Affirmed by published opinion. Chief Judge GREGORY wrote the opinion, in which Judge NIEMEYER and Judge HARRIS joined.
Michael Rand was convicted of conspiracy, in violation of 18 U.S.C. §§371 and 1349, and obstruction of justice, in violation of 18 U.S.C. § 1512(b)(3), (c)(1), and (c)(2), following his involvement in earnings mismanagement and improper accounting transactions while acting as chief accounting officer at Beazer Homes USA, Inc. Rand appeals several aspects of his convictions and sentence. Finding no error, we affirm.
I.
In 2010, the government charged Michael Rand with accounting fraud based on his work at Beazer Homes USA, Inc. (“Beazer”), a home-building company, from 2000 to 2007 and with obstructing an investigation into Beazer’s mortgage origination practices. Rand, a certified public accountant, was Beazer’s controller and later its chief accounting officer from 1999 to 2007. He reported to Beazer’s CEO and CFO.
The government’s accounting charges concerned earnings management: it believed that Rand attempted to adjust Beazer’s reported earnings over time so that Beazer would hit consensus — that is, the quarterly earnings amount that Wall Street predicted. This practice involved “cookie jar” accounting with respect to Beazer’s reserve accounts, where funds are set aside for future expenditures or revenue. It is generally accepted that the amount put into a reserve account is what the company reasonably anticipates needing to meet the expected expense. It is not appropriate to increase or decrease funds in reserve accounts to understate or inflate its actual earnings. Instead, if a company determines that it does not need the reserve funds, those funds “are to be taken back as income as soon as [the company] know[s] that they are no longer required.” J.A. 1260.
The government attempted to prove that Rand manipulated the accounting to reduce earnings when Beazer was beating consensus. E.g., J.A. 3720 (“If you have more than 100k extra, hide it.”); id. at 3722 (“To achieve the ‘goal’ $ for this year, let’s squirrel $ away in places which will turn around in the next year; not be so ‘open.’ ”); id. at 1982-83 (“We may have $5 million to squirrel away, so if you have ant [sic] ideas, let me know. Joavan’s cookie jar has no more room.”). This practice resulted in a misrepresentation of Beazer’s earnings in many quarters, including each quarter in fiscal year 2006.
The government also alleged that Rand improperly accounted for transactions involving model homes Beazer sold to and leased back from GMAC, an investment company. In 2005, Beazer sought to enter into model-home sale-leaseback agreements. Under these agreements, Beazer would sell model homes to investors and rent the homes back from the investors until the subdivision was complete and the model home could be sold to a third party.
Generally, a seller cannot count the transaction as a sale and recognize revenue until “all risks and rewards of ownership” are transferred to the buyer. J.A. 2056. A seller may not have any “continuing involvement” with the property for it to be counted as a sale. Id. A transaction is not counted as a sale if the seller retains *457the ability to share in the appreciation of the home after it is sold.
Deloitte & Touche (“Deloitte”) served as Beazer’s auditors. Rand consulted with Deloitte senior manager, Corbin Adams, about a potential sale-leaseback arrangement with GMAC. In December 2005, Rand sent Adams a draft Master Sale and Rental Agreement (“MSRA”) that did not include any provision for Beazer to benefit from later appreciation in the value of the homes. He later confirmed that Beazer would not be able to “participate in appreciation of [the] leased assets.” Id. at 2074. Meanwhile, Rand was assuring Beazer’s employees that Beazer would share in the upside — the future profits from appreciation in value before GMAC eventually sold them. The same day Beazer entered into the MSRA, a Property Management Agreement (“PMA”) between GMAC entities was executed, providing that Beazer would share in the upside of any consumer transactions. In the next nine months, Beazer entered into two more MSRAs, followed by PMAs, agreeing that Beazer would share in appreciation when the model homes sold. Beazer received $117 million for the model homes it sold and reported $24.8 million in total profit.
Finally, Rand was charged with obstruction of justice stemming from his allegedly deleting emails following a grand jury subpoena. In March 2007, the FBI began investigating Beazer for mortgage fraud. On March 23, 2007, a federal grand jury issued a subpoena requiring Beazer to retain all documents, including emails, related to mortgages or home sales.
On March 28, Beazer initiated an “email dumpster,” which would save all deleted emails from permanent deletion. Beginning March 29, all deleted emails were caught in this dumpster without the employee’s knowledge. At 2:58 p.m. on March 30, Beazer’s CEO Ian McCarthy sent a memorandum to Rand and other senior management notifying them that Beazer was providing documents in response to the subpoena and would be providing an updated document-retention memorandum. Around 4:20 p.m., Deborah Danzig, an in-house attorney, sent an email to all employees in the corporate office, including Rand, with this memorandum, instructing them not to destroy any records. Danzig also testified that she told Rand directly that “he was required to keep everything and destroy nothing.” Id. at 975.
Between 5:55 p.m. on March 29 and 5:45 p.m. on March 30, 2007, Rand deleted nearly 6,000 emails dating back to 1999. Some of the emails were responsive to the grand jury’s subpoena and contained evidence of mortgage fraud. Other emails that Rand deleted were related to the cookie-jar accounting scheme. Others still appeared irrelevant to either set of charges.
