MARCUS, Circuit Judge:
After a trial by jury, Nivis Martin was convicted of multiple counts of bank and wire fraud, sentenced to a prison term of 46 months, and ordered to pay $963,400.47 in restitution. On appeal, Martin challenges her convictions, contending that the district court erred in denying her motion for judgment of acquittal on all counts because the evidence was insufficient.
See
Fed.R.Crim.P. 29. She also challenges her sentence, arguing that her offense level should have been reduced pursuant to United States Sentencing Guidelines § 3B1.2 to reflect her minimal or minor role in the offense. Finally, she says that the trial court erred when it imposed an order of restitution. After thorough review, we affirm her convictions and sentence, but we vacate the restitution award and remand for a new determination of the victims’ losses.
I.
The essential facts are these.
On December 13, 2006, Martin and her ex-husband purported to sell their apartment in Miami, Florida, to Martin’s father for $495,000 (the “Miami Apartment”).
Her father applied for two mortgage loans from First Franklin Financial Corp. (a division of National City Bank Corp.), totaling $495,000 to fund his purchase. His applications misrepresented just about everything that an underwriter would want to know about him. Among other things, the applications grossly misstated his monthly income (they claimed that he made $13,500 per month when, in reality, he made only $34,168 per year) and bank account balance (they claimed that his bank account balances totaled over $106,000 when his true balance was $484.37). The loan applications also falsely claimed that he earned monthly rental income on another piece of residential property that he owned and continued to live in (a residential home at 10621 SW 27th St. Miami, FL). He submitted a fraudulent lease in support of his loan purporting to show that Martin’s childhood friend was renting the property from him. To verify the bogus lease, Martin paid that childhood friend to purchase a
cashier’s check made payable to Martin’s father — to reflect a supposed deposit on the lease.
The applications also omitted several important details. Thus, for example, there was no disclosure that Martin’s sister (the buyer’s daughter) prepared them; that Martin’s sister worked for a mortgage company owned by Martin’s ex-husband (a seller); that the buyer and the sellers were related; and that Martin had actually given her father the $1,500 that he paid as an initial deposit. Based on these misrepresentations and omissions, Martin’s father obtained the loans. After the sale closed, Martin and her ex-husband paid off their existing mortgage and pocketed about $216,000 in cash, from which they paid Martin’s sister $35,064 and' her father $64,074.
After the purported sale, First Franklin Financial Corp. sold the two mortgages on the Miami Apartment to a successor lender. Martin and her ex-husband continued to collect rent from a tenant in the Miami Apartment, despite her father’s representation that the property would serve as his primary residence. For approximately a year-and-a-half, Martin and her ex-husband also made the mortgage payments for her father (he never made any payments himself). Eventually, however, the loans went into default and the property was sold in a short sale.
Two days after the sham sale of their Miami Apartment to Martin’s father, Martin and her ex-husband closed on a residential home in Miami Beach with a price tag of $1,550,000 (the “Beach House”). To fund this purchase, they applied for two mortgage loans from LoanCity, Inc. (a mortgage company that the government failed to establish was a financial institution) totaling that amount. Martin’s ex-husband’s company prepared the loan applications, which, among other things, grossly misstated Martin’s monthly income (they claimed that Martin made $15,500 per month when, in reality, her W-2 reported only $18,400 in income for the year) and bank account balance (they claimed that Martin had a bank account balance that totaled $121,304.84 when, in reality, the total balance in that account was only $100). The mortgages were approved. At the closing, Martin and her ex-husband negotiated a $190,831.70 credit from the seller to cover needed repairs resulting from previous hurricane damage. Notably, Martin and her ex-husband refused to reflect that credit in the closing settlement statement, opting instead to memorialize it only in an addendum that never became part of the loan file and was never reported to the lenders. The credit was then wired to the account of Acqualina Corp., a company opened in Martin’s name. Martin would later tell a Special Agent from the United States Secret Service that the transfer to Acqualina Corp. was “possibly to hide something.” Rather than paying for the alleged repairs, Martin wrote three sequential checks to her ex-husband from the Acqualina Corp. account for over $100,000.
