Bruce E. Reppert (argued), Suzanne M. Wissmann, Office of U.S. Atty., Criminal Division, Fairview Heights, IL, for U.S.

J. William Lucco, Lucco, Brown & Mudge, Edwardsville, IL, Sheldon T. Zenner, Ted S. Helwig, Robert K. Niewijk (argued), Katten, Muchin & Zavis, Chicago, IL, Samuel C. Ebling, St. Louis, MO, for Edward L. Morris.

David H. Baris, Victor Simon, Kennedy & Baris, L.L.P., Washington, DC, for American Association of Bank Directors.

Steven S. Zaleznick, Deborah M. Zuckerman, Dorothy Siemon, American Association of Retired Persons, Washington, DC, for American Association of Retired Persons.

David M. Fahrenkamp, Edwardsville, IL, Michael Dwyer (argued), Office of the Federal Public Defender, Lincoln, NE, for Steven M. Gardner.

Before CUMMINGS, KANNE, and ROVNER, Circuit Judges.

IIANA DIAMOND ROVNER, Circuit Judge.

The defendants in this mail and wire fraud prosecution are former officers of Germania Bank, a St. Louis-based savings and loan association. After a lengthy trial, a jury convicted the bank's former chief executive officer, Edward L. Morris, and its former chief operating officer, Steven M. Gardner, on two counts of mail (18 U.S.C. § 1341) and one count of wire fraud (18 U.S.C. § 1343) in connection with Germania's $10 million offering of subordinated capital notes ("Schnotes") between October 1987 and March 1988.1 The district court sentenced Morris and Gardner to prison terms of forty-six months, but stayed those sentences during the pendency of their appeals. In challenging their convictions, Morris and Gardner argue that the government's evidence at trial was insufficient to establish violations of the mail and wire fraud statutes and, alternatively, that they are entitled to new trials due to the government's failure to produce exculpatory evidence under Brady v. Maryland, 373 U.S. 83, 83 S. Ct. 1194, 10 L. Ed. 2d 215 (1963). Defendants also challenge the sentences they received under the Sentencing Guidelines, contending that the district court erred in calculating the amount of loss attributable to their fraud under U.S.S.G. § 2F1.1(b)(1), and in refusing to depart downward from their sentencing range to account for other alleged causes of that loss. For the reasons that follow, we affirm defendants' convictions and sentences.

Prior to the deregulation of the savings and loan industry in the early 1980s, Germania had primarily been involved in residential real estate loans. After deregulation, however, the bank expanded its loan portfolio, involving itself in larger multi-family residential and commercial projects. This case involves problems that developed in the bank's loan portfolio in 1986 and 1987, and particularly the bank's decisions in taking and in failing to take loss reserves on that portfolio.2 The crux of the government's case is that the Executive Committee of Germania's Board of Directors failed to approve an additional $9.3 million in loan loss reserves recommended by bank management in September 1987 pursuant to an in-depth quarterly review of the bank's loan portfolio (the "September Analysis" or "SA"). The government maintained that the recommended reserves were rejected not because the Executive Committee decided they were unnecessary, but because Germania was about to embark on the $10 million Schnote offering and its controlling shareholder insisted that the bank show a profit for the quarter ending September 30, 1987. Because the reserves recommended by management would have caused a significant third-quarter loss, the Executive Committee authorized only those reserves necessary to cover identified losses in the portfolio, with gradual increases to follow over time. Germania then proceeded with the Schnote offering, using an offering circular that described current loan loss reserves as "adequate." It is management's dissemination of this offering circular to potential investors that is at the heart of the government's mail and wire fraud charges.

Almost five months after the Schnote offering began, Peat, Marwick, Main & Co. ("Peat Marwick"), in connection with its year-end audit of Germania's financial statements, recommended loan loss reserves of a magnitude similar to those recommended by management in the September Analysis. Germania eventually took an additional $9.4 million in reserves in February 1988, near the conclusion of the Schnote offering. The bank's financial condition deteriorated from that point, and it was finally placed in conservatorship by the Resolution Trust Corporation (the "RTC") in June 1990. With Germania's demise, the Schnotes became worthless, and investors recouped nothing on their investments.

The evidence presented through a multitude of witnesses to the jury in this case was replete with contradictions.3 The most numerous and significant concerned the nature and purpose of the SA, as well as the substance of the Executive Committee's discussion of its recommendations. Mindful that Morris and Gardner primarily raise a sufficiency challenge to the evidence supporting their convictions, we must resolve such conflicts in the evidence as the jury apparently resolved them--in the government's favor. We thus describe in somewhat abbreviated form below the evidence presented at trial as construed in the light most favorable to the government. See United States v. Biesiadecki, 933 F.2d 539, 545 (7th Cir.1991).

The government's case depended heavily upon the testimony of Jimmy Wayne New, a former bank officer who pled guilty to his role in the Schnote offering and agreed to cooperate with the government. New served as the bank's chief financial officer during the relevant time period and was a member of its Executive Committee. New was therefore present, along with Morris and Gardner, at the central meeting that concerns us here. Before discussing that meeting, however, we describe the events preceding it that apparently prompted management's decision to compile the in-depth analysis of the bank's loan portfolio that came to be known as the September Analysis.

Approximately one year before the Schnote offering, Germania conducted a public offering of its stock. Before the offering began, however, Morris became concerned about the bank's loan portfolio and the adequacy of its allowance for loan losses. Morris therefore decided to postpone the stock offering to allow for further investigation. New was on vacation in Mexico at the time, and Morris called him there to tell him that the offering would be postponed. Morris explained that in the course of inspecting certain properties that were the subject of Germania loans, he had become concerned about the level of the bank's reserves. Morris wanted a complete analysis of the bank's loan portfolio compiled before he would proceed with the public stock offering. When New returned from his vacation, he oversaw an in-depth inspection of a large segment of the bank's loan portfolio. As a result of that investigation, Germania took additional reserves, disclosed those reserves in the offering circular, and then proceeded with the stock offering. The 1986 stock offering eventually raised in excess of $5 million, but the per share price was lower than Germania had expected prior to taking the additional reserves.

In February 1987, the bank's auditors issued their report on Germania's financial statements for the year ending December 31, 1986. In connection with that report, Peat Marwick sent a "management letter" to Germania's Board of Directors expressing concern about the bank's control over its loan portfolio:

During our review of large loans, we noted no documentation evidencing ongoing monitoring of several large commercial loans. As noted in our prior year management letter, this practice exposes the Association to the risk that it may not become aware of declines in collateral value or other problems until a loan becomes delinquent. Subsequent to year-end, the Association has implemented a system to review selected large loans on an annual basis. We recommend strict adherence to this policy. In addition, we recommend that non-real estate commercial loans be reviewed on at least a quarterly basis, and that current financial data be obtained and reviewed in connection with this ongoing review. For both types of commercial loans, we recommend that loan files contain documentation of the ongoing review and monitoring procedures performed, including the conclusions as to whether any action by the Association is required.

(Govt.Ex. 27.) The management letter recommended that the bank determine "the adequacy of its allowance for loan losses on a quarterly basis." (Id.)

Germania's Audit Committee and its Board of Directors reviewed in a June 18, 1987 meeting the recommendations made in Peat Marwick's management letter. The Board then responded to the concerns of its auditor in a letter to the Federal Home Loan Bank of Chicago ("FHLB"), which had federal regulatory authority over Germania:

Subsequent to year-end, management of the Association and representatives of Peat, Marwick, Main & Co. developed procedures to determine the adequacy of the Association's allowance for loan losses in response to recent accounting and SEC literature. The analysis attempts to classify the Association's portfolio into three categories (1) not a problem, (2) potential problem, and (3) high risk. Management of the Association, on a quarterly basis, will utilize this classification to assign a probable loss for loans included in each of these categories. Management of the Association will continue to improve the procedures utilized during Peat, Marwick's year-end evaluation of the Association's loan portfolio and real estate owned in determining the adequacy of its allowance for loan losses. It is management's intent to submit an analysis of its loan loss provision review to the Loan Committee on a quarterly basis for their review and discussion.

(Govt.Ex. 28.) The Board offered a similar response on August 20, 1987 to a "report of examination" and "comment letter" that the bank had received from the FHLB. (Govt.Ex. 32.) That response reviewed the status of specific problem loans referenced in the report and then described procedures the bank had implemented to more effectively monitor its loan portfolio. The Board reported that it had developed enhanced procedures "to more effectively monitor and identify problem and potential problem loans." (Id.) Moreover, the Board explained that on a quarterly basis, "a report is submitted by management to the Bank's Board of Directors and its outside auditors which evaluates the adequacy of the Bank's allowance for loan losses." (Id.) The Board also represented that as of August 11, 1987, its management had completed a review of fifty percent of those loans in its portfolio that exceeded $1 million.