Shortly after the subpoena was issued, Beazer’s audit committee hired the law firm Alston & Bird to conduct an internal investigation. Mike Brown, a partner at Alston & Bird, interviewed Rand as part of that investigation. On June 15, 2007, during their first interview, Rand told Brown that he had not destroyed or deleted any documents or emails since the investigation had begun. On June 26, 2007, Brown met with Rand again. Brown had learned • that the email dumpster had recovered thousands of emails that Rand had attempted to delete. At that meeting, Rand initially provided that he did not delete any emails, but he eventually admitted that he might have deleted “a couple of emails” to reduce the size of his mailbox. Id. at 1072. On further questioning, Rand said that he deleted “a series of emails” from one particular coworker on March 30. Id. at 1073.
Beginning July 2008, the FBI conducted between six and eight interviews with *458Rand as part of a proffer. During these interviews, conducted by FBI Agent Douglas Curran and others, Rand admitted to manipulating Beazer’s earnings, admitted that that was illegal, and expressed remorse. Curran testified that he also asked Rand about the GMAC transaction, and Rand admitted that there was a “verbal side agreement to share in the appreciation of the model homes when they were ultimately sold.” Id. at 2780.
Curran also asked Rand about the email deletions. Curran testified that Rand admitted that “he was certain that by March 27th he was for sure at the latest aware that there was a federal investigation in Charlotte.” Id. at 2784. Rand also admitted that he had spoken with Danzig and understood that the document-retention memorandum applied to him, when he “went back to his office and started performing mass deletions of emails.” Id. at 2784-85. Explaining that he was “essentially in a state of panic,” he deleted the emails because “[t]here were a lot of stressful events going on in his life at that time, and on top of that he was aware of the federal grand jury investigation that was focused in Charlotte and he did not want to be associated with that investigation in any way.” Id. at 2785. Rand admitted that he “understood that he was deleting evidence pertinent to the investigation” and “[h]e knew it was wrong.” Id. at 2786.
Rand went to trial twice. Before the first trial, Rand sought leave to subpoena computer forensic evidence of Rand’s email deletions and records from Beazer’s accounting system to show Rand’s accounting was reasonable and justified and to contextualize and refute the prosecution’s accounting records. The district court denied both requests.
In the first trial, the government presented evidence of emails relevant to the grand jury’s investigation into Beazer’s mortgage division and that Rand deleted from his Beazer email account. Aaron Phi-lipp, a computer forensics expert, testified that based on Beazer’s backup tapes, 3,272 emails were deleted between March 23 and 28, while another 5,936 were deleted on March 30, after the email dumpster was put into place.1
The jury deliberated for twenty hours and returned a split verdict, convicting Rand on seven counts and acquitting on four. A new trial was later granted due to juror misconduct.
In advance of the second trial, Rand again sought to subpoena Beazer to obtain records from its accounting system. Again, the district court denied the request. Rand also tried again to get the backup tapes from Beazer of the March 23-28 email deletions, and this time, the court granted the request. Rand’s expert examined the data on the backup tapes and concluded that approximately 2,500 of the approximately 3,200 emails that Philipp testified during the first trial were deleted between March 23 and March 28, 2007 (prior to the dumpster being in place), were not, in fact, deleted, explaining that “there [were] various technical explanations why Mr. Philipp could not find them on the tape the first time.” Id. at 719.
The government dropped Philipp from its witness list, halted all efforts to prove the March 23-28 deletions, and moved to strike parts of the indictment relating to those deletions. The government also *459moved to preclude Rand from introducing evidence or mentioning the false accusations at the retrial. The court granted the prosecution’s request ruling that the evidence was irrelevant and excludable under Federal Rule of Evidence 403 as distracting or confusing because the prosecution was no longer seeking to prove the March 23-28 deletions.
In addition to dropping the count tied most closely to the March 23-28 deletions, the government also abandoned its effort to prove Rand had committed securities fraud. It thus proceeded only with the conspiracy counts (counts 1 and 2), in violation of 18 U.S.C. § 371 (conspiracy) and 18 U.S.C. § 1349 (wire fraud conspiracy), and three obstruction of justice counts (counts 6, 9, and 11), in violation of 18 U.S.C. § 1512(b)(3), (c)(1), and (c)(2). Rand was ultimately convicted on all five counts.
Prior to sentencing, Rand’s probation officer calculated a total offense level of 43 and an advisory guideline range of life based, in part, on the finding that the loss reasonably foreseeable to Rand was between $100 and $200 million. Rand objected to this loss calculation, and the district court conducted a full-day sentencing hearing. During the hearing, both parties presented expert testimony on the appropriate calculation of loss under U.S. Sentencing Guideline §2B1.1 and, in particular, the effect on the value of Beazer’s stock of three separate announcements Beazer made to the market related to Rand’s offense conduct.
The district court adopted the government’s expert’s most “conservative methodology” and found a loss of $135 million. Id. at 3279. Based in part upon that finding, the district court calculated the total offense level of 51, resulting in an adjusted offense level of 43 and an advisory guideline sentence of life in prison. After considering the appropriate sentencing factors under 18 U.S.C. § 3553(a), the district court sentenced Rand to 120 months in prison.
This appeal followed.
H.
Rand first argues that the exclusion of evidence surrounding the false email accusations hampered his constitutional right to present a defense in three distinct ways: he was prevented from explaining the circumstance of his confession; he was unable to show that certain statements were not misleading; and he could not effectively cross-examine certain witnesses.
