Full Opinion

In the United States Court of Appeals For the Seventh Circuit ____________________ No. 24-1822 UNITED STATES OF AMERICA, Plaintiff-Appellee, v. THOMAS E. DUNCAN, Defendant-Appellant. ____________________ Appeal from the United States District Court for the Northern District of Illinois, Eastern Division. No. 1:21-cr-00086-1 — Steven C. Seeger, Judge. ____________________ ARGUED SEPTEMBER 11, 2025 — DECIDED AUGUST 6, 2026 ____________________ Before BRENNAN, Chief Judge, and KIRSCH and JACKSON- AKIWUMI, Circuit Judges. JACKSON-AKIWUMI, Circuit Judge. Thomas Duncan, a for- mer supervisor at the Department of Veterans Affairs (VA), participated in a scheme to defraud the VA for over six years. For his part, Duncan submitted phantom orders for blood pressure cuffs from his co-defendant’s medical supply com- pany in exchange for kickbacks when the VA paid for the or- ders. Duncan eventually pleaded guilty to wire fraud. At 2 No. 24-1822 sentencing, he argued that he received one bribe with a stream of benefits and was responsible for only half of the VA’s nearly $1.9 million loss. The district court disagreed and imposed enhancements for multiple bribes and a loss calcula- tion of over $1.5 million, before sentencing Duncan to 84 months in prison. Seeing no error in the court’s Sentencing Guidelines calculation, and because any error would be harmless, we affirm. I In Thomas Duncan’s role at the Jesse Brown VA Medical Center in Chicago, he was responsible for purchasing medical supplies for and monitoring inventory at various VA hospi- tals. Duncan also had authority to direct his subordinates to place orders and document that orders had been received. In 2012, Duncan and his co-defendant, Daniel Dingle, de- vised a plan to defraud the VA in exchange for kickbacks. Pur- suant to the plan, Duncan submitted (or directed others to submit) purported orders for items, mostly blood pressure cuffs, from Dingle’s medical supply company. Most of the or- ders followed one of two patterns for 600 or 680 cuffs (Pattern Orders). To avoid detection, Duncan submitted orders that re- sulted in total costs just below his preauthorization threshold, which allowed him to avoid obtaining additional approval. Over the course of the scheme, Duncan’s pre-authorization threshold increased from $3,000 to $10,000, enabling him to place larger orders without detection. Dingle received the orders and the VA’s payment for the orders, but he did not fulfill them. Instead, he pocketed the money and paid Duncan a portion as a kickback. And because Duncan knew the supplies from Dingle would never arrive, No. 24-1822 3 he marked (or instructed other employees to mark) the sup- plies as received even though they had not been. In practice, the VA repeatedly paid Dingle for medical supplies it did not receive, and Duncan ensured that the scheme could continue by placing fake orders and marking them received in ex- change for payment from Dingle. In total, the VA paid Din- gle’s company almost $1.9 million for blood pressure cuffs over the course of the scheme. Over $1.7 million of that sum corresponded to Pattern Orders. After an investigation uncovered the scheme, the govern- ment charged Duncan with, among other charges not relevant here, five counts of wire fraud in violation of 18 U.S.C. § 1343. Duncan pleaded guilty to one of those counts. Before sentencing, the probation office prepared a Presen- tence Investigation Report (PSR) recommending 84 months’ imprisonment. In their sentencing memoranda, the parties agreed to the facts in the PSR but disagreed on whether Dun- can’s Sentencing Guideline range should include enhance- ments related to the number of bribes and the loss calculation. With respect to the number of bribes, Duncan argued that the fraud involved a single bribe pursuant to which Dingle paid him a series of benefits. The government argued that the fraud involved multiple bribes because the two men did not have an upfront agreement and negotiated the terms of their ar- rangement over time. As to the loss calculation, the parties did not dispute how many orders the VA paid Dingle for, or what portion of the payments related to Pattern Orders. They disputed only what portion of the Pattern Orders were attributable to Duncan. The government contended that Duncan was responsible for all Pattern Orders and thus a loss amount to the VA of $1.7 4 No. 24-1822 million. Duncan asserted that he was responsible for only $854,833.84 of the VA’s loss—the portion of the Pattern Or- ders for which there was evidence that he submitted