Stinn v. United States

25-293 Stinn v. United States of America In the United States Court of Appeals For the Second Circuit ________ AUGUST TERM 2025 ARGUED: MARCH 16, 2026 DECIDED: AUGUST 17, 2026 No. 25-293 BRADLEY J. STINN, Petitioner-Appellant, v. UNITED STATES OF AMERICA, Respondent-Appellee. ________ Appeal from the United States District Court for the Eastern District of New York ________ Before: WALKER, SULLIVAN, and BIANCO, Circuit Judges. ________ Bradley Stinn appeals from an order denying his petition for a writ of error coram nobis. In 2009, Stinn was convicted of securities fraud, mail fraud, and conspiracy to commit the same. At trial, the jury was permitted to rely on either the traditional fraud theory or the right-to-control theory of fraud liability. After the Supreme Court  No. 25-293 invalidated the right-to-control theory in Ciminelli v. United States, 598 U.S. 306 (2023), Stinn filed a coram nobis petition in the Eastern District of New York to vacate his sentence. The district court (Chen, J.) denied the petition, holding that the right-to-control instruction given at trial was harmless error. On appeal, Stinn argues that his conviction is not supported by either the traditional fraud theory or the now-invalidated right-to-control theory, and that the district court erred by applying the wrong test for harmless error. We disagree. For the reasons explained below, we hold that the standard articulated in Kotteakos v. United States, 328 U.S. 750 (1946), governs review of harmless error in coram nobis petitions. Stinn failed to satisfy that test. Thus, we AFFIRM the judgment of the district court. ________ GIL WALTON (David W. Shapiro, on the brief), The Norton Law Firm PC, Oakland, CA, for Petitioner- Appellant VICTOR ZAPANA (David C. James, on the brief), Assistant United States Attorneys, for Joseph Nocella, Jr., United States Attorney, Eastern District of New York, Brooklyn, NY, for Respondent-Appellee ________ JOHN M. WALKER, JR., Circuit Judge: Bradley Stinn appeals from an order denying his petition for a writ of error coram nobis. In 2009, Stinn was convicted of securities fraud, mail fraud, and conspiracy to commit the same (18 U.S.C. §§ 1341, 1348, 1349). At trial, the district court instructed the jury that it could rely on either the traditional fraud theory or the right-to-control theory of fraud liability. The jury voted to convict on 2  No. 25-293 all three counts, and Stinn served a sentence of 144 months of imprisonment and three years of supervised release. After Stinn completed his sentence and his term of supervised release, the Supreme Court invalidated the right-to-control theory in Ciminelli v. United States, 598 U.S. 306 (2023). Stinn then filed a coram nobis petition to vacate his sentence. The district court (Chen, J.) denied the petition, holding that the right-to-control instruction given at trial was harmless error, and that the traditional fraud theory was a valid basis for Stinn’s conviction. On appeal, Stinn argues that his conviction is not supported by either the traditional fraud theory or the now-invalidated right-to-control theory. He also argues that the district court erred by applying the test for harmless error used to evaluate habeas petitions, Kotteakos v. United States, 328 U.S. 750 (1946), rather than the test used on direct review, Chapman v. California, 386 U.S. 18 (1967). We disagree. For the reasons explained below, we hold that Kotteakos is the appropriate standard for evaluating harmless error in coram nobis petitions. Under that standard, the relevant inquiry is whether the error “had substantial and injurious effect or influence in determining the jury’s verdict.” Kotteakos, 328 U.S. at 776. Stinn failed to satisfy that test. Thus, we affirm the judgment of the district court. BACKGROUND From 1992 to 2003, Bradley Stinn served as the chief executive officer of Friedman’s Inc. (“Friedman’s”). Friedman’s was the third- largest specialty retailer of fine jewelry in the United States, operating over six hundred stores in twenty states. Beginning in August 2003, the company’s fortune took a turn for the worse when it was sued in a civil lawsuit for purported accounting fraud. The following month, the Securities and Exchange 3  No. 25-293 Commission and the Department of Justice launched investigations into the allegations made in the civil lawsuit. By December 2003, Stinn had resigned as CEO and from