We “review[ ] evidentiary rulings implicating constitutional claims de novo.” United States v. Williams, 632 F.3d 129, 132 (4th Cir. 2011). Thus we review Rand’s claim regarding the circumstances of a confession under this standard. See Crane v. Kentucky, 476 U.S. 683, 106 S.Ct. 2142, 90 L.Ed.2d 636 (1986).
Nevertheless, “a defendant’s right to present a defense is not absolute: criminal defendants do not have a right to present evidence that the district court, in its discretion, deems irrelevant or immaterial.” United States v. Prince-Oyibo, 320 F.3d 494, 501 (4th Cir. 2003); see also Crane, 476 U.S. at 689-90, 106 S.Ct. 2142 (noting that the “Constitution leaves to the judges who must make these decisions wide latitude to exclude evidence that is repetitive ..., only marginally relevant or poses an undue risk of harassment, prejudice, [or] confusion of the issues” (citation and internal quotation marks omitted)). Thus, as to Rand’s other two arguments, we review for abuse of discretion, as they are “better fi’amed” as “evidentiary argu-mentes].” See United States v. Malloy, 568 F.3d 166, 177 (4th Cir. 2009). While his argument is “couched in terms of his due process right to defend himself, the crux of *460his complaint is that he was not allowed to present a particular defense.” Id. (citing United States v. Uzenski, 434 F.3d 690, 708-09 (4th Cir. 2006)).
Ultimately, harmless error review applies: “[Arizona v.] Fulminante[, 499 U.S. 279, 111 S.Ct. 1246, 113 L.Ed.2d 302 (1991),] enumerated the wide variety of constitutional errors subject to harmless error analysis,” including the “erroneous exclusion of a defendant’s testimony regarding the circumstances of a confession.” Sherman v. Smith, 89 F.3d 1134, 1137 (4th Cir. 1996) (en banc) (citing Crane, 476 U.S. at 691, 106 S.Ct. 2142). “That analysis requires a reviewing court to quantitatively assess the effect of the error ‘in the context of other evidence presented’ at trial.” Id. at 1138 (quoting Fulminante, 499 U.S. at 308, 111 S.Ct. 1246).
Here, the district court did not permit testimony that some of the March 23-28 email deletion accusations turned out to be false, concluding that such evidence was irrelevant or confusing or distracting for the jury, as the government would not be presenting evidence as to that timeframe. Rand argues that this unconstitutionally impinged his ability to present a complete defense. Had he been allowed, Rand would have introduced the following:
At the June 26 interview, Rand acknowledged deleting some junk emails, as well as innocuous emails with a colleague. He also truthfully denied deleting emails with high-level Beazer personnel. However, Brown, having just learned of the forensic analysis by Philipp, accused Rand of “falsefly]” denying mass email deletions. Beazer in turn fed that information to the prosecution. In a later reverse proffer, the prosecution told Rand it had forensic evidence that he had deleted accounting-related emails. It was that reverse proffer and the weight of alleged evidence purportedly showing Rand had mass-deleted deleted emails the week following the subpoena that prompted Rand’s statements to the government.
Rand Br. 23 (internal citation omitted) (alteration in original). We find no error, constitutional or otherwise, in the district court’s ruling to exclude this evidence.
First, Rand was not “stripped of the power to describe to the jury the circumstances that prompted his confession.” Crane, 476 U.S. at 683, 106 S.Ct. 2142.2 While a case may “stand or fall on [the defendant’s] ability to convince the jury that the manner in which the confession was obtained casts doubt on its credibility,” id. at 689, 106 S.Ct. 2142, the district court did permit Rand to “testify as to why he was induced into proffering,” J.A. 2664. He was allowed to “truthfully respond to what was in his mind at the time.” Id. The only thing Rand could not do was make “reference to the fact that years later some of the information he was confronted with turned out to be false.” Id. Rand did ultimately delete nearly 7,000 emails, including the 800 emails deleted between March 23 and 28, despite their not being at issue in the second trial. Further, as the government argues, only a few slides of the PowerPoint that they presented to *461Rand during the reverse proffer dealt with email deletion, and even fewer contained any incorrect information.
Rand also argues that the district court’s, ruling specifically impacted Count 9, which charged him with “knowingly and corruptly engaging] in misleading conduct” during the June 26, 2007, interview with Brown.3 Rand Br. 31-32 (citing J.A. 52). Rand relies on his view of the government’s theory of Count 9 from the first trial — that is, before the government discovered that the bulk of the March 23-28 emails were not deleted.' Under this view, Rand explains that the government’s “original theory of count 9” was that Rand told Brown that he had deleted some emails that “he believed were largely junk, ‘similar to advertisements for the drug Viagra,’ as well as some emails from particular ‘non-essential’ Beazer employees.” Id. at 32 (citing J.A. 38). After the interview, Brown received information that Rand had deleted “a large number of e-mails involving” the current and former CFOs and the CEO. J.A. 322-323. On receiving this information, Brown returned and confronted Rand with this information, but Rand “offered no corrections.” Id. As Counts 9 and 11 required proof of corrupt intent, 18 U.S.C. § 1512(b)(3), (c)(2), Rand argues that “[h]ad the jury learned that [he] truthfully denied the March 23-28 email deletions at the interview, the jury may have concluded, in view of Rand’s overall conduct at the interview, that Rand did not intend to mislead Brown or to hinder a grand jury investigation.” Rand Br. 33.