the order or instructed another employee to do so. After asking the parties for briefing related to the VA’s loss and holding a two-day sentencing hearing, the district court agreed with the government on both issues. The court found that the fraud involved multiple bribes because Dingle made several payments to Duncan over many years and the scheme was “ad hoc and intermittent.” As to the loss calculation, the court found that although the government did not have writ- ten documentation linking Duncan to each phantom pur- chase, Duncan’s repeated instructions to others to submit or- ders and the momentum of the scheme suggested that he was responsible for all Pattern Orders. The district court calculated the applicable Guidelines range beginning with a base offense level of 14 because Dun- can was a public official. U.S.S.G. § 2C1.1(a)(1). Relevant here, the court applied two enhancements: (1) a two-level enhance- ment under § 2C1.1(b)(1) because the offense involved more than one bribe; and (2) a 16-level enhancement pursuant to §§ 2C1.1(b)(2) and 2B1.1(b)(1)(I)–(J) because the loss amount was greater than $1.5 million but less than $3.5 million. The court made other adjustments to Duncan’s offense level based on findings not relevant on appeal and landed at a total of- fense level of 29. With a criminal history category of I, the dis- trict court calculated a Guidelines range of 87 to 108 months’ imprisonment. The court sentenced Duncan to 84 months, which it deemed appropriate irrespective of the Guidelines calculation. Duncan now appeals the court’s inclusion of the challenged enhancements. No. 24-1822 5 II We review the district court’s legal interpretations at sen- tencing and application of the Guidelines de novo, and its fac- tual findings for clear error. United States v. White, 883 F.3d 983, 986 (7th Cir. 2018). A factual finding is not clearly errone- ous if it is consistent with a permissible inference from the ev- idence. United States v. Kyereme, 127 F.4th 702, 706 (7th Cir. 2025). We will deem the court’s findings clearly erroneous only if “after considering all of the evidence, [we are] left with the definite and firm conviction that a mistake has been made.” United States v. Dickerson, 42 F.4th 799, 804 (7th Cir. 2022) (citation modified). Duncan challenges the district court’s application of the multiple bribe enhancement as well as its conclusions related to the amount of the VA’s loss that was attributable to him. We address each issue in turn. A. Number of Bribes Pursuant to U.S.S.G. § 2C1.1(b)(1), a two-level sentencing enhancement is warranted if the offense a defendant was con- victed of involved multiple bribes. Commentary to that sec- tion instructs courts to treat “[r]elated payments that, in es- sence, constitute a single incident of bribery” as one bribe. U.S.S.G. § 2C1.1, app. n.2. Drawing on that language, Duncan maintains that, although he received multiple payments, he and Dingle had one agreement pursuant to which “his posi- tion as a government official was utilized on an as needed ba- sis—whenever Dingle came calling.” On that basis, he argues that he received a single bribe “that involve[d] a stream of benefits over time, not just singly negotiated deals.” United States v. Solomon, 892 F.3d 273, 277 (7th Cir. 2018). 6 No. 24-1822 To be sure, Duncan’s theory is plausible on this record. But the record also supports the district court’s conclusion that Duncan’s fraud involved more than one bribe. The scheme occurred over six years, at inconsistent times, for inconsistent amounts, and was negotiated over time. Given these facts, the court’s finding of multiple bribes in this case was not unrea- sonable. Nevertheless, Duncan accuses the district court of com- mitting two errors: (1) concluding that an upfront agreement is the only way to show a single bribe and (2) treating the lack of such an agreement here as dispositive. Neither accusation holds water. Duncan is correct that circumstances can evince a single bribe despite the lack of an upfront agreement. See U.S.S.G. § 2C1.1, app. n.2 (referencing “a number of install- ment payments for a single action” as one indicia of a single bribe). But nothing in the court’s analysis suggests that it mis- interpreted or misapplied that rule. Rather, the record demonstrates that the court found Duncan accepted multiple bribes based on the scheme’s duration, the co-defendants’ continued negotiations, and the absence of common indicia of a single bribe. On this record, we cannot say that finding was clearly erroneous. See United States