the company’s board of directors. Approximately two years later, the company filed for bankruptcy. In 2007, Stinn was charged by superseding indictment in the Eastern District of New York on three counts of mail fraud, securities fraud, and conspiracy to commit the same. The indictment centered on Friedman’s credit-extension program, which allowed low- and middle-income customers to purchase jewelry on installment plans. Although the program was lucrative, comprising over half of the company’s revenues, these customers were at greater risk of defaulting on their payments. As alleged in the superseding indictment, Stinn conspired with Friedman’s executives to defraud investors by masking that heighted risk of default, through falsifying Friedman’s accounting data and misrepresenting the company’s financial condition in public reports. At trial, the government presented evidence regarding Stinn’s financial motivations for the scheme, including his $352,000 bonus for 2002 and $300,000 salary increase for 2003, and argued that both “result[ed] [from] the lies the defendant told” regarding Friedman’s targeted earnings. App’x at 2061. The government presented testimony from Friedman’s former chief financial officer, Victor Suglia, who testified that Stinn would not have received his $352,000 bonus “[a]bsent [his] manipulation[s.]” Id. at 1203. Former director Robert Cruickshank also testified that Stinn received a $300,000 salary raise, a “much larger increase than anybody else” at the company, 4  No. 25-293 because the compensation committee “thought that the company was doing very well[.]” Id. at 1366. After the close of evidence, the district court (Gershon, J.) issued the jury instructions relevant to this appeal. The jury was instructed that to convict, it must find that the alleged scheme to defraud “contemplated or intended some harm to property rights of another[,]” as defined under either of two theories. Special App’x at 3 (internal quotation marks omitted). Under the traditional theory of fraud liability, Stinn must have “planned to obtain or actually obtained money from Friedman’s by materially fraudulent representations[.]” Id. (internal quotation marks omitted). Under the right-to-control theory, Stinn must have “intended that other individuals would make investment decisions . . . based on materially fraudulent misrepresentations.” Id. (internal quotation marks omitted). Following these instructions, the jury voted to convict Stinn on all counts. Because the jury returned a general verdict, there was no indication as to which of the two theories the jury relied on. The jury also returned a forfeiture verdict requiring Stinn to forfeit $1,019,000, including his $352,000 bonus received in 2002, his $300,000 raise in 2003, and other funds improperly obtained from expensed tax liabilities. The district court then sentenced Stinn to 144 months of imprisonment and three years of supervised release and ordered him to pay approximately $5.5 million in restitution and forfeiture. During his incarceration, Stinn made direct and collateral attacks on his conviction, resulting in two decisions by our court. On direct appeal, we rejected Stinn’s argument that the district court erred in giving a conscious avoidance instruction and a purportedly coercive charge pursuant to Allen v. United States, 164 U.S. 492 (1896). United States v. Stinn (“Stinn I”), 379 F. App’x 19, 20-21 (2d Cir. 2010) 5  No. 25-293 (summary order). On appeal of Stinn’s habeas petition, pursuant to 28 U.S.C. § 2255, we rejected the argument that Stinn was convicted under an honest services theory of fraud purportedly rendered unconstitutional by Skilling v. United States, 561 U.S. 358 (2010). See Stinn v. United States (“Stinn II”), 515 F. App’x 4, 5 (2d Cir. 2013) (summary order). In 2023, the Supreme Court decided Ciminelli v. United States, holding that “the right-to-control theory is not a valid basis for liability” for federal fraud. 598 U.S. at 309. 1 Ten months after the Supreme Court’s decision, Stinn filed a coram nobis petition, and for Fed. R. Civ. P. 60(b) relief from the denial of his habeas petition in Stinn II, in the Eastern District of New York. He argued that his conviction should be vacated because the legal basis for his conviction rested on the now invalidated right-to-control theory. The district court (Chen, J.) denied Stinn’s petition. 