United States v. Rainford

Full title: United States of America, Appellee, v. Ryan Rainford, Bryan Duncan, Robert…

Court: United States Court of Appeals, Second Circuit

Date published: Aug 2, 2024

110 F.4th 455 (2d Cir. 2024)

Opinion

Nos. 20-359 20-2695 20-2993 21-1753

08-02-2024

UNITED STATES of America, Appellee, v. Ryan RAINFORD, Bryan Duncan, Robert Locust, Defendants-Appellants, Peter Kalkanis, Kerry Gordon, aka Curry, Defendants.*

Alexandra N. Rothman, Assistant United States Attorney (Nicholas W. Chiuchiolo, Nicholas S. Folly, David Abramowicz, Assistant United States Attorneys, on the brief), for Damian Williams, United States Attorney for the Southern District of New York, New York, NY, for Appellee. Donna R. Newman, Law Offices of Donna R. Newman, PA, New York, NY, for Defendant-Appellant Ryan Rainford. Bryan Duncan, pro se, for Defendant-Appellant Bryan Duncan. Randall Douglas Unger, Law Offices of Randall Douglas Unger, Kew Gardens, NY, for Defendant-Appellant Robert Locust.

Menashi, Circuit Judge

On Appeal from the United States District Court for the Southern District of New York

Alexandra N. Rothman, Assistant United States Attorney (Nicholas W. Chiuchiolo, Nicholas S. Folly, David Abramowicz, Assistant United States Attorneys, on the brief), for Damian Williams, United States Attorney for the Southern District of New York, New York, NY, for Appellee.

Donna R. Newman, Law Offices of Donna R. Newman, PA, New York, NY, for Defendant-Appellant Ryan Rainford.

Bryan Duncan, pro se, for Defendant-Appellant Bryan Duncan.

Randall Douglas Unger, Law Offices of Randall Douglas Unger, Kew Gardens, NY, for Defendant-Appellant Robert Locust.

Before: Jacobs, Menashi, and Merriam, Circuit Judges.

Judge Jacobs concurs in a separate opinion. Judge Merriam concurs in part and dissents in part in a separate opinion.

Menashi, Circuit Judge:

Defendants-Appellants Ryan Rainford, Robert Locust, and Bryan Duncan appeal their convictions and sentences for conspiracy

to commit mail and wire fraud in violation of 18 U.S.C. § 1349.^{1 1} Duncan proceeds pro se on appeal.

1 Duncan was convicted of two counts of conspiracy to commit mail and wire fraud as well as one substantive count of mail fraud and one substantive count of wire fraud.

1 The District Court imposed these adjustments on all three defendants, but only Rainford expressly challenges the adjustments on appeal.

The convictions arose from a fraudulent slip-and-fall scheme that the defendants and others orchestrated. The scheme involved recruiting poor and homeless people to fake accidents at properties around the New York area. The recruit would stage an accident and then seek unnecessary medical treatment—sometimes including surgery—from doctors who were part of the scheme. The organizers of the scheme would then refer the recruit to a lawyer, who would sue the property owner or the owner's insurance company for damages. The proceeds from the lawsuits, which often settled, were then divided among the co-conspirators, with the recruits receiving relatively little.

The defendants raise several arguments on appeal. We affirm with respect to each issue relating to the trial and convictions. See infra Part I. We affirm the judgment with respect to the sentencing guidelines calculations, but we remand for factfinding as to the number of fraudulent accidents orchestrated by the conspiracy while Rainford and Locust were members for the purpose of performing a loss calculation under U.S.S.G. § 2B1.1. See infra Part II. Finally, we vacate and remand Duncan's forfeiture order, affirm but modify the restitution order for Rainford and Locust, and affirm Rainford's sentence but remand for reconsideration in the interest of justice. See infra Part III.

BACKGROUND

The conduct underlying this appeal involves two fraudulent slip-and-fall schemes. The first scheme began around 2013 and included Rainford, Duncan, and Locust. Peter Kalkanis was the principal organizer of the first scheme, and Rainford, Duncan, and Locust were lower-level co-conspirators known as "runners." The runners would seek out people who were often poor or homeless. They would then find suitable locations for slip-and-fall accidents and instruct a recruit to stage a fall at the location and to seek medical attention for nonexistent injuries. Sometimes, the unnecessary medical attention included surgery.

The recruit was then referred to a lawyer who would pursue a personal injury lawsuit on his or her behalf. Kalkanis would typically sit in on a recruit's meeting with the attorney. At the meeting, Kalkanis would record pertinent information on an "intake sheet," which included the name of the recruit as well as others involved in the "accident," including the runner who referred the recruit to the scheme. Notably, some of the intake sheets processed in this way were not fraudulent but documented genuine slip-and-fall accidents and legitimate legal claims. When asked at trial how many of the intake sheets involved fraudulent slip-and-falls, Kalkanis initially testified that "[a]t least 80 percent" were fraudulent. Rainford App'x 884. Kalkanis then backtracked, saying that "practically all of them" were fraudulent. Id. When asked to clarify, Kalkanis said "the majority of them" were fraudulent. Id. The district court noted that Kalkanis had given different answers to the same question; Kalkanis then reiterated that "[t]he majority of them" were fraudulent. Id. When asked how many cases he managed during the first conspiracy, Kalkanis estimated that there were "[a]pproximately 300, if not more." Id. at 883.

The runners also ensured that the recruits attended medical and legal appointments

by transporting them to those appointments. The organizers of the scheme would arrange for litigation funding companies to underwrite the medical expenses and litigation. While litigation proceeded, the organizers would often arrange "loans" to the recruits from the litigation funding companies to pay expenses. One witness testified that he used his loan to pay for "anything, whether it be rent, bills" as well as "to pay the medical facilities for the surgeries that [he] would need." Id. at 751. The lawsuits frequently resulted in settlements, often for six figures. The proceeds were distributed among the organizers, doctors, lawyers, litigation funders, and others involved in the scheme. The recruit would receive what was left.

One recruit who testified at trial was Yvette Battle. After staging a fraudulent slip-and-fall, Battle underwent knee surgery for which she received anesthesia. She was compensated with $1,000 along with cookies and juice. See id. at 685. A "couple of months later," she underwent a shoulder surgery and in exchange for the surgery she was paid $1,000. Id. at 685-88. She filed an action against the owner of the property where she staged the accident. That action was "[d]ismissed" and she received nothing of value from that lawsuit. Id. at 689. During the government's examination of Battle, the prosecutor referred to these $1,000 payments as "loans," id., and Battle did not correct that characterization. Some recruits whose cases settled received larger payouts. One recruit who underwent shoulder surgery received $19,000 out of a $100,000 settlement, and another who underwent back and knee surgery received $35,000 out of a $225,000 settlement.

In 2015, Duncan and Kerry Gordon—another co-conspirator in the Kalkanis scheme—began a spin-off scheme. That scheme was substantially similar to the Kalkanis scheme, often using the same attorneys, doctors, and low-level co-conspirators. Duncan and Gordon created a business entity—D&G Premier Solutions LLC ("D&G")—to operate the scheme. D&G would connect recruits with litigation funding companies. D&G would receive a referral fee from a funding company after the company contracted with a recruit to provide payments in exchange for the recruit's future settlement amount.

In 2019, a six-count superseding indictment was returned against the co-conspirators in the two schemes. Rainford, Duncan, and Locust were each charged with three counts in connection with the Kalkanis scheme: conspiracy to commit mail and wire fraud (Count One), mail fraud (Count Two), and wire fraud (Count Three). Id. at 60-64. Duncan was indicted on three additional counts relating to the spin-off scheme: conspiracy to commit mail and wire fraud (Count Four), mail fraud (Count Five), and wire fraud (Count Six). Id. at 64-67.

The government presented extensive evidence of guilt at trial. Almost a dozen recruits testified that they had participated in the schemes by staging accidents and receiving unnecessary medical treatment. Kalkanis also testified. He stated that he "directed the traffic" in the first scheme and "was a manager in th[e] whole thing." Id. at 819. Kalkanis elaborated that the scheme was fraudulent because "these weren't real accidents." Id. And he identified Locust, Rainford, and Duncan as members of the conspiracy. Id. at 819-20. The government also introduced the intake sheets, medical records, and communications between the co-conspirators.

The jury found Rainford, Locust, and Duncan guilty of conspiracy to commit mail and wire fraud in connection with the Kalkanis scheme (Count One). See id. at

1226.^{2 2} The jury also found Duncan guilty of conspiracy to commit mail and wire fraud (Count Four), mail fraud (Count Five), and wire fraud (Count Six) in connection with the spin-off scheme. This appeal followed.

2 The district court declared a mistrial as to Counts Two and Three; the government then moved to dismiss those counts, and the district court granted the motion. See Rainford App'x 1353.

2 Furthermore, financial losses suffered by criminally liable co-conspirators should not be considered in calculating the Guidelines. Cf. United States v. Harris, 821 F.3d 589, 606 (5th Cir. 2016) (finding that losses suffered by "willing participants in the fraudulent scheme" are not countable in calculating the loss amount under the Guidelines).

DISCUSSION

Rainford, Duncan, and Locust raise several arguments on appeal. We begin with the arguments relating to the trial and convictions, and we affirm the judgment of the district court with respect to those issues. See infra Part I. We next consider the defendants' challenges to their sentencing guidelines calculations. We affirm the judgment with respect to the calculations, but we remand for additional fact-finding relating to the loss enhancements for Rainford and Locust. See infra Part II. Finally, we address the defendants' arguments about their sentences. We vacate and remand Duncan's forfeiture order because the district court relied only on representations by the government, not on evidence, in calculating the forfeiture amount. We affirm the district court's restitution order for Rainford and Locust, but we modify the order as the parties agree. And we affirm Rainford's sentence of imprisonment but remand with instructions to reconsider it in the interest of justice. See infra Part III.

I

First, we consider the defendants' challenges to their convictions.

A

Duncan argues that his due process rights were violated because the government introduced false testimony by Alvin Martin, Reginald Dewitt, and Tina Nichols. The government may not knowingly introduce false evidence or testimony to obtain a conviction. See United States v. Alston, 899 F.3d 135, 146 (2d Cir. 2018). A witness does not perjure himself merely by giving incorrect, confusing, or mistaken testimony: "Simple inaccuracies or inconsistencies in testimony do not rise to the level of perjury." United States v. Monteleone, 257 F.3d 210, 219 (2d Cir. 2001). Rather, a witness commits perjury "if he gives false testimony concerning a material matter with the willful intent to provide false testimony." United States v. Aquart, 912 F.3d 1, 20 (2d Cir. 2018) (quoting Monteleone, 257 F.3d at 219).

Duncan has identified no false testimony by Martin, Dewitt, or Nichols that was material to his conviction and that the government knew was false. Duncan points to (1) Martin's testimony that Kalkanis was his attorney; (2) Martin's confusion as to whether his attorney's office was in Astoria, Queens; (3) Martin's testimony as to who entered a meeting with the lawyer at a certain time; (4) Martin's recollection of a person's name; and (5) Martin's testimony as to whether his mother lied to the government. But Duncan has not established that any of this testimony—assuming it was false—was material to his conviction or that the government knew the testimony was false. The same is true of Duncan's claim that Dewitt perjured himself by stating that he was not involved in any cases in 2014, despite an intake form from 2014 listing him as a runner for the conspiracy.

Finally, Duncan notes that Nichols testified that she had been recruited by Duncan, but other pieces of evidence indicated

that Dewitt made the referral. This may not be a contradiction—recruitment and referral may be different concepts—but in any event Duncan has not established how the testimony was material. He has identified, at most, a "[s]imple ... inconsistenc[y]." Monteleone, 257 F.3d at 219.

Because Duncan has not identified a material falsity known to the government, we affirm with respect to this issue.

B

Duncan also argues that the district court erred by admitting intake sheets—documents Kalkanis created that recorded pertinent information about each slip-and-fall accident—into evidence. Duncan's brief does not make clear the basis of his argument. But because "[a] document filed pro se is to be liberally construed," Boykin v. KeyCorp, 521 F.3d 202, 214 (2d Cir. 2008) (quoting Erickson v. Pardus, 551 U.S. 89, 94, 127 S.Ct. 2197, 167 L.Ed.2d 1081 (2007)), we interpret Duncan's brief as (1) arguing that the intake sheets were not relevant because Kalkanis testified he had no personal knowledge of which cases were fraudulent and which were legitimate and (2) raising a Confrontation Clause claim.

We review a district court's evidentiary decisions for abuse of discretion. See United States v. Persico, 645 F.3d 85, 99 (2d Cir. 2011). Duncan's argument that the intake sheets were irrelevant is mistaken. Kalkanis testified that he personally filled out the intake sheets and that he did so while meeting with the recruit. See Rainford App'x 857 ("These are combination intake sheets that I did with each individual patient."). Even if Kalkanis himself could not say which intake sheets involved fraudulent slip-and-falls and which involved legitimate accidents, we cannot conclude that the intake sheets documenting the accidents in the scheme had no "tendency to make a fact more or less probable than it would be without the evidence." Fed. R. Evid. 401(a).

Separately, Duncan contends that intake sheets were admitted for recruits who did not testify, which violated his Sixth Amendment right "to be confronted with the witnesses against him." U.S. Const. amend. VI. The Confrontation Clause prohibits the admission of "testimonial statements of a witness who did not appear at trial unless he was unavailable to testify, and the defendant had had a prior opportunity for cross-examination." Crawford v. Washington, 541 U.S. 36, 53-54, 124 S.Ct. 1354, 158 L.Ed.2d 177 (2004). In this case, Duncan had the opportunity to cross-examine Kalkanis, who drafted the intake sheets, so there was no violation of the Confrontation Clause.

We affirm the district court's judgment with respect to the admission of the intake sheets into evidence.

C

Locust argues that the district court's summary denial of his motion for appointment of new counsel deprived him of his right to effective assistance of counsel. We disagree.

We review the denial of a motion to substitute counsel for abuse of discretion. United States v. Simeonov, 252 F.3d 238, 241 (2d Cir. 2001). We consider four factors: (1) whether the motion for new counsel was timely; (2) whether the district court adequately inquired into the matter; (3) whether the conflict between the defendant and his attorney was so great that it caused a lack of communication and prevented an adequate defense; and (4) whether the defendant substantially and unjustifiably contributed to the breakdown in communication. United

States v. Hsu, 669 F.3d 112, 122-23 (2d Cir. 2012).

Locust's argument fails because he waived the argument before the district court. One trial day after Locust's motion was denied, the district court directly asked Locust: "[Y]ou've indicated that you don't want to go pro se here and you want to continue with Dinnerstein and Cecutti [Locust's trial counsel], correct?" Supp. App'x 25-26. Locust replied: "Yes, sir." Id. at 26. That statement was an "intentional relinquishment or abandonment of a known right, and ... permanently extinguishe[d] the right to raise the claim" on appeal. United States v. Polouizzi, 564 F.3d 142, 153 (2d Cir. 2009) (internal quotation marks omitted).

Locust responds that he did not waive the argument because the district court was clear that any motion for new counsel would have been futile. But the cases on which Locust relies for that proposition—despite using the term "waiver"—each address the failure to object, not the intentional relinquishment of a right. See Anderson v. Branen, 17 F.3d 552, 556-57 (2d Cir. 1994); Ostrowski v. Atl. Mut. Ins. Cos., 968 F.2d 171, 177-79 (2d Cir. 1992). Locust misses this critical distinction. "Waiver is different from forfeiture. Whereas forfeiture is the failure to make the timely assertion of a right, waiver is the intentional relinquishment or abandonment of a known right." United States v. Olano, 507 U.S. 725, 733, 113 S.Ct. 1770, 123 L.Ed.2d 508 (1993) (emphasis added) (internal quotation marks omitted). We have "discretion to correct errors that were forfeited because not timely raised in the district court, but no such discretion applies when there has been true waiver." United States v. Spruill, 808 F.3d 585, 596 (2d Cir. 2015). Locust's response when the district court asked if he wished to proceed was intentional and affirmative—and it qualified as a waiver. Accordingly, Anderson and Ostrowski provide no basis for applying a discretionary exception. Locust has waived this argument.