In early 1987, as the bank was in the process of responding to the concerns of its auditor and of federal regulators, Morris proposed to Germania's Board that the bank undertake the Schnote offering. He initially explored such an offering with outside underwriters but ultimately determined that Germania could conduct the offering itself. New was given the task of overseeing preparation of the Schnote offering circular, which ultimately disclosed the following about Germania's loan loss reserves:

It is not presently possible to determine the actual losses that will ultimately be incurred by the Savings Bank in connection with the collection of its loan portfolio and liquidation of real estate owned; however, the Savings Bank believes that, based on its present analysis, adequate provision has been made for estimated losses from these portfolios. However, no assurance can be given that additional provisions for losses on loans and real estate owned will not be required as the result of future economic conditions or other factors.

(Gardner Ex. 7A, at 10 (emphasis added).) Consistent with the recommendation of Peat Marwick and Germania's own representations to the FHLB, the circular also revealed that bank management was submitting to the Board of Directors on a quarterly basis a report evaluating the adequacy of its allowance for loan losses. (Id. at 43.) The first Schnote sales in connection with this offering circular were recorded in October 1987, and the $10 million offering continued into March of the following year.

At the same time that management was preparing for the Schnote offering, Morris asked New to compile the September Analysis, which New characterized as "management's quarterly risk analysis assessment of the portfolio" as of August 31, 1987. (See, e.g., Tr. vol. III, at 383.) Morris' request came in response to Peat Marwick's recommendation that a quarterly analysis be done, and the bank's own representation to federal regulators that such an analysis would be performed. New testified that the SA also would enable management to ensure that the Schnote offering circular accurately reported the bank's financial condition. That circular, as we have noted, disclosed to investors that bank management was submitting such an analysis to the Board on a quarterly basis. The overall purpose of the SA, according to New, was "to make adequate provision for loan losses." (Tr. vol. III, at 385.)

New testified that as the bank's first quarterly analysis of its loan portfolio, the SA was a much more detailed analysis than anything the bank had previously undertaken. He described it as more detailed than the analysis done the previous year in connection with the public stock offering. To gather information for the analysis, New directed all of the bank's commercial loan officers and residential lenders to analyze their segments of the portfolio and to determine probable or expected losses from individual loans. Much of the supporting documentation used for this purpose had been compiled during the first two quarters of 1987, but the most intensive work on the September Analysis itself occurred in a period of approximately three weeks in early September.

On September 22, 1987, New met with Morris and Gardner to review the SA and to discuss the recommendation they would make to the Executive Committee about expected losses in the loan portfolio. New, Morris, and Gardner reviewed all of the supporting documentation underlying the analysis and also discussed its methodologies and conclusions. According to New, the three unanimously agreed in the course of this meeting that they would recommend to the Executive Committee an expected loss figure of approximately $18.6 million, which would exceed the bank's August 31, 1987 loss reserves by approximately $9.3 million. New indicated that the $18.6 million figure represented neither a best nor a worst case scenario, but the scenario that was "expected" at the time. The worst case scenario reflected in the SA would have required loss reserves in the range of $22 to $23 million.

Management presented its recommendations to the Executive Committee of Germania's Board at a September 26, 1987 meeting. At the time, the committee was comprised of Morris, Gardner, New, Joseph L. Mason, and Weldon Rogers. The SA had been disseminated to Mason and Rogers prior to the meeting,4 and Morris asked New to discuss its details for the committee. New testified that he had just begun to summarize the report when Mason, a real estate developer who was Germania's controlling shareholder and who had named Morris to the Board, "threw the document on the table" and said that "there is no fucking way we are going to take any more of these reserves at this time." (Tr. vol. III, at 395.) According to New, a silence descended over the room, and everyone seemed to wait for someone else to respond. New finally told Mason that management was recommending additional reserves in part because over $1 million in specific losses had already been identified in the portfolio by the bank's commercial loan officers. Mason retorted that he had "forgotten more about real estate than [New would] ever know." (Id.)5

In articulating his opposition to the SA's recommendations, Mason cited, among other things, a recent audit report by Peat Marwick and an examination by Kenneth Levanthal & Co., both of which had found the bank's reserves to be within an acceptable range. New testified that these examinations were far less detailed than the SA, however; indeed, Kenneth Levanthal had examined only thirteen of the thousands of loans in Germania's portfolio. The documentation supporting the SA's recommendation of $9.3 million in additional reserves was not specifically discussed at the meeting. Mason also provided little specific direction as to what should be done to improve the SA, although he did direct New to obtain additional appraisals in connection with the next quarterly analysis.6 Later in the meeting, Mason emphasized that the bank had to consider the impact on its shareholders (of which he was the largest) of taking such a large loan loss reserve at any one time. At another point, the committee's discussion turned to Germania's level of profitability for the quarter ending September 31, 1987. Due to the impending Schnote offering, it was important to Mason that the bank record a profit for that quarter. Had the committee approved management's loss reserve recommendation, the bank would have recorded a third-quarter loss of approximately $9 million.

At the conclusion of the discussion, the committee approved only $1.2 million in additional reserves, which enabled the bank to record a quarterly profit of slightly over $100,000. Those reserves were required to cover existing losses that the SA had identified in the portfolio. The committee directed management to continue to recommend additional reserves over time. New testified at trial that in the course of the meeting, neither Morris nor Gardner had attempted to defend the SA's recommendation once Mason voiced his objections. He also testified that Mason said nothing in the meeting to convince him, as the bank's chief financial officer, that management's "expected loss" recommendation had been erroneous. Indeed, New told his wife after the meeting that the bank had made a very bad mistake that day.

The following week, Morris, Gardner, and New convened to discuss the results of the September 26 meeting. According to New, Gardner initially suggested that the SA and its underlying documentation be destroyed. New objected, observing that destruction of the analysis would be improper and that too many employees already had access to its supporting materials. Morris apparently agreed, and New then asked whether he should discuss the analysis with Peat Marwick's auditors, who would soon be reviewing the bank's third-quarter financial statements. New was told not to disclose the SA to Peat Marwick because bank management had ultimately adopted a much lower estimate of expected losses than that reflected in the SA; thus, the SA's recommendations were no longer those of bank management.7

In connection with Peat Marwick's review of the bank's third-quarter financial statements, however, Morris, Gardner, and New signed a representation letter affirming to Peat Marwick that the bank had made available to it all financial records and related data relevant to the quarterly review of the bank's financial statements. New testified that Peat Marwick should have received the SA in connection with that review because the SA reflected "management's recommendation going into the September 26th meeting of the expected losses [in the loan portfolio], and it would have had a significant impact on their analysis during the quarterly review." (Tr. vol. III, at 411.) The Peat Marwick partner in charge of the Germania engagement agreed, testifying that the SA was the type of document he would have expected to receive from the bank to support its financial statements. Having no knowledge of the SA, however, Peat Marwick approved Germania's third-quarter financial statements, and they were then placed in the back cover of the Schnote offering circular.

Bank management similarly represented in a debenture purchase agreement between Germania and Laclede Development Company ("Laclede") that "to the best knowledge of the bank, there is no fact relating to the bank or its subsidiaries which the bank has not disclosed to you in writing which materially affects adversely or will materially affect adversely the properties, profits, or condition of the bank and its subsidiaries taken as a whole." (Govt. Ex. 44, at 6.) New testified that this was not a truthful statement at the time because the bank had not disclosed the SA to Laclede in the course of its due diligence examination. Laclede's chief executive officer testified with assurance that the level of additional reserves recommended by the SA had not been disclosed to his company because "that would have blown the deal." (Tr. vol. VI, at 1022.) Germania's debenture purchase agreement with Laclede was disclosed in the Schnote offering circular and was a component of the bank's campaign to market its Schnotes.

In January 1988, prior to Peat Marwick's year-end audit for 1987, Germania's Board approved an additional $1.4 million in loan loss reserves. In a memo to Morris and Gardner dated February 2, 1988, New observed that he was "still very uncomfortable" with the level of the bank's reserves, and he indicated that Germania would have difficulty getting Peat Marwick to approve the reserve levels recommended by management. (Govt.Ex. 36.) After auditing the bank's year-end financial statements, Peat Marwick recommended that Germania add between $6.5 and $13 million to its existing loan loss reserves. Although the need for these reserves was partially attributable to regulatory changes and to changes in certain of Peat Marwick's policies for valuing portions of the portfolio, the auditor's overall assessment of those portions of the portfolio analyzed by bank management in the SA were similar to the SA's recommendations.

When Peat Marwick's recommendations were presented to the bank's Audit Committee, its members interrogated bank management and the auditors about the reasons for such a dramatic increase in loan loss reserves. Committee members were primarily interested in learning what had caused the need for additional reserves and when management had learned that the bank was under-reserved. In the course of these discussions, neither Morris nor Gardner disclosed that bank management had earlier recommended comparable reserves to the Executive Committee.

