method consistent with the Private Securities Litigation Reform Act. Grabske, 260 F.Supp.2d. at 871-74. In Bakhit, the court adopted a variant of the Snyder approach. .

The Government relied on the Heil testimony and on the Gunderson report, an expert analysis that, while attempting to account for market losses due to extrinsic effects on Dynegy's stock price, nevertheless measured gross market capitalization loss rather than actual losses to shareholders because of Olis's offense. The Government’s alternative theory, which purported to measure, as "gain to the defendant”, Dynegy’s tax benefit from Project Alpha, need not be considered unless actual loss proves impossible to measure. Cf. Snyder, 291 F.3d at 1296.