2 See Stinn v. United States (“Stinn III”), No. 11-CV-2071 (PKC), 2024 WL 4989241 (E.D.N.Y. Dec. 5, 2024). The district court held that Stinn satisfied the first two requirements for coram nobis relief: (1) “sound reasons . . . for failure to seek appropriate earlier relief,” and (2) continuing collateral “legal consequences from his conviction that may be remedied” by a grant of coram nobis. Id. at *3 (internal quotation marks omitted). But the court rejected Stinn’s petition on the last requirement: (3) the need to show “circumstances compelling [a grant 1 Ciminelli interpreted the meaning of “money or property” under the wire fraud statute (18 U.S.C. § 1343), not the securities or mail fraud statutes (18 U.S.C. §§ 1341, 1348). Nonetheless, we assume without deciding that Ciminelli’s “money or property” has a uniform meaning across all three statues. See Ciminelli, 598 at 312 n.2 (“[W]e have construed identical language in the wire and mail fraud statutes in pari materia.” (internal quotation marks omitted)). 2 Judge Gershon recused herself and the case was then reassigned to Judge Chen. See Stinn III, 2024 WL 4989241, at *2 n.4. 6  No. 25-293 of coram nobis] to achieve justice.” Id. (internal quotation marks omitted). In this regard, the district court ruled that the jury did not necessarily rely on the right-to-control theory and that Stinn’s salary raise and bonus were valid property interests under the traditional fraud theory. Id. at *5–6. Stinn timely appealed. 3 DISCUSSION On an appeal from the denial of a coram nobis petition, “we review de novo the question of whether a district judge applied the proper legal standard, but review the judge’s ultimate decision to deny the writ for abuse of discretion.” United States v. Mandanici, 205 F.3d 519, 524 (2d Cir. 2000). Stinn makes two principal arguments. First, he contends that his salary raise and bonus were not traditional property rights protected under the federal fraud statutes. Second, he argues that even if the traditional fraud theory is valid, and his salary raise and bonus are traditional property interests, his conviction must still be vacated because the jury was permitted to convict on the now- invalidated right-to-control theory, “and it is impossible to tell which ground the jury selected.” Yates v. United States, 354 U.S. 298, 312 (1957). We address each argument in turn. I. Traditional Fraud Theory The federal fraud statutes require that the government prove beyond a reasonable doubt that “money or property” was the object of a defendant’s scheme to defraud. 18 U.S.C. §§ 1341, 1348. The Supreme Court has explained that the money or property requirement extends to both “tangible” and “intangible” property 3 Stinn did not appeal the district court’s denial of his Rule 60(b) relief, and we thus consider that argument abandoned. See Tereshchenko v. Karimi, 102 F.4th 111, 123 n.5 (2d Cir. 2024). 7  No. 25-293 rights. Carpenter v. United States, 484 U.S. 19, 25 (1984). But the object of the fraud must still be a “traditional property interest” that has “‘long been recognized as property.’” Ciminelli, 598 U.S. at 309, 314 (quoting Carpenter, 484 U.S. at 26). On appeal, Stinn argues that “neither [] maintenance of [his] salary nor [] [his] bonus and salary raise satisfies” that test. Appellant Br. at 49-50. Regarding the “salary maintenance” argument, Stinn relies principally on a Ninth Circuit case, United States v. Yates, which held that “a scheme whose object is to deceive an employer while continuing to draw an existing salary” does not fall within the sweep of the federal fraud statutes. 16 F.4th 256, 266 (9th Cir. 2021). 4 At the outset, we note that we previously considered identical arguments in Stinn II. There, Stinn argued that the district court erred when it held that a “salary raise and a bonus constitute money or property in a traditional mirror-image fraud case” and “the government did not rely on other forms of compensation (such as salary maintenance) to prove the money/property element.” Brief for Petitioner-Appellant, Stinn v. United States, No. 12-1930, at *36 (2d Cir. Oct. 12, 2012). We rejected Stinn’s arguments then, see Stinn II, 515 F. App’x at 5 (“We have considered all of Stinn’s arguments on appeal and find them to be without merit.”), and we reject those same arguments now. The law-of-the-case doctrine “forecloses relitigation of issues expressly or impliedly decided by the appellate court.” United States v. Frias, 521 F.3d 229, 234 (2d Cir. 2008) (internal quotation marks omitted). Although there are narrow exceptions to the rule, they do not apply here. Stinn has not pointed to the “availability of new 4 The D.C. Circuit declined to follow the Ninth Circuit’s approach in Yates. See United States v. Guertin, 67 F.4th 445, 453 (D.C. Cir. 2023), abrogated on other grounds by Kousisis v. United States, 605 U.S. 114, 145 (2025). 