Germania's Board ultimately approved $9.4 million in additional reserves for the year ending December 31, 1987, and Germania publicly announced those reserves in February 1988, after suspending the ongoing Schnote offering. Bank management and the Audit Committee discussed rescinding the Schnote sales made prior to the announcement, but they ultimately decided to amend the offering circular and to proceed with the offering. When it renewed the offering, the bank assured investors that the 1987 losses did not affect the safety of their Schnotes. Germania employees were encouraged, moreover, to tell customers that the bank had not actually lost the $9.4 million, but that it had merely moved the funds to a reserve account, which served to protect the bank and its Schnotes. Germania raised an additional $1 million upon renewal of the Schnote offering.

After a lengthy investigation, Morris and Gardner were charged with two counts of mail and one count of wire fraud in connection with the Schnote offering. The wire fraud charge concerned a toll free "hot line" the bank had established in marketing the Schnotes, whereas the mail fraud counts concerned, respectively, an unsolicited mass mailing promoting the Schnotes to approximately 18,000 Germania customers in January 1988, and a February 1988 sale of an $18,000 Schnote to a local chapter of the Disabled American Veterans. A jury convicted Morris and Gardner on all charges, and the district court subsequently imposed prison sentences on each of forty-six months.

The mail and wire fraud statutes are addressed to "any scheme or artifice to defraud" executed through the appropriate communications medium. 18 U.S.C. §§ 1341 & 1343. To establish a violation of either statute, the government must prove the existence of such a scheme and a use of the mails or communication wires, as the case may be, for the purpose of executing or attempting to execute the scheme. See, e.g., Schmuck v. United States, 489 U.S. 705, 721, 109 S. Ct. 1443, 1453, 103 L. Ed. 2d 734 (1989); United States v. Ross, 77 F.3d 1525, 1542 (7th Cir.1996); United States v. Briscoe, 65 F.3d 576, 583 (7th Cir.1995). Morris and Gardner argue that the government proved neither mail nor wire fraud in connection with the Schnote offering because it never established the existence of an intentional scheme to defraud. In considering that argument, which essentially is a challenge to the sufficiency of the supporting evidence at trial, we construe the evidence in the light most favorable to the government and consider whether any rational jury could have found the existence of such a scheme beyond a reasonable doubt. See Jackson v. Virginia, 443 U.S. 307, 316, 99 S. Ct. 2781, 2787, 61 L. Ed. 2d 560 (1979); United States v. Catalfo, 64 F.3d 1070, 1076 (7th Cir.1995).

As used in the mail and wire fraud statutes, the term "scheme to defraud" encompasses "any scheme to deprive another of money or property by means of false or fraudulent pretenses, representations, or promises." Carpenter v. United States, 484 U.S. 19, 27, 108 S. Ct. 316, 321, 98 L. Ed. 2d 275 (1987); see also Richards v. Combined Ins. Co. of Am., 55 F.3d 247, 252 (7th Cir.1995). It therefore encompasses fraudulent schemes premised on false statements or factual misrepresentations. See McNally v. United States, 483 U.S. 350, 359, 107 S. Ct. 2875, 2881, 97 L. Ed. 2d 292 (1987); United States v. Keplinger, 776 F.2d 678, 698 (7th Cir.1985), cert. denied, 476 U.S. 1183, 106 S. Ct. 2919, 91 L. Ed. 2d 548 (1986). As we recently had occasion to explain in a civil RICO suit, the mail and wire fraud statutes broadly apply to any scheme "where in order to get money or something else of monetizable value from someone you make a statement to him that you know to be false, or a half truth that you know to be misleading, expecting him to act upon it to your benefit and his detriment." Emery v. American Gen. Fin., Inc., 71 F.3d 1343, 1346 (7th Cir.1995). We reiterated, moreover, that the statutes apply not only to false or fraudulent representations, but also to the omission or concealment of material information, even where no statute or regulation imposes a duty of disclosure. Id. at 1346-47; see also Biesiadecki, 933 F.2d at 543; Keplinger, 776 F.2d at 697-98. Such a misleading omission is actionable as part of a scheme to defraud "if it is intended to induce a false belief and resulting action to the advantage of the misleader and the disadvantage of the misled." Emery, 71 F.3d at 1348.

The government's theory at trial was that the entire Schnote offering had been designed by bank management to defraud potential investors. At the heart of the government's case was the representation in the Schnote offering circular that based upon its present analysis, bank management believed that it had made adequate provision for losses in the loan portfolio. That statement was false or misleading, according to the government, because it omitted any mention of the approximately $8 million in additional losses that bank management fully expected after compiling the September Analysis. Indeed, management recommended to Germania's Executive Committee that reserves be taken in the third quarter of 1987 to cover those losses. The committee rejected that recommendation, according to the government, not because it considered the recommendation to be erroneous, but because the bank's majority shareholder wanted Germania to show a profit for the quarter immediately preceding the Schnote offering. To conceal its fraud, then, bank management allegedly failed to disclose the SA to auditors conducting a review of the bank's quarterly financial statements or to Laclede's due diligence examiners in connection with the Germania-Laclede debenture purchase agreement.

In attacking the sufficiency of the evidence supporting this fraud theory, defendants make a series of arguments. First, they attempt to discredit the SA itself, calling it a "rough first draft" and a "worst-case scenario" that was never intended to reflect management's current position on necessary reserves. Yet the jury's verdict necessarily precludes us from accepting this argument. Although Morris and Gardner testified that the SA was indeed a hastily prepared first draft filled with typographical errors, that it was merely a discussion document, and that it reflected a worst-case assessment of possible losses in the portfolio over time, those assertions were flatly contradicted by Germania's chief financial officer Jimmie New, who had overseen the SA's preparation. New told the jury that the SA was "management's quarterly risk analysis assessment of the portfolio" as of August 31, 1987 (see, e.g., Tr. vol. III, at 383), and that the bank had prepared the report to comply with its auditor's recommendation and the bank's own representation to federal regulators that it performed a quarterly analysis of its loan portfolio.8 New indicated that the $18.6 million loss figure in the SA reflected neither a best nor a worst case scenario; instead, he described it as management's best estimate of "expected" losses in Germania's portfolio. New explained to the jury, moreover, that he had met with Morris and Gardner on September 22, 1987 to discuss the SA's recommendations, and that the three had unanimously agreed to recommend to the Executive Committee that it approve additional loss reserves of $9.3 million. Morris then disseminated the SA to committee members along with an agenda for the September 26 meeting. In convicting defendants on all charges, the jury obviously believed New, not Morris and Gardner. We must accept the jury's resolution of this disputed factual issue, as we are in no position to reassess the credibility of New vis-a-vis that of defendants. See Ross, 77 F.3d at 1543-44; Folger Adam Co. v. PMI Indus., Inc., 938 F.2d 1529, 1534 (2d Cir.) (resolution of dispute over proper characterization of internal reports is for the jury), cert. denied, 502 U.S. 983, 112 S. Ct. 587, 116 L. Ed. 2d 612 (1991). We must therefore assume that the SA was indeed management's risk analysis of the loan portfolio for the third quarter of 1987. We must also assume, consistent with New's testimony, that Morris and Gardner agreed with the expected loss figure reflected in that analysis and in fact recommended to the Executive Committee that reserves be taken immediately to cover those losses.9

For this reason, defendants' reliance on such cases as Wielgos v. Commonwealth Edison Co., 892 F.2d 509, 516 (7th Cir.1989), and Panter v. Marshall Field & Co., 646 F.2d 271, 293 (7th Cir.), cert. denied, 454 U.S. 1092, 102 S. Ct. 658, 70 L. Ed. 2d 631 (1981), is unavailing. In those cases, we agreed with the defendants that a company need not disclose a tentative internal estimate that conflicts with published figures unless the internal estimate is made with such a reasonable degree of certainty that it reveals the published figures to be materially misleading. Wielgos, 892 F.2d at 516; Panter, 646 F.2d at 292. We have reaffirmed that rule in a series of recent cases. See, e.g., In re HealthCare Compare Corp. Sec. Litig., 75 F.3d 276, 282-83 (7th Cir.1996) (following Wielgos and Panter); Searls v. Glasser, 64 F.3d 1061, 1067 (7th Cir.1995) (same). In the cited cases, however, the character of the internal reports at issue was not seriously disputed, as the plaintiffs had been unable to produce evidence sufficient to contradict the "tentative" characterization placed on those estimates by the company itself. We could thus determine as a matter of law that the defendants need not have disclosed the tentative estimates because they were not so certain as to reveal the company's published figures to be materially misleading. See Wielgos, 892 F.2d at 516; Panter, 646 F.2d at 292-93. Here, however, the government was able to show that the SA's estimate of "expected" losses in the loan portfolio was reasonably certain. The government established that the SA's estimates had been embraced by Morris and Gardner, who recommended to Germania's Executive Committee that reserves be taken to cover those expected losses. It established, moreover, that the Executive Committee rejected the recommended reserves not because of any defect in the SA's estimates, but because Germania's majority shareholder insisted that the bank show a profit in the quarter preceding the Schnote offering. The government's evidence apparently persuaded the jury that the SA in fact contained estimates of expected losses in the loan portfolio that bank management considered to be reasonably certain. That was an issue for the jury to decide, and its decision cannot be considered erroneous as a matter of law. See Folger Adam Co., 938 F.2d at 1534 & n. 8 (distinguishing Wielgos and Panter and leaving proper characterization of internal reports to the jury); see also Schwartz v. System Software Assoc., Inc., 32 F.3d 284, 287-88 (7th Cir.1994) (jury resolved dispute over whether three internal reports were true forecasts or worst-case simulations).

