RECOMMENDED FOR PUBLICATION Pursuant to Sixth Circuit I.O.P. 32.1(b) File Name: 25a0043p.06

UNITED STATES COURT OF APPEALS FOR THE SIXTH CIRCUIT

┐ UNITED STATES OF AMERICA, │ Plaintiff-Appellee, │ > No. 22-5603 │ v. │ │ CALVIN COGDILL, │ Defendant-Appellant. │ ┘

On Remand from the United States Supreme Court United States District Court for the Eastern District of Tennessee at Chattanooga. No. 1:20-cr-00087-1—Curtis L. Collier, District Judge.

Argued: October 2, 2024

Decided and Filed: March 3, 2025

Before: CLAY, GRIFFIN, and DAVIS, Circuit Judges.

_________________

COUNSEL

ARGUED: Jennifer Niles Coffin, FEDERAL DEFENDER SERVICES OF EASTERN TENNESSEE, INC., Knoxville, Tennessee, for Appellant. Luke A. McLaurin, UNITED STATES ATTORNEY’S OFFICE, Knoxville, Tennessee, for Appellee. ON SUPPLEMENTAL BRIEF: Jennifer Niles Coffin, FEDERAL DEFENDER SERVICES OF EASTERN TENNESSEE, INC., Knoxville, Tennessee, for Appellant. Luke A. McLaurin, UNITED STATES ATTORNEY’S OFFICE, Knoxville, Tennessee, for Appellee. GRIFFIN, J., delivered the opinion of the court in which DAVIS, J., concurred. CLAY, J. (pp. 12–26), delivered a separate dissenting opinion. No. 22-5603 United States v. Cogdill Page 2

_________________

OPINION _________________

GRIFFIN, Circuit Judge.

This case returns to us after the Supreme Court vacated our earlier decision and remanded for our reconsideration following Erlinger v. United States, 602 U.S. 821 (2024). After defendant Calvin Cogdill pleaded guilty to being a felon in possession of a firearm, the district court, over Cogdill’s objection, determined that Cogdill committed three prior drug offenses “on occasions different from one another,” subjecting him to an enhanced sentence under the Armed Career Criminal Act. Erlinger later clarified that it was error for the judge, instead of a jury, to make that occasions decision. Here, we must review that error for harmlessness, see United States v. Campbell, 122 F.4th 624, 629–31 (6th Cir. 2024), and we hold that the district court’s error was not harmless. We thus vacate Cogdill’s sentence and remand to the district court for further proceedings consistent with this opinion.

I.

A.

In February 2022, without a plea agreement, Cogdill pleaded guilty to one count of being a felon in possession of a firearm, in violation of 18 U.S.C. § 922(g)(1). At the time of his plea, that offense ordinarily carried a maximum penalty of 10 years’ imprisonment. 18 U.S.C. § 924(a)(2) (2018). But under the Armed Career Criminal Act (ACCA), a defendant convicted of a § 922(g) offense who “has three previous convictions . . . for a violent felony or a serious drug offense, or both, committed on occasions different from one another” is subject to a minimum sentence of 15 years’ imprisonment. 18 U.S.C. § 924(e)(1).

The Presentence Investigation Report (PSR) concluded that Cogdill was “subject to an enhanced sentence” under ACCA based on three prior offenses: (1) a Georgia conviction for manufacturing methamphetamine, committed December 18, 2003; (2) a Tennessee conviction for selling methamphetamine, committed June 12, 2014; and (3) a Tennessee conviction for No. 22-5603 United States v. Cogdill Page 3

possessing methamphetamine with intent to sell or deliver, committed September 15, 2014. The latter two convictions—those listed in PSR paragraphs 66 and 67—are central to this appeal. The ACCA designation increased Cogdill’s statutory penalty range from a maximum of 10 years to a minimum of 15 years, and increased his effective Guidelines range from 10 years, see U.S.S.G. § 5G1.1(a), to a range of 15 to 17.5 years, id. § 5G1.1(c)(2).