8  No. 25-293 evidence,” and the main “intervening change of controlling law” that Stinn relies on (Ciminelli) did not concern whether salary raises and bonuses are traditional property interests protected by the federal fraud statutes. Id. at 235 n.6 (internal quotation marks omitted). Nor do we find any “clear error” or “manifest injustice” in the district court’s adherence to Stinn II. Id. (internal quotation marks omitted). But even if we were to consider Stinn’s salary maintenance arguments on the merits—and even assuming that salary maintenance does not alone suffice to support a fraud conviction— those arguments would still fail. 5 Stinn points to examples in the trial record where the government referenced his “handsome salary” and that he was “paid handsomely to honor” investors’ “trust” in him. App’x at 924, 936. Based on these references, he argues that we should infer that the jury’s verdict rested on the impermissible right-to- control and salary-maintenance theories. That inference is not supported by the trial record. “[A] court should not lightly infer that a prosecutor intends an ambiguous remark to have its most damaging meaning or that a jury, sitting through lengthy exhortation, will draw that meaning from the plethora of less damaging interpretations.” Donnelly v. DeChristoforo, 416 U.S. 637, 647 (1974). The stray remarks that Stinn cites are 5 The Supreme Court’s recent decision in Kousisis endorsed the “fraudulent- inducement” theory of federal fraud, which criminalizes schemes that “trick a victim into . . . handing over her money or property” (regardless of whether the victim receives “something in return”). 605 U.S. at 118. Under this straightforward definition, lying so as to receive a salary would appear to qualify as fraud. Furthermore, while Yates worried that such a flexible conception would “criminalize a wide range of commonplace conduct,” 16 F.4th at 267, Kousisis explained that courts should “distinguish[] everyday misstatements from actionable fraud” based on their materiality, 605 U.S. at 131—not on broad, categorical exceptions to otherwise clearly applicable criminal statutes. 9  No. 25-293 outweighed by the litany of the government’s references at trial to Stinn’s bonus and salary raise. For example: • App’x at 934 (“[T]he defendant’s yearly bonus at Friedman’s was tied directly to Friedman’s making its earnings expectation.”) • App’x at 935 (“[A]s a result of hitting these earnings targets, this defendant received [a] $352,000 bonus.”) • App’x at 2061 (Stinn “rigged the mathematical equation that gave him that 352,000 dollar bonus in 2002”); id. (“As a result of the lies the defendant told, he got an ever-increasing salary and an ever more generous compensation package of bonuses . . . .”) • App’x at 2073 (Stinn “insisted on the 10 percent to boost reported earnings and not too coincidentally to secure his bonus. That was intent to [commit] fraud.”) • App’x at 2077 (Stinn was “looking for another big bonus . . . [t]he reason he did it was because of his intent to defraud”) Two district judges have examined the voluminous trial record and correctly concluded that “the record does not support th[e] assertion” that “the government attempted to satisfy the money or property element by showing that [the] petitioner maintained the salary he was already receiving[.]” Stinn v. United States, 856 F. Supp. 2d 531, 543 (E.D.N.Y. 2012); Stinn, 2024 WL 4989241, at *6 (same). The record is “replete with instances in which the government proffered that [P]etitioner’s fraud resulted directly in his receiving a substantial raise and bonus.” Stinn, 856 F. Supp. 2d at 543. Hence, Stinn’s case is dissimilar to Yates, where the “prohibited” “[right-to-control] and salary-maintenance theories” did “not make up just a few stray lines on a PowerPoint slide at closing argument [but] were the focus of the 10  No. 25-293 entire prosecution from beginning to end.” Yates, 16 F.4th at 269 (internal quotation marks omitted). In sum, we agree with the