The first mortgage on the Beach House was later sold to IndyMac Bank (which was still later acquired by OneWest Bank). The second mortgage was eventually sold to Saxon Mortgage Services, Inc., but that company went out of business and no successor-in-interest was identified at trial. Martin and her ex-husband later defaulted, and the Beach House was also sold in a short sale.
In January 2007, less than a month after he purportedly purchased the Miami Apartment for $495,000, Martin’s father bought still another residential property in Miami for $490,000 (the “Miami Property”). He funded that purchase with two mortgage loans obtained from Bear
Stearns totaling $488,798.94. Martin’s ex-husband again prepared the loan applications, which still again grossly misstated Martin’s father’s monthly income and bank account balance, and again falsely stated that he earned monthly rental income on another residential property that he owned. The same fictitious rental agreement and deposit check from- Martin’s childhood friend that were used in connection with the Miami Apartment loan applications were once again used to verify the fraudulent lease. Martin’s ex-husband did not disclose that he was the buyer’s son-in-law; that the buyer had just bought the Miami Apartment from Martin and him; and that the buyer had taken out mortgage loans for $490,000 less than a month earlier to fund that purchase. Nor was it disclosed that Martin, the buyer’s daughter, had issued the required insurance on the property. Each of those facts, if properly disclosed, would have prompted further questions regarding the legitimacy of the transaction. After the closing, Bear Stearns sold the two mortgages on the Miami Property; ultimately, JPMorgan Chase & Co. acquired both of them. Martin’s father later defaulted on this loan, too, and the property was sold in a short sale.
Eventually, the Secret Service investigated. That investigation led to the June 2013 indictment of Martin, her ex-husband, her sister, and two other codefendants. Martin was charged in four counts: one count of conspiracy to commit bank and wire fraud, in violation of 18 U.S.C. § 1349 (Count 1); one count of bank fraud, in violation of 18 U.S.C. § 1344 (Count 2); and two counts of wire fraud, in violation of 18 U.S.C. § 1343 (Counts 4 and 5). All of the codefendants entered guilty pleas except Martin, who proceeded to trial. The jury found her guilty as charged on all counts, and the district court later sentenced her to 46 months of imprisonment on each count, to run concurrently, followed by five years of supervised release. It also ordered her to pay $963,400.47 in restitution.
Martin timely appealed to this court.
II.
We start with Martin’s challenge to the sufficiency of the evidence supporting her convictions. She argues that the government failed to prove that she had any knowledge of a conspiracy to defraud, or that she made any intentional misrepresentations in the three transactions at issue. Martin also claims that the government failed to demonstrate that she defrauded any financial institutions. We disagree.
We review
de novo
the denial of a motion for judgment of acquittal, viewing the evidence in the light most favorable to the government and drawing all reasonable inferences in favor of the jury’s verdict.
See Hernandez,
743 F.3d at 814. “[T]he issue is not whether a jury reasonably could have acquitted but whether it reasonably could have found guilt beyond a reasonable doubt.”
United States v. Thompson,
473 F.3d 1137, 1142 (11th Cir.2006). We will not overturn a jury’s verdict if there is “any reasonable construction of the evidence [that] would have allowed the jury to find the defendant guilty beyond a reasonable doubt.”
United States v. Friske,
640 F.3d 1288, 1291 (11th Cir.2011) (internal quotation marks and citation omitted). “The test for sufficiency of the evidence is identical^] regardless of whether the evidence is direct or circumstantial,”
United States v. Mieres-Borges,
919 F.2d 652, 656-57 (11th Cir.1990), but if the government relied on circumstantial evidence, “reasonable inferences, not mere speculation, must support the conviction,”
United States v. Mendez,
528 F.3d 811, 814 (11th Cir.2008) (per curiam).
To convict Martin of conspiracy to commit bank and wire fraud under 18 U.S.C. § 1349, the government had to prove beyond a reasonable doubt that (1) two or more persons agreed to a common and unlawful plan to commit bank and wire fraud as alleged in the indictment; (2) Martin knew of the unlawful plan; and (3) she knowingly and voluntarily joined the plan.
See United States v. Moran,
778 F.3d 942, 960 (11th Cir.2015). “A scheme to defraud requires proof of a material misrepresentation, or the omission or concealment of a material fact calculated to deceive another out of money or property.”