Cogdill objected to his ACCA designation. First, he objected “to the convictions captured at paragraphs 66 and 67 of the PSR both counting as [ACCA] predicate offenses” because “the government cannot prove those convictions occurred on occasions different from one another.” He asserted that Wooden v. United States, 595 U.S. 360 (2022), requires “a holistic, multi-factored approach to interpreting the Occasions Clause,” considering “time,” “proximity of location,” and the degree to which “the conduct giving rise to the offenses” is “similar or intertwined.” See id. at 369. Applying those factors, he argued that “his convictions in Paragraphs 66 and 67 were of a common scheme or course of conduct, were not separated by an intervening arrest, and should not count as predicates that occurred on occasions different from one another.” Second, he objected that, based on Apprendi v. New Jersey, 530 U.S. 466, 490 (2000), and Alleyne v. United States, 570 U.S. 99, 116 (2013), “[t]he Occasions Clause factual analysis must be charged in the indictment and proved to a jury beyond a reasonable doubt.”

The government responded by arguing that court records from Cogdill’s prior convictions show that he “committed the three drug-trafficking offenses not just days, but months—and in some cases years—apart from one another.” The government attached these records (which we refer to as Shepard documents1), including:

1) An indictment alleging that Cogdill “possess[ed] with intent to sell or deliver” methamphetamine “on or about the 15th day of September, 2014, in Bradley County, Tennessee.”

1 Such documents can include charging documents, plea agreements, transcripts in which the defendant confirmed the factual basis for the plea, or “comparable judicial record[s]” revealing facts about the prior offenses. See Shepard v. United States, 544 U.S. 13, 26 (2005); see also Erlinger, 602 U.S. at 839–42 (discussing use of Shepard documents in ACCA context). No. 22-5603 United States v. Cogdill Page 4

2) A January 2016 judgment for that possession-with-intent-to-sell-or-deliver offense. The judgment states the offense date was September 15, 2014, and the offense occurred in Bradley County. 3) A separate January 2016 judgment for an offense of selling/delivering methamphetamine. The judgment states the offense date was June 12, 2014, and the offense occurred in Bradley County.

The district court overruled Cogdill’s objections, finding that his three prior offenses happened on separate occasions for purposes of Wooden. Accordingly, the court sentenced Cogdill to a 180-month (15-year) mandatory-minimum term of imprisonment.

B.

Cogdill appealed, arguing that the district court erred by concluding that he committed his offenses on different occasions. Applying then-controlling precedent, see, e.g., United States v. Williams, 39 F.4th 342, 350–51 (6th Cir. 2022), we rejected his arguments and affirmed his sentence, United States v. Cogdill, 2023 WL 4030069, at *3–4 (6th Cir. June 15, 2023). Cogdill then petitioned for certiorari. While his petition was pending, the Supreme Court decided Erlinger, 602 U.S. at 835, holding that that the Sixth Amendment requires a jury to decide whether past offenses were committed on different occasions for ACCA purposes. It then vacated our decision and remanded for reconsideration in light of Erlinger. Cogdill v. United States, 144 S. Ct. 2710 (2024) (order). On remand, the parties filed supplemental briefs, and we heard oral argument.

II.

We begin with the occasions issue. First, all agree that, under Erlinger, it was error for the district judge, instead of a jury, to decide whether Cogdill committed his prior offenses on different occasions under ACCA. And second, we and the parties agree that such an error is not structural so we must review it for harmlessness. Campbell, 122 F.4th at 630–31 (holding that “the constitutional violation identified in Erlinger” is not structural and is subject to harmless- error review); see also Neder v. United States, 527 U.S. 1, 14 (1999) (explaining that we take a “categorical approach to structural errors,” under which “a constitutional error is either structural or it is not”). No. 22-5603 United States v. Cogdill Page 5