Even if he had not waived it, the argument would fail. Locust's motion for new counsel was made during the trial. "[O]nce trial has begun, a defendant has no unbridled right to reject assigned counsel and demand another" because defendants may "manipulat[e] ... the right so as to obstruct the orderly procedure in the courts or to interfere with the fair administration of justice." United States v. John Doe No. 1, 272 F.3d 116, 122 (2d Cir. 2001) (internal quotation marks omitted). Locust contends that the motion was timely because it responded to his counsel's performance at trial. Indeed, Locust identifies several comments by the trial judge indicating dissatisfaction with Locust's counsel. That may affect our analysis of the timeliness prong. See Hsu, 669 F.3d at 122. But it cuts against Locust overall. If Locust's belief that his counsel was deficient was based on his counsel's performance at trial, then the district court did not need to make a special inquiry into his counsel's performance; the district court had witnessed the counsel's performance during the trial. And the district court evaluated that performance. In denying Locust's motion, the district court stated that it "f[ou]nd the representation of [Locust's trial counsel] to be quite good." Locust App'x 37. Moreover, Locust confirmed to the district court that the purported deficiency was "due to the conduct of the trial," id. at 38, and was not based on a lack of communication, see Hsu, 669 F.3d at 123.

Locust suggests that the district court should have asked more targeted questions. But, on these facts, we see no abuse of discretion in failing to inquire further about the counsel's conduct at trial. Locust's

challenge on appeal therefore would fail on the merits even if it had not been waived.

We affirm the district court's judgment with respect to the denial of Locust's motion for new counsel.

D

Locust further argues that the prosecutor committed "several serious improprieties" at closing argument, prejudicing his right to a fair trial. Locust Br. 32.

Prosecutorial remarks "do not amount to a denial of due process unless they constitute 'egregious misconduct.'" United States v. Shareef, 190 F.3d 71, 78 (2d Cir. 1999) (quoting Donnelly v. DeChristoforo, 416 U.S. 637, 647, 94 S.Ct. 1868, 40 L.Ed.2d 431 (1974)). So long as a prosecutor does not "misstate the evidence," he or she is entitled to "wide latitude during closing arguments." United States v. Tocco, 135 F.3d 116, 130 (2d Cir. 1998). Even if a prosecutor's remarks were improper, a defendant will succeed on a misconduct claim only when "the remarks, taken in the context of the entire trial, resulted in substantial prejudice." United States v. Thomas, 377 F.3d 232, 244 (2d Cir. 2004) (quoting United States v. Perez, 144 F.3d 204, 210 (2d Cir. 1998)). When evaluating prejudice, we consider three factors: (1) "the severity of the misconduct," (2) "the measures adopted to cure the misconduct," and (3) "the certainty of conviction absent the misconduct." United States v. Elias, 285 F.3d 183, 190 (2d Cir. 2002). A new trial is ordered only in a "rare case." United States v. Rodriguez, 968 F.2d 130, 142 (2d Cir. 1992) (quoting Floyd v. Meachum, 907 F.2d 347, 348 (2d Cir. 1990)). When the defendant failed to object to the challenged remarks, we review for plain error and will not reverse unless the remarks "amount to flagrant abuse which seriously affects the fairness, integrity, or public reputation of judicial proceedings, and causes substantial prejudice to the defendant." United States v. Williams, 690 F.3d 70, 75 (2d Cir. 2012) (internal quotation marks omitted).^{3 3}

3 "The Supreme Court has identified four prongs of plain error analysis: (1) there must be an error; (2) the error must be plain, meaning it must be clear or obvious, rather than subject to reasonable dispute; (3) the error must have affected the appellant's substantial rights in that it affected the outcome of the proceedings; and (4) if these other three prongs are satisfied, the court of appeals has the discretion to remedy the error if the error seriously affects the fairness, integrity or public reputation of judicial proceedings." United States v. Montague, 67 F.4th 520, 528 (2d Cir. 2023) (internal quotation marks and alteration omitted), judgment vacated on other grounds, No. 23-959, ___ U.S. ___, ___ S.Ct. ___, ___ L.Ed.2d ___, 2024 WL 3014465 (U.S. June 17, 2024).

3 The majority opinion focuses on the alleged vulnerability of the co-conspirators. See Maj. Op. at 483-84 ("Even if they engaged in the scheme voluntarily, the organizers of the conspiracy took advantage of their economic vulnerability to lead them to undertake serious physical and legal risks."). But being a "victim" and being "vulnerable" are two separate inquiries. The question of whether one is vulnerable should not impact whether one is a victim. It is an unfortunate reality that the vast majority of criminal defendants, particularly lower-level members of criminal conspiracies, are vulnerable and desperate in various ways. They are often in dire financial straits, mentally ill, and/or suffering from addiction. Nonetheless, we do not treat those coconspirators as victims of their own crimes, even when they suffer bodily harm, financial harm, or other serious consequences as a result of their participation in an offense.

Locust identifies four purported improprieties. None require reversal.

First, Locust contends that the government vouched for four witnesses: Tucker, Dewitt, Kalkanis, and Martin. Locust did not object to the comments that allegedly vouched for those witnesses, so we review for plain error. Locust explains that the prosecutor vouched when she made statements such as "Dewitt's telling the truth." Locust Br. 35. But the government surrounded these statements with references to items submitted into evidence, so the statements "turn out on closer examination to be permissible reference to the evidence in the case" rather than vouching. Perez, 144 F.3d at 210. In context, the government's statements that the witnesses were telling the truth "did not imply the existence of extraneous proof and cannot be characterized as improper vouching." Williams, 690 F.3d at 76 (internal quotation marks omitted). Even if the statements did amount to vouching, such

statements do not constitute "flagrant abuse." Id. at 75. The district court did not plainly err by failing to intervene.

Second, Locust argues that the prosecutor "denigrated" the defense by saying, for example, that its case was a "total sideshow." Locust Br. 33; see Locust App'x 113. Locust did not object to these statements, so we again review for plain error. One component of Locust's defense was that the government should have focused on the corrupt lawyers rather than low-level co-conspirators such as Locust. See Supp. App'x 1, 61. In describing that strategy as a "sideshow," the prosecutor responded to the defense's arguments, which is permissible in a closing argument. See United States v. Salameh, 152 F.3d 88, 139 (2d Cir. 1998) ("[T]he Government is ordinarily permitted to respond to arguments impugning the integrity of its case and to reply with rebutting language suitable to the occasion.") (quoting United States v. Bagaric, 706 F.2d 42, 60 (2d Cir. 1983)). We again see no plain error.

Third, Locust claims that the prosecutor misstated the evidence when she said that Locust orchestrated three slip-and-fall accidents. Locust did not object when the government stated in closing that Locust brought three recruits into the scheme—Gilford, Roberts, and Wright—so we review for plain error. Locust notes that those recruits were not called to testify. Locust argues that, even assuming that he introduced the three recruits to the other members of the scheme, there was no evidence that their claims were fraudulent because Kalkanis testified that he could not specify which claims were fraudulent and which were genuine. For that reason, Locust says, the government misstated the evidence.

But there was evidence that Locust brought in these recruits to make fraudulent claims. The intake sheets for these recruits listed the name "Robert" or "Rob" as the runner. Supp. App'x 110-11, 113. Locust's first name is Robert. Dewitt testified that Locust had reported staging fake slip-and-fall accidents at a Wendy's restaurant, at a bike shop, and at a Domino's restaurant—the locations at which Gilford, Roberts, and Wright had accidents. Rainford App'x 199-200. While the government's description of Dewitt's testimony and the intake sheets may have been somewhat conclusory, we again see no "flagrant abuse," Williams, 690 F.3d at 75, and the district court did not plainly err in failing to intervene.

Fourth, Locust argues that the prosecutor lowered the burden of proof by asserting that it did not matter whether Locust knew for certain that the slip-and-fall accidents were staged. The prosecutor stated at closing that "[e]ven if Locust did not know for certain, he is still guilty. He was aware there was a high probability that these patients had staged accidents." Locust App'x 117. Locust's counsel promptly objected, saying "[t]hat's not the standard." Id. The district court sustained the objection and reminded the jury immediately that "the government's burden is proof beyond a reasonable doubt. Remember that." Id. Accordingly, even if the government's statement constituted misconduct, Locust has not established prejudice. The jury was immediately reminded that the standard is beyond a reasonable doubt, and we presume that juries follow the instructions. See United States v. Becker, 502 F.3d 122, 130 (2d Cir. 2007). We are not persuaded that any misconduct affected the verdict. See Elias, 285 F.3d at 190.

For these reasons, Locust has not shown that the government committed misconduct or that, if it did, the misconduct prejudiced the outcome of his trial. We affirm

the district court's judgment with respect to this issue.

II

Second, we consider the defendants' challenges to their sentencing guidelines calculations. We affirm the district court's judgment with respect to the calculation for each defendant. However, we remand for factfinding as to the number of fraudulent accidents the conspiracy orchestrated while Rainford and Locust were members for the purpose of performing a loss calculation under U.S.S.G. § 2B1.1(b)(1).^{4 4}

4 The practice of leaving in place a judgment but "remand[ing] partial jurisdiction to the district court to supplement the record on a discrete factual or legal issue while retaining jurisdiction over the original appeal" is known in this circuit as a Jacobson remand. United States v. Rosa, 957 F.3d 113, 121 n.29 (2d Cir. 2020) (quoting Corporación Mexicana De Mantenimiento Integral, S. De R.L. De C.V. v. Pemex-Exploración y Producción, 832 F.3d 92, 115 (2d Cir. 2016) (Winter, J., concurring)); see United States v. Jacobson, 15 F.3d 19, 22 (2d Cir. 1994).

4 The majority opinion also relies on the Black's Law Dictionary definition of "victim" as a "person harmed by a crime, tort, or other wrong." Maj. Op. at 483. But this definition is too broad to apply in this context; surely, a person who suffers harm during the commission of his own crime cannot be considered a victim of that crime. A getaway driver for a robbery crew who suffers serious injuries in a car crash fleeing the scene has been harmed by the crime — but do we consider him, legally, a victim of it? The majority opinion's citation to other dictionary definitions of "victim" as "a 'person subjected to oppression, deprivation, or suffering,' 'someone tricked, duped, or subjected to hardship,' or 'someone badly used or taken advantage of'" is likewise misplaced. Id. Here, the claimants testified that they joined the scheme willingly and with knowledge of what they would be doing to make money; they were not tricked or duped. And though the claimants may have endured hardship and suffering, the same can, sadly, be said of many criminal defendants. That a criminal act may cause its perpetrator hardship does not transform the perpetrator from a criminally liable participant into a victim, in a legal sense.

We review a district court's application of the guidelines de novo, but factual determinations are reviewed for clear error. See United States v. Cramer, 777 F.3d 597, 601 (2d Cir. 2015). However, when a defendant fails to object to a procedural error in the district court's guidelines calculation, we review for plain error. See United States v. Verkhoglyad, 516 F.3d 122, 128 (2d Cir. 2008).

A

All three defendants argue that the district court erred when it included a guidelines enhancement due to the "loss" associated with the schemes. We agree, but only insofar as the argument applies to Rainford and Locust.

U.S.S.G. § 2B1.1(b)(1) imposes an enhancement when the "loss" exceeds certain levels. As relevant here, if the loss is between $9.5 million and $25 million, the defendant receives a twenty-level increase; if the loss is between $25 million and $65 million, the defendant receives a twenty-two-level increase. U.S.S.G. § 2B1.1(b)(1)(K)-(L). The application note clarifies that the "loss is the greater of actual loss or intended loss." Id. § 2B1.1, comment. (n.3(A)). And the "intended loss" means "the pecuniary harm that the defendant purposely sought to inflict," even including "pecuniary harm that would have been impossible or unlikely to occur." Id. § 2B1.1, comment. (n.3(A)(ii)). In Stinson v. United States , the Supreme Court explained that "commentary in the Guidelines Manual that interprets or explains a guideline is authoritative unless it violates the Constitution or a federal statute, or is inconsistent with, or a plainly erroneous reading of, that guideline." 508 U.S. 36, 38, 113 S.Ct. 1913, 123 L.Ed.2d 598 (1993). Here, the application note defining loss is neither inconsistent with nor a plainly erroneous reading of the guideline. "[T]he term 'loss' in § 2B1.1 has no one definition and can mean different things in different contexts," so the guideline does not contradict the understanding expressed in the commentary that "loss" encompasses intended loss. United States v. You, 74 F.4th 378, 397 (6th Cir. 2023) (internal quotation marks omitted).^{5 5} To apply § 2B1.1(b)(1), the sentencing court "is only required to make a 'reasonable estimate of the loss.'" United States v. Lacey, 699 F.3d 710, 719 (2d Cir. 2012) (quoting U.S.S.G. § 2B1.1, comment. (n.3(C))). Even so, the sentencing court must "make findings that are sufficiently specific to permit meaningful appellate review." United States v. Flores, 945 F.3d 687, 721 (2d Cir. 2019). "A district court satisfies its obligation to make findings sufficient to permit appellate review if the court indicates, either at the sentencing hearing or in the written judgment, that it is adopting the recommendations in the PSR"—that is, the Presentence Report. United States v. Wagner-Dano, 679 F.3d 83, 90 (2d Cir. 2012) (internal quotation marks and alterations omitted) (quoting United States v. Prince, 110 F.3d 921, 924 (2d Cir. 1997)). But "adoption of the PSR does not suffice if the PSR itself does not state enough facts to permit meaningful appellate review." United States v. Ware, 577 F.3d 442, 452 (2d Cir. 2009). When the sentencing court makes findings adequate to permit appellate review, "its findings of fact will be overturned only if they are clearly erroneous." Flores, 945 F.3d at 721.

5 The continuing vitality of Stinson is subject to debate. See John S. Acton, The Future of Judicial Deference to the Commentary of the United States Sentencing Guidelines, 45 Harv. J.L. & Pub. Pol'y 349, 355 (2022) (describing "four discrete issues" that have "complicated... Stinson deference's scope"). The holding in Stinson rested on the comparison of the guidelines commentary to "an agency's interpretation of its own legislative rule." 508 U.S. at 44, 113 S.Ct. 1913 (citing Bowles v. Seminole Rock & Sand Co., 325 U.S. 410, 414, 65 S.Ct. 1215, 89 L.Ed. 1700 (1945)). After the Supreme Court modified the framework for reviewing an agency's interpretation of its own rule, see Kisor v. Wilkie, 588 U.S. 558, 574-75, 139 S.Ct. 2400, 204 L.Ed.2d 841 (2019), circuit courts have disagreed as to whether the application note defining "loss" to include the intended loss should continue to receive deference. Compare United States v. Banks, 55 F.4th 246, 258 (3d Cir. 2022) ("[T]he ordinary meaning of the word 'loss' is the loss the victim actually suffered. ... Because the commentary expands the definition of 'loss' by explaining that generally 'loss is the greater of actual loss or intended loss,' we accord the commentary no weight."), with You, 74 F.4th at 397 ("Applying Kisor's framework, we defer to the Sentencing Commission's interpretation of 'loss.'").
We adhere to Stinson and defer to the application note for two reasons. First, the Supreme Court has not overruled Stinson. "[I]f a precedent of [the Supreme] Court has direct application in a case, yet appears to rest on reasons rejected in some other line of decisions, the Court of Appeals should follow the case which directly controls, leaving to [the Supreme] Court the prerogative of overruling its own decisions." Agostini v. Felton, 521 U.S. 203, 237, 117 S.Ct. 1997, 138 L.Ed.2d 391 (1997) (quoting Rodriguez de Quijas v. Shearson/Am. Exp., Inc., 490 U.S. 477, 484, 109 S.Ct. 1917, 104 L.Ed.2d 526 (1989)). Second, the guidelines commentary would meet the Kisor standard in any event. Because the Sentencing Commission adopts the commentary alongside the guidelines, see United States v. Moses, 23 F.4th 347, 353 (4th Cir. 2022) ("[T]he Commission, in practice, generally follows the same process for adopting and amending policy statements and commentary as it uses for the promulgation and amendment of the Guidelines themselves."), the commentary necessarily reflects the Commission's "authoritative, expertise-based, fair, or considered judgment," Kisor, 588 U.S. at 573, 139 S.Ct. 2400 (internal quotation marks and alteration omitted). Indeed, the guidelines and the commentary "operate together as a reticulated whole," Moses, 23 F.4th at 355, and accordingly "the two are to be read together," United States v. Pedragh, 225 F.3d 240, 244 (2d Cir. 2000).

5 The majority opinion concludes that the claimants were "duped" because some of them "testified that their signatures had been forged on ... loan agreements." Maj. Op. at 483-84 & n.10. This is not evidence that the claimant conspirators were duped into joining the conspiracy. Indeed, it is not even evidence that they were unaware that someone else might sign documents on their behalf. To the contrary, the testimony cited by the majority opinion describes events that appear to be part and parcel of the scheme the claimants voluntarily joined. See, e.g., Rainford App'x at 526 ("I signed the papers and then I received a check."); id. at 659 (A claimant "got [$] 6,000 ... because of the paperwork that was signed" on her behalf.). In one instance cited by the majority opinion, the claimant conceded that signing documents without reading or fully understanding them was not something she did only in the context of this conspiracy. See id. at 481 ("So do you have a habit of signing things you don't read? ... Sometimes I do, sometimes I don't."). In the other instance cited by the majority opinion, the claimant testified that "somebody called [his mom] and said somebody forged her signature on a piece of paper," but the issue was quickly resolved. Id. at 1095-96 ("He said he'll get back to me, and then couple days later, him and my mother resolved the issue.").