United States v. Maxwell,
579 F.3d 1282, 1299 (11th Cir.2009). If a misrepresentation has “a natural tendency to influence, or [is] capable of influencing, the decision maker to whom it is addressed,” it is material.
United States v. Hasson,
333 F.3d 1264, 1271 (11th Cir.2003). Because conspiracies are secretive by nature, the jury must often rely on “inferences from the conduct of the alleged participants or from circumstantial evidence of a scheme.”
United States v. Vernon,
723 F.3d 1234, 1273 (11th Cir.2013) (internal quotation marks and citation omitted). Although the government must prove that the defendant had knowledge of the scheme, it need not show that she had “all of the details” of the conspiracy.
Moran,
778 F.3d at 960 (internal quotation marks and citation omitted). Rather, it need only prove that she knew of “the essential nature of the conspiracy.”
Vernon,
723 F.3d at 1273 (internal quotation marks and citation omitted).
To convict Martin of bank fraud in violation of 18 U.S.C. § 1344(2), the government had to prove beyond a reasonable doubt that (1) a scheme existed to obtain money or property in the custody of a federally insured financial institution by fraud; (2) she participated in the scheme by means of material false pretenses, representations, or promises; and (3) she acted knowingly.
United States v. McCarrick,
294 F.3d 1286, 1290 (11th Cir.2002) (citing
United States v. Goldsmith,
109 F.3d 714, 715 (11th Cir.1997)). As with other fraud crimes, “circumstantial evidence may prove [a defendant’s] knowledge.”
United States v. Williams,
390 F.3d 1319, 1325 (11th Cir.2004). Finally, to convict Martin of wire fraud under 18 U.S.C. § 1343, the government had to prove beyond a reasonable doubt that she (1) “participated in a schememr artifice to defraud; (2) with the intent to defraud; and (3) used, or caused the use of, interstate wire transmissions for the purpose of executing the scheme or artifice to defraud.”
United States v. Williams,
527 F.3d 1235, 1240 (11th Cir.2008). Wire fraud, likewise, “can be proved by circumstantial evidence.”
United States v. Robertson,
493 F.3d 1322, 1331 (11th Cir.2007).
Viewed in the light most favorable to the government and the jury’s verdict, the evidence sufficiently established that Martin knew of and voluntarily joined a mortgage fraud conspiracy and actively participated in obtaining millions of dollars in fraudulent loans from various banking institutions by making material misrepresentations to the lenders involved. For starters, Martin and her ex-husband en- ’ gaged in the sham sale of their Miami Apartment to her father without disclosing their familial relationship to the lenders. If Martin had made that disclosure, which is typically required in the sales contract, the underwriter would have required additional verification of the information on the loan application because the sale was not at arm’s length. The sale was also facilitated by Martin’s role in creating a fraudulent lease and procuring a fraudulent deposit check from a childhood friend — a
lease relied on by the underwriters when they approved the mortgages. Martin also gave her father the money he used for his $1,500 deposit. An underwriter explained, however, that deposit money was “required to have come from [Martin’s father].” Almost immediately after Martin and her ex-husband received over $216,000 from the sham Miami Apartment sale, Martin distributed a portion of the proceeds to the other participants . in the fraud: $64,074 to her father and $35,064 to her sister. Tellingly, after the sale, Martin and- her ex-husband made her father’s mortgage payments on his behalf (for approximately 21 months), arguably, as the government put it in closing, because they knew that her father could not afford to make the payments himself.
Only two days after the sham sale of the Miami Apartment, Martin and her ex-husband purchased the $1,550,000 Beach House. Even though Martin had swindled a lender out of over $200,000 in proceeds from the Miami Apartment sale, she lied still again in her mortgage applications to purchase this home, greatly exaggerating her income (Martin claimed that she made $15,500 per month when, in reality, her W-2 reported only $18,400 in income for the year) and savings information (Martin claimed she had a bank account balance that totaled $121,304.84 when, in reality, the total balance in that account was only $100). Although she argued to the jury that her signature on the loan application may have been forged because it was “the world’s easiest signature to forge” (it was characterized as an “upside down infinity fish”), that was a matter for the jury to decide. Having reviewed numerous documents signed by Martin, including signatures that were identified by Martin’s childhood friend, the jury had a sufficient basis to make that determination.