On this latter point, our colleague disagrees. He instead suggests that we should disregard Campbell because that opinion “contravenes the Supreme Court’s holding in Erlinger.” But our court’s published opinions bind us. United States v. Calvetti, 836 F.3d 654, 670 n.4 (6th Cir. 2016); United States v. Moody, 206 F.3d 609, 615 (6th Cir. 2000); see also 6 Cir. R. 32.1(b). Indeed, we are “bound to follow the prior published decisions of this court, without regard for whether they might be mistaken in their reading of the Supreme Court’s decisions.” United States v. Conces, 507 F.3d 1028, 1039 n.10 (6th Cir. 2007). True, we may revisit prior published opinions “if intervening Supreme Court caselaw requires modification,” Smith v. Cook, 956 F.3d 377, 391 n.3 (6th Cir. 2020) (emphasis added), but that is not the situation here—Campbell interpreted Erlinger, and the Supreme Court has issued no further opinions since that would permit us to disregard Campbell. Regardless of how we might have applied Erlinger were we writing on a blank slate, Campbell is binding precedent and now controls.

The main issue, then, is whether the district court’s error was harmless. See Campbell, 122 F.4th at 630–33. That is, rather than automatically remand for a jury to make the different- occasions decision, we must first make the “case-by-case” determination of whether the Erlinger error here affected the outcome. See Neder, 527 U.S. at 14; see also Campbell, 122 F.4th at 630–33. In this context, harmlessness turns on whether the government proved beyond a reasonable doubt—through “relevant and reliable information” in the record—that, absent the error, any reasonable jury would have found that Cogdill committed the prior offenses on different occasions. See Campbell, 122 F.4th at 630–33 (quoting Greer v. United States, 593 U.S. 503, 511 (2021)); see also Neder, 527 U.S. at 18.

A.

Recall that ACCA’s occasions clause asks whether the defendant “has three previous convictions . . . for a violent felony or a serious drug offense, or both, committed on occasions different from one another.” 18 U.S.C. § 924(e)(1). Within the past few years, the Supreme Court has issued two opinions clarifying that clause: Wooden, 595 U.S. at 360, and Erlinger, 602 U.S. at 821. No. 22-5603 United States v. Cogdill Page 6

First came Wooden. There, the Court explained the factors and considerations that constitute the occasions inquiry. At issue was whether the defendant’s 10 prior offenses— burglarizing 10 storage units in one building—occurred on different occasions. The defendant committed all 10 break-ins in one evening and one location, “break[ing] through wall after wall dividing 10 separate storage units to complete his crimes.” Wooden, 595 U.S. at 384 (Gorsuch, J., concurring). The Court explained that the occasions inquiry is “multi-factored in nature,” and it described the “range of circumstances [that] may be relevant to identifying episodes of criminal activity.” Id. at 369 (majority opinion). The relevant factors include:

• Timing. “Offenses committed close in time, in an uninterrupted course of conduct, will often count as part of one occasion; not so offenses separated by substantial gaps in time or significant intervening events.” Id. • Proximity of location. “[T]he further away crimes take place, the less likely they are components of the same criminal event.” Id. • Character and relationship of the offenses. “The more similar or intertwined the conduct giving rise to the offenses—the more, for example, they share a common scheme or purpose—the more apt they are to compose one occasion.” Id.

Applying these factors to the storage-unit burglaries, Wooden held that the defendant “burglarized ten storage units on a single occasion.” Id. at 370. He thus was ineligible for an ACCA-enhanced sentence. Id. at 376.

Then came Erlinger. As in Wooden, at issue was whether the defendant committed his prior offenses on different occasions. But the Court considered a more specific question: Who decides? It held that—based on Apprendi, 530 U.S. at 490, and Alleyne, 570 U.S. at 103—only a jury may decide the occasions question, so it was error for the sentencing judge to do so. Erlinger, 602 U.S. at 835. The majority was silent on whether that error was harmless, and it provided few details about the prior crimes at issue, noting only that “within a span of days Mr. Erlinger burglarized a pizza shop, a sporting goods store, and two restaurants.” Id. at 826. But the Court opined that it was not clear how a jury would have decided the multi-factored occasions question, stating: “Presented with evidence about the times, locations, purpose, and character of those crimes, a jury might have concluded that some or all occurred on different occasions. Or it might not have done so.” Id. at 835. No. 22-5603 United States v. Cogdill Page 7

B.