Having cleared those preliminary hurdles, we get to the heart of defendants' argument--that bank management did not engage in an intentional scheme to defraud because it had a reasonable basis for rejecting the SA's recommendations and for describing the bank's existing reserves as "adequate." According to defendants, the government proved only that the offering circular's "prediction" about the adequacy of loan loss reserves was erroneous in light of later events, not that its prediction was the primary component of an intentional scheme to defraud. Because defendants allegedly had a reasonable basis for their description of loss reserves in the offering circular, they deride the government's case as a classic example of "fraud by hindsight."

In considering similar fraud claims based upon management's characterization of the adequacy of loan loss reserves, our colleagues in other circuits have been guided by the Supreme Court's decision in Virginia Bankshares, Inc. v. Sandberg, 501 U.S. 1083, 111 S. Ct. 2749, 115 L. Ed. 2d 929 (1991). See, e.g., In re Wells Fargo Sec. Litig., 12 F.3d 922, 927 (9th Cir.1993), cert. denied, --- U.S. ----, 115 S. Ct. 295, 130 L. Ed. 2d 209 (1994); Shapiro v. UJB Fin. Corp., 964 F.2d 272, 282-83 (3d Cir.), cert. denied, 506 U.S. 934, 113 S. Ct. 365, 121 L. Ed. 2d 278 (1992). Sandberg did not involve a bank's characterization of the adequacy of its loan loss reserves, but we agree that its analysis is applicable to such a statement. See, e.g., In re Westinghouse Sec. Litig., 832 F. Supp. 948, 971 (W.D.Pa.1993) ("characterizations of reserves as adequate are necessarily a matter of opinion and belief" under Sandberg).

The Court in Sandberg considered whether a statement of opinion or belief could be materially misleading and thus actionable under section 14(a) of the Securities Exchange Act of 1934, 15 U.S.C. § 78n(a), and SEC Rule 14a-9. The statement at issue appeared in a company's proxy materials and delineated for shareholders management's reasons for approving a merger proposal. The proxy materials explained that the proposal offered shareholders the opportunity "to achieve a high value for their shares" and that its terms were "fair." Id. at 1090, 111 S.Ct. at 2756-57 (internal quotation omitted). When a disgruntled shareholder alleged that these statements were fraudulent, the defendants argued that such statements of opinion and belief are not actionable under the securities laws because they are not "factual" statements subject to verification. The Supreme Court disagreed, finding defendants' statements to be "factual in two senses: as statements that the directors do act for the reasons given or hold the belief stated and as statements about the subject matter of the reason or belief expressed." Id. at 1092, 111 S.Ct. at 2758. Although the Court agreed that management's statements were conclusory statements of opinion, it explained that such statements "are reasonably understood to rest on a factual basis that justifies them as accurate, the absence of which renders them misleading." Id. at 1093, 111 S.Ct. at 2758. Whether a conclusory statement is accurate or misleading, moreover, is subject to proof by reference to historical facts:

Provable facts either furnish good reasons to make a conclusory commercial judgment, or they count against it, and expressions of such judgments can be uttered with knowledge of truth or falsity just like more definite statements, and defended or attacked through the orthodox evidentiary process that either substantiates their underlying justifications or tends to disprove their existence.

Id. The Court therefore concluded that plaintiffs in such cases must be permitted to prove that a conclusory statement of opinion was "knowingly false or misleadingly incomplete" when made. Id. at 1095, 111 S.Ct. at 2759.

In Shapiro v. UJB Fin. Corp., 964 F.2d 272, 282-83 (3d Cir.), cert. denied, 506 U.S. 934, 113 S. Ct. 365, 121 L. Ed. 2d 278 (1992), the Third Circuit likened a bank's characterization of the adequacy of its loan loss reserves to management's conclusory reasons for approving the merger proposal in the Sandberg proxy statement. The court believed that by affirmatively characterizing a management practice as "adequate," a defendant puts that particular subject matter "in play" under the securities laws:

[I]f a defendant represents that its lending practices are "conservative" and that its collateralization is "adequate," the securities laws are clearly implicated if it nevertheless intentionally or recklessly omits certain facts contradicting these representations. Likewise, if a defendant characterizes loan loss reserves as "adequate" or "solid" even though it knows they are inadequate or unstable, it exposes itself to possible liability for securities fraud. By addressing the quality of a particular management practice, a defendant declares the subject of its representation to be material to the reasonable shareholder, and thus is bound to speak truthfully.

Id. at 282. The Third Circuit reaffirmed that holding in Hayes v. Gross, 982 F.2d 104, 106-07 (3d Cir.1992), concluding there that the plaintiff had stated a "paradigm case" of securities fraud when he alleged that officers and directors had known of the inadequacy of a bank's loss reserves but had represented in an annual report that reserves were maintained at a level considered "adequate" by management to provide for potential losses in the portfolio.

In both Shapiro and Hayes, however, the Third Circuit was careful to distinguish claims based upon management's characterization of the adequacy of reserves from those claims alleging merely a failure to provide for adequate reserves, a form of mismanagement that generally is not considered fraudulent. Hayes, 982 F.2d at 106; see also Shapiro, 964 F.2d at 281. As Hayes explained:

[A]n allegation of mismanagement on the part of a defendant will not alone support a claim [for securities fraud]; nor will an allegation that a defendant failed to disclose the existence of mismanagement.... [A] complaint does allege an actionable misrepresentation if it alleges that a defendant was aware that mismanagement had occurred and made a material public statement about the state of corporate affairs inconsistent with the existence of mismanagement. A statement of opinion or belief, if known to be false, may be the basis of such a claim.

982 F.2d at 106.

The Ninth Circuit subsequently followed Shapiro and Hayes in In re Wells Fargo Sec. Litig., 12 F.3d 922 (9th Cir.1993), cert. denied, --- U.S. ----, 115 S. Ct. 295, 130 L. Ed. 2d 209 (1994). There, the plaintiff shareholders grounded their securities fraud claim against bank officers and directors on the representation in financial reports that management considered its present allowance for loan losses to be "adequate." Id. at 930. Although the reserves later proved grossly inadequate, the court explained that management's earlier statement would be actionable only if the contingencies that rendered the reserves inadequate "were foreseen or foreseeable." Id. at 927. The plaintiffs satisfied this test by alleging that the bank in fact knew of certain contingencies that rendered its reserves inadequate but failed to disclose that information to investors. The plaintiffs' claim, then, was neither one of "second-guessing decisions by management nor one alleging 'fraud by hindsight' " because the plaintiffs had identified specific facts known to the defendants that had been omitted from a report which represented existing reserves as "adequate." Id. As in Shapiro, such a representation declares its subject to be material to the reasonable shareholder, binding the bank to speak truthfully. Id. at 930.

The First Circuit reached a similar conclusion in Serabian v. Amoskeag Bank Shares, Inc., 24 F.3d 357 (1st Cir.1994). There, too, a class of shareholders alleged that bank officers and directors had violated federal securities laws in describing the bank's loan review capabilities as "strong" and its allowance for loan losses as "sufficient" when those representations were inconsistent with internal information then available to bank management. Id. at 364-65. Plaintiffs' allegations were actionable, the court held, because they suggested that "the bank knew, or should have known, that its public statements were inconsistent with the actual conditions then being reported to them." Id. at 365 (emphasis in original).