district court that the object of Stinn’s fraud was his 2003 salary increase and 2002 bonus, not his pre- existing salary. Under Yates, “a raise or a bonus” is covered by the federal fraud statutes if the remuneration is “tied to some specific performance metric,” and the employee “lies about having achieved that metric.” Yates, 16 F.4th at 268. Stinn’s “rigg[ing] [of] the mathematical equation” behind Friedman’s financial reports to “hit[] . . . earnings target[s]” and “secure his bonus” easily meets that test. App’x at 935, 2061. II. Harmless Error We next consider the issue of the Yates instructional error. Under the Supreme Court’s Yates test (not to be confused with the Ninth Circuit’s decision of the same name), a “verdict [is] to be set aside in cases where the verdict is supportable on one ground, but not on another, and it is impossible to tell which ground the jury selected.” Yates, 354 U.S. at 312. Yates errors are “not structural and are subject to harmlessness review.” Johnson v. United States, 144 F.4th 133, 142 (2d Cir. 2025). The parties do not dispute that there is a Yates error. It is “impossible to tell” from the jury’s general verdict whether the jurors relied on the traditional fraud theory or right-to-control theory of fraud liability. Yates, 354 U.S. at 312. But the parties disagree on 11  No. 25-293 whether that error was harmless, and also dispute which standard would apply to determining harmless error on coram nobis petitions. 6 Stinn argues in favor of Chapman, 386 U.S. at 18, the standard typically applied on direct review of constitutional errors. See generally Brecht v. Abrahamson, 507 U.S. 619 (1993). Under the Chapman standard, “before a federal constitutional error can be held harmless,” the government must demonstrate that it was “harmless beyond a reasonable doubt.” Chapman, 386 U.S. at 24. The government bears the burden of proving that a Chapman error is harmless. See Brown v. Davenport, 596 U.S. 118, 126 (2022). The government argues in favor of the standard articulated in Kotteakos, 328 U.S. at 776, which governs in habeas corpus review of state convictions. See Brecht, 507 U.S. at 637. Under Kotteakos, the test is whether the error had a “substantial and injurious effect or influence in determining the jury's verdict.” Kotteakos, 328 U.S. at 776; accord Johnson, 144 F.4th at 142. Unlike Chapman, the Kotteakos burden of persuasion is for the petitioner to bear. See Brown, 596 U.S. at 126 (observing that Brecht, a case applying the Kotteakos standard, “inverted Chapman’s burden”). 7 The choice between these two standards is a question of first impression for our court. Neither the Supreme Court nor our court 6 The parties also disagree on whether the government has waived, or is estopped from, asserting harmless error review. Stinn raised that exact argument in Stinn II, which we rejected then, and thus the argument is barred by the law-of-the-case doctrine. 7 There is language in Johnson suggesting that the burden under Kotteakos is for the government to bear. See Johnson, 144 F.4th at 143 (“[U]nder both Chapman and Kotteakos the government bears the burden of persuasion.”). But the Johnson court held that the petitioner would have “prevail[ed] under either standard.” Id. Hence, we regard that observation as dictum, “not necessary to the holdings of the decisions in which they were made.” Cotto 12  No. 25-293 has “[]ever squarely stated which [of the two] standard[s] of harmlessness review applies in coram nobis.” Johnson, 144 F.4th at 142. For the reasons stated below, we hold that Kotteakos is the proper standard. First, coram nobis is more akin to habeas than to direct review. In Brecht v. Abrahamson, the Supreme Court considered whether to apply Chapman or Kotteakos in cases of harmless error raised on a habeas petition brought pursuant to 28 U.S.C. § 2254. 