To prove that the Erlinger error here was harmless, the government must show, beyond a reasonable doubt, that any rational jury would have found that all three of Cogdill’s predicate offenses were “committed on occasions different from one another.” 18 U.S.C. § 924(e)(1); see Campbell, 122 F.4th at 630–33; see also Neder, 527 U.S. at 18. Cogdill does not dispute that his 2003 Georgia methamphetamine offense, committed over a decade before and in a different state from the other two offenses, was committed on a different occasion. Our focus, then, is on whether any rational jury would have found that the June and September 2014 Tennessee methamphetamine offenses—both of which he pleaded guilty to in January 2016—were committed on occasions different from one another. We cannot “confidently say, based on the whole record, that the government has shown that the constitutional error at issue here was harmless beyond a reasonable doubt.” Campbell, 122 F.4th at 631 (internal quotation marks omitted).

1.

The record by which to make this determination is scant. Under Cogdill’s plea agreement, he admitted only that he was guilty of the unenhanced § 922(g)(1) charge and that “he had sustained felony convictions including, but not limited to, the following: Possession of Methamphetamine for Resale (2016) [and] Sale/Delivery of Methamphetamine (2016).” As far as his prior offenses are concerned, his plea confirmed only that he was convicted of those two methamphetamine-related offenses in 2016; he did not admit to any details regarding the underlying conduct, like when, where, or how the offenses occurred. So, to build its case for harmlessness, the government relies on post-plea documents in the record—specifically the PSR and Shepard documents it filed—which provide the offense dates and few other details about the prior offenses. But the government’s evidence comes up short.

First consider what little evidence we have for two of the three Wooden factors. For proximity, Wooden, 595 U.S. at 369, the record reveals only that Cogdill’s offenses were committed in the same county—Bradley County, Tennessee. No documents provide addresses No. 22-5603 United States v. Cogdill Page 8

or more specific location information for the offenses. For all we know, Cogdill committed these methamphetamine offenses in the exact same place.

As for the “character and relationship of the offenses,” id., both offenses related to selling—or possessing with intent to sell—methamphetamine. No one arrested Cogdill between the June offense and the September offense, and he ultimately was convicted of both offenses on the same date. For all the record shows, “it is certainly possible”—as Cogdill’s attorney argued before the district court—that the drugs for these two offenses came from the same source. Cogdill might have obtained a large quantity of methamphetamine, sold some of that stash in June, and then “was busted” for possessing with intent to sell the rest in September. Without any other information about how these crimes took place, it is reasonable to think that these crimes may have been “similar or intertwined” and “share[d] a common scheme or purpose.” Id.

Remaining then is the sole factor on which the government relies: timing. As the government sees it, the three-month gap (June 12, 2014, to September 15, 2014) between Cogdill’s Tennessee methamphetamine offenses is decisive—no rational jury could find that those offenses took place on the same occasion. True, a three-month gap is strong evidence to tip this factor in favor of different occasions. But on its own, a gap in time between offenses— even one as long as three months—does not prove beyond a reasonable doubt how a rational jury, properly instructed under Wooden and Erlinger, would have resolved the multi-factored occasions inquiry. As Erlinger states in no uncertain terms: “no particular lapse of time . . . automatically separates a single occasion from distinct ones.” 602 U.S. at 841 (citing Wooden, 595 U.S. at 369–70).

