United States v. Roberta Sheffield
Opinion
[PUBLISH]
IN THE UNITED STATES COURT OF APPEALS
FOR THE ELEVENTH CIRCUIT
No. 17-13682
D.C. Docket No. 1:15-cr-00403-MHC-AJB-2
UNITED STATES OF AMERICA, Plaintiff - Appellee,
versus ROBERTA SHEFFIELD, Defendant - Appellant.
Appeal from the United States District Court for the Northern District of Georgia
(October 1, 2019)
Before TJOFLAT, JORDAN, and ANDERSON, Circuit Judges. JORDAN, Circuit Judge:
Roberta Sheffield pled guilty to numerous criminal charges relating to her involvement with others in a fraudulent tax credit scheme. Generally speaking, the
scheme involved the submission of thousands of tax returns falsely claiming a $1,000 tax credit. The returns were fraudulent because the taxpayers had not incurred the $4,000 in educational expenses needed to qualify for the tax credit. Based on the fraudulent returns, the IRS issued a $1,000 refund to each taxpayer for each tax credit claimed.
The district court sentenced Ms. Sheffield to 37 months’ imprisonment followed by a term of supervised release. The court fixed the amount of loss from the fraudulent tax credit scheme at $3,461,638, and ordered Ms. Sheffield to pay restitution in that amount jointly and severally with her co-defendants. In this appeal, Ms. Sheffield contests only the restitution order.
We agree with Ms. Sheffield that the restitution order must be set aside. The refunds issued by the IRS due to the fraudulent tax credit scheme were all for the same exact amount — $1,000 — and the evidence the government submitted in support of its restitution request was admittedly and demonstrably inaccurate. Where, as here, the appropriate restitution amount is definite and easy to calculate, the government cannot satisfy its burden of proof by relying on the oft-stated (but not always applicable) principle that restitution can be based on a reasonable estimate of loss. Cf. Fred R. Shapiro, The Yale Book of Quotations 639 (Yale University Press 2006) (quoting Frank Robinson: “Close only counts in horseshoes and grenades.”).
I
At sentencing, the government bore the burden of demonstrating the loss sustained by the IRS by a preponderance of the evidence. See 18 U.S.C. § 3664(e); United States v. Baldwin, 774 F.3d 711, 728 (11th Cir. 2014). Without objection from Ms. Sheffield, the government introduced a spreadsheet purportedly listing each fraudulent tax return, along with the name of the taxpayer and the amount of the tax credit refund issued by the IRS. According to the spreadsheet, the loss from the tax credit scheme totaled $3,461,638.
Although Ms. Sheffield did not object to the introduction of the spreadsheet, she argued that the restitution amount should be lower than $3,461,638 because the spreadsheet contained duplicative entries:
I believe that [the figure] included not just the actual loss but some of the duplicates, like, some of the things were counted twice. There were some entries that I reviewed that appeared to be duplicates. So I’m not sure what the exact restitution amount should be because that’s, frankly, the government’s burden, but I believe it was lower than $3.4 million.
D.E. 225 at 7.
In response to Ms. Sheffield’s objection, the government asserted that “restitution does not have to be calculated with absolute precision,” and maintained that “the burden lies on the defendant, if this is an inaccurate number, to point out why it’s inaccurate.” Id. at 8. The government acknowledged that the spreadsheet
might contain duplicate entries, and conceded that one taxpayer was listed twice for the same tax year. But it again argued that “the burden is on the defendant to point out what the restitution amount should be if this is inaccurate.” Id.
The district court said that it could not tell whether there were duplicate entries, and invited Ms. Sheffield to provide a list of the entries she believed were duplicative. It also noted that, “unless Ms. Sheffield hits the lottery sometime between now and the end of her life, the odds of her and her co-defendants paying this money back are below slim.” Id. at 9. When Ms. Sheffield said she did not have such a list, the district court overruled her objection and ordered restitution in the amount set out in the spreadsheet.
II
We review the district court’s restitution order for clear error. See United States v. Martin, 803 F.3d 581, 595 (11th Cir. 2015). That means that we will reverse only if we are left with a “definite and firm conviction that a mistake has been committed.” Anderson v. City of Bessemer City, 470 U.S. 564, 573 (1985) (citation and internal quotation marks omitted). Despite this deferential standard of review, we cannot affirm the restitution order because the government’s spreadsheet was admittedly inaccurate.
“[R]estitution seeks to make victims whole by reimbursing them for their losses[.]” United States v. Joseph, 743 F.3d 1350, 1354 (11th Cir. 2014). Because
“[r]estitution is not designed to punish the defendant[,] . . . the amount . . . owed . . . must be based on the amount of loss actually caused by the defendant’s conduct.” Martin, 803 F.3d at 595 (citation and internal quotation marks omitted). See also 18 U.S.C. § 3663(a)(1)(B)(i)(I) (directing courts to consider the “amount of the loss sustained by each victim as a result of the offense”).
The “use of estimation” is permitted because “it is sometimes impossible to determine an exact restitution amount.” United States v. Futrell, 209 F.3d 1286, 1291–92 (11th Cir. 2000). In some cases, therefore, the government is allowed to submit, and the district court is permitted to use, a “reasonable estimate” of that amount. See, e.g., United States v. Gushlak, 728 F.3d 184, 196 (2d Cir. 2013) (explaining that “a ‘reasonable approximation’ will suffice, especially in cases in which an exact dollar amount is inherently incalculable”).
For example, in Martin we laid out a formula to calculate loss in situations where a successor lender is injured by a defendant’s mortgage fraud—“[i]n simple terms, how much it paid minus how much it made.” 803 F.3d at 595–96. But we left open the possibility that a district court might lack evidence regarding a loan’s actual purchase price, given that “today’s banking realities,” such as “the bundling of mortgages into securities . . . may make it difficult to identify precisely the proceeds a lender received for a specific . . . loan.” Id. (citation and quotation marks omitted). In such a situation, some “reasonable estimate[s]” would be, by necessity,
acceptable. See id. We cautioned, however, that district courts cannot simply make baseless presumptions, and remanded the restitution portion of the defendant’s sentence because the government had not introduced evidence regarding the actual prices paid or evidence supporting the estimates it had put forth. See id.
Like the Second Circuit, we have “never adopted a one-size fits all standard of precision for application in restitution cases.” Gushlak, 728 F.3d at 195. In our view, Martin does not stand for the broad proposition that a “reasonable estimate” of restitution will always suffice. Indeed, before Martin we explained that “where difficulties arise, a district court may accept a ‘reasonable estimate’ of the loss based on the evidence presented.” Baldwin, 774 F.3d at 728 (emphasis added and citation omitted).
Here the losses suffered by the IRS due to the tax credit scheme were definite and easy to calculate. Because each fraudulent tax credit triggered a refund of $1,000 — not a penny more, not a penny less — figuring out the loss only required a simple mathematical exercise, i.e., multiplying each false tax credit by $1,000. Given that the government must put forth “reliable and specific evidence” to support a restitution award, see United States v. Stein, 846 F.3d 1135, 1156 (11th Cir. 2017), it is not too much to require precision in a case like this one.
Free access — add to your briefcase to read the full text and ask questions with AI
939 F.3d 1274 (United States v. Roberta Sheffield) — published by Counsel Stack Legal Research, free access to 12M+ legal documents.