We agree with the government that Rand’s argument falls short here. At the second trial, Brown was only asked to testify about the deletions on March- 30, which “unquestionably occurred.” Gov’t Br. 49 (referring to J.A. 1070). And again, even if not as in as great a number as previously thought, Rand did still delete a certain number of emails between March 23-28. Thus we do not find that the district court abused its discretion in excluding this evidence based on its relevance weighed against potential juror confusion — the evidence was “irrelevant to the crime charged.” See Malloy, 568 F.3d at 177. Moreover, during both trials, Brown testified about Rand’s constantly shifting position. E.g., J.A. 1072-73 (“[H]e initially said that he did not delete any emails); id. at 1073 (“When we asked him about the week of the 23rd, I think he told us that he might have deleted a couple of emails in this manner [to clear out his inbox]. Mostly related to Viagra or some type of — sort of spam emails, but nothing else.”); id. at 1074 (describing how Rand “revise[d] his answers” and eventually told Brown that “he deleted a series of emails from Cory Boydston during one of those two weeks”). We can hardly say with any certainty that a jury would have found Rand’s “overall conduct” during the interview did not show an intent to mislead.4
Finally, we find any error harmless. While any harmlessness assessment *462should be sensitive to the “indelible impact a full confession may have on the trier of fact,” Fulminante, 499 U.S. at 318, 111 S.Ct. 1246 (Kennedy, J., concurring), we do not find that Rand’s explanation as to the circumstances of his confession would have countered the charges in light of the overwhelming evidence at trial. Rand did in fact delete the vast majority of the emails he was accused of deleting, including one quarter of those during March 23 to 28. Rand had the opportunity to present a vigorous defense, cross-examine the government’s witnesses, and, albeit in a slightly limited way, explain “what was in his mind” during his various interviews. J.A. 2664. We do not find that Rand was ultimately prejudiced by the omission.
III.
Rand next argues that several of the district court’s other evidentiary rulings were improper. The Court reviews these decisions for abuse of discretion. United States v. Richardson, 607 F.3d 357, 368 (4th Cir. 2010) (citing United States v. Fowler, 932 F.2d 306, 311 (4th Cir. 1991)). An abuse of discretion can occur “when the court uses an erroneous legal standard or bases its decision on clearly erroneous facts.” Id. (citing United States v. Under Seal (In re Grand Jury), 478 F.3d 581, 584 (4th Cir. 2007)).
Rand first argues that the district court abused its discretion in quashing his Federal Rule of Criminal Procedure 17(c) subpoena to Beazer. Rule 17(c) permits a defendant to issue a subpoena duces tecum to compel the production at trial of “books, papers, documents, data, or other objects.” Fed. R. Crim. P. 17(c)(1). A'district court “may quash or modify” the subpoena “if compliance would be unreasonable or oppressive,” Fed. R. Crim. P. 17(c)(2).
Rule 17(c) “is not intended to provide a means of pretrial discovery; rather, its primary purpose is simply ‘to expedite the trial by providing a time and place before trial for the inspection of subpoenaed materials.’ ” Richardson, 607 F.3d at 368 (quoting United States v. Nixon, 418 U.S. 683, 689-99, 94 S.Ct. 3090, 41 L.Ed.2d 1039 (1974)). In United States v. Nixon, the Supreme Court held that the requesting party bears the burden of showing
(1) that the documents are evidentiary and relevant; (2) that they are not otherwise procurable reasonably in advance of trial by exercise of due diligence; (3) that the party cannot properly prepare for trial without such production and inspection in advance of trial and that the failure to obtain such inspection may tend unreasonably to delay the trial; and (4) that the application is made in good faith and is not intended as a general “fishing expedition.”
418 U.S. at 699-700, 94 S.Ct. 3090 (footnote omitted). The Court distilled this showing into three requirements: “(1) relevancy; (2) admissibility; [and] (3) specificity.” Id. at 700, 94 S.Ct. 3090; see Richardson, 607 F.3d at 368.
Rand argues that the Nixon test applies only to subpoenas issued to the prosecution not to those issued to third parties. Instead, Rand contends that the standard explicit in the rule itself — unreasonable or oppressive — is the proper standard. While the Nixon Court noted that the special prosecutor suggested that the “evidentiary requirement” of the heightened standard did “not apply in its full vigor” for subpoenas to third parties, 418 U.S. at 700 n.12, 94 S.Ct. 3090, the Court determined that it “need not decide whether a lower standard exists” because the district court’s refusal to quash the subpoena was proper regardless. Id.
We have previously applied the Nixon test to third-party subpoenas, e.g., *463Richardson, 607 F.3d 357; In Re Martin Marietta Corp., 856 F.2d 619, 621 (4th Cir. 1988), but have not specifically considered how the evidentiary requirement should apply in that context. Thus, the issue appears to be one of first impression. See Legal Servs. Corp. v. Velazquez, 531 U.S. 533, 557, 121 S.Ct. 1043, 149 L.Ed.2d 63 (2001) (“Judicial decisions do not stand as binding ‘precedent’ for points that were not raised, not argued, and hence not analyzed.”). No circuit court appears to have applied the explicit standard apart from the Nixon standard, and Rand cites only a handful of district courts that have done so. See United States v. Al-Amin, No. 1:12-CR-50, 2013 WL 8865079, at *8 (E.D. Tenn. July 25, 2013) (noting that application of the lower standard “is a distinct minority view”). Nonetheless, Rand argues that the purpose for the heightened standard as to the government — to not allow bypass of Rule 16 through Rule 17 — does not apply in the case of third parties. And application of the higher standard is inconsistent with Rule 17(e)’s basic purpose of “implement[ing] the Sixth Amendment guarantee that an accused have compulsory process to secure evidence in his favor.” In Re Martin Marietta Corp., 856 F.2d at 621 (citing California v. Trombetta, 467 U.S. 479, 485, 104 S.Ct. 2528, 81 L.Ed.2d 413 (1984)). Rand contends that in document-intensive cases such as this one, requiring a defendant to specify precisely what he wants hinders this guarantee.