507 U.S. at 619. The Court opted for the latter, reasoning that “[t]he Kotteakos standard is . . . better tailored to the nature and purpose of collateral review,” including the extraordinary nature of habeas relief, and the sovereign and society’s interest in the finality of criminal sentences. Id. at 637- 38. Those same considerations apply here. As with habeas, “a petition for writ of error coram nobis is a collateral attack on a criminal conviction,” granted only where “extraordinary circumstances are present.” Foont v. United States, 93 F.3d 76, 78-79 (2d Cir. 1996) (internal quotation marks omitted). And like habeas, coram nobis petitions are filed long after direct proceedings have concluded, when the “erosion of memory and dispersion of witnesses that accompany the passage of time” have occurred. Brecht, 507 U.S. at 637 (internal quotation marks omitted). This “imbalance of costs v. Herbert, 331 F.3d 217, 250 n.20 (2d Cir. 2003). Additionally, in a later decision written by the author of Johnson, our court clarified that under “Brecht [a Kotteakos case] . . . the burden is on the petitioner to show harm.” Hernandez v. McIntosh, 146 F.4th 142, 163 (2d Cir. 2025), rev'd and remanded on other grounds sub nom. McCarthy v. Hernandez, 146 S. Ct. 1873 (2026). 13  No. 25-293 and benefits counsels in favor of [applying] the less onerous Kotteakos standard” in the coram nobis context. Id. Faced with this chain of logic, Stinn attempts to distinguish Brecht on the ground that it analyzed challenges to state convictions under section 2254—not challenges to federal convictions under section 2255. Stinn points to our opinion in Johnson, where we explained, in dicta, that “the concerns about comity, federalism, and state interests cited by [Brecht] for preferring Kotteakos in collateral review are not present . . . [when] the Petitioner asks us to overturn a federal conviction originating in this Circuit.” 144 F.4th at 143 (cited in Reply Br. at 17-18). But Brecht itself did not draw so fine a line. Instead, the Supreme Court (1) relied heavily on the “resound[ing]” “principle that collateral review is different from direct review”; (2) emphasized general principles of habeas jurisprudence (both federal and state)—including that “[r]etrying defendants whose convictions are set aside . . . imposes significant social costs”; and (3) expressly applied parts of its rationale to “federal [and] state” trial courts, both of which are responsible for “fully performing their sworn duty” in criminal proceedings. Brecht, 507 U.S. at 633–37 (internal quotation marks omitted). Brecht also invoked the importance of finality—an interest that matters just as much for federal convictions as for state ones. Id. at 635; see United States v. Frady, 456 U.S. 152, 166 (1982) (“[T]he Federal Government, no less than the States, has an interest in the finality of its criminal judgments.”). 8 8 We also note that while we have “not yet” formally “decided what harmless error standard applies in the context of a [section] 2255 proceeding” (as opposed to a section 2254 proceeding), Tavarez v. United States, 81 F.4th 234, 240 n.8 (2d Cir. 2023), we have, in practice, repeatedly turned to Kotteakos in section 2255 cases, see Colotti, 71 F.4th at 115–16 (applying Kotteakos standard); Stone, 37 F.4th at 829 (same). Our practice accords with the consensus of nearly every other circuit court that has 14  No. 25-293 Second, and reinforcing this conclusion, the consequences from the denial of coram nobis are less severe for the petitioner than those from the denial of a habeas petition. In a coram nobis petition “the only consequences still being inflicted, namely, civil disabilities and moral stigma,” are “considerably less severe than continued imprisonment.” United States v. Keogh, 440 F.2d 737, 741 (2d Cir. 1971) (Friendly, J.). It makes little sense to impose a more onerous standard for coram nobis petitions, which by definition do not involve incarcerated petitioners, than for habeas petitions. Third, the practical “unlikelihood of a new trial” after a grant of coram nobis also counsels in favor of the Kotteakos standard. United States v. Keogh, 391 F.2d 138, 148 (2d Cir. 1968). Unlike in habeas, petitioners seek coram nobis after they have already completed their sentence. Thus, although retrial is theoretically possible, the likely “consequence[] of granting a writ of coram nobis . . . would be to expunge [a sentence] without possibility of reinstatement[.]” Mandanici, 205 F.3d at 532 (Kearse, J., concurring). Before obtaining such irreversible relief, a coram nobis petitioner should be required to clear at least as rigorous a bar as would a petitioner seeking habeas relief. Finally, adopting Stinn’s proposed test would undercut the higher bar Brecht erected for habeas review, and create a perverse incentive for delay. Suppose a petitioner serving the final year of his sentence files a habeas petition that is denied under the Kotteakos standard. He then waits one year and files a coram nobis petition considered whether the Brecht/Kotteakos standard applies in the section 2255 context. See United States v. Bentley, 49 F.4th 275, 289 n.9 (3d Cir. 2022) (joining the Fourth and Eighth Circuits in extending Brecht to section 2255 cases); United States v. Smith, 723 F.3d 510, 517 (4th Cir. 2013) (same, joining the Sixth, Ninth, Tenth, and Eleventh Circuits, and observing that the Seventh Circuit is the sole outlier). 