To illustrate this point, a recent jury verdict in a similar case demonstrates that a jury, properly instructed under Wooden and Erlinger, will not necessarily decide that two offenses were committed on different occasions even when there is a weeks-long gap between the offenses. In United States v. Willis, No. 4:21-cr-548 (E.D. Mo. July 16, 2024), a jury had to determine whether the defendant’s three prior offenses—“committed on or about” July 15, 2010; August 1, 2010; and August 19, 2010—were “committed on occasions different from one another.” (Phase II Jury Instructions, Willis, ECF 217, PageID 1250.) There, as here, the government relied on Shepard documents to prove the dates and circumstances of the prior No. 22-5603 United States v. Cogdill Page 9

crimes, and there, as here, those documents revealed that the defendant was prosecuted for two of those prior offenses at the same time. (See generally Phase II Jury Trial Tr., Willis, ECF 226, PageID 1281–95 (closing arguments); Prior Conviction Documents, Willis, ECF 99-2, PageID 408–09 (Indictment, Counts I and V).) In fact, the documents there confirmed that two of the offenses—separate robberies of the same restaurant—occurred in the exact same place. (Prior Conviction Documents, Willis, ECF 99-2, PageID 408–09 (Indictment, Counts I and V); Phase II Jury Trial Tr., Willis, ECF 226, PageID 1292–95.) Again, the documents here leave that possibility open.

Tracking Wooden and Erlinger, the Willis jury instructions correctly explained that a single occasion may “encompass multiple, temporally distinct activities,” see Wooden, 595 U.S. at 367, and that no “particular lapse of time or distance between offenses automatically separates a single occasion from distinct ones,” see Erlinger, 602 U.S. at 841. (Phase II Jury Instructions, Willis, ECF 217, PageID 1251–52.) Given those instructions, the jury did not find that the defendant committed “three or more criminal offenses on different occasions”—even though evidence suggested that the offenses were committed at least two weeks apart from one another. (Phase II Jury Trial Tr., Willis, ECF 226, PageID 1299–1301.) And in the context of Cogdill’s case, where only the timing factor favors the government, that jury verdict illustrates more than “a theoretical possibility of acquittal.” Cf. Campbell, 122 F.4th at 633. It shows that a jury, properly instructed, may not necessarily treat a long gap in time between offenses as conclusive.

In sum, given the scant evidence of Cogdill’s prior offenses and how they fare under the relevant factors, a rational jury could reasonably conclude that Cogdill committed his latter two offenses “during a single criminal episode.” Erlinger, 602 U.S at 835.

2.

Indeed, contrast the government’s lack of evidence here with the “overwhelming evidence” it presented in Campbell, where we recently held that the Erlinger error was harmless. 122 F.4th at 632. There, we focused on whether any rational jury would have found that the defendant’s prior drug-related offenses—an August 1992 Tennessee drug-trafficking offense and a March 1993 Virginia drug-trafficking offense—were committed on different occasions. See id. No. 22-5603 United States v. Cogdill Page 10

at 627, 632. On that score, the “undisputed facts” for all factors favored different occasions. See id. at 632. In terms of time, those offenses were separated by over six months. Id. For proximity, the offenses occurred nearly “eight hundred miles” apart. Id. And as for character and relationship, the convictions, we stated, “were for wholly distinct crimes—possession with intent to distribute crack cocaine and trafficking in powder cocaine.” Id. Because “overwhelming evidence” on all factors “showed Campbell committed his predicate offenses ‘on occasions different from one another,’” the government succeeded in proving “beyond a reasonable doubt that a jury’s failure to consider the different-occasions question had no effect on Campbell’s sentence.” Id. (citation omitted).

The government argues that Campbell compels the same conclusion here, but we cannot agree. Contrary to the government’s suggestion, Campbell did not hold that the harmlessness inquiry turns “singularly” on “whether the offenses were committed a ‘day or more apart.’” Id. at 629 (quoting Wooden, 595 U.S. at 370). Rather, Campbell simply acknowledged the Wooden Court’s reference to a “‘day’ of separation” between crimes as a “benchmark” for the time factor. Id. at 632 (citing Wooden, 595 U.S. at 370).