We decline to adopt a lower standard for third-party subpoenas under Rule 17(c) and find that the district court applied the correct standard. In Nixon, the Supreme Court reiterated that the subpoena duces tecum “was not intended to provide a means of discovery for criminal cases.” 418 U.S. at 698, 94 S.Ct. 3090 (citing Bowman Dairy Co. v. United States, 341 U.S. 214, 220, 71 S.Ct. 675, 95 L.Ed. 879 (1951)). Importantly, the Court did not cabin this purpose to discovery from the government. See id. Moreover, Rule 17(c) is available to both the defense and prosecution. As to Rand’s argument that the defense is hampered by the application of the Nixon standard, one court has noted, “The right to defend oneself does not extend to using the power of the Court to compel third parties to provide information that may not even be admissible at trial or at a hearing or that is merely ‘investigatory.’ ” Al-Amin, 2013 WL 3865079, at *7 n.3. Finally, the Nixon standard is not at' odds with our interpretation of the explicit standard in Rule 17. A subpoena should be quashed as unreasonable or oppressive if it is “irrelevant; abusive or harassing; overly vague; or excessively broad.” In re Grand Jury, John Doe No. G.J.2005-2, 478 F.3d at 585 (internal citations omitted) (considering a grand jury subpoena and citing in part United States v. Loe, 248 F.3d 449, 466 (5th Cir. 2001), which applied this standard in the context of a third-party trial subpoena). These map on quite well to the Nixon standard of relevance, admissibility, and specificity. See Nixon, 418 U.S. at 700, 94 S.Ct. 3090.
As we find that the district court applied the correct standard, we also find that the ruling was not an abuse of discretion. Rand’s request to Beazer was to produce “accounting entries, budgets, budget entries, and financial reports for seven categories of reserve accounts over an eight-year period — the timeframe of the alleged conspiracy.” Rand Br. 43-44. Rand argues that these reports would have enabled him to show the reasonableness of the reserve adjustments looking more broadly at Beazer resources and over a period of time. We find that the district court did not abuse its discretion in finding that Rand failed to limit his request to entries in issue in the prosecution’s case and to justify his broad *464request, instead finding it more of a fishing expedition, frowned upon by Nixon.
We further decline to find error as to Rand’s other two evidentiary claims. First, the district court did not abuse its discretion in prohibiting Rand’s accounting expert from testifying about work papers prepared by Beazer’s independent auditors at Deloitte. The district court permitted him to “offer his own opinion as to the legitimacy of [the] entries,” but that he could not “bootstrap that opinion by the Deloitte and Touche work papers which the court [did] not find reliable,” as the court found that the Deloitte witness, Adams, said that “he was not provided sufficient information” to make those conclusions himself. J.A. 2475, 2479-80. The district court did not abuse its discretion, especially where not all evidence relating to the Deloitte work papers was excluded and Rand was “free to call any of the auditors who did the work” to have them testify about their findings. See Gov’t Br. 55.
Additionally, the district court did not err in allowing the government to have Beazer employees testify as lay witnesses about the propriety of complex accounting transactions without calling an accounting expert to testify. Citing Federal Rules of Evidence 701 and 702, Rand contends that lay witnesses may not offer opinions about matters of “technical” or “other specialized knowledge” requiring expert proof and that the government needed an expert to explain accounting principles such as setting reserve levels and analyzing historical costs and projections of future costs.
We have previously affirmed the admission of lay-opinion testimony in a securities-fraud case. United States v. Offill, 666 F.3d 168, 177 (4th Cir. 2011). Under plain error review, we held that the district court “acted well within its broad discretion” in admitting testimony including that the defendant’s actions were “fraudulent,” “securities manipulation,” and “illegal.” Id. at 177-78. While Rand attempts to distinguish Offill by pointing out that the government presented expert testimony in addition to the lay witnesses who testified, here, Rand had opportunity to cross-examine these lay witnesses to expose the apparent falsity of their testimony, as well as having his own expert testify. Furthermore, “[i]f the government proves that a defendant was responsible for financial reports that intentionally and materially misled investors, the statute is satisfied. The government is not required in addition to prevail in a battle of expert witnesses over the application of individual [Generally Accepted Accounting Practice] rules.” United States v. Ebbers, 458 F.3d 110, 125-26 (2d Cir. 2006).
rv.