From these cases, it is clear that Germania's characterization of the adequacy of its loss reserves is not immune from the government's fraud charge simply by virtue of the fact that it offers only the opinion of bank management on a debatable and indefinite matter. As Shapiro explained, "[t]here is nothing unique about representations and omissions regarding loan loss reserves that removes them from the purview of the antifraud provisions of the federal securities laws." 964 F.2d at 281; see also Wells Fargo, 12 F.3d at 927. Although we can agree with defendants that the estimation of probable losses in a large loan portfolio like Germania's is more an art than a science, and that any two analyses of probable losses in the same portfolio are unlikely to exactly correspond, representations by management about the adequacy of reserves still may be fraudulent if they are knowingly false or misleadingly incomplete. See, e.g., Sandberg, 501 U.S. at 1093, 1095, 111 S. Ct. at 2758-59, 2759-60; Shapiro, 964 F.2d at 281, 283; In re Westinghouse Sec. Litig., 832 F. Supp. at 970-71. In that sense, the government must show that bank management did not truly believe in the adequacy of Germania's loss reserves and that its statement to the contrary was not supported by the available facts. See Sandberg, 501 U.S. at 1092, 1095-96, 111 S. Ct. at 2757-58, 2759-60; Mayer v. Mylod, 988 F.2d 635, 639 (6th Cir.1993); Hayes, 982 F.2d at 106-07.

It is true, as Morris and Gardner stress, that the government cannot establish fraud in the offering circular merely by pointing to the bank's announcement a short time later that an additional $9.4 million in reserves would be required. Our oft-cited decision in DiLeo v. Ernst & Young, 901 F.2d 624, 627 (7th Cir.), cert. denied, 498 U.S. 941, 111 S. Ct. 347, 112 L. Ed. 2d 312 (1990), establishes that the government's case would fail in that circumstance because nothing in its proof would suggest that the difference between management's assessment of probable losses at point A and point B was attributable to fraud, as opposed to mere mismanagement or some unforeseeable market force. See also, e.g., Grassi v. Information Resources, Inc., 63 F.3d 596, 599 (7th Cir.1995); In re GlenFed, Inc. Sec. Litig., 42 F.3d 1541, 1548-49 (9th Cir.1994) (en banc) ("The fact that an allegedly fraudulent statement and a later statement are different does not necessarily amount to an explanation as to why the earlier statement was false." (emphasis in original)); Serabian, 24 F.3d at 361, 367; Wells Fargo, 12 F.3d at 927-28; Eckstein v. Balcor Film Investors, 8 F.3d 1121, 1132 (7th Cir.1993), cert. denied, --- U.S. ----, 114 S. Ct. 883, 127 L. Ed. 2d 78 (1994); Arazie v. Mullane, 2 F.3d 1456, 1467-68 (7th Cir.1993). Had that been the sum of the government's case here, we could agree that at most, it would have proven mismanagement, which is not the equivalent of fraud. See, e.g., Hayes, 982 F.2d at 106; Shapiro, 964 F.2d at 281.

But the government did not merely contrast the characterization of reserves in the offering circular with the auditor's recommendations in February of the following year. Rather, consistent with the case law, the government pointed to information known to bank management at the time of the Schnote offering which indicated that the offering circular's characterization of the adequacy of reserves was without a legitimate factual basis. See Sandberg, 501 U.S. at 1093, 111 S. Ct. at 2758-59; GlenFed Sec. Litig., 42 F.3d at 1548-49; Serabian, 24 F.3d at 365; Wells Fargo, 12 F.3d at 927-28; DiLeo, 901 F.2d at 626-27. That is precisely the sort of internal information the courts look to in assessing whether a statement of opinion or belief in a public document could be considered fraudulent. The Ninth Circuit has observed, for example, that an allegedly fraudulent statement could be proven untrue or misleading when made "by pointing to inconsistent contemporaneous statements or information (such as internal reports) which were made by or available to the defendants." GlenFed Sec. Litig., 42 F.3d at 1549; see also Serabian, 24 F.3d at 365. The government pointed to both types of evidence here, as it relied on the SA's conclusion that a dramatic increase in reserves was necessary to cover expected losses in the loan portfolio, and on New's testimony that both Morris and Gardner had agreed with the SA's recommendations prior to the September 26, 1987 meeting of Germania's Executive Committee. This evidence could suggest to a reasonable jury that bank management did not truly believe that existing reserves were adequate and that its statement to the contrary in the offering circular was not supported by the available facts. See Sandberg, 501 U.S. at 1092, 1095-96, 111 S. Ct. at 2757-58, 2759-60, Mayer, 988 F.2d at 639.

Morris and Gardner argue, however, that even if they had agreed to recommend the level of reserves reflected in the SA, that recommendation was rejected by the Executive Committee in good faith after it had examined and discussed the SA in the course of its September 26 meeting. Defendants' good faith in describing existing reserves as adequate would constitute a complete defense to the mail and wire fraud charges here because if they genuinely believed in the truth of the offering circular's statement, the intent to defraud required for a conviction on such charges would be absent. See, e.g., United States v. Dunn, 961 F.2d 648, 650 (7th Cir.1992); United States v. Alexander, 743 F.2d 472, 478-79 (7th Cir.1984).

But the nature of the Executive Committee's discussions at its September 26 meeting was, along with the proper characterization of the SA, one of the primary factual issues disputed at defendants' trial. Although Morris and Gardner insisted that the SA's recommendations were discussed and rejected on their merits, with Mason criticizing the SA's methodology and calling for additional information, New told a far different story. He indicated that there was little if any discussion of the SA's recommendations once Germania's majority shareholder voiced vehement opposition to taking reserves of the recommended magnitude at a time when the bank was preparing for a debt offering. Because the bank's financial statements for the third quarter of 1987 would be included in the Schnote offering circular, it was imperative to Mason that Germania record a profit for that quarter. Approval of the recommended reserves would have caused the bank to record a third-quarter loss of approximately $9 million. New also testified that neither Morris nor Gardner attempted to defend the level of reserves they had agreed to recommend once Mason voiced his opposition.

If New's version of the September 26 meeting is credited, a reasonable jury could conclude that the Executive Committee, which included Morris and Gardner, did not reject the SA's recommendations in good faith, but because Germania's controlling shareholder had insisted that the bank show a quarterly profit for purposes of its impending debt offering. Indeed, the jury was instructed at defendants' request that there was no fraud if defendants merely made an honest mistake in judgment. (See Defendants' Instruction 7A.)10 The jury, however, apparently concluded that the SA's recommendations had not been rejected in good faith and that the offering circular's characterization of the adequacy of reserves was therefore fraudulent. The evidence we have detailed amply supports those conclusions. See Ross, 77 F.3d at 1543-44; United States v. Ashman, 979 F.2d 469, 480-81 (7th Cir.1992), cert. denied, --- U.S. ----, 114 S. Ct. 62, 126 L. Ed. 2d 32 (1993); Keplinger, 776 F.2d at 698-99 (whether omission of consultant's contrary finding from a laboratory report was fraudulent was for the jury); United States v. Sylvanus, 192 F.2d 96, 105 (7th Cir.1951) ("The question was whether, upon the whole record, the representations were fraudulently made, and this was a question for the jury."), cert. denied, 342 U.S. 943, 72 S. Ct. 555, 96 L. Ed. 701 (1952).11

Defendants make one final argument addressed to the legal sufficiency of their convictions. They point to other language in the offering circular which allegedly modifies the loan loss reserve statement and contend that this language neutralized any fraud. They point, for example, to the sentences that preceded and that followed the "adequacy" statement:

It is not presently possible to determine the actual losses that will ultimately be incurred by the Savings Bank in connection with the collection of its loan portfolio and liquidation of real estate owned; however, the Savings Bank believes that, based on its present analysis, adequate provision has been made for estimated losses from these portfolios. However, no assurance can be given that additional provisions for losses on loans and real estate owned will not be required as a result of future economic conditions or other factors.

(Gardner Ex. 7A, at 10 (emphasis added).) They also highlight the offering circular's disclosure of 1986 loan loss reserves (id. at 10, 43), of the "greater risk" present in the bank's commercial and multi-family residential loans (id. at 11), and of the impact of a potential regulatory change on the bank's classification of certain loans (id. at 12). Defendants maintain that these disclosures, which reveal the uncertainty of predicting losses in any portfolio and of the risks inherent in Germania's particular portfolio, modify the circular's "adequacy" statement and thus "bespeak caution" to the ordinary investor. They argue that under the "bespeaks caution" doctrine, the circular's characterization of the adequacy of reserves was not materially misleading because, when considered with the offering circular's other disclosures, that statement could neither have been designed to mislead nor could it have misled the ordinarily prudent investor.