See Vernon,
723 F.3d at 1253. Moreover, at the closing on the Beach House, Martin also tunneled an additional $190,000 in “credits for repairs” from the seller to a company that she owned. That credit was not disclosed to the lenders, who believed the seller received the full purchase price without offering a discount. Nor were any corresponding repairs ever performed.
Finally, Martin facilitated her father’s fraudulent purchase of yet another property that he could not afford, the Miami Property, by issuing a required home insurance policy without disclosing her relationship to the buyer and by allowing him to use the same falsified lease deposit that she procured for the Miami Apartment fraud. Taking the evidence as a whole, a jury reasonably could have found beyond a reasonable doubt that Martin was aware of and knowingly and voluntarily participated in an agreement to commit fraud, and that Martin made numerous material misrepresentations that induced the lenders to grant (or later acquire) the mortgages.
Martin’s claim that her conduct did not involve federally insured financial institutions likewise fails. The term “financial institution” includes federally insured' depository institutions.
See
18 U.S.C. § '20(1). The bank fraud charge (Count 2) arose out of Martin’s sham sale of her Miami Apartment to her father. In the process of that sale, her father submitted falsified loan applications concerning material matters to First Franklin Financial Corp., a division of National City Bank, a federally insured financial institution. Even though First Franklin Financial Corp. sold the mortgages before the short sale, resulting in a possibility that it suffered no loss, it was nonetheless defrauded because it had granted mortgages based on applications that materially overstated the buyer’s ability to repay them. This scheme to defraud is all that the government must prove.
See Loughrin v.
United States,
— U.S. -, 134 S.Ct. 2384, 2395 n. 9, 189 L.Ed.2d 411 (2014) (stating that bank fraud convictions under § 1344(2) do not require proof of loss — or even risk of loss — to a financial institution because “the gravamen of § 1344 is the 'scheme, rather than the completed fraud” (internal quotation marks and citation omitted)).
The evidence adduced at trial likewise demonstrated that the scheme sufficiently “affected” a financial institution to support Martin’s wire fraud convictions. The indictment charged Martin with two counts of “wire fraud affecting a financial institution.” The district court instructed the jury, without objection, that an element of the wire fraud offenses included the requirement that “the scheme affected a financial institution.” Accordingly, the government was required to prove that the fraudulent scheme affected a financial institution.
United States v. Spletzer,
535 F.2d 950, 954 (5th Cir.1976) (explaining that when the trial judge instructs the jury on an element of an offense and there is no objection, the element becomes “a necessary element for conviction pursuant to the ‘law of the case’ doctrine”).
The first wire fraud count (Count 4) also involved the Miami Apartment sale. First Franklin Financial Corp. — a federally insured financial institution — granted mortgages on that apartment based on falsified representations of the buyer’s ability to repay, plainly resulting in an increased risk of loss through default. That increased risk sufficiently affected a financial institution.
See United States v. Stargell,
738 F.3d 1018, 1022 (9th Cir.2013) (“The increased risk of loss presented by fraudulent terms is sufficient to ‘affect’ a financial institution [in the context of wire fraud].”),
cert. denied,
— U.S. -, 134 S.Ct. 2289, 189 L.Ed.2d 178 (2014);
United States v. Serpico,
320 F.3d 691, 694-95 (7th Cir.2003) (same).
The second wire fraud count (Count 5) involved Martin’s purchase of the Beach House, where she and her ex-husband applied for and received mortgages from LoanCity, Inc., which the government did not prove was a financial institution. However, IndyMac Bank — a federally insured financial institution — later relied on her fraudulent applications when it decided to purchase one of the Beach House mortgages. The Supreme Court has held, albeit in the context of bank fraud, that a defendant need not even intend to defraud a financial institution so long as she engages in a fraudulent scheme that, “in the ordinary course of business,” triggers the disbursement of bank funds.