Rand takes issue with the government’s statements in its rebuttal closing argument concerning Rand’s wealth, Rand’s decision not to testify, and the government’s vouching for its own witness. See Griffin v. California, 380 U.S. 609, 615, 85 S.Ct. 1229, 14 L.Ed.2d 106 (1965) (unconstitutional to comment on defendant’s silence); United States v. Socony-Vacuum Oil Co., 310 U.S. 150, 239, 60 S.Ct. 811, 84 L.Ed. 1129 (1940) (class prejudice); United States v. Lewis, 10 F.3d 1086, 1089 (4th Cir. 1993) (bolstering).
As this issue raises questions of law, we review it de novo. United States v. Collins, 415 F.3d 304, 307 (4th Cir. 2005).5 Nevertheless, the claims are still *465subject to harmless error review. Chapman v. California, 386 U.S. 18, 22, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967); Sherman, 89 F.3d at 1137 (noting that harmless error review applies to errors such as improper comment on defendant’s silence). That is, “[w]ith respect to claims of prosecutorial misconduct, an appellant must show that the remarks were improper and that they ‘prejudicially affected the defendant’s substantial rights so as to deprive the defendant of a fair trial.’ ” United States v. Baptiste, 596 F.3d 214, 226 (4th Cir. 2010) (alteration in the original) (quoting United States v. Adam, 70 F.3d 776, 780 (4th Cir. 1995)). We have previously laid out factors to consider in determining whether improper remarks require reversal:
(1) the degree to which the prosecutor’s remarks have a tendency to mislead the jury and to prejudice the accused; (2) whether the remarks were isolated or extensive; (3) absent the remarks, the strength of competent proof introduced to establish the guilt of the accused; (4) whether the comments were deliberately placed before the jury to divert attention to extraneous matters [; ] ... (5) whether the prosecutor’s remarks were invited by improper conduct of defense counsel [;] ... and (6) whether curative instructions were given to the jury.
Id. (quoting United States v. Wilson, 135 F.3d 291, 299 (4th Cir. 1998)) (alterations in original). On all three of these potential misconduct claims, Rand makes little attempt to contend with any of these factors and fails to argue that any error was not harmless. Accordingly, we affirm the district court’s denial of a mistrial.
As to Rand’s class prejudice argument, Rand acknowledges that in closing, his counsel made reference to a bag of gold:
“You remember the story about the Emperor With No Clothes? In that story a weaver said to the Emporer [sic], .if you give me a bag of gold, I will make you an invisible suit of clothes.... That’s what this case is like, accounting entries, after accounting entries whizzing by that you can’t put together, and don’t add up.”
J.A. 3008-09. Rand argues, however, that the government impermissibly expanded the analogy in rebuttal.
At the outset of its rebuttal, the government referred to Rand’s wealth: “You just heard a story alright. It took a lot of gold. A lot of gold. The defendant’s lawyers, all of them, his experts, a lot of gold.” Id. at 3010; see also id. at 3012 (referring to Rand as “rich” in discussing testimony from Rand’s wife); id. at 3019 (referring to Rand’s experts as being paid $650 an hour to provide testimony helpful to Rand); id. at 3023 (“So you can go live in the story like the Emporer [sic] with no clothes and listen to the story that was bought with gold, or you can look at the facts.”).
We cannot find, as the government urges, that no .error occurred as the remarks were not a reference to Rand’s wealth, only to the fact that Rand’s case was “built on the testimony of compensated expert witnesses, none of whom had any personal knowledge of what Rand did or did not do.” Gov’t Br. 62. These conclusions are but two sides of the same coin. See United States v. Farinella, 558 F.3d *466695, 700-01 (7th Cir. 2009) (finding prose-cutorial misconduct where the government argued to the jury about not “let[ting] the defendant and his high-paid lawyer buy his way out of this”).
Nonetheless, considering the factors above, we conclude that any error did not affect Rand’s substantial rights. The remarks may not have been “isolated,” Baptiste, 596 F.3d at 226, and no curative instructions were given, id. But we find no evidence that the remarks were deliberately made to focus on “extraneous matters,” id.; instead they were founded in, and even inspired by, Rand’s closing — even if they ultimately exceeded the scope of that context. Furthermore, given the strength of the evidence presented throughout the trial, we do not find that the comments overly misled the jury or prejudiced Rand. Id.
Considering Rand’s second argument that the prosecution improperly commented on his decision not to testify, the right of a defendant in a criminal trial “to remain silent unless he chooses to speak in the unfettered exercise of his own will” is guaranteed by the Fifth Amendment, Malloy v. Hogan, 378 U.S. 1, 8, 84 S.Ct. 1489, 12 L.Ed.2d 653 (1964); see U.S. Const. amend. V; and the Constitution “forbids either comment by the prosecution on the accused’s silence or instructions by the court that such silence is evidence of guilt,” Griffin, 380 U.S. at 615, 85 S.Ct. 1229. We ask, “Was the language used manifestly intended to be, or was it of such character that the jury would naturally and necessarily take it to be a comment on the failure of the accused to testify?” United States v. Francis, 82 F.3d 77, 78 (4th Cir. 1996) (quoting United States v. Anderson, 481 F.2d 685, 701 (4th Cir. 1973)).
Here again, we affirm the district court. In explaining Rand’s earlier confession, the government described Rand’s argument as “[t]he FBI is lying. And I . [Rand] lied to the FBI because I was desperate.” J.A. 3010. Returning to this idea, the government said,
“But then he also said, but wait, I also lied to the FBI because I was desperate. How those things fit together I [the prosecutor] didn’t understand, maybe you did.