We recently had occasion to address a similar invocation of the "bespeaks caution" doctrine in a civil securities fraud case. See Harden v. Raffensperger, Hughes & Co., 65 F.3d 1392, 1404-1406 (7th Cir.1995). There, we characterized the doctrine in this way: " '[W]hen forecasts, opinions, or projections in a disclosure statement are accompanied by meaningful warnings and cautionary language, the forward-looking statements may not be misleading. The substantial disclosure of specific risks may render alleged misrepresentations concerning soft information immaterial and thus nonactionable as securities fraud.' " Id. at 1404 (quoting 3B Harold S. Bloomenthal, Securities and Federal Corporate Law § 8.26, at 8-110 (1995)). We also observed that when the doctrine is properly construed, it " 'merely represents the pragmatic application of two fundamental concepts in the law of securities fraud: materiality and reliance.' " Id. at 1404-05 (quoting In re Worlds of Wonder Sec. Litig., 35 F.3d 1407, 1414 (9th Cir.1994), cert. denied, --- U.S. ----, ----, 116 S. Ct. 185, 277, 133 L. Ed. 2d 123, 197 (1995)); see also Rubinstein v. Collins, 20 F.3d 160, 167 (5th Cir.1994) ("[T]he 'bespeaks caution' doctrine merely reflects the unremarkable proposition that statements must be analyzed in context."). We noted that a number of other circuits have approved some variation on the doctrine, and that the Supreme Court has implicitly accepted its logic. See Harden, 65 F.3d at 1405 & nn. 9-10; see also Sandberg, 501 U.S. at 1097, 111 S.Ct. at 2760-61; In re Donald J. Trump Casino Sec. Litig., 7 F.3d 357, 372 (3d Cir.1993), cert. denied, --- U.S. ----, 114 S. Ct. 1219, 127 L. Ed. 2d 565 (1994). Although Harden implicitly recognizes the viability of the "bespeaks caution" doctrine in this circuit as a defense in securities fraud cases, we have yet to encounter a case where the doctrine applied to negate the materiality of a misleading statement as a matter of law. See Harden, 65 F.3d at 1405-06 & n. 11; Roots Partnership v. Lands' End, Inc., 965 F.2d 1411, 1417 (7th Cir.1992).

Assuming that the "bespeaks caution" doctrine could provide a valid defense to mail and wire fraud charges, it does not provide a basis for overturning the convictions of these defendants. Harden explained that the effect of cautionary language on an otherwise misleading statement must be assessed on a case-by-case basis. 65 F.3d at 1404. Vague or boilerplate disclaimers advising that a particular investment has risks generally will not suffice; instead, the cautionary language "must be substantive and tailored to the specific future projections, estimates or opinions" that are alleged to be misleading. In re Trump, 7 F.3d at 371-72; see also Harden, 65 F.3d at 1404. In Sandberg, moreover, the Supreme Court observed that "not every mixture with the true will neutralize the deceptive. If it would take a financial analyst to spot the tension between the one and the other, whatever is misleading will remain materially so...." Id. It is only "when the inconsistency would exhaust the misleading conclusion's capacity to influence" the reasonable investor that the conclusion will be rendered immaterial. Id. at 1097-98, 111 S.Ct. at 2760-61; see also Fecht v. Price Co., 70 F.3d 1078, 1082 (9th Cir.1995), cert. denied, --- U.S. ----, 116 S. Ct. 1422, 134 L. Ed. 2d 547 (1996).

The language relied upon by defendants here does nothing more than to apprise potential investors that actual losses in the portfolio cannot now be predicted with absolute certainty. Even while noting the possibility that "future economic conditions or other factors" such as a regulatory change may require the bank to take additional reserves in the future, the offering circular still advises that "based on its present analysis," bank management believes that "adequate provision" has been made for estimated losses in the loan portfolio. The uncertain nature of predicting future loan losses does not neutralize the misleading character of this statement when the government proved to the jury's satisfaction that management's "present analysis" (the SA) actually recommended the doubling of existing reserves. The qualifying statements thus do not exhaust the capacity of the adequacy statement to mislead. See Sandberg, 501 U.S. at 1097-98, 111 S.Ct. at 2760-61; Harden, 65 F.3d at 1405-06; Roots Partnership, 965 F.2d at 1417 (declining to apply doctrine to forward-looking statements about corporation's goals).12

Morris and Gardner alternatively argue that they are entitled to a new trial because the government suppressed material exculpatory information in violation of its obligations under Brady v. Maryland, 373 U.S. 83, 83 S. Ct. 1194, 10 L. Ed. 2d 215 (1963). Defendants did not raise any Brady issues prior to or during their trial but instead argued in a series of post-trial motions that the government had suppressed certain materials they considered to be exculpatory. After two separate evidentiary hearings, the district court found that no Brady violation had occurred, and the court therefore refused to grant defendants a new trial. In so ruling, the district court did not determine whether the items referenced by defendants were in fact exculpatory; instead, the court concluded that defendants had failed to show that the government's prosecution team had been aware of the existence of the allegedly suppressed documents. In that respect, the court emphasized that there were literally hundreds of boxes of documents potentially relevant to this case and that the government had provided defendants and their counsel free and open access to all of those documents.13 To the extent defendants were alleging that certain documents had not been specifically identified by the government as exculpatory, the court believed that the government also had been unaware of the existence of those documents. The court emphasized, moreover, that defendants had been given the same opportunity as the government to discover the identified documents.

In Brady, the Supreme Court held that "the suppression by the prosecution of evidence favorable to an accused upon request violates due process where the evidence is material either to guilt or to punishment, irrespective of the good faith of the prosecution." 373 U.S. at 87, 83 S.Ct. at 1196-97; see also Kyles v. Whitley, --- U.S. ----, ----, 115 S. Ct. 1555, 1565, 131 L. Ed. 2d 490 (1995). Before they would be entitled to a new trial under Brady, however, Morris and Gardner must establish: "(1) that the prosecution suppressed evidence; (2) that the evidence was favorable to the defense; and (3) that the evidence was material to an issue at trial." United States v. Silva, 71 F.3d 667, 670 (7th Cir.1995); see also United States v. Young, 20 F.3d 758, 764 (7th Cir.1994). The district court determined that the government had not "suppressed" the evidence identified by defendants even if that evidence could be considered both favorable and material. We agree with that conclusion and thus with the district court's decision to deny defendants a new trial.

As a general rule, the government's obligation to disclose exculpatory or impeaching information under Brady is limited to that information which is then known to the government. See United States v. Moore, 25 F.3d 563, 569 (7th Cir.), cert. denied, --- U.S. ----, 115 S. Ct. 341, 130 L. Ed. 2d 297 (1994); Young, 20 F.3d at 764; Mendoza v. Miller, 779 F.2d 1287, 1297 (7th Cir.1985), cert. denied, 476 U.S. 1142, 106 S. Ct. 2251, 90 L. Ed. 2d 697 (1986). Although we have not interpreted Brady as requiring prosecutors to affirmatively seek out information not presently in their possession (see Moore, 25 F.3d at 569; United States v. Romo, 914 F.2d 889, 898 (7th Cir.1990), cert. denied, 498 U.S. 1122, 111 S. Ct. 1078, 112 L. Ed. 2d 1183 (1991)), we have found it improper for a prosecutor's office to remain ignorant about certain aspects of a case or to compartmentalize information so that only investigating officers, and not the prosecutors themselves, would be aware of it. See Carey v. Duckworth, 738 F.2d 875, 877-78 (7th Cir.1984); see also Young, 20 F.3d at 764. The district court concluded from these cases that Brady did not require the government here to seek out allegedly exculpatory information in the hands of the Office of Thrift Supervision ("OTS"), the Securities Exchange Commission ("SEC"), or the Internal Revenue Service ("IRS") when it had been unaware of the existence of that information.14 Because none of those agencies were part of the team that investigated this case or participated in its prosecution, the district court would not impute their knowledge of potentially exculpatory information to the present prosecutors.

Morris and Gardner contend that the district court's decision is inconsistent with Kyles v. Whitley, --- U.S. ----, ----, 115 S. Ct. 1555, 1567, 131 L. Ed. 2d 490 (1995), because there, the Court imposed a duty on individual prosecutors "to learn of any favorable evidence known to the others acting on the government's behalf in the case, including the police." We do not agree, however, that Kyles is inconsistent with the rule articulated by the district court here, or for that matter with the previous decisions of this court. Indeed, this circuit has long followed a rule similar to that approved in Kyles; we held in United States ex rel. Smith v. Fairman, 769 F.2d 386, 391 (7th Cir.1985), for instance, that the Brady rule is violated even where police, rather than prosecutors, are responsible for the suppression of exculpatory information. But neither Kyles nor Fairman can be read as imposing a duty on the prosecutor's office to learn of information possessed by other government agencies that have no involvement in the investigation or prosecution at issue. Cf. Carey, 738 F.2d at 878 (Brady obligation extends to members of the prosecution "team," which includes DEA and police). That is the only rule that would help defendants here, as they argue that the prosecutors had a duty to discover information in the hands of other government agencies that at various times conducted independent investigations of Germania's affairs. The district court was satisfied, as are we, that the prosecution team, which included investigating officers and agents, had no knowledge of the specific documents identified by defendants. The prosecutors therefore had no affirmative duty to discover those documents and to disclose them to defendants.