Loughrin,
— U.S. -, 134 S.Ct. at 2389-90, 2393-95. Here, Martin was undoubtedly aware that a financial institution could be affected by her fraudulent scheme. She was expressly warned in her mortgages that they may be subsequently sold to another lender that would rely on her representations. Her wire fraud offense “affected” a financial institution. This remains true even if the successor lender suffered no loss — the increased risk of loss is sufficient.
See Stargell,
738 F.3d at 1022.
The long and the short of it is that sufficient evidence supported the jury’s findings of guilt beyond a reasonable doubt on each of the counts. The district court did not err in denying the motions for judgment of acquittal.
III.
Martin also challenges her sentence, claiming first, that she should have been
given a role reduction for her modest involvement in the crimes, and second, that a restitution order was improvidently entered.
A.
Martin claims that the district court clearly erred by refusing to grant her a role reduction pursuant to U.S.S.G. § 3B1.2. We remain unconvinced. This section of the guidelines provides for a four-level reduction to a defendant’s base offense level if she was a “minimal participant,” a two-level reduction if she was a “minor participant,” and a three-level reduction in “cases falling between” those two categories. U.S.S.G. § 3B1.2. A “minimal participant” is “plainly among the least culpable of those involved in the conduct of a group” and generally “lack[s] ... knowledge or understanding of the scope and structure of the [criminal] enterprise and of the activities of others.”
Id.
cmt. n. 4. A “minor participant” is someone who is “less culpable than most other participants, but whose role could not be described as minimal.”
Id.
cmt. n. 5.
The defendant, as the proponent of the downward adjustment under § 3B1.2, bears the burden of proving her mitigating role in the offense by a preponderance of the evidence.
United States v. Alvarez-Coria,
447 F.3d 1340, 1343 (11th Cir.2006) (per curiam). The sentencing court’s determination of a defendant’s role is a factual finding that we review only for clear error.
United States v. Rodriguez De Varon,
175 F.3d 930, 937 (11th Cir.1999) (en banc). We “will not find clear error unless our review of the record leaves us with the definite and firm conviction that a mistake has been committed.”
United States v. White,
335 F.3d 1314, 1319 (11th Cir.2003) (internal quotation marks and citation omitted). The district court is guided by two principles when it determines whether a defendant’s role in the offense falls within § 3B1.2’s continuum. First, and most importantly, the court must consider the defendant’s role “in relation to the relevant conduct attributed to [her] in calculating her base offense level.”
Rodriguez De Varón,
175 F.3d at 941. The defendant must show “that she played a relatively minor role in the conduct for which she has already been held accountable — -not a minor role in any larger criminal conspiracy.”
Id.
at 944. Second, the court may choose to “measure the defendant’s culpability in comparison to that of other participants in the relevant conduct.”
Id.
But “not all participants [are] relevant to this inquiry.”
Id.
Only “those participants who were involved in the relevant conduct attributed to the defendant” may be considered.
Id.
Even if a defendant played a lesser role than the other participants, that fact does not entitle her to a role reduction “since it is possible that none are minor or- minimal participants.”
United States v. Stanley,
739 F.3d 633, 654 (11th Cir.) (internal quotation marks and citation omitted),
cert. denied sub nom. Harris v. United States,
— U.S. -, 134 S.Ct. 2317, 189 L.Ed.2d 195 (2014).
The district court did not clearly err in finding that Martin played more than a “minor” role in the mortgage fraud conspiracy. Although the conspiracy itself involved seven properties, Martin was held accountable for only three of them. In those three transactions, however, the court found by a preponderance of the evidence that Martin recruited her father to serve as a straw buyer; she grossly misstated material information that enabled her to purchase a $1.55 million home; and she provided false documents and deposit money that enabled her father to purchase two other properties for a total sum of nearly $1 million. The district
court did not clearly err in finding that Martin’s involvement in these transactions, even if less serious than her ex-husband’s, was serious enough to place her outside the scope of § 3B1.2.
Martin’s corollary argument that she deserves a role reduction because her code-fendant, Juan Godoy, received one is similarly misplaced. Godoy was not implicated in any of the transactions for which Martin was held accountable at sentencing. Thus, he is irrelevant to the inquiry.