We have held that a sentencing court's methodology was not "too crude" when it calculated a loss amount based on two factors: the total profits of the scheme and testimony as to the underlying percentage that was fraudulent. United States v. Moseley, 980 F.3d 9, 29 (2d Cir. 2020); see also United States v. Uddin, 551 F.3d 176, 180 (2d Cir. 2009) ("A district court may make a reasonable estimate by extrapolating the average amount of loss from known data and applying that average to transactions where the exact amount of loss is unknown.") (internal quotation marks omitted). That is the method the district court used in this case to calculate the loss, and the defendants do not challenge the district court's methodology. We therefore proceed to evaluate for clear error the district court's factual determinations regarding (1) the intended loss of the schemes and (2) the number of fraudulent accidents in which each defendant was involved.^{6 6}

6 As explained below, Locust objected specifically to the loss calculation, so we review for clear error, but Rainford and Duncan did not. See infra Part II.A.2. Because we conclude that the factual finding about the intended loss of $100,000 per fraudulent case was not clearly erroneous, we need not conduct a separate plain error analysis for Rainford and Duncan.

6 The majority opinion goes to great lengths to conclude that Stinson has not been overruled, and thus applies here, see Maj. Op. at 475 & n.5, but that issue was not presented in this appeal. None of the parties have suggested that Stinson should be ignored or that it has been overruled. As a practical matter, it makes no difference to the outcome. Even if "Stinson deference" has survived recent developments in the law, the Application Note is not authoritative as to the meaning of "victims" because reading the Application Note to § 2B1.1 to classify the claimants as victims is both inconsistent with, and results in a plainly erroneous reading of, the guideline. Because the issue of the continuing validity of Stinson has not been presented here, and does not alter the outcome, I would not reach it, and I neither concur in nor dissent from footnote 5 of the majority opinion.

1

The defendants' PSRs indicated that the intended loss for each fraudulent accident was $100,000 and that this estimate is a "conservative" one because it represents a "low settlement amount for a... Fraudulent Case." Rainford PSR ¶ 43; Locust PSR ¶ 38; Duncan PSR ¶ 32. In support of that conclusion, the PSRs referenced (1) Clarence Tucker's testimony that his case settled for $100,000, (2) Kasheem Jones's testimony that his case settled for $225,000 and that his girlfriend's case settled for $250,000, (3) Carol White's testimony that that her case settled for $80,000, and (4) Alvin Martin's testimony that his case settled for $120,000. See, e.g., Rainford PSR ¶ 24. The district court adopted these findings for each defendant. See Rainford App'x 1281 (Locust); id. at 1299 (Rainford); id. at 1234 (Duncan). It thereby satisfied its obligation to adopt findings of fact "sufficient to permit appellate review." Wagner-Dano, 679 F.3d at 90; see also Ware, 577 F.3d at 452.

Even if it had not adopted the facts in the PSRs, the district court made several statements at the sentencing hearings indicating that it had independently found the intended loss for each fraudulent accident was, conservatively, about $100,000. The district court observed that "sometimes it was held out to people that they could make up to a hundred thousand dollars." Rainford App'x 1278. And the district court showed that it was focused on intended loss—as opposed to actual loss—when it asked the government, "where does the foreseeable loss per recruit of each victim of a hundred thousand come from?" Id. at 1279. The government responded with specific evidence: "Clarence Tucker testified at trial ... that he was told by Mr. Locust he could make a hundred thousand dollars or better in staging an accident." Id. These figures were conservative; if the recruit were told to expect compensation of about $100,000, then the intended loss to the insurance company would need to be substantially higher in order to pay runners, organizers, doctors, lawyers, and others who were involved in the conspiracy.

Moreover, to conduct a clear error review, we must review the "entire evidence." United States v. Mattis, 963 F.3d 285, 291 (2d Cir. 2020) (quoting United States v. Sabhnani, 493 F.3d 63, 75 (2d Cir. 2007)). Here, there was evidence introduced at trial indicating that many cases settled for six figures. See Rainford App'x 528, 562, 1087. And the district judge emphasized several times at the sentencing hearings that he presided over the trial and had a command of the evidence. See, e.g., id. at 1255 ("I am quite comfortable, as the judge who presided over your trial, that there was extensive evidence, extensive credible evidence."). We have never required a sentencing court to have stated all the facts in the record on which it based its findings, only enough to "permit meaningful appellate review." Flores, 945 F.3d at 721. And on clear error review, even if there are "two permissible views"

of the facts, "the factfinder's choice between them cannot be clearly erroneous." United States v. Norman, 776 F.3d 67, 76 (2d Cir. 2015) (quoting United States v. Abiodun, 536 F.3d 162, 170 (2d Cir. 2008)).

In this case, the settlement numbers fell along a range. It may have been possible for a reasonable factfinder to conclude that the intended loss was less than $100,000. But there was also enough evidence to conclude that the intended loss was $100,000 or more. For that reason, the district court's choice of the latter view "cannot be clearly erroneous," Norman, 776 F.3d at 76, and we will not disturb the district court's finding that the intended loss for each fraudulent accident was $100,000.

2

The number of fraudulent cases is a more complicated question. The sentencing court relied on the PSRs for the finding that the fraud involved more than 400 recruits. See Rainford PSR ¶ 24; Duncan PSR ¶ 24; Locust PSR ¶ 24 ("During the Fraud Scheme, more than 400 Patients were referred by LOCUST, DUNCAN, RAINFORD and their co-conspirators to the lawyers in order to initiate fraudulent cases.") (emphasis omitted). But that reliance was misplaced because the PSRs merely asserted—without reference to any admitted evidence—the claim of 400 fraudulent cases. Accordingly, the sentencing court's "adoption of the PSR does not suffice" because "the PSR itself does not state enough facts to permit meaningful appellate review." Ware, 577 F.3d at 452.

The question then is whether there are facts "sufficiently specific to permit meaningful appellate review" elsewhere in the PSR or that the district court recognized at the sentencing hearings. Flores, 945 F.3d at 721. We must consider whether the sentencing court erred when it concluded that those facts were established by a preponderance of the evidence. See United States v. Thorn, 317 F.3d 107, 117 (2d Cir. 2003). We proceed defendant by defendant.

a

Duncan "objected to the entirety" of his PSR on the ground that the Probation Office "solely adopted [facts] from whatever the government said the facts were." Rainford App'x 1233. Although the government acknowledges that Duncan lodged a timely general objection to the PSR and concedes that we should review for clear error, Appellee's Br. 46, "[t]o preserve an objection for appellate review, a defendant must articulate it to the trial court 'with sufficient distinctness to alert the court to the nature of the claimed defect,'" United States v. Dorvee, 616 F.3d 174, 179 (2d Cir. 2010) (quoting United States v. Gallerani, 68 F.3d 611, 617 (2d Cir. 1995)). Duncan did not articulate an objection with distinctness; he did not alert the district court to the "nature" of his objection to the extent that he challenged the loss calculation for any particular reason. We therefore review for plain error. See Verkhoglyad, 516 F.3d at 128.

Duncan received a twenty-two-level loss enhancement on the ground that the intended loss from the schemes was about $30 million. See Duncan PSR ¶ 52. We resolve Duncan's challenge to the loss calculation based on the third prong of plain error review, prejudice. Assuming arguendo that the PSR's estimation that the Kalkanis scheme involved 400 cases was plainly erroneous, it would not have been erroneous had it estimated the number to be 300. That is because Kalkanis estimated that his scheme involved "[a]pproximately 300" cases. Rainford App'x 883, 1018. To be sure, Kalkanis testified inconsistently about the percentage of

those cases that were fraudulent, starting with "[a]t least 80 percent," then "practically all," and finally "[t]he majority of them." Id. at 884. But even if we assume conservatively that only 50 percent of the cases were fraudulent, that would yield a total of 150 fraudulent cases. And in light of the district court's factual determination that each fraudulent accident had an intended loss of $100,000, Duncan's loss from the Kalkanis scheme would be $15,000,000.

That is just the Kalkanis scheme. Duncan also participated in a spin-off scheme that he conducted with Gordon. Duncan's PSR estimated that the spin-off scheme involved "at least 300 Patients," Duncan PSR ¶ 28, a fact on which the district court expressly relied, Rainford App'x 1340. The district court acknowledged that, although there was evidence that not all of these accidents were fraudulent, "it's still what I think amount to hundreds of accidents." Id. Even if the district court reduced the 300-accident figure by two-thirds, down to 100 accidents, and if each of those accidents had an intended loss of $100,000, that would generate an additional intended loss of $10 million.

That means—reading the evidence as to the number of fraudulent accidents in Duncan's favor as much as possible—there was a $15 million intended loss from the Kalkanis scheme and a $10 million intended loss from the spin-off scheme. That is $25 million, which is the threshold under § 2B1.1(b)(1)(L) for a twenty-two-level enhancement.

Given these conservative estimates, we conclude that any error did not prejudice Duncan's substantial rights. We affirm the judgment with respect to the twenty-two-level loss enhancement.

b

Locust specifically objected to his loss calculation, id. at 1280-81, which estimated a loss between $9.5 million and $25 million. We therefore review for clear error, but we conclude that the district court did not "make findings that are sufficiently specific to permit meaningful appellate review." Flores, 945 F.3d at 721. So we remand with instructions for the district court to make a factual finding as to the number of fraudulent accidents that the Kalkanis conspiracy orchestrated during Locust's involvement with the conspiracy.

Locust received a twenty-level loss enhancement. Locust objected to the loss calculation on the ground that it was not "based in evidence." Rainford App'x 1281. The district court simply "den[ied] the objection to loss amount" and "adopt[ed] the loss amount of between [$]9.5 and $25 million." Id. But the district court never made a factual finding; it said only that the range described in the PSR was "appropriate." Id. In other words, the district court never determined how many fraudulent accidents implicated Locust.

The district court's reliance on the PSR does not save the loss enhancement. The PSR said only that Locust "joined the scheme in 2015, and there were at least 200 Patients recruited into the scheme after LOCUST joined." Locust PSR ¶ 38 (emphasis omitted). The PSR cites no testimony or other admitted evidence supporting that figure.

In sum, the PSR did not state, the district court did not find, and we have located no statement in the record indicating that Locust participated in the conspiracy for a total of 200 cases. We cannot say whether that is a permissible or a clearly erroneous reading of the evidence. Accordingly, we remand for factfinding.

c

Rainford did not object to his loss calculation, so we review for plain error, as

Rainford concedes we should on appeal. Rainford Br. 36. But we conclude that the district court again did not "make findings that are sufficiently specific to permit meaningful appellate review." Flores, 945 F.3d at 721. We therefore remand with instructions for the district court to make a factual finding as to the number of fraudulent accidents the Kalkanis conspiracy orchestrated during Rainford's involvement with the conspiracy.

Rainford received a twenty-two-level loss enhancement. Rainford PSR ¶ 51. Rainford's PSR explained the loss enhancement by stating that "[t]he value of the intended loss was estimated in the amount of at least $30,000,000." Id. While the PSR acknowledged the $100,000 intended loss per fraudulent accident, see id. ¶ 43, the Rainford PSR never provided evidence for the number of fraudulent accidents touching the conspiracy during his tenure. The PSR offered three statements regarding the number of fraudulent accidents in which Rainford was involved. But each of those statements is insufficient to sustain the loss enhancement.

First, the Rainford PSR asserted that the Kalkanis conspiracy involved "more than 400 Patients," id. ¶ 24, but it did so without reference to evidence. We have located no evidence in the Rainford PSR, in the transcripts of the sentencing hearings, or in the record indicating that there were 400 cases orchestrated by the conspiracy while Rainford was a member.

Second, the PSR stated that "RAINFORD participated in the scheme from approximately 2012 through 2018, and there were at least 300 Patients recruited into the scheme during that time." Id. ¶ 43 (emphasis omitted). That appears to refer to the Kalkanis testimony. But the district court never determined the percentage of those 300 cases that were fraudulent. Moreover, Kalkanis testified inconsistently. As explained above, see supra Part II.A.2.a., Kalkanis indicated at one point that at least 50 percent of those cases were fraudulent. If so, then Rainford's intended loss would be $15 million.^{7 7} That factual finding would support only a twenty-level loss enhancement, not a twenty-two-level loss enhancement. Accordingly, the statement in the PSR that Rainford participated in the scheme for 300 cases is not sufficient to sustain the twenty-two-level loss enhancement.

7 Kalkanis testified that there were approximately 300 cases and that the majority were fraudulent, leading to a total of at least 150 fraudulent cases. Because each fraudulent case had an intended loss of $100,000, the total intended loss would be at least $15 million.

7 On remand, it is entirely possible that the government would be able to offer evidence of an additional insurance company victim, such that the "ten or more victims" adjustment pursuant to § 2B1.1(b)(2)(A) would be warranted. But it was error for the sentencing court to count the claimants in that analysis.

Third, the PSR noted that Rainford "recruited patients" for the spin-off conspiracy, but it puts the number only at "at least two." Rainford PSR ¶ 41. That would not be enough to salvage the loss calculation.

Because (1) the PSR did not cite evidence for the statement that Rainford was involved in 400 cases and (2) neither the PSR nor the district court explained what percentage of cases was fraudulent, we remand with instructions to make sufficient findings of fact.

d

In sum, we conclude that the district court's finding that the intended loss for each fraudulent accident was $100,000 was not clearly erroneous. We affirm the judgment with respect to Duncan's twenty-two-level loss enhancement. But we remand for factfinding regarding the number of fraudulent cases in which Locust and Rainford were involved. See supra note 4. B

The district court added a four-level enhancement to Duncan's guidelines calculation for his leadership role in the conspiracy. The guidelines state that a leadership enhancement is appropriate for an "organizer or leader of a criminal activity that involved five or more participants or was otherwise extensive." U.S.S.G. § 3B1.1(a).

On appeal, Duncan argues correctly that the PSR and the district court did not make specific findings as to the five individuals involved in the spin-off scheme. See Duncan PSR ¶¶ 26-28. But he does not argue that the scheme was not "otherwise extensive." Nor could he make such an argument. The PSR explained that Duncan's spin-off scheme lasted for three years, covered several hundred cases, and obtained over a million dollars in profit. That is extensive. For these reasons, we see no error in the district court's application of the four-level leadership enhancement.

C

Duncan and Locust each argue that they should have received a guidelines reduction on account of minor or minimal participation in the Kalkanis conspiracy. We disagree.

The guidelines provide for a four-level decrease if the defendant was a "minimal participant" and a two-level decrease if the defendant was a "minor participant." For participants in between these categories, the guidelines provide for a three-level decrease. U.S.S.G. § 3B1.2. A "minimal participant" is one who had "lack of knowledge or understanding of the scope and structure of the enterprise and of the activities of others" and who is "plainly among the least culpable of those involved." Id. § 3B1.2, comment. (n.4). A "minor participant" is one who is "less culpable than most other participants in the criminal activity, but whose role could not be described as minimal." Id. (n.5). The guidelines also explain that assessing whether a defendant was a minimal or a minor participant involves the consideration of five factors.^{8 8}

8 Those factors are "(i) the degree to which the defendant understood the scope and structure of the criminal activity; (ii) the degree to which the defendant participated in planning or organizing the criminal activity; (iii) the degree to which the defendant exercised decision-making authority of influenced the exercise of decision-making authority; (iv) the nature and extent of the defendant's participation in the commission of the criminal activity, including the acts the defendant performed and the responsibility and discretion the defendant had in performing those acts; [and] (v) the degree to which the defendant stood to benefit from the criminal activity." U.S.S.G. § 3B1.2, comment. (n.3(C)).

8 Canons of statutory construction also apply to the Sentencing Guidelines. See, e.g., United States v. Herrera, 974 F.3d 1040, 1047 (9th Cir. 2020) ("Canons of statutory construction can also guide the interpretation [of the Sentencing Guidelines]."); United States v. Jones, 15 F.4th 1288, 1291 (10th Cir. 2021) ("When interpreting the Guidelines, we ... apply traditional canons of statutory construction." (citation and quotation marks omitted)); United States v. Bryant, 996 F.3d 1243, 1259-60 (11th Cir. 2021) ("Our interpretation of the Sentencing Guidelines is governed by traditional rules of statutory construction." (citation and quotation marks omitted)); United States v. Cortez-Gonzalez, 929 F.3d 200, 203 (5th Cir. 2019) ("When interpreting the Sentencing Guidelines, we apply the ordinary rules of statutory construction. When the language of the guideline is unambiguous, the plain meaning of that language is controlling unless it creates an absurd result." (citation and quotation marks omitted)).