v. Cruz–Rea, 626 F.3d 929, 938 (7th Cir. 2010) (“[W]hen a district court chooses between two permissible inferences from the evidence, the factual findings cannot have been clearly erroneous.”). Duncan’s challenge therefore fails. B. Loss Calculation Duncan’s second challenge is to the district court’s conclu- sions about the amount of the VA’s loss attributable to him. We review the district court’s methodology for calculating No. 24-1822 7 loss de novo and its ultimate calculation for clear error. United States v. Miller, 153 F.4th 531, 537 (7th Cir. 2025). A defendant’s offense level is subject to an enhancement corresponding to the amount of loss the government suffered because of his conduct. U.S.S.G. § 2C1.1(b)(2) (citing U.S.S.G. § 2B1.1). A defendant may be held accountable only for losses directly attributable or reasonably foreseeable to him. United States v. Newton, 76 F.4th 662, 672 (7th Cir. 2023). If the loss exceeds $550,000, the offense level is increased by 14; if the loss is greater than $1.5 million but less than $3.5 million, a 16- level enhancement applies. U.S.S.G. § 2B1.1(b)(1)(H)–(J). The government must establish the loss amount by a pre- ponderance of the evidence, but the district court need only make a “reasonable estimate” of loss based on the evidence. Miller, 153 F.4th at 537. That is because, as Duncan concedes, courts “must invariably estimate the resulting loss amount” when determining the amount attributable to a defendant in a fraud scheme. United States v. Collins, 949 F.3d 1049, 1053 (7th Cir. 2020) (“Bearing in mind that the court had to make the best estimate it could, and that Collins had no incentive to make his conduct transparent, we find that the court did not clearly err in its calculations.”). Accordingly, we will not disturb a district court’s findings related to the loss amount attributable to a defendant where the court carefully considers the scope of a defendant’s par- ticipation in the scheme. Newton, 76 F.4th at 672. The same holds true where the court carefully considers the scope of the scheme and resulting loss. United States v. Gumila, 879 F.3d 831, 834 (7th Cir. 2018). Thus, to succeed on appeal, Duncan must establish that “the court’s loss calculations were not only inaccurate but outside the realm of permissible 8 No. 24-1822 computations.” United States v. White, 737 F.3d 1121, 1142 (7th Cir. 2013) (citation modified). Duncan claims in his brief that the government relied “on the absence of evidence to substantiate its loss amount calcu- lation” and thus failed to “meet its burden of proof with re- spect to the $370,144 difference” between its proposed loss amount and the losses linked to him. To support his position, Duncan highlights the government’s concession that it had tangible evidence linking him to Pattern Orders totaling only approximately $1.3 million. But the lack of tangible evidence for a portion of the loss did not prevent the district court from concluding that the government met its burden. Indeed, we have said that a district court “need only make ’a reasonable estimate of the loss’ in applying the enhancement.” White, 737 F.3d at 1142 (quoting U.S.S.G. § 2B1.1, app. n.3(C), which is now app. n.3(B)). Thus, although the district court was not permitted to speculate that Duncan was responsible for all the VA’s losses, United States v. Erving, 164 F.4th 953, 964 (7th Cir. 2026), it could attribute losses from the scheme to Duncan based on reasonable inferences drawn from the record. See White, 737 F.3d at 1141–42. The district court followed that rule here. The parties presented plausible theories of loss at sentenc- ing leaving the district court to decide whether written con- firmation of Duncan’s involvement in each Pattern Order was necessary to attribute the loss from that order to him, or if the pattern alone was sufficient. To aid in its decision, the court ordered supplemental briefing and permitted extensive argu- ment on the issue during the two-day sentencing hearing. That the court ultimately sided with the government does not show clear error. Collins, 949 F.3d at 1054 (“The district court No. 24-1822 9 accepted the government’s interpretation of the evidence, and we cannot say that it clearly erred in doing so.”). Moreover, Duncan’s expert concluded that $854,833.84 of the VA’s losses from Pattern Orders linked to Duncan was “a reasonably certain loss amount attributable to” Duncan. Even accepting this expert’s conclusion, it is still plausible that Duncan also could have been responsible for the remaining Pattern Orders