Rodriguez De Varon,
175 F.3d at 944. And even if the district court had considered Godoy, he was involved in only one consummated fraudulent transaction as opposed to the three that Martin played key roles in. By any measure, we simply cannot say that the district court clearly erred when it determined that Martin did not play a minimal or minor role in the offense.
B.
Martin’s final challenge concerns the district court’s order of restitution. She was ordered to pay a total of $963,400.47 to three banks under the Mandatory Victims Restitution Act of 1996, 18 U.S.C. § 3663A (“MVRA”). Specifically, the court ordered Martin to pay $229,619.35 to JPMorgan Chase & Co. in connection with the Miami Property mortgages, $358,781.12 to Bank of America Home Loans in connection with the Miami Apartment mortgages, and $375,000 to OneWest Bank in connection with the first mortgage on the Beach House (the second mortgage on the Beach House was not included in the restitution award because the government produced no evidence demonstrating who owned it when Martin defaulted and the property was sold at a short sale).
The three identified victim banks are not the original lenders, but instead are successor lenders who subsequently purchased the fraudulently procured mortgages and owned them when the properties were sold in short sales. Over Martin’s objection, the district court awarded each victim bank an amount totaling the difference between the outstanding principal balance of the mortgages they owned and the amount the successor lenders recouped from the short sales. Martin argues that the district court’s award was improper since it could not have known whether the successor lenders suffered actual losses absent evidence showing how much they had paid to acquire the mortgages. According to Martin, the successor lenders might have acquired the loans for a price far lower than their outstanding principal balances — potentially even lower than the amount they recouped from the short sales. If so,, the lenders would have already reaped a profit.
As a result, she claims, the government failéd to prove that the successor lenders were victims entitled to restitution under the MVRA.
Martin’s challenge to the district court’s restitution order then raises two issues: (1) whether the successor lenders were “victims” of her offenses within the meaning of the MVRA; and (2) whether the district court’s restitution award accurately reflected any losses. We agree with the
district court that the successor lenders were victims, but we find that the court clearly erred in its determination of the amount of loss. We discuss each issue in turn.
1.
The MVRA requires the district court to award restitution to identifiable victims of certain crimes, including crimes of fraud, without regard to the defendant’s ability to pay. 18 U.S.C. § 3663A(a)(1), (c)(1)(A)(ii);
id.
§ 3664(f)(1)(A);
see United States v. Singletary,
649 F.3d 1212, 1220 (11th Cir.2011). To be considered a “victim” under the MVRA, an entity must have been “directly and proximately harmed as a result of the commission of [the defendant’s] offense.” 18 U.S.C. § 3663A(a)(2);
see Robertson,
493 F.3d at 1334. Thus, a victim must have suffered harm, and the defendant must have proximately caused that harm. The government has the burden of proving by a preponderance of the evidence that a particular entity was a victim of the defendant’s offense.
See Robertson,
493 F.3d at 1334. We review the district court’s factual finding regarding proximate cause only for clear error, and its legal conclusion that an entity is a victim
de novo. Id.
Here, the lenders were not only harmed, but Martin’s conduct was also the direct and proximate cause. To be sure, Martin correctly notes that the successor lenders might have actually profited from the short sales. Nonetheless, they suffered harm for the very reasons we previously discussed. Their decisions to purchase these mortgages were based, in large part, on the fraudulent loan applications that Martin played a significant role in — purchases that were made before the fraud came to light. As a result, these financial institutions purchased the mortgages based upon false perceptions of the buyers’ ability to repay them. That was the precise harm the Ninth Circuit highlighted when it reach the same result. That court addressed a challenge by an appellant who had been convicted of a similar fraudulent real estate scheme involving the use of falsified information on loan applications.
United States v. Yeung,
672 F.3d 594, 596 (9th Cir.2012),
abrogated on other grounds by Robers v. United States,
— U.S. -, 134 S.Ct. 1854, 1859, 188 L.Ed.2d 885 (2014). The appellant was ordered to pay restitution to a successor lender that purchased some of the mortgages on the secondary market, despite the lack of evidence demonstrating the price the successor lender paid.
Id.
at 598-600. Although the Ninth Circuit vacated the restitution award and remanded for a new determination of loss, the court nonetheless rejected the appellant’s contention that the successor lender was not a “victim” for purposes of the MVRA.