Locust argues that he was a minor participant and entitled to a two-level reduction. Locust Br. 49. Locust notes that the government "repeatedly conceded that Locust was the least culpable of the defendants who proceeded to trial." Id. at 50. Perhaps. But we have explained that the relevant inquiry is not whether a conspirator "played a lesser role than his co-conspirators" but whether a conspirator had a "'minor' or 'minimal' [role] as compared to the average participant in such a crime." United States v. Carpenter, 252 F.3d 230, 235 (2d Cir. 2001) (emphasis added) (quoting United States v. Rahman, 189 F.3d 88, 159 (2d Cir. 1999)). Locust fails to explain how his role in the Kalkanis conspiracy fits that description.

Duncan argues that he was a minimal participant and, in the alternative, that he was a minor participant. But the evidence suggested that Duncan was integral to the first scheme—that is, the Kalkanis

scheme—for which he identified recruits, transported them to facilitate the scheme, referred doctors and lawyers, and so on. Duncan PSR ¶¶ 17-24. We are not persuaded that the district court erred in failing to reduce his guidelines calculation. Duncan's arguments focus on his culpability compared to his co-conspirators. As noted, however, the relevant inquiry is whether the conspirator's role was "'minor' or 'minimal' as compared to the average participant in such a crime." Carpenter, 252 F.3d at 235. The district court did not err in failing to apply a minor or minimal role reduction for either Duncan or Locust.

D

Rainford argues that the recruits to the scheme were co-conspirators rather than victims. See Rainford Br. 26. For that reason, he contends, the district court erroneously added three victim-related enhancements to each defendant's guidelines calculation: a vulnerable victim enhancement under U.S.S.G. § 3A1.1(b)(1); a ten-or-more victims enhancement under § 2B1.1(b)(2)(A)(i); and a risk of death or serious bodily injury enhancement under § 2B1.1(b)(16). We disagree.

1

Section 3A1.1(b)(1) provides for a two-level enhancement when "the defendant knew or should have known that a victim of the offense was a vulnerable victim." U.S.S.G. § 3A1.1(b)(1). The commentary clarifies that a vulnerable victim "is [1] a victim of the offense of conviction ... who [2] is unusually vulnerable due to age, physical or mental condition, or who is otherwise particularly susceptible to the criminal conduct." Id. § 3A1.1(b)(1), comment. (n.2). Rainford argues that it was a mistake to conclude that the "coconspirator[s]... qualified as 'victims.'" Rainford Br. 27.

This issue is controlled by United States v. Borst, 62 F.3d 43 (2d Cir. 1995). In that case, we considered a fraud scheme in which Borst "handle[d] all of the paperwork in obtaining loans" for several couples hoping to purchase mobile homes. Id. at 45. The documentation that Borst submitted to the bank "contained false statements regarding the year, model, purchase price, and serial number of the mobile homes." Id. On appeal, Borst contended that a vulnerable victim enhancement was improper.

We noted that at least one of the couples appeared to know that Borst entered fraudulent information on the loan application forms:

Mrs. Russell noticed a discrepancy between the sales agreement Borst had prepared, which indicated that they were purchasing a 1988 Skyline mobile home, and the bank's promissory note, which stated that the mobile home was a 1992 model. Borst explained away the discrepancy as a means of expediting the transaction. Mrs. Russell later informed the probation officer that she accepted Borst's explanation and overlooked the discrepancy because her diabetic husband required constant medical care and they were temporarily homeless.

Id. The facts supported—and we recognized—that at least one of the borrowers knew of the fraud and inquired about it but was sufficiently vulnerable that she acquiesced rather than acting to "thwart" it. Id. at 46 (quoting United States v. Kaye, 23 F.3d 50, 54 (2d Cir. 1994)).

We then turned to the question most relevant to the present appeal: whether the "§ 3A1.1 enhancement was inappropriately applied because the three couples were not the actual victims of Borst's crimes." Id. at 47. We said that the couples

were still victims for purposes of § 3A1.1 and announced a rule: "Whether or not the three couples ... were 'unwitting instrumentalities' of Borst's criminal conduct in light of their apparent knowledge of Borst's misrepresentations to the bank, they were exploited and suffered harm as a result of his actions." Id. at 48. In other words, even if the couples knew that Borst was making fraudulent statements to banks and thereby acted as instrumentalities of the fraud, the couples were victims of Borst's scheme nonetheless.

That rule applies straightforwardly in our case. Regardless of whether the recruits in the slip-and-fall scheme knew of the fraud, "they were exploited and suffered harm as a result" of the fraud. Id. As in Borst, these victims were exploited precisely because their economic vulnerability made them willing to engage in conduct that harmed them. The partial dissent argues that this case differs from Borst because the recruits were "knowing, willing, and voluntary, co-conspirators." Post at 495. But in both Borst and this case, the supposed victims knew of the fraud and willingly participated in it in the hope of obtaining a benefit. Yet they nevertheless were victims. The defendants "knew or should have known that these individuals were susceptible to [their] criminal conduct and less likely to thwart the crime." Borst, 62 F.3d at 46.

Even if we could distinguish Borst from this case, the § 3A1.1 enhancement still would properly apply. We begin with the text of § 3A1.1 and its commentary. See Stinson, 508 U.S. at 38, 113 S.Ct. 1913; see also supra note 5. Section 3A1.1 does not define "victim," saying only that the defendant "knew or should have known that a victim of the offense was a vulnerable victim." U.S.S.G. § 3A1.1(b)(1). The commentary defines a "vulnerable victim" in part as someone "who is a victim of the offense of conviction and any conduct for which the defendant is accountable under § 1B1.3." Id. § 3A1.1(b)(1), comment. (n.2). That does not provide a definition of "victim" either.^{9 9}

9 Apart from Borst, we have never precisely defined a "victim" for purposes of § 3A1.1. In United States v. McCall, we discussed "limits on the kinds of conduct that will justify a vulnerable victim enhancement." 174 F.3d 47, 50 (2d Cir. 1998). We identified three limits: (1) the "vulnerability of the victim must bear some nexus to the criminal conduct," (2) "the defendant generally must have singled out the vulnerable victims from a larger class of potential victims," and (3) "broad generalizations about victims based upon their membership in a class are disfavored where a very substantial portion of the class is not in fact particularly vulnerable to the crime in question." Id. The limits we identified concerned the victim's vulnerability rather than the status of being a victim in the first place.

9 In fact, Duncan's offense level was increased four levels under § 3B1.1(a), because he "organize[d] all of the lawyers, doctors and funders for the Patients he and his coconspirators recruited into the scheme." Appellee's Br. at 62 (quoting Duncan PSR ¶26).

As a matter of plain meaning, a victim is a "person harmed by a crime, tort, or other wrong." Victim, Black's Law Dictionary (11th ed. 2019). We think the dictionary definition of victim supports the conclusion that the recruits here were victims. A victim is a "person subjected to oppression, deprivation, or suffering," "someone tricked, duped, or subjected to hardship," or "someone badly used or taken advantage of." Victim, Webster's Third New International Dictionary (2002). The recruits were victims because the adverse burdens of the conspiracy fell on their shoulders—they were subjected to hardship by the doctors. The organizers of the conspiracy charged in this case, by contrast, were not. It is beyond dispute that some of the recruits were injured, and therefore endured a hardship, while the organizers were not. And the record indicates that the recruits were duped.^{10 10} Even

10 The record indicates, for example, that recruits were duped into borrowing high-interest loans from litigation funding companies that consequently collected large portions of any settlements. The litigation funding companies would underwrite the recruit's medical expenses, small personal loans to recruits, and referral fees, at interest rates between 25 and 50 percent. Recruits testified that their signatures had been forged on the loan agreements or that they were not aware that they had borrowed high-interest loans. See, e.g., Rainford App'x 480 ("I don't know what it was that signed. All I know, [Kalkanis] said it's to start the lawsuit."); id. at 526 ("No, I really didn't [read the papers], I just signed it. I just wanted to, you know, do whatever they told me to do to sign it up and that's it."); id. at 658 ("I called Bryan and told him I see my signature on some stuff that I took out loans for surgery and stuff and that's not my signature."); id. at 1095 ("[My mom] said somebody forged her signature and she got a loan out in her name and she didn't know anything about it.").

10 The concurrence's reliance on 42 U.S.C. § 274e is misplaced. That statute does not define or indeed even mention "victims," and a search of all federal cases reported on Westlaw reveals no decisions holding that a willing seller of an organ qualifies as a "victim" under that law simply because he or she undergoes surgery. The statute provides no insight into the issue presented here.

if they engaged in the scheme voluntarily, the organizers of the conspiracy took advantage of their economic vulnerability to lead them to undertake serious physical and legal risks.

Moreover, the definition of "victim" used in § 2B1.1—though inapplicable to § 3A1.1^{11 11}—suggests the scope of the term. Section 2B1.1 says that a victim is "any person who sustained any part of the actual loss determined under subsection (b)(1)" or "any individual who sustained bodily injury as a result of the offense." U.S.S.G. § 2B1.1, comment. (n.1). It does not mention voluntariness or willfulness at all. It does not exclude co-conspirators. Under this definition, the recruits were victims.

11 Application note 1 in the commentary to § 2B1.1, which provides several definitions including "victim," states at the very beginning that the definitions apply "[f]or purposes of this guideline." U.S.S.G. § 2B1.1, comment. (n.1) (emphasis added). As a result, the definition may be persuasive but it is not controlling for purposes of § 3A1.1.

11 While the majority opinion limits its remand on this basis to only Rainford and Locust, I would also remand as to Duncan. The failure to adequately support the loss calculation extends to all three defendants. Furthermore, it is impossible to know whether a recalculation would affect Duncan's sentence because we do not know what the revised calculations will be. The majority opinion too quickly assumes that Duncan was not "prejudiced" because it assumes the number of claims the District Court would deem fraudulent on remand, and also credits the unsupported $100,000 intended loss figure. Notably, the District Court reduced Duncan's sentence in light of new information about "the increasing percentage ... of nonfraudulent accidents from D&G." Rainford App'x at 1340. This suggests that further reconsideration of the loss calculation might impact Duncan's sentence.

The partial dissent does not dispute that the plain meaning of "victim" includes co-conspirators.^{12 12} But it argues that applying this plain meaning "would lead to absurd outcomes" in which "active and essential participants in a conspiracy could be considered victims." Post at 499. Not so. The plain meaning is consistent with the purpose of the guideline and results in no "genuine ... absurdity" that would "justify departure from the plain text." Gibbons v. Bristol-Myers Squibb Co., 919 F.3d 699, 706 (2d Cir. 2019). The purpose of the sentencing enhancement is to account for conduct that is "more culpable than that of the typical perpetrator of that crime." United States v. Morrill, 984 F.2d 1136, 1137 (11th Cir. 1993). A defendant who commits an offense that causes harm to others—including others who are exploited as participants in the conspiracy—is more culpable than a defendant who commits the same offense without such harm to others. In this case, the defendants profited by recruiting economically desperate people to undergo unnecessary surgeries. Such conduct is more culpable than that of a defendant who, say, commits fraud by undergoing the unnecessary surgery himself.^{13 13} Our interpretation of the Mandatory Victims Restitution Act ("MVRA") supports this conclusion. As the partial dissent observes, "[u]nder the plain text of the MVRA ... co-conspirators have just as much right to restitution as do innocent victims." Post at 501 (quoting United States v. Lazarenko, 624 F.3d 1247, 1250 (9th Cir. 2010)). We have departed from that plain-text reading because it would "require[ ] 'restitutionary' payments to the perpetrators of the offense of conviction," which would not make sense in the context of statute that aims to compensate the victims of a crime. United States v. Reifler, 446 F.3d 65, 127 (2d Cir. 2006); see also Lazarenko, 624 F.3d at 1251 ("Congress could not have intended that result. Otherwise, the federal courts would be involved in redistributing funds among wholly guilty co-conspirators, where one or more co-conspirators may have cheated their comrades."). But that context is absent here. While the MVRA aims at compensation, the sentencing enhancements aim at assigning culpability. Even the partial dissent acknowledges that the distinction makes a substantive difference in the reach of each provision.^{14 14}

12 Rainford similarly acknowledges that a "literal reading," Rainford Br. 27, or a "broad reading," Rainford Reply Br. 10, of the guidelines would lead to that conclusion.

12 The majority opinion addresses the continuing validity of Stinson in its discussion of the District Court's findings as to loss amount. But the Application Note relied upon by the majority opinion — which discusses the use of intended loss rather than actual loss for calculation of the offense level — has no impact on the problem actually presented. The District Court's error was not in relying on intended loss versus actual loss. Its error was in failing to accurately calculate either actual or intended loss, based on evidence in the record, and in failing to explain how it calculated the $100,000 per claim loss amount. Again, I would not reach the question of whether Stinson deference remains valid, because it has not been argued and does not affect the outcome of this appeal.

13 The partial dissent argues that, if the Sentencing Commission intended for the enhancements to apply to co-conspirators, it would have expressly said so. See post at 499-500. It points to U.S.S.G. § 2D1.1, which provides for a sentencing enhancement when a defendant "distributed a controlled substance" to a vulnerable individual or otherwise "involved that individual in the offense." U.S.S.G. § 2D1.1(b)(16)(B). But § 2D1.1 imposes an enhancement for merely "involv[ing]" a vulnerable person in the offense without regard to whether that person was harmed, and therefore § 2D1.1 sweeps more broadly than a guideline that applies only to victims. We do not see why the Commission would have been expected to "use that same language if it wished to impose an adjustment" for conduct that harmed others. Post at 500.

13 Indeed, the District Judge appeared to recognize the lack of an evidentiary basis in the PSR for the loss amount, prompting him to ask, upon review of the PSR, where "the foreseeable loss per recruit of each victim of a hundred thousand" came from. Rainford App'x at 1279. The District Court did not inquire further, however, suggesting that it accepted the government's explanation. If that is so, then the District Court based its entire loss calculation on the testimony of a single co-conspirator claimant about what he was told he might be able to make.

14 See post at 505 n.15 ("[T]he MVRA is intended to compensate victims, while sentencing enhancements for loss amount under the Guidelines are predicated on the notion that fraudsters who cause more monetary harm are more culpable and should therefore generally be given longer terms of imprisonment. As we have held, because a defendant's culpability will not always equal the victim's injury, an amount-of-loss calculation for purposes of sentencing does not always equal such a calculation for restitution.") (quoting United States v. Niebuhr, 456 F. App'x 36, 39 (2d Cir. 2012)).

14 The record on appeal includes the trial testimony from the testifying claimants as well as a number of spreadsheets, apparently submitted by insurance companies identified as potential victims. "Based on the information provided by the insurance companies," the government, in the Restitution Letter, compiled a chart of "Patients for which the insurance companies paid out settlements or judgments." Rainford Sealed App'x at 5. It includes numerous claims that are below the $100,000 "conservative" intended loss per claim amount. See id. at 6-7 (listing claim payments of $90,000, $82,500, $25,000, $70,000, $30,000, $41,667, $75,000, $50,000, $65,000, $2,000, $5,000, $85,000, $89,500, $80,000). These broadly varying payment amounts only further demonstrate that the District Court needed to provide more of an explanation as to how it concluded that the intended loss per claim was $100,000.

The partial dissent further suggests that under our purportedly "literal reading" of the application note, the term "victim" would include not only co-conspirators but also "a lone defendant ... who suffered bodily harm as a result of his own offense." Post at 499, But here it is the partial dissent that is being unreasonably literalistic. At least in law, a person generally cannot be his own victim. A person does not commit a crime against himself or recover from himself in tort. See Riggs v. Palmer, 115 N.Y. 506, 511-12, 22 N.E. 188 (1889) ("[A]ll laws as well as all contracts may be controlled in their operation and effect by general, fundamental maxims of the common law. No one shall be permitted to profit by his own fraud, or to take advantage of his own wrong, or to found any claim upon his own iniquity, or to acquire property by his own crime. These maxims are dictated by public policy, have their foundation in universal law administered in all civilized countries, and have nowhere been superseded by statutes."); Sieffer v. McLean, 7 Tex.Civ.App. 158, 26 S.W. 315, 315 (1894) ("The general rule may be found expressed in the maxim that no man can make his own misconduct the ground for an action in his own favor. If he suffers because of his own wrongdoing, the law will not relieve him. The law cannot recognize equities as springing from a wrong, in favor of one concerned in committing it.") (quoting Thomas M. Cooley, A Treatise on the Law of Torts 167 (2d ed. 1888)); see also People v. Latzman, 153 Mich.App. 270, 395 N.W.2d 56, 58 (1986) (explaining that the "definition of victim" under state sentencing guidelines "cannot be construed so as to include defendant as a victim of his own crime"). Given that legal background,^{15 15} we do not agree that either § 3A1.1 or § 2B1.1 would

15 Cf. United States v. Scott, 990 F.3d 94, 128 (2d Cir. 2021) (Menashi, J., concurring in part and concurring in the judgment) ("The ultimate objective is to determine the meaning the law assigns to the text and therefore its legal effect.").