To be sure, Wooden does state that “a single factor” (like time) “can decisively differentiate occasions.” 595 U.S. at 370 (emphasis added). But it does not go as far as to hold that a day’s separation between crimes is dispositive. Indeed, given its facts—one-after-another burglaries in a single evening—Wooden did not face the question of how long is too long between occasions. And if there were any doubt that Wooden preserved the possibility of an occasion spanning several days or longer, the Supreme Court has since made clear—in a case involving several crimes taking place “within a span of days,” Erlinger, 602 U.S. at 826—that “no particular lapse of time . . . automatically separates a single occasion from distinct ones,” id. at 841 (citing Wooden, 595, U.S. at 369–70).

Thus, whether the government can meet its burden of proving harmlessness here depends on whether the government can prove beyond a reasonable doubt that Wooden’s factors (not just one) favor different occasions. And here, the government did not. No. 22-5603 United States v. Cogdill Page 11

C.

Wooden and Erlinger make clear that the prior offenses’ timing, proximity, and relationship with one another all bear on whether they were committed on the same or different occasions. Here, the “whole record”—Shepard documents included2—leaves reasonable doubt about whether a jury, properly instructed under Wooden and Erlinger, would have found that Cogdill committed the prior offenses on different occasions. See Campbell, 122 F.4th at 631–34 (citation omitted). Because the government has not satisfied its harmless-error burden, we must vacate Cogdill’s sentence.

III.

Finally, in his supplemental brief, Cogdill argues that, following Erlinger, the Double Jeopardy Clause bars the government from trying him for the ACCA-enhanced version of his § 922(g)(1) offense and from punishing him as if he had pleaded guilty to it. But “[l]ike the Supreme Court, we are a court of review, not first view,” United States v. Houston, 792 F.3d 663, 669 (6th Cir. 2015), so we leave it to the district court to address this issue in the first instance, if it needs addressing at all. On remand, the government might decide not to seek an ACCA-enhanced sentence, in which case there would be no double-jeopardy question to resolve.

IV.

For these reasons, we vacate Cogdill’s sentence and remand for further proceedings consistent with this opinion.

2 Some have reasonably questioned whether we ought to consider these types of documents when evaluating harmlessness in this context, see Campbell, 122 F.4th at 635–37 (Davis, J., concurring) (critiquing use of Shepard documents), but whether we consider them here makes no difference. No. 22-5603 United States v. Cogdill Page 12

_________________

DISSENT _________________

CLAY, Circuit Judge, dissenting. When a Defendant’s right to a jury trial is violated, that infringement is so profound that it requires the highest scrutiny from this Court. Defendant Calvin Cogdill’s right to a jury trial was infringed when the district court applied the Armed Career Criminal Act (“ACCA”) mandatory minimum sentence, pursuant to 18 U.S.C. § 924(e)(1), without a jury determination. Under the Supreme Court’s precedent in Erlinger v. United States, 602 U.S. 821 (2024), that violation was so significant that it requires this Court to automatically reverse the district court. Because the majority fails to faithfully apply Erlinger, I respectfully dissent.

I

In 2022, Defendant Calvin Cogdill pleaded guilty to being a felon in possession of a firearm and ammunition in violation of 18 U.S.C. § 922(g)(1). Cogdill’s offense carried a maximum penalty of ten years imprisonment. See 18 U.S.C. § 924(a)(2) (2020) (amended 2022). But prosecutors claimed that Cogdill qualified for a sentencing enhancement under the Armed Career Criminal Act, which imposes a mandatory minimum sentence of fifteen years on § 922 defendants who have “three previous convictions . . . for a violent felony or a serious drug offense, or both, committed on occasions different from one another.” 18 U.S.C. § 924(e)(1). This requirement is often referred to as the “occasions inquiry.” Erlinger, 602 U.S. at 828.