Id.
at 603. It found that because the successor lender “purchased the loan without an awareness of its true value due to [the appellant’s] fraud, the district court could reasonably conclude that [the appellant’s] fraudulent conduct proximately harmed the [successor lender].”
Id.
Most importantly, our conclusion that the successor lenders were harmed is supported by the text of the MVRA itself, which distinguishes between the concepts of harm and loss.
Compare
18 U.S.C. § 3663A(a)(2) (defining the. “victim” in terms of the “harm” suffered),
with id.
§ 3663A(b)(l) (defining the amount of restitution owed in terms of “value” and “loss”).
It is also consistent with the line
of cases we already discussed, which find that a financial institution is the victim of a defendant’s fraudulent conduct even if the financial institution does not suffer an actual loss.
See supra
Part II. For these reasons, we agree that the successor lenders suffered harm.
That leaves the question of whether Martin’s conduct proximately caused the harm. In order to do so, the government must establish that the defendant is the “but[-]for” cause of the harm and that the connection is “not too attenuated (either factually or temporally).”
Robertson,
493 F.3d at 1334 (quotation marks omitted);
see Robers,
— U.S. -, 134 S.Ct. at 1859 (“The basic question ... is whether the harm alleged has a sufficiently close connection to the conduct at issue.”) (internal quotation marks and citation omitted).
' The district court did not clearly err in determining that Martin’s conduct proximately caused the successor lenders’ harm. But for the fraudulent applications, the lenders would not have been misled about the buyers’ ability to repay the mortgages. It was entirely foreseeable to Martin not only that the original lenders would rely on the fraudulent applications, but that the mortgages would be resold to other lenders that would rely on the applications as well. Her own mortgage provided her with express warning that “[t]he Note or a partial interest in the Note (together with this Security Instrument) may be sold one or more times without prior notice to the Borrower.”
See Robertson,
493 F.3d at 1334-35 (noting that the intervening actions of others — such as lawsuits — break the chain of causation only where they are unforeseeable). Martin cannot point to an intervening act that broke the causal chain. Thus, the district court properly deemed the successor lenders “victims” within the meaning of the MVRA.
See Yeung,
672 F.3d at 603.
2.
Having decided that the successor lenders may recover restitution under the MVRA, we turn to the issue of whether the district court correctly calculated the amount of the award. Two principles drive our inquiry. The first is that “[t]he purpose of restitution ‘is not to provide a windfall for crime victims but rather to ensure that victims, to the greatest extent possible, are made whole for their losses.’ ”
United States v. Cavallo,
790
F.3d 1202, 1238 (11th Cir.2015) (alteration adopted) (quoting
United States v. Huff,
609 F.3d 1240, 1249 (11th Cir.2010));
see Hughey v. United States,
495 U.S. 411, 416, 110 S.Ct. 1979, 109 L.Ed.2d 408 (1990) (“[T]he ordinary meaning of ‘restitution’ is restoring someone to a position he occupied before a particular event.”). Restitution is not designed to punish the defendant.
United States v. Bane,
720 F.3d 818, 828 (11th Cir.2013). Thus, the amount of restitution owed to each victim “must be based on the amount of loss
actually
caused by the defendant’s conduct.”
Huff,
609 F.3d at 1247 (internal quotation marks and citation omitted).
The second principle is that “the determination of the restitution amount is by nature an inexact science.”
Id.
at 1248 (internal quotation marks and citation omitted). Thus, the government need not calculate the victim’s actual loss with laser-like precision, but may instead provide a “reasonable estimate” of that amount.
United States v. Futrell,
209 F.3d 1286, 1290 (11th Cir.2000) (per curiam) (internal quotation marks and citation omitted);
see also
18 U.S.C. § 3664(e) (providing that the government must prove the loss amount only by a preponderance of the evidence). With those two principles in mind, we review for clear error the district court’s factual finding regarding the amount of restitution owed.