15 "[T]he MVRA is intended to compensate victims, while sentencing enhancements for loss amount under the Guidelines are predicated on the notion that fraudsters who cause more monetary harm are more culpable and should therefore generally be given longer terms of imprisonment. As we have held, because a defendant's culpability will not always equal the victim's injury, an amount-of-loss calculation for purposes of sentencing does not always equal such a calculation for restitution." United States v. Niebuhr, 456 F. App'x 36, 39 (2d Cir. 2012) (summary order) (citations and quotation marks omitted). Indeed, we may require more precision in calculating restitution than we do in calculating loss under the Guidelines; an estimate that may "suffice[ ] for purposes of Guidelines calculations," may nonetheless be inadequate to support "an order of restitution." United States v. Reifler, 446 F.3d 65, 127 (2d Cir. 2006).

include a defendant who suffered harm as a result of his own offense.

The remaining question is whether these victims were "vulnerable" within the meaning of § 3A1.1. We conclude that they were. The application note defines a vulnerable victim as a person "who is a victim of the offense of conviction and any conduct for which the defendant is accountable under § 1B1.3" and who is "unusually vulnerable due to age, physical or mental condition, or who is otherwise particularly susceptible to the criminal conduct." U.S.S.G. § 3A1.1, comment. (n.2). The schemes in this case targeted the "down and out," including the homeless. Rainford App'x 1277. We have come close to saying that homelessness necessarily renders a victim vulnerable. See United States v. Irving, 554 F.3d 64, 75 (2d Cir. 2009) ("The facts that Irving's victims in Mexico and Honduras were children who were homeless and were without parental or other appropriate guidance made them unusually vulnerable, independently of their ages."). The recruits here were vulnerable too.

We reject Rainford's arguments that the recruits were not victims within the meaning of § 3A1.1 because they were co-conspirators. We affirm the judgment with respect to the vulnerable victim enhancements.

2

Based on a similar argument, Rainford challenges the enhancement under § 2B1.1(b)(2)(A)(i), which provides for a two-level enhancement when the offense "involved 10 or more victims." U.S.S.G. § 2B1.1(b)(2)(A)(i). Rainford concedes that the nine insurance companies were victims, but he contends that the recruits were not victims because they were co-conspirators. See Rainford Br. 32. For that reason, he says, his crime did not implicate ten victims, so the § 2B1.1(b)(2)(A)(i) enhancement is inappropriate. We again disagree.

The application note explains that a victim is "any person who sustained any part of the actual loss determined under subsection (b)(1)" or anyone who "sustained bodily injury as a result of the offense." U.S.S.G. § 2B1.1, comment. (n.1). It further explains that "person" includes a natural person as well as a business entity. Id. The "loss determined under subsection (b)(1)" is a financial loss. See id. § 2B1.1, comment. (n.3(A)) (noting that "loss is the greater of actual loss or intended loss" and that "'[a]ctual loss' means the reasonably foreseeable pecuniary harm that resulted from the offense"). It is undisputed that the nine insurance companies that submitted claims for restitution were victims within the meaning of § 2B1.1(b)(2)(A)(i).

In determining that the recruits were victims for purposes of § 2B1.1(b)(2)(A)(i), the district court applied the application note, which is subject to Stinson deference. We see no error in the district court's analysis. The guideline provision and the commentary "are to be read together" because "[n]o threshold test of ambiguity need be passed before the commentary can be consulted." Pedragh, 225 F.3d at 244. Only when "the commentary contradict[s] the provision's text" does "the provision's plain language... control[ ]." United States v. Lewis, 93 F.3d 1075, 1080 (2d Cir. 1996). As discussed above, a straightforward application of the application note indicates that the recruits were victims: a victim is anyone who "sustained bodily injury as a result of the offense," and several of the recruits sustained such injuries. Accordingly, they are victims.

To reach the opposite conclusion, one must decide that the application note is a "plainly erroneous" interpretation of the

guideline insofar as it understands the word "victim" to include co-conspirators. But for the reasons stated above, the word "victim" does not necessarily exclude co-conspirators. Moreover, we have previously applied the application note without issue. See Lacey, 699 F.3d at 716.

We conclude that at least one of the recruits was a victim, meaning there were ten or more victims implicated by the schemes. We affirm the judgment with respect to the enhancement under § 2B1.1(b)(2)(A)(i).

3

Rainford makes a similar challenge to a third victim-related enhancement: a two-level enhancement for committing an offense that involved the "conscious or reckless risk of death or serious bodily injury" under U.S.S.G. § 2B1.1(b)(16). Rainford argues that his conduct "did not risk or cause bodily injury to the insurance companies, the only victims here." Rainford Br. 31. He says that the recruits were willing participants who "put themselves at risk by agreeing to stage a slip-and-fall accident or to undergo surgery." Id.

That is incorrect. Section 2B1.1(b)(16) is not limited to nonparticipants. As the government points out, § 2K1.4(a)(1) already provides for an enhancement when the offense "created a substantial risk of death or serious bodily injury to any person other than a participant in the offense." Appellee's Br. 56. We do not understand § 2B1.1(b)(16) to be inapplicable when the person subject to the risk of injury was a participant, as the recruits were here. Their consent is immaterial.

Rainford does not say that the staged accidents or the surgeries did not contain a substantial risk of death or serious bodily injury. Unnecessary surgery is "the type of risk that is obvious to a reasonable person and for which disregard of said risk represents a gross deviation from what a reasonable person would do." United States v. Lucien, 347 F.3d 45, 56-57 (2d Cir. 2003). We affirm the judgment with respect to the two-level enhancement for risk of death or serious bodily injury.

E

Duncan suggests in passing that the district court erred by applying a two-level enhancement for obstruction of justice. Duncan Br. 3. He does not offer an argument, however, so the issue is waived. See United States v. Brennan, 650 F.3d 65, 137 (2d Cir. 2011). Even so, the record contained evidence that Duncan pressured a co-conspirator to encourage his mother to state that her slip-and-fall claims were not fraudulent. See Supp. App'x 134 et seq. We see no reason to disturb the enhancement.

III

Third, the defendants offer three challenges to their sentences apart from the guidelines calculations. Duncan argues that his forfeiture order cannot be sustained. We agree, vacate the order, and remand for reconsideration. Rainford and Locust argue that their restitution order cannot be sustained. We disagree, but we modify the amount of the restitution order. Finally, Rainford argues that his term of imprisonment must be vacated because it is inconsistent with Duncan's and Locust's sentences. We affirm the judgment with respect to Rainford's sentence but remand for reconsideration in the interest of justice.

A

The district court imposed a forfeiture order of $644,056 on Duncan for his involvement

in the spin-off scheme. See Rainford App'x 1347.^{16 16} On appeal, Duncan challenges that order as lacking an evidentiary basis and says that it violates the Excessive Fines Clause of the Eighth Amendment. We agree with his first argument and do not reach the Excessive Fines Clause claim.

16 The forfeiture order was joint and several with Gordon, Duncan's partner in D&G. Rainford App'x 1347.

16 It is notable that the government did not seek restitution on behalf of the criminally liable co-conspirator claimants, though it argues that they suffered serious harms. This reinforces the reality that the claimants are not victims under the law. See, e.g., United States v. Archer, 671 F.3d 149, 173 (2d Cir. 2011) (observing that any person who knowingly participated in a fraud would not be entitled to restitution for any losses suffered in the fraudulent scheme).

"When a forfeiture award is challenged on appeal, this Court reviews the district court's legal conclusions de novo and its factual findings for clear error." United States v. Treacy, 639 F.3d 32, 47 (2d Cir. 2011). Because forfeiture is part of sentencing, a sentencing court must "determine forfeiture amounts by a preponderance of the evidence." United States v. Capoccia, 503 F.3d 103, 116 (2d Cir. 2007). Computing a forfeiture amount is not an "exact science." Treacy, 639 F.3d at 48. The district court must make only a "reasonable estimate of the loss, given the available information." Uddin, 551 F.3d at 180. It may "use general points of reference as a starting point for calculating the losses or gains from [the criminal activity] and may make reasonable extrapolations from the evidence established by a preponderance of the evidence at the sentencing proceeding." Treacy, 639 F.3d at 48.

Some additional background is necessary to understand Duncan's argument. In 2020, when Duncan first appeared for sentencing, the government requested forfeiture of about $1.6 million. That figure represented the entire amount that Duncan's company, D&G, received in referral fees from when Duncan started the company in 2015 until his arrest in 2018. Duncan objected, and the district court adjourned for more time to decide the issue. When Duncan's sentencing resumed six months later, the government had reduced the request to $644,056.

At the hearing, the government stated that it had reduced the amount based on information from a "cooperating witness who's very familiar with the D&G's operation and its revenue." Rainford App'x 1332-33.^{17 17} The government also represented that the notes from the Gordon interview were produced to Duncan's counsel. Rainford App'x 1333. Purportedly, Gordon had told the government that each entry in D&G's books would need to be analyzed to determine whether it was fraudulent. Id. But he had also said that, generally speaking, only about "40 percent of our cases were trip and falls and the rest were non[-]trip and falls. In total, around 80 percent of all cases were fraudulent, but of all the cases, 40 percent were related to trip-and-fall cases." Id.

17 On appeal, the government notes that the witness was Gordon and that Duncan was aware at the time that Gordon was the cooperating witness. Appellee's Br. 78 n.18.

17 Claimants who participated in the scheme but did not testify at trial are referred to herein only by their initials.

The government argued to the district court that its forfeiture request did not need to be limited to trip-and-fall cases because "the indictment doesn't limit the allegations to trip and falls." Id. However, the government acknowledged that it did not introduce evidence at trial regarding any non-trip-and-fall cases, so it decided to "cut[ ] the 60 percent he estimated were non-trip-and-fall cases." Id. In other words, the government decided to confine the forfeiture request to the 40 percent of D&G's revenue that was attributable to slip-and-fall cases. The government observed that this was a conservative request because Gordon had also stated that the "trip-and-fall cases are funded easier ... so the ledger is going to understate revenues for [those claims]." Id. In other words, even though slip-and-falls were

only 40 percent of D&G's claims, slip-and-falls would compose more than 40 percent of D&G's revenue because it was easier to receive financing for those claims. For that reason, the government said, its forfeiture request "understate[d] considerably what the total amount of criminal proceeds would be." Id. at 1334.

In response, Duncan noted that several items on the ledger were from "entities that aren't even [litigation] funding companies," so the government was "just pulling numbers from everywhere." Id. at 1339.

The district court agreed with the government and imposed a restitution order in the amount of $644,056. Id. at 1347. After describing the spin-off scheme in detail, the district court said: "I do conclude that the government has proven the much-reduced figure of $644,000 by a preponderance of the evidence, which is a conservative estimate of the proceeds traceable to Duncan's criminal conduct." Id. at 1343. The district court acknowledged that Duncan had pointed to some entries in the ledger used to calculate the forfeiture amount that had nothing to do with the slip-and-fall scheme. But it concluded that "[w]ithout specific evidence showing that certain claims were fraudulent, his general challenge is unsuccessful" and the "significant reduction from 1.6 million down to 644,000, by a preponderance, ... does represent the proceeds Duncan derived from fraud scheme 2." Id. at 1344.

We vacate and remand the forfeiture order. The district court's determination 'was clearly erroneous for two reasons. First, the district court based its forfeiture order only on the representations of the government about Gordon's statements. The notes from the meeting were never presented to the district court, and Gordon never testified as to those figures. So the only evidence on which the district court relied to reach the $644,056 figure was the government's word. But the government's word is not evidence. True, the district court may take "general points of reference as a starting point for calculating the losses or gains," Treacy, 639 F.3d at 48, but an unsubstantiated government claim is not a "point[ ] of reference."

Second, even if the government's allegations about Gordon's statements had been corroborated, the government arrived at the forfeiture amount by crediting only part of Gordon's statement. The government reduced the forfeiture request by 60 percent because Gordon said that only 40 percent of D&G's cases were slip-and-falls. Yet the government entirely disregarded his other statement: that only 80 percent of all cases were fraudulent. Neither the government nor the district court has explained why the initial forfeiture request of $1.6 million should be cut by 60 percent but the resulting $644,000 should not be cut by another 20 percent to account for slip-and-fall claims that were not fraudulent.

For these reasons, we vacate the district court's forfeiture order and remand for reconsideration consistent with this opinion.

B

Rainford and Locust were ordered to pay approximately $3.9 million in restitution to defrauded insurance companies. They bear that responsibility jointly and severally with Kalkanis, Gordon, and Dewitt. On appeal, Rainford argues that (1) the district court relied on incorrect loss numbers from several insurance companies, and (2) the district court used a faulty computation methodology. We disagree, so we affirm the order with a modification.

We review an order of restitution for abuse of discretion. See United

States v. Ojeikere, 545 F.3d 220, 222 (2d Cir. 2008). Review of restitution orders is "extremely deferential," United States v. Grant, 235 F.3d 95, 99 (2d Cir. 2000) (quoting United States v. Giwah, 84 F.3d 109, 114 (2d Cir. 1996)), because "ordering restitution requires a delicate balance of diverse, sometimes incomparable, factors, some of which not only lack certainty but may indeed be based on mere probabilities, expectations, guesswork, even a hunch," United States v. Rossi, 592 F.3d 372, 376 (2d Cir. 2010) (internal quotation marks and alteration omitted).

The MVRA provides for mandatory restitution for offenses "committed by fraud or deceit" in which "an identifiable victim" has suffered "pecuniary loss." 18 U.S.C. § 3663A(c)(1). In each case, the district court orders restitution "in the full amount of each victim's losses as determined by the court and without consideration of the economic circumstances of the defendant." Id. § 3664(f)(1)(A). The "calculation of these losses need not be mathematically precise." United States v. Rivernider, 828 F.3d 91, 115 (2d Cir. 2016) (internal quotation marks omitted). The district court need make only a "reasonable estimate" of the loss "based on the evidence before it." United States v. Milstein, 481 F.3d 132, 137 (2d Cir. 2007).

Here, nine insurers submitted a total of thirty-eight claims for restitution, totaling about $4.9 million. The government sought to hold the defendants accountable for 80 percent of that figure because Kalkanis stated that "[a]t least 80 percent" of the cases he managed were fraudulent. Rainford App'x 883-84. That yields the figure of $3.9 million.

Rainford makes two arguments, both of which concern minor errors. First, he says that two claims by insurance companies used to compute the restitution amount were incorrect. Rainford suggests that one of the AmTrust claims is off by $2,500. Rainford Br. 44. But the figure on which the government relied is substantiated elsewhere in a sealed portion of the record. Rainford also suggests that a Hartford claim for $30,000 was erroneously included. He argues that the claim should not have been included because Hartford's own records list the claim as open. Because listing the claim as "open" and unpaid may have been an administrative mistake by Hartford, we are not "left with the definite and firm conviction that a mistake has been committed" by the district court. Easley v. Cromartie, 532 U.S. 234, 242, 121 S.Ct. 1452, 149 L.Ed.2d 430 (2001) (quoting United States v. U.S. Gypsum Co., 333 U.S. 364, 395, 68 S.Ct. 525, 92 L.Ed. 746 (1948)). We see no clear error and thus no abuse of discretion. See Garcia v. Garland, 64 F.4th 62, 69 (2d Cir. 2023) (explaining that a decision that rests on a "clearly erroneous factual finding" or that "cannot be located within the range of permissible decisions" constitutes an abuse of discretion) (quoting Morgan v. Gonzales, 445 F.3d 549, 551-52 (2d Cir. 2006)).

Second, Rainford and Locust contend that there was no evidence that the claims were in fact fraudulent. Rainford Br. 45; Locust Br. 43. For that reason, they say that the district court should have evaluated each claim by an insurance company to determine whether it was fraudulent. The district court adopted a different method. It credited Kalkanis's testimony that 80 percent of the cases he managed were fraudulent. The district court thereby treated all the claims as fraudulent but applied a 20 percent reduction to the total in recognition of Kalkanis's statement. Because restitution may be based on "guesswork" or "even a 'hunch,'" Rossi, 592 F.3d at 376 (quoting United States v. Atkinson, 788 F.2d 900, 902 (2d Cir. 1986)), we do not think the district

court's methodology relied on clearly erroneous facts or that it "cannot be located within the range of permissible decisions" such that it would amount to an abuse of discretion. Garcia, 64 F.4th at 69.