that were not explicitly linked to him. This is especially so considering that Duncan sometimes directed his subordinates to place orders and mark them delivered. The district court could have inferred from this that Duncan’s sub- ordinates failed to record his name alongside each order placed at his direction. See Miller, 153 F.4th at 537 (“A defend- ant challenging a loss calculation bears a heavy burden: he must show that the court’s loss calculations were not only in- accurate but outside the realm of permissible computations.”) (citation modified). Furthermore, the government presented circumstantial evidence that supported the court’s conclusion. The record in- dicates that (1) the Pattern Orders were fake, (2) the type of blood pressure cuffs at issue were rarely used, and (3) a nurse at one of the hospitals Duncan supplied told investigators that her hospital usually had only between three and five of that type of blood pressure cuff available at a time. Although this evidence fell short of conclusively demonstrating that Dun- can orchestrated all Pattern Orders, it was sufficient for the court to calculate “a reasonable estimate of the loss, which is all that is required.” Miller, 153 F.4th at 538. As such, the dis- trict court did not clearly err by attributing $1.7 million of the VA’s loss to Duncan. 10 No. 24-1822 C. Harmless Error Even if we agreed with Duncan that either or both en- hancements were improper, reversal would not be warranted in this case. To be sure, Duncan would have had a lower Guidelines range without the enhancements. And errors that impact the Guidelines range are presumptively prejudicial. Molina-Martinez v. United States, 578 U.S. 189, 201 (2016) (“In- deed, in the ordinary case a defendant will satisfy his burden to show prejudice by pointing to the application of an incor- rect, higher Guidelines range and the sentence he received thereunder.”). However, a “court’s statement that it would have imposed the same sentence, even if it had improperly calculated the defendant’s Guidelines range, renders the error harmless.” United States v. Cook, 108 F.4th 574, 583 (7th Cir. 2024) (citation modified). That is the case here. The district court expressly stated that it would have ren- dered the same sentence without the enhancements. And to support its view that an 84-month sentence was appropriate regardless of its rulings on the enhancements, the court refer- enced the duration of the fraud, the number of payments Duncan received, and the breach of trust occasioned by Dun- can’s misuse of his position. The court’s pronouncement was a far cry from the “generic disclaimer and conclusory” state- ment Duncan rightly observes would merit a remand. It was enough to render any alleged error harmless. United States v. Caraway, 74 F.4th 466, 468 (7th Cir. 2023) (“Where a district court unambiguously states that it would have imposed the same sentence regardless of any potential error, the error is harmless.”). We also reject Duncan’s contention that the court was nec- essarily “anchored by the Guidelines” because it imposed the No. 24-1822 11 sentence recommended in the PSR. The case Duncan relies on to support this argument, United States v. Bravo, 26 F.4th 387, 396 (7th Cir. 2022), is inapposite. There, the district court ex- plained its choice of a sentence by way of mathematical cal- culations that used the top of the Guidelines range as a start- ing point. We held that the court’s calculation was, therefore, “firmly anchored” in the Guidelines, necessitating a remand for resentencing once we ruled that the Guidelines range had been miscalculated. Id. at 397. Here, by contrast, in sentencing Duncan, the district court acknowledged that the Guidelines are “a helpful yardstick, but they’re only a yardstick,” and stated that it had “looked at the totality of the evidence, all the facts in the case.” The court then explicitly referenced evi- dence in the record before concluding that Duncan’s “con- duct, taken as a whole, deserve[d] a sentence of 84 months.” On this record, Duncan’s challenge cannot survive the harm- less error standard. III The district court’s factual findings did not align with Duncan’s theory of the case. But that does not make them er- roneous. In addition, the court believed that 84 months’ im- prisonment was an appropriate sentence for Duncan’s crime. That the court’s view was consistent with the PSR’s recom- mendation does not show impropriety. Because the district court’s findings are supported by the record and any error would be harmless, the judgment is AFFIRMED.