See Futrell,
209 F.3d at 1289.
As Martin concedes, “some entity ... suffered an actual loss” since — over the life of each mortgage — more money was loaned out than was recouped. But no evidence demonstrated what portion of that loss, if any, was suffered by the successor lenders. If a successor lender had paid less to acquire the mortgage than it recouped in a short sale, it would have already profited regardless of the outstanding principal balance at the time (then a prior lender would have suffered the loss). Any restitution award to the successor lender in such a case would simply add to its profit rather than making it whole. This we cannot allow.
We agree with Martin that a successor lender’s restitution award should turn on how much it paid to acquire the mortgage. Our sister circuits have reached the same conclusion.
See United States v. Howard,
784 F.3d 745, 750-51 (10th Cir.2015) (finding that the measure of a successor lender’s loss must be based on the amount it paid to acquire the mortgage to prevent overcompensating downstream noteholders);
United States v. Beacham,
774 F.3d 267, 278-79 (5th Cir.2014) (holding that the district court abused its discretion when it used the original loan amounts to calculate restitution for a successor lender);
United States v. Chaika,
695 F.3d 741, 748-49 (8th Cir.2012) (observing that, for purposes of awarding restitution, a successor lender’s “actual loss will turn on its purchase price in the secondary market ... and perhaps other factors”);
Yeung,
672 F.3d at 601-06 (reasoning that, where a successor lender pays only a fraction of the outstanding principal balance, awarding restitution based on that balance would overcompensate and constitute plain error).
Thus, restitution to a successor lender must typically equal the sum that lender paid to acquire the mortgage less the principal payments it received and the amount it recouped in the short sale.
In simple
terms, how much it paid minus how much it made.
See Howard,
784 F.3d at 750-51 (“[T]he MVRA limits restitution to actual, out-of-pocket losses.” (internal quotation marks and citation omitted));
United States v. James,
592 F.3d 1109, 1115 (10th Cir.2010) (“The successor lenders’ actual loss [under the guidelines], then, is the difference between what they paid the original lenders for the loans (less principal repayments by borrowers, if any) and what they received for the properties at the foreclosure sales, plus reasonably foreseeable expenses relating to the foreclosure proceedings.”). This calculation serves the MVRA goal of making victims whole while avoiding windfall awards. It also corresponds to the loss calculation where the
original
lenders still hold the mortgage at default: the outstanding balance of the loan (by definition, the original lender’s out-of-pocket cost) less the amount it recouped from short sale. Both formulas simply place the lenders in the position they would be in if the defendant never committed the fraud and, accordingly, the loan had never been made.
See Hughey,
495 U.S. at 416, 110 S.Ct. 1979.
But we do not go so far as to say that a restitution award to a successor lender is always improper if the district court lacks evidence on the loan’s actual purchase price. As we have said, “the determination of the restitution amount is by nature an inexact science.”
Huff,
609 F.3d at 1248 (internal quotation marks and citation omitted). This is especially true in mortgage fraud cases where “today’s banking realities — the bundling of mortgages into securities, for example — may make it difficult to identify precisely the proceeds a lender received for a specific mortgage loan.”
James,
592 F.3d at 1116. But the district court cannot simply presume that the successor lenders paid the outstanding principal balance to acquire the mortgages.
See Yeung,
672 F.3d at 602 (noting that the value of a mortgage on the secondary market may vary with the market value of the property, the credit rating of the borrower, and other market conditions). Rather,, in some circumstances, the government must present a “reasonable estimate” of the loan’s purchase price,
Futrell,
209 F.3d at 1290, which may very well be the outstanding principal balance in some cases.
Here, however, the government introduced no evidence regarding the actual price the successor lenders paid for the mortgages. Nor did it introduce evidence to support a conclusion that the outstanding principal balances were reasonable estimates of what the successor lenders paid to acquire the fraudulently obtained loans. Given this dearth of evidence on matters of considerable importance, we are obliged to vacate the restitution order and remand to the district court for a new determination of the amounts of restitution, if any, to be awarded to the three victim banks.
See, e.g., Yeung,
672 F.3d at 606.
We, therefore, affirm Martin’s convictions and sentence, but remand for a new determination of the restitution award consistent with this opinion. We leave to the district court the decision of whether to consider additional evidence or conduct an evidentiary hearing to expand the record on remand.
See Howard,
784 F.3d at 752.
AFFIRMED in part, VACATED and REMANDED in part.