So I heard that the FBI lied, I also heard that Mr. Rand lied because he was desperate. Why was he desperate? He didn’t say. Nor could he really, because your lawyer can’t talk about your own beliefs, they can just make arguments.”
Id. at 3012. The district denied the motion for a mistrial, as it found that the jury could conclude that the “he” was referring to Rand’s counsel’s silence, as counsel had just spoken and claimed in argument that Rand made the confession only because Rand was desperate. We do not find this conclusion in error. While Rand argues that the jury would have understood that the “he” was referring to Rand as “[t]he rebuttal argument repeatedly referred to Rand’s theory about why Rand was desperate,” Rand Br. 55 (referencing the government’s repeated use of “I”), we do not find this to be the “necessarfy]” conclusion the jury would draw. See Francis, 82 F.3d at 78.
Finally, Rand argues that the government improperly vouched for Curran’s credibility. In rebuttal, the government argued that if Curran lied, “he’s risking perjury. He’s risking the loss of his career. There’s a federal judge sitting right there going to put him in jail — .” J.A. 3011. Rand immediately objected to this argument, and the district court sustained the objection, but Rand contends that this was insufficient as the court did not direct the jury to disregard it or give a curative *467instruction. Cf. United States v. Forlorma, 94 F.3d 91, 95 (2d Cir. 1996) (considering one of various factors and finding “we cannot be confident that the judge’s unexplained ruling dispelled the misperception that was likely caused by the baseless argument”). The government importantly notes, however, that although he objected, Rand did not ask for a curative instruction, either at the time or later in the jury charge.
Again, we find any error harmless. Rand himself acknowledges that Curran was “extensively cross-examined about discrepancies between his trial testimony, his contemporaneous notes of Rand’s proffer sessions, and his later write-up of those notes.” Rand Br. 56. While Rand argues that, had the jury doubted Curran’s testimony, it might have found reason to doubt Rand’s confession as well, we do not find that the jury’s determination of guilt or innocence “hinged entirely on the credibility” of Curran such that any improper remarks affected Rand’s substantial rights. See United States v. Gracia, 522 F.3d 597, 606 (5th Cir. 2008). Instead, given the overwhelming evidence, including Curran’s testimony, supported in part by Brown’s testimony and the physical evidence in the record, we find the context of the error harmless.
V.
Rand finally challenges his sentence as procedurally unreasonable, arguing that the district court erred in determining the loss calculation by failing to apply the principles from Dura Pharmaceuticals, Inc. v. Broudo, 544 U.S. 336, 125 S.Ct. 1627, 161 L.Ed.2d 577 (2005). As Rand objected at sentencing, the Court reviews improper calculation of a guideline range de novo. United States v. McManus, 734 F.3d 315, 318 (4th Cir. 2013). Meanwhile, “[t]he determination of loss attributable to a fraud scheme is a factual issue for resolution by the district court, and we review such a finding of fact only for clear error.” United States v. Keita, 742 F.3d 184, 191 (4th Cir. 2014) (citation omitted).
U.S. Sentencing Guideline §2B1.1 sets the offense level for certain fraud offenses and requires an increase based on the loss caused by the offense conduct, in accordance with a table in § 2Bl.l(b)(l). An application note instructs that “in a case involving the fraudulent inflation or deflation in the value of publicly traded security,” loss should be calculated based on how the price of a security changed, “after the fraud was disclosed to the market.” U.S.S.G. § 2B1.1 Application Note 3(F)(ix).
At sentencing, the parties debated which of Beazer’s three public disclosures qualified as the date on which the “fraud was disclosed to the market”:
June 27, 2007: Beazer announced that Rand had been fired for “destroying] documents” and that an investigation was ongoing involving mortgage origination and “related matters.” J.A. 4624. August 10, 2007: Beazer announced that its investigation “has discovered that [Rand] may have caused reserves ... to have been recorded ... in excess of amounts that would have been appropriate,” but that the “investigation is ongoing” and that Beazer did not “believe that the amounts ... are quantitatively material.” Id. at 4626.
October 11, 2007: Beazer summarized the findings of its investigation, quantified the effects of Rand’s reserve adjustments, explained the GMAC issue, and informed the public that Beazer would restate its financials. Id. at 4606-07.
The court determined that the fraud was disclosed in June and August and that the loss to investors following those dates was $135 million. Accordingly, the district court *468calculated an offense level of 51 for a guidelines range of life imprisonment, capped by the statutory maximum. The parties agreed that if the October date were used, the resulting loss would be $0. Had the district court used the loss amount following the October disclosure, Rand’s offense level would have been 19, with a range of 30 to 37 months. The court ultimately varied downward from the guidelines range of life imprisonment and imposed a ten-year sentence.
Drawing from principles in civil securities cases, Rand argues that the proper date to consider was the October disclosure. In the civil context, the Supreme Court has held that to sustain a damages claim for civil securities fraud under 15 U.S.C. §§ 78j(b) and 78u-4, a plaintiff must show “a causal connection between the material misrepresentation and the loss.” Dura Pharms., 544 U.S. at 342, 125 S.Ct. 1627. In so holding, the Dura Court rejected the notion that stock overvaluation resulting from so-called “fraud-on-the-market” may form the basis for a plaintiffs damages award in a private securities action. Id. at 341-43, 125 S.Ct. 1627. That is, a shareholder’s claim that he bought stock at a price that was artificially inflated due to fraud does not state a claim for loss. Id.