Defendants are entitled to no relief under Brady for one further reason. As our cases have consistently emphasized, the government will not be found to have suppressed material information if that information also was available to a defendant through the exercise of reasonable diligence. See United States v. White, 970 F.2d 328, 337 (7th Cir.1992); see also United States v. Rodriguez-Andrade, 62 F.3d 948, 952 (7th Cir.1995); United States v. Dimas, 3 F.3d 1015, 1018-19 (7th Cir.1993); United States v. Hedgeman, 564 F.2d 763, 768 (7th Cir.1977), cert. denied, 434 U.S. 1070, 98 S. Ct. 1253, 55 L. Ed. 2d 773 (1978). Here, Morris and Gardner knew as much about the OTS, SEC, and IRS investigations as the prosecution team itself did. It appears, for example, that both Morris and Gardner had counsel present during New's OTS deposition. (See supra, n. 14.) Moreover, that deposition was referenced in New's separate SEC deposition, which defendants admit they possessed prior to trial. Defendants also were involved in litigation with the SEC before the instant prosecution was initiated, and the government established below that the survey of Schnote holders had been available to defendants in connection with that litigation. Where, as here, alleged Brady material was available to defendants through the exercise of reasonable diligence, they cannot obtain a new trial by insisting that the government should have conducted their investigation for them. See United States v. Baker, 1 F.3d 596, 598 (7th Cir.), cert. denied, --- U.S. ----, 114 S. Ct. 412, 126 L. Ed. 2d 359 (1993); White, 970 F.2d at 337.

For all of these reasons, we agree with the district court that Morris and Gardner have failed to show that the government "suppressed" the identified information. We need not determine, then, whether that information is favorable to the defense and whether "there is a reasonable probability that, had the evidence been disclosed to the defense, the result of the proceeding would have been different." Kyles, --- U.S. at ----, 115 S. Ct. at 1565 (internal quotation omitted); see also Wood v. Bartholomew, --- U.S. ----, ----, 116 S. Ct. 7, 10, 133 L. Ed. 2d 1 (1995) (per curiam).

Morris and Gardner also challenge the forty-six month prison terms they received under the Sentencing Guidelines. They first contend that the district court erred in calculating the amount of the "loss" under U.S.S.G. § 2F1.1(b)(1). They alternatively argue that the district court should have departed downward to account for other possible causes of the loss here.

Morris and Gardner were assigned a base offense level of six pursuant to U.S.S.G. § 2F1.1(a). Finding that the "estimated, probable or intended loss" from their conduct exceeded $5,000,000, the district court then added eleven points to this base offense level pursuant to section 2F1.1(b)(1)(L).15 Likening the purchase of a subordinated debt instrument to the issuance of a loan, the court concluded that the actual loss to the victims of defendants' scheme was the face amount of the Schnotes, as the Schnotes were rendered worthless by Germania's demise. That amount, which exceeded $5,000,000 for Schnote purchasers in the Southern District of Illinois, reflected both the loss to the victims as well as the gain to the bank. In settling on that figure, the court rejected defendants' argument that they should not be allocated the entire loss because factors other than their fraud had contributed to Germania's failure. Because the district court's calculation of the amount of loss under section 2F1.1(b)(1) is a finding of fact, we review it only for clear error. Ross, 77 F.3d at 1551-52; United States v. Johnson, 16 F.3d 166, 170 (7th Cir.1994). The meaning of "loss" in section 2F1.1, however, presents a question of law that we consider de novo. United States v. Barrett, 51 F.3d 86, 89 (7th Cir.1995); United States v. Holiusa, 13 F.3d 1043, 1045 (7th Cir.1994).

Defendants have not contested that the present circumstances can be likened to the more familiar scenario we encounter in fraudulent loan cases. Such cases generally involve a material misrepresentation on a loan application that aids the borrower in procuring a loan. Here, Schnote purchasers effectively loaned to Germania the face value of the Schnotes at a specified rate of interest, and the jury found that these "loans" had been obtained by the bank in connection with a scheme to defraud. In fraudulent loan cases, we define "loss" under section 2F1.1(b)(1) to include the actual or intended loss to the victim, and we have calculated a victim's actual loss by subtracting from the face amount of the loan both the amount repaid prior to discovery of the fraud and any amount to be recovered from assets pledged as collateral to secure the loan. See, e.g., United States v. Channapragada, 59 F.3d 62, 66 (7th Cir.1995); Johnson, 16 F.3d at 171-72; United States v. Chevalier, 1 F.3d 581, 585-87 (7th Cir.1993); United States v. Mount, 966 F.2d 262, 265 (7th Cir.1992). This rule has been adopted by the Sentencing Commission in Application Note 7(b), which was added in November 1991, although our decisions pre-dating the application note had employed a similar methodology. See United States v. Schneider, 930 F.2d 555 (7th Cir.1991); United States v. Miller, 962 F.2d 739, 748 (7th Cir.1992) (Flaum, J., concurring); see also United States v. Kopp, 951 F.2d 521, 531-36 (3d Cir.1991) (discussing Schneider). Indeed, in Holiusa, 13 F.3d at 1047-48, we employed an analogous net loss approach under the version of section 2F1.1 at issue here. See also Kopp, 951 F.2d at 535-36. Thus, we agree that such an approach would be an appropriate means of calculating the loss attributable to the fraud of Morris and Gardner.

It turns out, however, that the actual loss in this case is equivalent to the face value of the Schnotes. Germania made no principal payments on the Schnotes prior to its collapse, and because the Schnotes were uninsured, investors recovered nothing thereafter. The actual loss to the victims, then, exceeds $5,000,000.

Defendants argue that they should not be charged with all or even any portion of this loss, however, because factors other than the failure to take recommended reserves in 1987 actually caused Germania's demise. We rejected a similar argument in United States v. Miller, 962 F.2d 739 (7th Cir.1992). There, the defendants had been charged with making false statements in order to obtain mortgage loans insured by the Department of Housing and Urban Development ("HUD"). After procuring the HUD-insured mortgages, the defendants sold the mortgaged buildings to a third party, who eventually defaulted on the mortgages. After purchasing the delinquent loans and acquiring the properties from a successor mortgage company, HUD resold them at a sheriff's sale at a significant loss. Id. at 742-43. The defendants argued that HUD's losses were not attributable to their fraud under section 2F1.1(b)(1) because those losses had been caused by the defaulting third party and HUD itself. Id. at 743. The Miller panel rejected that argument, finding that the defendants' fraud had put HUD's money at risk in the first place. Thus, neither the intervening acts of a negligent third party nor the victim's own failure to mitigate damages could provide a basis for reducing section 2F1.1(b)(1) loss in that circumstance. Id. at 744; see also id. at 748 (Flaum, J., concurring) ("As noted, however, § 2F1.1(b)(1) loss is measured by the net loss to the injured party"). The existence of other possible causes of the loss could be accounted for only by way of a downward departure pursuant to Application Note 11. Id. at 744, 748-49; see also United States v. Carey, 895 F.2d 318, 323 (7th Cir.1990).16

Other circuits have similarly rejected the notion that intervening causes would serve to reduce the amount of loss attributable to a defendant's fraud. Kopp observed, for example, that in Application Note 11, the Sentencing Commission, "definitively rejected adjusting the 'loss' itself downward to reflect other causes beyond the defendant's control.... To the extent [the] actual loss had other, more proximate causes, a discretionary downward departure--but not a mandatory 'loss' adjustment--might be appropriate." 951 F.2d at 531; see also, e.g., United States v. Sarno, 73 F.3d 1470, 1500-01 (9th Cir.1995); United States v. Neadle, 72 F.3d 1104, 1110 (3d Cir.1995) ("it is not appropriate to reduce the amount of the loss, as computed under the Guidelines, in order to reflect other causes of the loss which were beyond the defendant's control."); United States v. Shattuck, 961 F.2d 1012, 1016-17 (1st Cir.1992) ("The victim loss table in U.S.S.G. § 2F1.1(b)(1) presumes that the defendant alone is responsible for the entire amount of victim loss specified in the particular loss range selected by the sentencing court." (internal quotation omitted)); see also United States v. Brandon, 17 F.3d 409, 459 n. 83 (1st Cir.), cert. denied, --- U.S. ----, ----, 115 S. Ct. 80, 81, 130 L. Ed. 2d 34 (1994).