There is, however, one flaw in the restitution order. The government concedes that one of the thirty-eight claims by an insurance company was not related to a fraudulent accident. The claim was submitted to the insurance company by Alvin Martin, and Martin testified at trial that the claim was legitimate. Rainford App'x 1221. That claim settled for $150,000. Because of the 20 percent reduction on account of Kalkanis's testimony, Martin's claim contributed only $120,000 to the restitution order. The government therefore recognizes that it would be appropriate either to vacate and remand to correct this issue or to affirm the judgment with a modification. Appellee's Br. 74 & n.16. We affirm the restitution order but reduce it by $120,000.

C

Finally, Rainford argues that we should remand his term of imprisonment for reconsideration in the interest of justice pursuant to 28 U.S.C. § 2106. That statute permits us to "affirm, modify, vacate, set aside or reverse any judgment, decree, or order ... and ... remand the cause and direct the entry of such appropriate judgment, decree, or order, or require such further proceedings to be had as may be just under the circumstances." 28 U.S.C. § 2106; see also United States v. Jones, 878 F.3d 10, 24 n.6 (2d Cir. 2017) (Calabresi, J., concurring) (noting that "28 U.S.C. § 2106 permits affirmances and remands for further proceedings in the interest of justice, and has been applied in criminal situations"). We are persuaded by Rainford's arguments. Accordingly, we affirm his sentence but remand for reconsideration "as may be just under the circumstances."

Some background is necessary to understand Rainford's claim. The first sentencing hearing in this case occurred on January 7, 2020. At that hearing, the district court sentenced Duncan to 80 months of imprisonment, but it reserved judgment on forfeiture. Rainford App'x 1260, 1262. It sentenced Rainford to 68 months of imprisonment, but it reserved judgment on restitution. Id. at 1316-18. And it sentenced Locust to 60 months of imprisonment, but it reserved judgment on restitution. Id. at 1289-92.

Duncan's forfeiture amount was determined at a hearing on July 27, 2020. As noted above, supra Part III.A., the government conceded at that hearing that a significant percentage of the proceeds from the spin-off scheme were not fraudulent, so it reduced the requested forfeiture amount. Rainford App'x 1333. After concluding that the spin-off scheme involved fewer fraudulent cases than originally thought, the district court reduced the forfeiture amount and also "decrease[d] the sentence of Mr. Duncan from 80 months to 72 months." Id. at 1340.

Locust's restitution hearing was held about a year later on July 9, 2021. The district court recalled that it had lowered Duncan's sentence on account of the government's concession that the spin-off scheme involved non-fraudulent claims. See id. at 1375 ("[M]y intention at this point is I need to take into consideration the change in Duncan so that the people are comparable for everything, with each other and in between each other."). In light of Duncan's reduced sentence, the district court found it "more appropriate" that Locust's sentence be "48 months rather than 60," so it made that reduction. Id. Rainford's sentence has not been reconsidered. That appears to be inconsistent with the district court's other decisions. The district court sentenced all three defendants on the same day in January 2020 in proportion to their relative culpability. It later reduced Duncan's sentence in light of new information regarding the spin-off scheme. It then reduced Locust's sentence in light of Duncan's reduction—even though Locust was not involved in the spin-off scheme. But Rainford has not had his sentence reconsidered in light of the Duncan and Locust reductions. This result is perplexing because Rainford—unlike Locust—was involved in the spin-off scheme. Rainford PSR ¶ 41. Thus, Rainford's case for a reduction based on Duncan's reduction would be stronger than Locust's—but only Locust received a sentence reduction. True, the defendants' current sentences align with their levels of culpability: Duncan at 72 months, Rainford at 68 months, and Locust at 48 months. But on these facts, we remand for the district court to consider whether it wants to adjust Rainford's sentence as it did for the other two defendants.

The government argues that we should not do so, saying that Rainford's reliance on United States v. Jones is misplaced. We disagree. In Jones, we affirmed a sentence but remanded for reconsideration pursuant to § 2106. See 878 F.3d at 13. The government says that Jones is inapplicable because Jones received a "very, very high sentence in contrast with almost every similarly situated defendant" as a result of "timing quirks" and that no similar facts are present here. Appellee's Br. 67-68 (quoting Jones, 878 F.3d at 24 (Calabresi, J., concurring)). But these are similar facts. Rainford received a higher sentence relative to his codefendants as a result of timing quirks—because the district court had a chance to reconsider Locust's sentence, but not Rainford's sentence, following its reconsideration of Duncan's sentence. Remand is warranted because of the district court's stated intention to adjust the sentences in accordance with the defendants' relative culpability, though the district court is not required to make any adjustment if it determines that the sentences are warranted.

We affirm Rainford's sentence but remand for reconsideration "as may be just under the circumstances." 28 U.S.C. § 2106.

CONCLUSION

We affirm the judgment of the district court except as follows. We affirm the judgment with respect to Rainford's and Locust's guidelines calculations but remand to the district court for factfinding as to the number of fraudulent accidents that were orchestrated by the scheme during Rainford's and Locust's tenures for the purpose of computing the loss enhancement. We vacate Duncan's forfeiture order and remand for further proceedings consistent with this opinion. We affirm Rainford and Locust's restitution order but reduce the amount of restitution by $120,000. And we affirm Rainford's sentence but remand for reconsideration under 28 U.S.C. § 2106.

In light of the Jacobson remand on the issue of the loss enhancement, see supra note 4, upon the district court's issuance of a new order, any party may restore the matter to the active docket of this court by letter without filing a notice of appeal. If further action is sought from this court, the matter will be referred to this panel.

Dennis Jacobs, Circuit Judge, concurring:

The majority opinion, which I join in full, persuasively explains why co-conspirators can sometimes be deemed victims

pursuant to U.S.S.G. §§ 2B1.1(b)(2)(A)(i) and 3A1.1(b)(1). See Majority Op. at 482-87. This interpretation is supported by precedent, the plain meaning of the Guidelines, and the Application Note to § 2B1.1, to which we owe a qualified deference. See id. at 475-76 n.5.

The partial dissent challenges this interpretation of the word victim because the co-conspirators here also played an active role in defrauding the insurance companies. But this assignment of victim status is a mere irony, and the law is full of them. Thus, a doctor who trafficked in the organs of persons willing to sell body parts has committed an offense that has real victims. See 42 U.S.C.A. § 274e. Mayhem is an ancient offense; and someone paid to mutilate a reluctant draftee committed a crime with a real victim. See Major Eugene R. Milhizer, Maiming As A Criminal Offense Under Military Law, ARMY LAW, 5 May 1991, at 5-13.

The undeniable irony in this case does not defeat our precedent, the plain text of the Guidelines, and § 2B1.1's Application Note.

Sarah A. L. Merriam, Circuit Judge, concurring in part and dissenting in part:

I concur with Parts I., II.A.2.b., II.A.2.c., II.B., II.C., II.D.3., II.E., III.A., and III.C. of the majority opinion. However, I dissent as to Parts II.A.1., II.A.2.a., II.D.1., II. D.2., and III.B. of the majority opinion. I would remand on three additional grounds. First, the District Court erred by treating criminally liable, active co-conspirators as "victims" for purposes of adjustments under § 3A1.1(b)(1) and § 2B1.1(b)(2)(A) of the United States Sentencing Guidelines. Second, the District Court failed to adequately explain how it calculated a $100,000 intended loss per claim for purposes of determining the loss amount under § 2B1.1(b)(1), and the record does not support such a calculation. Third, the District Court erred in its calculation of restitution under the Mandatory Victims Restitution Act.

I. Criminally Liable, Actively Involved Co-Conspirators Are Not "Victims" Under the Guidelines.

The District Court erred when it treated the co-conspirator claimants as "victims" and imposed, as a result of that classification, offense level adjustments under § 2B1.1(b)(2)(A)(i) and § 3A1.1(b)(1). criminally liable, active coconspirators cannot be considered "victims" under the Guidelines. The District Court imposed¹ a two-level increase under § 2B1.1(b)(2)(A)(i) because the offense supposedly involved ten or more "victims," and a two-level increase under §3A1.1(b)(1) because defendants "knew or should have known that a victim of the offense was a vulnerable victim." This was the result of an erroneous understanding of who constitutes a "victim" under the Guidelines.

The claimants were not victims. To the contrary, they were essential coconspirators. This slip-and-fall scheme necessarily relied on each claimant's knowing, willing, and voluntary participation. Each claimant willingly agreed to stage an accident, fake an injury, and undergo medical treatment, all in order to obtain fraudulent payments. They did so knowing that the scheme posed a risk of physical harm — after all, they each volunteered to stage an accident and in many cases to undergo unnecessary surgeries.

The claimants who staged the accidents were not "victims" because they knowingly

joined the conspiracy and willingly participated in the scheme from start to finish. For example, Wanda Diaz testified that she staged the accident by herself. See Rainford App'x at 376. Keona Norwood testified: "I was told from my friend ... about it because she was doing the same thing, and when she told me that it was a lawsuit and money coming involved, I definitely agreed to do it." Id. at 75. Willie Barbour testified someone told him "that he had staged an accident ... and how easy it was and how much money he was getting from a lawsuit and how I could do the same," so Barbour "agreed and said [he] would like to do it." Id. at 382, 383. Kasheem Jones, who had heard about the scheme previously, testified that he decided to lie about his injury and join the conspiracy after he was already at the hospital for injuring his hand by punching his TV during an argument with his wife. See id. at 553-54.

Several claimants testified at trial. Not one testified that he or she was duped into joining the conspiracy without understanding what he or she would be doing. For example, Norwood testified she had no problem telling her lawyer she had a physical injury when she did not have one. See id. at 131. Reginald Dewitt testified: "What [Duncan] said was, you go to the hospital, you receive the paperwork from the hospital, tell them that you hurt your knee and your back, and once you get that paperwork, give him a call, and he'll pick up the paperwork and he'll take care of everything from there." Id. at 162. Clarence Tucker testified that it was his idea to hold a bag of chips because he thought it would make his staged slip-and-fall look better. See id. at 543-44. Yvette Battle testified that she agreed to stage a slip-and-fall after she was asked if she wanted to make some money doing so. See id. at 677-78, 695.

Sustaining bodily injury was not merely a possible consequence of participation, but the very essence of each claimant's criminal agreement. The testimony made clear that the claimants agreed to stage the accidents and undergo unnecessary surgeries knowingly. Norwood testified that she was told: "If I do the slip-and-fall, I would undergo surgery, and because I would have surgery, the lawsuit would give me more money." Id. at 73. Tucker testified: "[Locust] explained to me that you have to fall — you have to fall a certain way. When you fall, you have to fall." Id. at 514.

Indeed, the only evidence regarding any sort of misrepresentation made to any claimant was related to the amount of money that each claimant was likely to receive for his or her participation. See, e.g., id. at 599 (Jasmond Cunningham testifying that he was upset because he was told he wasn't going to get the money he expected from the scheme). But the amount makes no difference; the government's theory is that these people are "victims" because they suffered physical, not financial, harm. If each claimant had received exactly the amount of money promised to them at the outset of the scheme, that would not alter the outcome under the government's theory. Any such misrepresentations are thus irrelevant to the question of whether the claimants were victims because some of them suffered bodily harm.²

The government and the majority opinion read "victim" under § 3A1.1 and

§2B1.1 too expansively; this reading risks broadening the definition of "victim" under the Guidelines beyond its natural limits.

A. Guidelines Section 3A1.1

The term "victim" is not defined within § 3A1.1. An Application Note to § 3A1.1 defines "vulnerable victim," but as the majority opinion observes, that definition speaks more to the meaning of "vulnerable" than "victim." See Maj. Op. at 483. Thus, in classifying the claimants as "victims" under § 3A1.1, the majority opinion relies on United States v. Borst, 62 F.3d 43 (2d Cir. 1995), in support of its contention that the claimants here were "vulnerable victims" under § 3A1.1. See Maj. Op. at 482-83.

Borst is persuasive — in how its facts differ from the facts of this case. In Borst, the defendant, who was convicted of mortgage fraud in connection with loan applications he processed for three couples, argued that the vulnerable victim "enhancement was inappropriately applied because the three couples were not the actual victims of Borst's crimes." 62 F.3d at 47. The Court disagreed, holding "that even though the harm Borst caused the three couples was not an element of any of the crimes of which he was convicted, the district court did not err in considering them vulnerable victims for purposes of section 3A1.1." Id. at 48 (citation and quotation marks omitted). The argument there focused on whether the harm that the three borrower couples suffered as a result of Borst's mortgage fraud scheme was tied to the elements of the offense of conviction. See id. at 47-48. Thus, the relevance of the Borst decision lies in how the role of the borrowers there contrasted with the role of the claimants in this case.

The question of whether the borrowers should properly be considered coconspirators was not addressed in Borst, probably because the borrower-victims so clearly were not co-conspirators. Borst lied to the borrowers, including to the one borrower who actually inquired of him about a discrepancy in the paperwork. See id. at 45. In one instance, he failed to pay loan proceeds to the seller, resulting in a borrower's eviction. See id. In another case, he obtained a promissory note for a different mobile home than had been agreed upon by the borrowers. See id. at 46. The only mention in Borst that the borrowers might have had any idea that fraud was afoot is this: "Whether or not the three couples in the case before us were 'unwitting instrumentalities' of Borst's criminal conduct in light of their apparent knowledge of Borst's misrepresentations to the bank, they were exploited and suffered harm as a result of his actions." Id. at 48. The basis for the allusion to "apparent knowledge" is not clear from the opinion. But at most, this language suggests that the borrowers might have been aware that Borst made misrepresentations to the banks in connection with their loans. The borrowers did not enter into the transactions knowing they would be fraudulent, and they certainly did not agree to incur the harm they eventually suffered. They did not collude with Borst so that they could jointly profit with Borst from the fraud. They did not take any affirmative steps to advance the fraud. They did not recruit additional participants into the fraud. They did not, themselves, make misrepresentations to the lenders. Perhaps they looked the other way when they suspected that Borst was not doing things quite right, but they were not knowing, willing, and voluntary, co-conspirators. The claimants in this case were.

Additionally, the bulk of the analysis — and the primary precedential holding — in Borst relates to the question of whether the borrowers targeted by the defendant

were vulnerable, not on whether they were criminally liable as coconspirators. In other words, Borst skipped to the question of whether the borrowers were vulnerable, without addressing the question of whether they were victims. See Borst, 62 F.3d at 46-47. After finding vulnerability, the Borst Court turned to the defendant's argument that the borrowers should not be considered victims "because the three couples were not the actual victims of Borst's crimes." Id. at 47. The question confronted by the Court was not whether a knowing, voluntary, active criminal co-conspirator could be considered a "victim" for Guidelines purposes. Rather, the question was whether the borrowers could be considered victims even though they did not suffer the primary economic losses of Borst's actions. See id. at 47. That is not the question presented here.

Borst did state that "the couples' precarious financial situation alone may serve as the sole basis of a § 3A1.1 enhancement," but it did so in the context of addressing the defendant's argument "that the vulnerability requirement of § 3A1.1 cannot be fulfilled solely on the basis of the couples' socioeconomic circumstances." Id. at 46-47. Here, the question is whether the co-conspirators were victims at all — not whether they were vulnerable. We do not need to reach the vulnerability³ inquiry because the analysis ends at the threshold question: Were the criminally liable co-conspirator claimants "victims" under the Guidelines? They were not. In stark contrast to the borrowers in Borst, the claimants here knowingly and willingly joined in and furthered the conspiracy by deliberately taking repeated, affirmative, criminal actions. These coconspirators, like many other criminal defendants, may have been vulnerable, in a colloquial sense, but they were not victims by any means.⁴

Borst also draws on United States v. Echevarria, 33 F.3d 175 (2d Cir. 1994), abrogated on other grounds by United States v. Hussey, 254 F.3d 428 (2d Cir. 2001). There, the vulnerable victim adjustment

applied because Echevarria posed as a doctor and tricked his unwitting patients, "luring them to his inadequate and dangerous medical attention for the purposes of defrauding third-party medical insurers." Id. at 180. Echevarria compared the factual circumstances there to the facts in United States v. Bachynsky, 949 F.2d 722 (5th Cir. 1991), noting that the patients in Bachynsky were vulnerable victims "both because they falsely believed that they were receiving effective medical attention, and because they were unwitting instrumentalities of the fraud." Echevarria, 33 F.3d at 180. The Court concluded that the patients in Echevarria "were similarly duped." Id. The same cannot be said of the claimants here.⁵

The differences between the claimants here and the borrowers in Borst and the patients in Echevarria are telling. The claimants took numerous affirmative steps to advance their joint venture with defendants. They were active participants, not mere instrumentalities, in the scheme. The conspiracy required the claimants to perform numerous overt acts in furtherance of its goals. Any claimant could have stopped participating at any time. But the claimants took part voluntarily step, after step, after step.