The Second and Fifth Circuits have suggested that the Dura loss-causation principles apply to criminal securities fraud cases. In United States v. Olis, 429 F.3d 540 (5th Cir. 2005), the Fifth Circuit indicated, “The civil damage measure should be the backdrop for criminal responsibility both because it furnishes the standard of compensable injury for securities fraud victims and because it is attuned to stock market complexities.” 429 F.3d at 546 (citing Dura Pharms., 544 U.S. at 341-43, 125 S.Ct. 1627). Olis cited several out-of-circuit cases, including various “cook the books” scenarios, and noted with approval that “each case takes seriously the -requirement to correlate the defendant’s sentence with the actual loss caused in the marketplace, exclusive of other sources of stock price decline.” Id. at 547 (citing United States v. Snyder, 291 F.3d 1291 (11th Cir. 2002); United States v. Bakhit, 218 F.Supp.2d 1232, 1238 (C.D. Cal. 2002); United States v. Grabske, 260 F.Supp.2d 866, 869-71 (N.D. Cal. 2002)). In United States v. Rutkoske, 506 F.3d 170 (2d Cir. 2007), the court stated, “[W]e see no reason why considerations relevant to loss causation in a civil fraud case should not apply, at least as strongly, to a sentencing regime in which the amount of loss caused by a fraud 'is a critical determinant of the length of a defendant’s sentence.” 506 F.3d at 179; cf. United States v. Nacchio, 573 F.3d 1062, 1078 (10th Cir. 2009) (“Courts in criminal cases have sought guidance from civil damage measures in considering an estimate of loss from the defendant’s unlawful conduct.” (citing Kevin P. McCormick, Untangling the Capricious Effects of Market Loss in Securities Fraud Sentencing, 82 Tul. L. Rev. 1145,1153 (2008) (“Faced with a myriad of new issues never encountered before in the criminal context, the courts have turned to civil jurisprudence for answers.”))) (considering profits for sentencing).6
Meanwhile, the Third, Sixth, and Ninth Circuits have declined to apply Dura in the context of criminal sentencing. The Ninth Circuit, for example, has found the Dura *469Court’s concern not relevant in criminal sentencing:
[ I]n a private civil fraud action, a court gauges loss from the perspective of the plaintiff-victim, i.e., whether the plaintiff can show the amount and cause of loss he sustained. Because a civil plaintiff bears the burden tó show loss, it is logical to require that the plaintiff show that any loss he sustained was attributable directly to devaluation caused by revelation of the defendant’s fraud. It likewise follows that a plaintiffs mere allegation that he purchased overvalued stock is insufficient to state a claim, because the allegation does not by itself establish that the plaintiff personally incurred loss commensurate with the overvaluation.
In criminal sentencing, however, a court gauges the amount of loss caused, i.e., the harm that society as a whole suffered from the defendant’s fraud. Whether and to what extent a particular individual suffered actual loss is not usually an important consideration in criminal fraud sentencing. Therefore, where the value of securities have been inflated by a defendant’s fraud, the defendant may have caused aggregate loss to society in the amount of the fraud-induced overvaluation, even if various individual victims’ respective losses cannot be precisely determined or linked to the fraud. As a result, the principle underlying the Dura Pharmaceuticals Court’s reluctance to allow mere overvaluation as a basis for establishing loss is generally not present in the criminal sentencing context, and we are not persuaded that it would be appropriate to expand the Dura Pharmaceuticals rule to the criminal sentencing context.
United States v. Berger, 587 F.3d 1038, 1044 (9th Cir. 2009) (internal citations omitted); see also United States v. Georgiou, 777 F.3d 125, 146 (3d Cir.), cert. denied, - U.S. -, 136 S.Ct. 401, 193 L.Ed.2d 312 (2015); United States v. Peppel, 707 F.3d 627, 644-45 (6th Cir. 2013).
We find the reasoning of the Berger court convincing and adopt it here. Accordingly, we decline to adopt the Dura principles in the criminal context. The district court thus committed no clear error in determining the loss amount, and we affirm the sentence.
Even assuming we found Dura’s principles applicable, the district court’s finding would still stand. In Peppel, the Sixth Circuit affirmed the amount-of-loss determination where class actions had been filed alleging false inflation, newspaper articles reported the allegations, and “less than a month later, an announcement was made informing the investing public as follows: ‘MCSi ... today announced that it has learned of an investigation of the Company by the [SEC] and has received a subpoena from the SEC seeking production of documents ....’” Peppel, 707 F.3d at 644 (alterations and omission in original). The court determined that “[information concerning Peppel’s fraud was thus generally available to the investing public.” Id. Similarly here, we find that Beazer’s June and August announcements sufficiently put investors on notice of fraud. We are unmoved by Rand’s invocation of Loos v. Immersion Corp., 762 F.3d 880, 890 (9th Cir. 2014), as amended (Sept. 11, 2014) (considering a motion to dismiss for failure to state a claim in a civil case and holding that “the announcement of an investigation, without more, is insufficient to establish loss causation”). As in Peppel, here “it does not take a strong inference to connect the publication of this information to the near-immediate” losses to investors. 707 *470F.3d at 644.7 Accordingly, we affirm the loss finding and Rand’s sentence as procedurally reasonable.
VI.
For the foregoing reasons, the district court is
AFFIRMED.