15  No. 25-293 asserting the same constitutional error, under the more-relaxed Chapman standard. Should he now gain the benefit of a more lenient standard, simply because he waited out the clock? We think not. In sum, we hold that a petitioner seeking coram nobis relief must demonstrate that the error had a “substantial and injurious effect or influence in determining the jury's verdict.” Kotteakos, 328 U.S. at 776. The burden of persuasion is on the petitioner. See Brown, 596 U.S. at 126. And to evaluate whether a petitioner has satisfied that burden, the district court must consider whether a “properly instructed jury would have found” the petitioner guilty, absent the invalid jury instruction. Colotti v. United States, 71 F.4th 102, 119 (2d Cir. 2023) (applying the “properly instructed jury” test in the habeas context); Stone v. United States, 37 F.4th 825, 832 (2d Cir. 2022) (same). In an effort to reframe this test, Stinn points to an aside in Johnson, where we mused—without citing any supporting authority—that “[t]he question” under Kotteakos “is not whether a jury could have or even would have convicted [the defendant] if presented only with a [valid] theory of fraud,” but “rather, whether [the defendant’s] jury did convict him on that basis.” Johnson, 144 F.4th at 145. But Johnson never actually relied on the distinction between what the jury did and what it would have done. On the contrary, the panel repeatedly analyzed whether the jury “would have found” the defendant guilty had it been properly instructed. 144 F.4th at 145 (emphasis added); see also id. at 143, 146. The throwaway line that Stinn quotes is thus textbook dicta. See Barclays Cap. Inc. v. Theflyonthewall.com, Inc., 650 F.3d 876, 899 (2d Cir. 2011) (“[S]ubsequent appellate panels are required to follow only . . . previous appellate legal holdings.” (internal quotation marks omitted)). 16  No. 25-293 And there is good reason to disregard that dicta here. As we have repeatedly explained—including in the context of Yates errors— “we will find [an] error harmless when the jury would have found” the defendant guilty if properly instructed. Gomez v. United States, 87 F.4th 100, 107 (2d Cir. 2023) (internal quotation marks omitted and emphasis added); Colotti, 71 F. 4th at 116 (same); Stone, 37 F.4th at 832 (same); Peck v. United States, 106 F.3d 450, 457 (2d Cir. 1997) (same). Because, as Johnson itself noted, the entire premise of a Yates error is that the reviewing court “cannot determine upon which basis the jury convicted,” 144 F.4th at 142 (emphasis added), Stinn’s reading of Johnson would, in effect, mean that a Yates error could never be harmless, since the standard itself would require the court to determine that the “jury did convict” on a legally permissible theory. Id. at 145. That is clearly an untenable position that finds no support in our caselaw. Applying Kotteakos’s “would-have-found” test here, we find that Stinn has failed to meet his burden. The government presented ample evidence that the objective of Stinn’s scheme was his $350,000 salary raise and $352,000 bonus. See supra op. n.3. The jury considered that evidence and returned not only a guilty verdict, but also a forfeiture verdict including the bonus and salary raise. See 18 U.S.C. § 981 (subjecting to forfeiture “[a]ny property, real or personal, which constitutes or is derived from proceeds traceable to a violation of” mail, wire, and securities fraud statutes). On this record, there can be no doubt that a properly instructed jury, without recourse to the right-to-control instruction, would have found the evidence of Stinn’s salary raise and bonus sufficient to convict Stinn under the traditional fraud theory. Thus, it was no abuse of discretion for the district court to deny Stinn’s coram nobis petition. 17  No. 25-293 CONCLUSION We have evaluated Stinn’s remaining arguments and find them to be without merit. For the reasons stated above, we AFFIRM the judgment of the district court. 18