They traveled to the accident site. They staged or play-acted an accident at the site. They called an ambulance. They lied to the responding emergency personnel about the accidents and their injuries. They checked into hospitals under their real names, then lied to hospital personnel about the accidents and their injuries. They continued to lie to medical personnel over many months. They met with lawyers and signed paperwork to bring fraudulent lawsuits. They underwent medical procedures that were either entirely unnecessary or necessitated by previous conditions rather than by any accident. Some even lied in depositions taken in the fraudulent lawsuits. See Rainford App'x at 366 (Diaz); id. at 562 (Jones); id. at 595-96 (Cunningham); id. at 639-41 (White). And a number of claimants testified that they were so satisfied with the experience that they referred close friends and family to the scheme as additional claimants, including their parents and significant others. See id. at 186 (Dewitt told his sister, his brother, his daughter, his wife and "a few other people."); id. at 458 (Nichols brought her friends into the conspiracy.); id. at 563 (Jones brought his girlfriend into the conspiracy.); id. at 1087 (Alvin Martin brought his mother and two others into the conspiracy.). These co-conspirators cannot be considered vulnerable victims under § 3A1.1. B. Guidelines Section 2B1.1

The term "victim" is not defined in § 2B1.1. Thus, the majority opinion relies on the following Application Note to classify the claimants as "victims": "'Victim' means (A) any person who sustained any part of the actual loss determined under subsection (b)(1); or (B) any individual who sustained bodily injury as a result of the offense." U.S.S.G. § 2B1.1 cmt. n.1. There is no contention here that any claimant suffered any part of the financial loss; the majority opinion contends that the claimants are "victims" under this definition because some of them sustained bodily injury as a result of their involvement in the offense.

However, the Application Note is not binding on the Court. As the majority opinion notes, that commentary is considered "authoritative" only if it does not violate "the Constitution or a federal statute," or is not "inconsistent with, or a plainly erroneous reading of, that guideline." Stinson v. United States, 508 U.S. 36, 38, 113 S.Ct. 1913, 123 L.Ed.2d 598 (1993).⁶ Applying this Application Note to classify the claimants as victims results in a "plainly erroneous reading" of the Guideline, under which an unindicted co-conspirator, or, indeed, a named defendant, may also be a victim under the Guidelines. Such an interpretation would lead to absurd results.⁷

We have long held that "[a] statute should be interpreted in a way that avoids absurd results." United States v. Dauray, 215 F.3d 257, 264 (2d Cir. 2000); see United States v. Venturella, 391 F.3d 120, 126-27 (2d Cir. 2004) ("The absurd results canon cited in Dauray ... is a rule of statutory construction that serves to help resolve ambiguity." (citation and quotation marks omitted)).⁸ The majority opinion's definition of "victim" as including any person who suffered bodily injury as a result of an offense would lead to absurd outcomes where individuals who had repeatedly

manifested their criminal intent and were active and essential participants in a conspiracy could be considered victims.

Reading the Application Note literally, as the majority opinion suggests, would mean that a lone defendant, participating in a crime on his own, who suffered bodily harm as a result of his own offense, would be deemed a "victim" under the Guidelines. A single individual could engage in a fraud like the one perpetrated here, but with no co-conspirators. He could fake an accident, lie about his injuries, seek unneeded medical care, and demand payment. If, in the course of committing that fraud, he suffered bodily injury, a literal reading of the Application Note would declare him — a solitary defendant — to be a "victim" of his own offense. That would be an unreasonable reading of the Application Note because, as the majority opinion points out, in law, "a person generally cannot be his own victim." Maj. Op. at 485.

Likewise, in a conspiracy in which several defendants are charged and those defendants suffer bodily harm as a result of their criminal conduct, those defendants would be counted as victims of themselves and their co-defendants. In a conspiracy charging ten defendants, all of whom suffered physical injuries as a result of their offense conduct, would all ten defendants be eligible for a two-level increase under § 2B1.1 because they were all serving dual roles as victims and perpetrators? What if there were eleven defendants, and only ten were hurt; would only the un-injured defendant be eligible for the increase? Or would they all be eligible? If a defendant sells drugs to nine customers, and takes some of the drugs himself, and all ten people suffer overdoses and serious medical consequences, are there ten victims of the offense? Or only nine?

Here, the government recognized the criminal liability of the claimants; it provided non-prosecution or immunity agreements to each claimant who testified. See, e.g., Rainford App'x at 567-69. The government chose not to prosecute the claimants, but it very well could have done so. Yet, had the government actually charged the claimants as co-conspirators, it is hard to imagine that the government would have sought to classify those claimants as victims. Is it, then, a prosecutor's charging decision that determines whether an acknowledged criminal co-conspirator is also a victim? The majority opinion appears to take that approach, asserting: "[W]e do not agree that either § 3A1.1 or § 2B1.1 would include a defendant who suffered harm as a result of his own offense." Maj. Op. at 485-86 (emphasis added). The claimants here were not defendants, because the government elected to give them immunity in exchange for their testimony, rather than charging them. But the fact that they are unindicted co-conspirators rather than defendants does not make it any more sensible to declare them "victims" under these Guidelines provisions. A person, after all, cannot be his own victim.

The government does not argue that Rainford, Locust, or Duncan performed the allegedly harmful surgeries, or that they physically harmed their co-conspirators by directly inflicting injuries on them. Nonetheless, the majority opinion contends that those three should be treated as "more culpable than ... a defendant who, say, commits fraud by undergoing the unnecessary surgery himself" because they "profited by recruiting economically desperate people." Maj. Op. at 484. That may well be true. But § 2B1.1 is not how the Guidelines impose increased penalties on those who recruit others to join a conspiracy; culpability for recruiting and supervising co-conspirators is addressed elsewhere. See, e.g., United States v. Burgos, 324

F.3d 88, 92 (2d Cir. 2003) (An increase in offense level for role in the offense under § 3B1.1 may be appropriate for a defendant who "played a significant role in the decision to recruit or to supervise lower-level participants." (citation and quotation marks omitted)).⁹

Moreover, had the Sentencing Commission wanted the adjustment under § 2B1.1 to apply to conspirators who were involved in the offense itself, and suffered bodily harm as a result of that involvement, the Guidelines would have explicitly said so. Indeed, the Commission did exactly that in § 2D1.1, which calls for an adjustment when a "defendant, knowing that an individual was (i) less than 18 years of age, (ii) 65 or more years of age, (iii) pregnant, or (iv) unusually vulnerable due to physical or mental condition or otherwise particularly susceptible to the criminal conduct, distributed a controlled substance to that individual or involved that individual in the offense." U.S.S.G. § 2D1.1(b)(16)(B) (emphasis added). Notably, § 2D1.1 does not classify individuals who were "involved" in the offense as victims, but still imposes an adjustment on a defendant who "involved" them in the offense by, for instance, recruiting them to participate in a conspiracy. The Commission used explicit language to impose an adjustment on a defendant who involved others in a drug offense; so why would the Commission not use that same language if it wished to impose an adjustment on a defendant who involved others in a fraud offense? Because it did not intend § 2B1.1 to impose an adjustment if the person who suffered harm was a co-conspirator.

The errors in the District Court's calculation of restitution under the Mandatory Victims Restitution Act ("MVRA") are addressed later, but it is notable that we do not treat co-conspirators as victims under the MVRA. The MVRA has a broad definition of victim:

For the purposes of this section, the term "victim" means a person directly and proximately harmed as a result of the commission of an offense for which restitution may be ordered including, in the case of an offense that involves as an element a scheme, conspiracy, or pattern of criminal activity, any person directly harmed by the defendant's criminal conduct in the course of the scheme, conspiracy, or pattern.

18 U.S.C. § 3663A(a)(2). But, under the MVRA, a co-conspirator may not receive compensation as a victim because courts have recognized the absurdity of that proposition. See United States v. Reifler, 446 F.3d 65, 127 (2d Cir. 2006) ("[A]ny order entered under the MVRA that has the effect of treating coconspirators as 'victims,' and thereby requires 'restitutionary' payments to the perpetrators of the offense of conviction, contains an error so fundamental and so adversely reflecting on the public reputation of the judicial proceedings that we may, and do, deal with it sua sponte."); see also United States v. Archer, 671 F.3d 149, 171 (2d Cir. 2011) ("[C]o-conspirators, who, by definition, know of the scheme, are not victims and may not receive restitution."); United States v. Ojeikere, 545 F.3d 220, 222-23 (2d Cir. 2008) (Victims who arguably had unclean hands were still entitled to restitution because they "were not involved in the offense of conviction, which was a fraudulent scheme to obtain money from them."). The Ninth Circuit, relying on Reifler, has made it clear:

Under the plain text of the MVRA ... co-conspirators have just as much right to restitution as do innocent victims.

But courts have recognized that Congress could not have intended that result. Otherwise, the federal courts would be involved in redistributing funds among wholly guilty co-conspirators, where one or more co-conspirators may have cheated their comrades.

United States v. Lazarenko, 624 F.3d 1247, 1250-51 (9th Cir. 2010) (emphasis added). This analysis applies equally to the definition of "victim" in the Application Note to § 2B1.1. The majority opinion would classify willing coconspirators as victims if they suffer bodily harm — but the Commission could not have intended that absurd result.

The claimants are co-conspirators who knowingly, willingly, and voluntarily participated in and even promoted the scheme.¹⁰ It was therefore error to treat them as "victims" for purposes of § 3A1.1(b)(l) and § 2B1.1(b)(2)(A), and I would remand for resentencing on that basis.

II. The District Court Erred in Determining that the Intended Loss as to Each Claim was $100,000.

The majority opinion correctly concludes that remand is necessary for the District Court to make factual findings regarding the number of fraudulent accidents the Kalkanis conspiracy orchestrated.¹¹ But remand is also necessary because the District Court did not "make findings that are sufficiently specific to permit meaningful appellate review" when it calculated an intended loss of $100,000 per claim. United States v. Flores, 945 F.3d 687, 721 (2d Cir. 2019). Additionally, a review of the record reveals that the District Court's intended loss figure of $100,000 is not "grounded in the evidence" in the record and instead impermissibly "derive[s] from mere speculation." United States v. Coppola, 671 F.3d 220, 249 (2d Cir. 2012).

Section 2B1.1(b)(1) provides for an adjustment to the Base Offense Level when the monetary loss from an offense exceeds certain thresholds. The District Court held Rainford and Duncan responsible for a loss between $25 and $65 million, which resulted in a twenty-two-point increase to their Base Offense Levels. See Duncan Pre-Sentence Report ("PSR") ¶52; Rainford PSR ¶51; U.S.S.G. § 2B1.1(b)(1)(L). The District Court held Locust responsible for a loss between $9.5 and $25 million, which resulted in a twenty-point adjustment to his Base Offense Level. See Locust PSR ¶51; U.S.S.G. § 2B1.1(b)(1)(K). These aggregate loss amounts were based on a $100,000 intended loss per claim — but the District Court failed to explain how it arrived at that number.¹² When calculating a Guidelines adjustment, the District Court "is required to make findings that are sufficiently specific to permit meaningful appellate review." Flores, 945 F.3d at 721; see also United States v. Patasnik, 89 F.3d 63, 69 (2d Cir. 1996) ("[I]t is clear under our precedents that an implicit finding is not enough."); United States v. Carter, 489 F.3d 528, 538 (2d Cir. 2007) ("Although this requirement of making specific factual findings may interfere with the smooth operation of the sentencing hearing, we require specific factual findings to permit meaningful appellate review." (citation and quotation marks omitted)).

To be sufficiently specific to permit meaningful appellate review, it is not enough for the court merely to repeat or paraphrase the language of the guideline and say conclusorily that the defendant meets those criteria. And although a sentencing court may sometimes satisfy its obligation to make findings by adopting the factual statements in the defendant's presentence report, adoption of the PSR does not suffice if the PSR itself does not state enough facts to permit meaningful appellate review.

United States v. Skys, 637 F.3d 146, 157 (2d Cir. 2011) (citations and quotation marks omitted).

On appeal, we are required to "determine whether the trial court's method of calculating the amount of loss was legally acceptable." United States v. Rutkoske, 506 F.3d 170, 178 (2d Cir. 2007) (citation and quotation marks omitted). The District Court failed to articulate the method and the numbers it relied on to come to the conclusion that the intended loss per claim was $100,000. The record is completely lacking in any meaningful analysis of the loss amount and thus there cannot be any meaningful appellate review of the District Court's decision.

The majority opinion acknowledges that the District Court did not make sufficiently specific factual findings on the number of fraudulent accidents that were part of the conspiracy. See Maj. Op. at 478, 480-81. But the District Court's failure to explain infects both aspects of the loss calculation because the District Court offered no more of an explanation for the intended loss per claim than it did for the number of fraudulent accidents. Instead, the majority opinion and the government rely on post-hoc rationalizations to justify the District Court's adoption of the $100,000 figure.

At Locust's sentencing, the District Court asked the government how it "came up with the loss amount here?" Rainford App'x at 1278. In response, the government "point[ed] the Court to paragraph 38 of the PSR, which set forth how the government calculated the loss amount for Mr. Locust." Id. The District Court then asked: "[W]here does the foreseeable loss per recruit of each victim of a hundred thousand come from?" Id. at 1279. In response, the government stated:

Clarence Tucker testified at trial — this is page 543 to 544 of the transcript —that he was told by Mr. Locust he could make a hundred thousand dollars or better in staging an accident. So in terms of the individual ability to recover, Mr. Locust had knowledge of that. And then in terms of the volume of patients, he also had knowledge of that because of the frequency of which he was driving patients every day to doctor's appointments, to lawyer's appointments, cars full of patients. I mean, he got a bigger car, with Mr. Kalkanis' assistance so he could transport more patients. So a conservative estimate here of a reasonably foreseeable loss amount is what the guidelines are as calculated in the PSR.

Id. Based on that response, the District Court, without further elaboration, adopted the "findings of fact in the presentence report" and found the "intended loss range of between 9.5 and 25 million [dollars]" to be "appropriate." Id. at 1281. It similarly adopted the findings of fact in the PSRs for Rainford and Duncan. See id. at 1299, 1234.

Neither the District Court nor the government offered any explanation of how they determined that $100,000 was a reasonable assessment of the intended loss per claim. Was one conspirator's testimony about the amount he was told he, personally, could make by participating in the fraud really the entire basis for calculating the intended loss for every claim brought in the entire scheme? The District Court never shared its reasoning, so we are left to speculate.

Furthermore, the District Court's "adoption of the PSR does not suffice" because the "PSR itself does not state enough facts to permit meaningful appellate review." Skys, 637 F.3d at 157 (citation and quotation marks omitted). The majority opinion takes a guess at how the District Court might have come up with the $100,000 figure:

In support of [the conclusion that the intended loss for each fraudulent accident was $100,000], the PSRs referenced (1) Clarence Tucker's testimony that his case settled for $100,000, (2) Kasheem Jones's testimony that his case settled for $225,000 and that his girlfriend's case settled for $250,000, (3) Carol White's testimony that that her case settled for $80,000, and (4) Alvin Martin's testimony that his case settled for $120,000. See, e.g., Rainford PSR ¶24. The district court adopted these findings for each defendant. See Rainford App'x 1281 (Locust), id. at 1299 (Rainford), id. at 1234 (Duncan).

Maj. Op. at 477. The District Court did adopt the factual findings in defendants' PSRs, but it provided no meaningful explanation as to how those factual findings supported the $100,000 figure. Indeed, the government has never articulated a coherent basis for the loss calculations. Simply listing the settlement amount for five cases mentioned in the PSRs, without further explanation, is not sufficient to permit appellate review. Further, the PSR's conclusion that "the attempted loss per recruited patient was only $100,000" makes no reference to the portion of the PSR cited by the majority opinion, and does not appear to rely on it. Compare, e.g., Rainford PSR ¶38, with Rainford PSR ¶24. Indeed, there is no indication that any of the PSR information relied upon on appeal actually served as the District Court's basis for its calculation of the intended loss.¹³ The majority opinion also points to the District Court's statement that "sometimes it was held out to people that they could make up to a hundred thousand dollars." Maj. Op. at 477 (quoting Rainford App'x at 1278). However, without further elaboration, this passing comment is not a form of factfinding sufficient to support meaningful appellate review. And, even if claimants were "sometimes" told they could make $100,000, that alone is not sufficient to support a finding that $100,000 was the intended loss for every single claim.