The Fifth Circuit also rejected an analogous argument made by two former bank officers in United States v. Stedman, 69 F.3d 737, 740-41 (5th Cir.1995). The defendants there had been convicted of removing material information from a bank's loan files in order to make the loans appear healthier to federal regulators. The defendants' scheme had the effect of concealing the bank's true financial condition from regulators and enabling it to postpone loan loss reserves that otherwise would have been required. Id. at 739. When their scheme was uncovered after the bank's failure, defendants argued that only those losses directly attributable to their fraud should be considered under the Guideline:

[Defendants] urge us to hold that the sentencing loss amount should be only that part of the loss for which their illegal conduct was the cause. They maintain that a portion of the loss was unavoidable, that the bank was financially troubled before they concealed information. Accordingly, they conclude that the sentencing court must determine the loss amount for which their wrongful conduct was the sole cause, and use only that amount in sentencing.

Id. at 740 (footnote omitted). The Fifth Circuit disagreed and affirmed the district court's decision to hold the defendants responsible for all losses suffered by the bank on the subject loans. The court explained that the defendants had "exposed the bank to the possibility of loss for the entire loan amount when they chose to impede regulators from considering information that could have led them to intercede to protect the bank." Id. at 741.

Under this circuit's Miller decision, then, which finds support in the cases of our sister circuits, Morris and Gardner must be held responsible for the total actual loss suffered by the victims of their fraud. The scheme to defraud found by the jury here included a material misrepresentation in the Schnote offering circular as well as the subsequent concealment of material information from the bank's auditors. If the recommendations of the September Analysis had been implemented by Germania, or if they at least had been disclosed to Germania's auditors, it is unlikely that the Schnote offering would have gone forward on the terms that it did. Defendants' scheme therefore facilitated the Schnote offering and caused depositors to put at risk the full value of their investments. In that respect, defendants' conduct here is similar to that of the defendants in Miller, as the fraudulent scheme in that case facilitated procurement of the HUD-insured mortgages, which put HUD at risk for the full amount of those mortgages. See also Sarno, 73 F.3d at 1501 (all losses on fraudulently-procured loan attributable to the defendant even where the default was not his fault; it was reasonably foreseeable from the defendant's conduct that the loan would be approved, putting the bank's money at risk); cf. Johnson, 16 F.3d at 172 (defendant's fraudulent scheme put the bank at risk for the full amount of its loan, although bank suffered no actual loss). Even if defendants' fraud may not have caused Germania's ultimate demise, just as the defendants' fraudulent representations in Miller did not actually cause the depreciation in value of the mortgaged property, the existence of intervening causes does not provide a basis for reducing the amount of loss under section 2F1.1(b)(1). Miller, 962 F.2d at 744; see also id. at 748 (Flaum, J., concurring); Kopp, 951 F.2d at 531. The existence of other possible causes of that loss would only provide a basis for a downward departure under the application note. U.S.S.G. § 2F1.1, App. Note 11; Miller, 962 F.2d at 744; Carey, 895 F.2d at 323.

Defendants argue, however, that Miller's intervening cause rule was upset by this circuit's subsequent decision in United States v. Marlatt, 24 F.3d 1005 (7th Cir.1994), which involved a fraud perpetrated against a title insurance company. The defendant in Marlatt had procured title insurance policies evidencing clear title for time-share condominium properties by concealing from the insurer that the properties were heavily encumbered by liens for unpaid taxes, judgments, and mortgages. The insurer eventually discovered the defendant's fraud and paid $476,000 to clear the titles. In the meantime, however, the value of the properties had plummeted, and third-party purchasers were threatening the insurer with a fraud suit. To avert that, the insurer decided to buy back the properties at the same price purchasers had paid the defendant--a total of $565,000. Id. at 1006; see also Barrett, 51 F.3d at 90 (discussing Marlatt). The district court included this amount as "loss" under section 2F1.1, but we reversed. By removing the title defects, we explained, the title insurer "had done all that it had promised to do in the insurance policies that it had issued." Marlatt, 24 F.3d at 1007. The insurer's subsequent decision to buy the properties back from disgruntled purchasers, we concluded, was not caused by the defendant's fraud and was therefore not chargeable to the defendant as a loss under section 2F1.1:

Application Note 7(b) distinguishes between loss on the one hand and consequential and incidental damages on the other and makes clear that with irrelevant exceptions the latter two items are not to be counted in computing loss for purposes of sentencing under this guideline. The reason for the distinction is no doubt to prevent the sentencing hearing from turning into a tort or contract suit. The distinction is nicely illustrated by this case. The defendant extracted from [the insurer] by fraud a bunch of insurance policies on which Ticor was required to make good to the tune of $476,000. This was the loss. In the wake of the loss, [the insurer] incurred other expenses, which were consequences, perhaps even foreseeable consequences, of the fraud, but were not the thing actually taken from Ticor, the loss; the thing taken was the promise to insure and the cost of honoring that promise was $476,000.

Id. at 1007-08.

This holding, which distinguishes between direct and consequential losses, does not conflict with Miller's intervening cause rule. Where defendants' fraud contributed to the victims' decision to put an amount of money at risk, that amount reflects the thing actually taken by the fraud--here, the face value of a Schnote, and in Marlatt, the promise to insure against title defects on the subject property. It was defendants' fraud in each case that enabled them to obtain what they needed from their victims. Here, for example, defendants' fraud prevented Schnote purchasers from making an informed investment decision, including an assessment of the likelihood that Germania would be unable to make good on its obligations under the Schnotes. See Barrett, 51 F.3d at 91-92. Defendants' scheme also served to conceal the bank's true financial condition from its auditors and federal regulators, either of whom might have interceded to protect the bank and its investors. See Stedman, 69 F.3d at 741. Even if intervening forces ultimately contributed to Germania's demise, the thing actually taken from defrauded investors was still the face amount of the Schnotes, and that amount is thus directly charged to defendants under section 2F1.1. That is the import of this circuit's Miller decision, which finds consistent support in the law of other circuits. See also U.S.S.G. § 2F1.1, App. Note 11 (where a defendant makes misrepresentations in a securities offering that enable securities to be sold at inflated prices, the loss to the victim is the subsequent loss in value even if attributable in substantial part to other causes). Marlatt applies, by contrast, where a victim loses not only the thing actually taken by virtue of the defendant's fraud (i.e., the direct loss), but also suffers consequential or incidental damages. Although a defendant's fraud may be a "but-for" cause of such damages, they are not generally included as loss under section 2F1.1(b)(1). See 24 F.3d at 1007-08; see also Barrett, 51 F.3d at 91.17 The present case is controlled by Miller, not Marlatt. The district court therefore properly interpreted the meaning of "loss" under section 2F1.1 and did not clearly err in estimating the amount of that loss.

Defendants finally contend that the district court erred in refusing to grant them a downward departure to account for other possible causes of the loss charged to them here. Application Note 11 to section 2F1.1 grants to a district court the discretion to depart downward upon finding that the total dollar loss resulting from a defendant's fraud overstates its seriousness, which may be the case "when a misrepresentation is of limited materiality or is not the sole cause of the loss." U.S.S.G. § 2F1.1, App. Note 11; see also Miller, 962 F.2d at 744; Carey, 895 F.2d at 323. Defendants insist that the district court thought it lacked any authority to depart on this basis, but the record belies defendants' assertion. In denying defendants a downward departure at the sentencing hearing, the district court provided the following reasons:

In this particular case, the application note that you refer to says in a few instances the total dollar los[s] that results [from] the offense may overstate its seriousness, and I think that's the key word. I have found that the total dollar loss is five million dollars. Now, I have found that that resulted from the offense, and I can't say that that overstates the seriousness of what's happened here.

Keep in mind that we are in a position where we have had a jury find that these defendants were guilty of certain crimes and the only question in front of us now is the application of these guidelines. I don't see how I can make a finding that this overstates the seriousness of the offense. I'm not sure exactly what that means. Certainly, it was a serious offense to the extent that even a small part of the victims lost in excess of five million dollars....

(July 8, 1994 Tr. at 623.) As these comments illustrate, the district court fully realized that it could depart if it believed that the loss overstated the seriousness of defendants' offense. That is not what the court believed, however, and it therefore denied a departure. This was a discretionary decision that is not subject to review in this court. See United States v. Reynolds, 64 F.3d 292, 298 (7th Cir.1995), cert. denied, --- U.S. ----, 116 S. Ct. 969, 133 L. Ed. 2d 890 (1996); United States v. Gaines, 7 F.3d 101, 105-06 (7th Cir.1993).V.

Having found the evidence sufficient to prove beyond a reasonable doubt the existence of an intentional scheme to defraud, and that Morris and Gardner failed to establish a violation of the government's Brady obligations, we affirm the mail and wire fraud convictions entered on the jury's verdict below. Moreover, we find no merit in defendants' argument that the district court erred in calculating the amount of loss under section 2F1.1 of the Sentencing Guidelines. We therefore also affirm the forty-six month prison sentences imposed below.

AFFIRMED.