The District Court's failure to make the requisite findings with sufficient specificity is enough to warrant remand. But even if we were to engage in review of the factual determination, we would be bound to find that the $100,000 figure is not supported by the evidence in the record. While a district court need not calculate the loss amount with "absolute precision," it must "make a reasonable estimate of the loss given the available information," and that reasonable estimate must be supported "by a preponderance of the evidence." United States v. Rivernider, 828 F.3d 91, 112 (2d Cir. 2016) (citation and quotation marks omitted). "In determining a loss amount for purposes of Guidelines calculation, a district court's findings must be grounded in the evidence and not derive from mere speculation." Coppola, 671 F.3d at 249.

Here, the loss calculations can only have been based on speculation because the evidence presented to the District Court does not support an intended loss per claim of $100,000. As discussed, the District Court asked one question about the basis for the $100,000 estimate, to which it received an utterly unsatisfactory answer. The government failed to explain why Clarence Tucker's testimony about the amount he was told he could make could be extrapolated to serve as an average amount for every claim in the conspiracy. The government contends that an intended loss of $100,000 per claim is a "conservative estimate." Appellee's Br. at 49. But any estimate — conservative or otherwise — must be supported by evidence in the record.¹⁴

In sum, the District Court failed to adequately explain the basis for its finding that the intended loss per victim was $100,000, and the record does not support that figure. Therefore, I would remand for the District Court to conduct a proper analysis of the intended loss per victim.

III. The District Court Erred in Calculating Restitution under the MVRA.

The District Court ordered Rainford and Locust to pay restitution in the total amount of $3,928,133.60 pursuant to the

MVRA, as requested by the government. See Rainford App'x at 1379, 1390. This award was not supported by specific findings as to which payments made by which insurance companies were based on fraudulent claims. Therefore, it must be vacated.

Calculating restitution under the MVRA is not meant to be a guessing game. Restitution is meant to compensate victims for their loss, their whole loss, and nothing but their actual loss. "[B]ecause the MVRA itself limits restitution to the full amount of each victim's loss,' a restitution order must be tied to the victim's actual, provable, loss." United States v. Zangari, 677 F.3d 86, 91 (2d Cir. 2012) (emphasis added) (quoting 18 U.S.C. § 3664(f)(1)(A)). "In determining the proper amount of restitution, a court must keep in mind that the loss must be the result of the fraud." United States v. Paul, 634 F.3d 668, 676 (2d Cir. 2011) (citations and quotation marks omitted); see also United States v. Marino, 654 F.3d 310, 319-20 (2d Cir. 2011) ("[R]estitution is authorized only for losses that were directly caused by the conduct composing the offense of conviction, and only for the victim's actual loss." (citations and quotation marks omitted)). "The Government bears the burden of proving a victim's actual loss by a preponderance of the evidence." Zangari, 677 F.3d at 92.

Even though the two calculations are independent, many of the defects discussed above in relation to the calculation of the Guidelines loss amount also infect the restitution orders.¹⁵ The government's method of calculating restitution in this matter — which was adopted without analysis by the District Court — is deeply flawed. It looks simple enough: The government started with a total of $4,910,167, which is the amount that nine insurance companies¹⁶ reported they "paid out on [38] cases that the Government ha[d] identified as being part of the Fraud Scheme." Rainford Sealed App'x at 5. Then, presumably relying on Peter Kalkanis's testimony, the government reduced the total amount by 20%, so that the "restitution orders [were] based on 80% of the total amount of money that was paid out by the insurance companies," because "approximately 80% of the cases in the Fraud Scheme were fraudulent." Id. at 7. That left a total restitution amount of $3,928,133.60. See Rainford App'x at 1379, 1390.

All parties acknowledge the most glaring error in that restitution calculation: the inclusion of Alvin Martin's legitimate $150,000 claim. While the majority opinion

and the government concede this error, they fail to acknowledge its broader significance — that it reflects the lack of evidence supporting a finding that each of the 38 cases aggregated by the government was actually a part of the charged fraud. A review of the record reveals that the only "evidence" of the allegedly fraudulent claims is the government's summary ("Restitution Letter") and a set of spreadsheets apparently prepared by the insurance companies.

The government's Restitution Letter indicates that the government identified the claims to be included in the restitution order "[b]ased on the information provided by the insurance companies." Rainford Sealed App'x at 5. The spreadsheets that were sent to the government by the insurance companies do not tell us much. They are not accompanied by affidavits or certifications explaining their contents. No restitution hearing took place, and thus, there was no testimony from any representatives of the companies explaining the spreadsheets. The spreadsheets are not consistent with each other, and do not provide all the information one would expect or hope to see. The Restitution Letter does not adequately explain why certain claims are included (or, potentially, excluded).

There is little other evidence indicating that any of these 38 claims were related to the fraud. There was testimony at trial from three of the conspirators — Kasheem Jones, Alvin Martin, and Clarence Tucker — about their own claims (and in the case of Jones, the claim of his girlfriend CF).¹⁷ These claims were included in the 38 claims on which the restitution order was based. See Rainford Sealed App'x at 3. But the only evidence the government offers to show that the remaining payouts were part of the fraud is that "the insurers' restitution claims related to cases involving an attorney, broker, runner, doctor, and/or funding company that participated in the conspiracy." Appellee's Br. at 76. This approach is problematic in several ways.

First, in one instance, we know that the assumption that a claim was fraudulent simply because a participant in the scheme was involved in it was false because the government has conceded that the "restitution totals should not have factored in AmTrust's $150,000 settlement for Alvin Martin's 2014 claim because, as Martin testified at trial, that claim — unlike Martin's fraudulent 2012 claim — was legitimate." Id. at 74.

There are other significant questions about the 38 claims on which the restitution order was based:

• As to at least one claim, the only connection that claim had to the scheme was that the litigation funding firm "Fast Trak" was involved. See Appellee's Br. at 76. But Fast Trak is a national company, and likely had involvement in many cases unrelated to this scheme.

• The Hanover claims in the record appear to have been presumed fraudulent based solely on the attorneys involved — though there was no evidence that Attorneys Elefant and Constantine brought only fraudulent claims related to the charged scheme during this time period. See Appellee's Br. at 76; Rainford Sealed App'x at 25.

• $30,000 in restitution was awarded to Hartford Financial Services Group for KG's claim, see Rainford Sealed App'x at 6, but that claim is reported as "open" rather than "paid" on the spreadsheet, see id. at 30, and thus presumably was not paid. The government dismisses this as a typo but it provides no basis for that assertion. See Appellee's Br. at 75.

• $89,500 in restitution was awarded to Nationwide for JT's claim, see Rainford Sealed App'x at 7, even though the Nationwide spreadsheet indicates that company paid only $2,500 and a co-defendant paid the remaining $87,000, see id. at 28. Similarly, $5,000 in restitution was awarded to Nationwide for BT's claim, see id. at 7, even though Nationwide reported that a co-defendant paid the "majority" of the $5,000 payout for BT, see id. at 28. In both instances, the government included the full amount of the claim in the restitution award, without taking into account the notes in the spreadsheet indicating that Nationwhide actually paid out a lower amount.

More importantly, the entire basis of the restitution calculation is erroneous. Kalkanis's estimate that 80% of claims were fraudulent is unreliable and insufficient to support the restitution order. He testified that "as high as 80 percent of those cases were fraudulent," Rainford App'x at 1018, but he also testified that "practically all" of the "intake sheets involved fraudulent slip-and-falls" and when he was asked to clarify that statement, he stated only that "[t]he majority of them" were fraudulent, id. at 884. But even assuming an 80% rate to be accurate, a 20% reduction to the total amount of restitution does not accurately account for a finding that 20% of the claims made were not fraudulent.

The District Court had an obligation to determine the exact amount of restitution owed to each victim and to award that amount, ensuring that it neither overpaid nor underpaid any victim. But this restitution award does both. It overpays insurance companies who did not actually suffer the full loss claimed and underpays insurance companies who did actually suffer the full loss claimed.

Our decision in United States v. Reifler is instructive. There, the sentencing court ordered restitution based on a list of payees that admittedly included persons who were not properly classified as victims under the MVRA. See 446 F.3d at 125. The government never "in fact identified the entries for" amounts that should not have been included in the restitution order, nor "removed them from" the exhibit on which the restitution order was based. Id. The exhibit on which the restitution order was based included payments to investors whom "the government conceded could not be considered victims," as well as to certain co-conspirators who the government conceded it knew were "not victims." Id. Instead of actually excluding those who were improperly classified as victims from the restitution calculation, the government "conceptually exclud[ed]" them "by requesting that Reifler be held accountable for losses of only $3 million, rather than the $6 million" in total claims appearing in the government's exhibit. Id. at 126.

We found the government's efforts to "estimate" the losses inadequate: "In sum, the Reifler Amended Judgment orders Reifler to pay restitution to persons listed in Government Exhibit 3, but Exhibit 3 includes persons who were not ... conspiracy victims within the meaning of the MVRA, either because their losses resulted from purchases they made after the conspiracy had ended or because they were coconspirators." Id. at 127. The district court's restitution order not only resulted in improper payments to ineligible parties, but in underpayment to true victims.

See id. at 133 ("But the [district court's] order that Reifler pay only $2 million to the persons listed in Government Exhibit 3, whose total losses are listed at more than $6 million, means that the maximum amount of restitution to be received from Reifler by each person on that list — victims and nonvictims alike — is less than one-third of the specified loss.").¹⁸ As a result, we vacated the restitution order. The government and the District Court took the same rough estimate approach here, and we should vacate the restitution order in this case as well.

18 The government suggested that the court award restitution of $3 million, that is, about half the losses listed in its exhibit, but the sentencing court reduced the percentage awarded further to only $2 million. Neither blanket reduction was supported by the evidence. See Reifler, 446 F.3d at 133.

The claims in Reifler were extremely complex, and the "best guess" approach was an understandable temptation. But here, it would have been simple enough for the government to evaluate each of the 38 claims; determine whether each was indeed fraudulent; calculate exactly how much was paid on each fraudulent claim; and then seek a proper restitution order based on that information. But it inexplicably failed to do so. The government concedes that at least 20% of the claims were not fraudulent, but rather than making an effort to determine which claims were fraudulent and which were not, it "conceptually" excluded the legitimate claims by reducing the total amount to be paid to each victim by 20%. We found that approach improper in Reifler, and it is improper here. See Reifler, 446 F.3d at 134 (observing that "the presence of nonvictims on the list of persons to whom restitution is to be paid has the effect of diluting the amount the victims will receive"); Zangari, 677 F.3d at 93 ("[T]he MVRA ... does not allow a sentencing court to substitute gain for loss...."). We simply do not know which claims constitute the 20% that the government appears to concede are legitimate. Are they the largest or the smallest claims? Which victims sustained those losses? What if all of the legitimate claims were paid by a single insurance company — why should that company be paid at all, and the other companies' payments be reduced by 20%?

"[E]ven where a defendant's complex fraud scheme results in many victims whose identities and losses are difficult to ascertain, the district court should identify the victims and their actual losses prior to imposing restitution under the MVRA." United States v. Catoggio, 326 F.3d 323, 329 (2d Cir. 2003) (emphasis added). This fraud was not complex. There were steps the District Court (and the government) could have taken to determine the actual restitution amount:

The court may, for example: "require additional documentation or hear testimony," 18 U.S.C. § 3664(d)(4); allow additional time "for the final determination of the victim's losses, not to exceed 90 days after the sentencing," id. § 3664(d)(5); and "refer any issue arising in connection with a proposed order of restitution to a magistrate judge or special master for proposed findings of fact and recommendation as to disposition," id. § 3664(d)(6).

Zangari, 677 F.3d at 93. The District Court could have declined to order restitution:

Ultimately, if the court finds that "complex issues of fact related to the cause or amount of the victim's losses would complicate or prolong the sentencing process to a degree that the need to provide restitution to any victim is outweighed by the burden on the sentencing process," then the court may, in the exercise of its sound discretion, decide not to order restitution at all. 18 U.S.C. § 3663A(c)(3)(B); see also USSG § 5E1.1(b)(2) (same).

Id.; see also Marino, 654 F.3d at 319 ("Congress explained [the MVRA's] causation standards as follows: ... 'The committee believes that losses in which the amount of the victim's losses are speculative, or in which the victim's loss is not clearly causally linked to the offense, should not be subject to mandatory restitution.'" (quoting S. Rep. No. 104-179, at 19 (1995))).

The majority opinion and the government, relying on United States v. Rossi, 592 F.3d 372 (2d Cir. 2010) (per curiam), contend that "restitution may be based on 'guesswork' or 'even a hunch.'" Maj. Op. at 490 (quoting Rossi, 592 F.3d at 376); see also Appellee's Br. at 73; Maj. Op. at 490 ("[O]rdering restitution requires a delicate balance of diverse, sometimes incomparable, factors, some of which not only lack certainty but may indeed be based on mere probabilities, expectations, guesswork, even a hunch." (quoting Rossi, 592 F.3d at 376)).¹⁹ This reliance on Rossi is misplaced, because the language about "guesswork" and "a hunch" was in the context of determining restitution under the Victim and Witness Protection Act ("VWPA"), not the MVRA. See Rossi, 592 F.3d at 374 ("The issue of restitution in this case is governed by the VWPA.").²⁰ And in Rossi, we found that the sentencing judge had been "cautious and restrained" in her restitution analysis, and observed that she had properly declined to award restitution based on "speculating and choosing numbers." Rossi, 592 F.3d at 376 (citation and quotation marks omitted).

19 The quoted language in Rossi in turn quotes United States v. Atkinson, 788 F.2d 900, 902 (2d Cir. 1986), which predates the adoption of the MVRA and is no longer good law. As this Court has noted, the MVRA wrought significant changes in the law applicable to restitution. See United States v. Walker, 353 F.3d 130, 131 (2d Cir. 2003).

20 "In contrast to the MVRA, ... restitution under the VWPA is discretionary." United States v. Battista, 575 F.3d 226, 230 (2d Cir. 2009).

The restitution order in this case was based on the MVRA, which requires: "In each order of restitution, the court shall order restitution to each victim in the full amount of each victim's losses as determined by the court and without consideration of the economic circumstances of the defendant." 18 U.S.C. § 3664(f)(1)(A). Under the MVRA, the District Court had an obligation to review the insurance company claims and determine which claims were part of the charged scheme, and to then award 100% of the losses on the fraudulent claims, and only the fraudulent claims, to the actual victims. See United States v. Walker, 353 F.3d 130, 133 (2d Cir. 2003). Guesswork was not permissible.

I agree with the majority opinion that a restitution calculation under the MVRA need not always be "mathematically precise." Maj. Op. at 489-90 (quoting Rivernider, 828 F.3d at 115). But, again, the context of that quoted language is important. Rivernider took the phrase "mathematically precise" from United States v. Gushlak, 728 F.3d 184, 195 (2d Cir. 2013) ("[W]e have never used the word 'actual' in this context to mean 'mathematically precise.'"). Gushlak explained that in calculating restitution, "a 'reasonable approximation' will suffice, especially in cases in which an exact dollar amount is inherently incalculable." Id. at 196 (emphasis added) (citations omitted). Here, the exact dollar amount was absolutely calculable. Furthermore, Gushlak held that a district court's "reasonable approximation" must be "supported by a sound methodology." Id. Gushlak presented a complex "pump-and-dump" investment scheme; calculating the restitution required determining "if and to what extent particular investors ha[d] been harmed by artificial prices that [were] the result of deliberate misinformation of one sort or another (including manipulative trading practices designed to inflate the price)." Id. And while Gushlak spoke of approximation of loss, the district court there actually undertook a detailed and involved inquiry to determine the loss amount. See id. at 197-201. Simply reducing the total restitution amount by 20% does not constitute a "reasonable approximation of losses [that was] supported by a sound methodology." Id. at 196.

Therefore, I would hold that remand is required for the District Court to determine "the full amount of each victim's losses" under the MVRA. 18 U.S.C. § 3664(f)(1)(A).

IV. Conclusion

I would remand for resentencing to address three errors beyond those identified by the majority opinion. First, the District Court erred by treating the criminally liable co-conspirators as "victims" for purposes of § 3A1.1(b)(1) and § 2B1.1(b)(2)(A) of the Guidelines. Second, the District Court failed to adequately explain its calculation of an intended loss per fraudulent claim of $100,000 for purposes of §2B1.1(b)(1) of the Guidelines, and that amount is not supported by the record. Third, the District Court erred by failing to make factual findings as to which payments made by which insurance companies were based on fraudulent claims when it ordered restitution under the MVRA. Accordingly, I respectfully dissent as to Parts II.A.1., II.A.2.a, II.D.1., II.D.2., and III.B. of the majority opinion.