Cogdill’s presentence report indicated that he had three prior drug-trafficking convictions that triggered ACCA’s mandatory minimum sentence: (1) a Georgia conviction for manufacturing methamphetamine, committed on December 18, 2003; (2) a Tennessee conviction for selling methamphetamine, committed on June 12, 2014; and (3) a Tennessee conviction for possessing methamphetamine with intent to sell or deliver, committed on September 15, 2014. But Cogdill objected to the presentence report, arguing that the two Tennessee convictions did not occur on “different occasions” and therefore did not trigger the ACCA mandatory minimum sentence. Further, Cogdill argued that, under Apprendi v. New Jersey, 530 U.S. 466 (2000), and No. 22-5603 United States v. Cogdill Page 13

Alleyne v. United States, 570 U.S. 99, 108 (2013), the Fifth and Sixth Amendments of the Constitution require that a jury, not a judge, determine whether the predicate offenses happened on multiple occasions such that the ACCA mandatory minimum applies.

The district court overruled Cogdill’s objections and imposed the ACCA mandatory minimum of fifteen years. Cogdill now appeals the district court’s application of the ACCA mandatory minimum.

II

In evaluating Cogdill’s case, we must first turn to the Supreme Court’s guidance in Erlinger. Erlinger considered whether the Fifth and Sixth Amendments required the ACCA’s occasions inquiry be performed by a jury. Erlinger, 602 U.S. at 834–35. The Supreme Court ultimately concluded that courts violate the jury trial right when judges, rather than juries, decide ACCA’s occasions inquiry for purposes of imposing ACCA’s mandatory minimum sentence. Id. at 849.

The Supreme Court reached its decision by conducting a two-track analysis. First, it reviewed the history of the jury trial right at common law and the founding. Second, it evaluated Supreme Court jurisprudence regarding the Sixth Amendment’s prohibition on judicial fact- finding that increases the minimum or maximum sentence a defendant may face. Id. at 835. The majority opinion observed that the Fifth and Sixth Amendments “ensure that a judge’s power to punish would derive wholly from, and remain always controlled by, the jury and its verdict,” thereby “mitigat[ing] the risk of prosecutorial overreach and misconduct.” Id. at 831–32 (cleaned up). These principles, the Erlinger majority noted, are what led the Court to hold previously that “[o]nly a jury may find ‘facts that increase the prescribed range of penalties to which a criminal defendant is exposed,’” id. at 833 (quoting Apprendi, 530 U.S. at 490) (cleaned up), and that “a fact that increases a defendant’s exposure to punishment, whether by triggering a higher maximum or minimum sentence, must be submitted to a jury and found unanimously and beyond a reasonable doubt,” id. at 833 (citing Alleyne, 570 U.S. at 111–113) (cleaned up).

Against this historical and legal backdrop, the majority found that Erlinger was controlled by Apprendi and Alleyne, and as a result, a jury must decide ACCA’s occasions No. 22-5603 United States v. Cogdill Page 14

inquiry. Id. at 835. In so holding, the majority rejected several arguments to the contrary. First, the Court disagreed with the contention that Erlinger fit within the recognized exception to Apprendi that judges may find, for sentencing purposes, the fact of a defendant’s prior conviction. Id. at 837 (citing Almendarez-Torres v. United States, 523 U.S. 224 (1998)). This exception, recognized in Almendarez-Torres, did not apply to Erlinger because “[t]o determine whether Mr. Erlinger’s prior convictions triggered ACCA’s enhanced penalties, the district court had to do more than identify his previous convictions and the legal elements required to sustain them. It had to find that those offenses occurred on at least three separate occasions. And, in doing so, the court did more than Almendarez-Torres allows.” Id. at 838–39.

The Supreme Court also rejected the idea that the sentencing judge can consult a “restricted set of materials often called Shepard documents, that include judicial records, plea agreements, and colloquies between a judge and the defendant” in order to decide the multiple occasions inquiry. Id. at 839 (internal quotation marks omitted). While the Court has previously held that a sentencing court may consult Shepard documents for purposes of determining whether a prior conviction existed, consistent with Almendarez-Torres, it may not do so for purposes that exceed this “limited function”—such as determining the time or manner in which the offense was committed, which the occasions inquiry requires. Id. at 839–40. “To sanction that practice would be to allow a sentencing court to do exactly what the Fifth and Sixth Amendments forbid.” Id. at 840.

In coming to its conclusion, the majority opinion emphasized the importance of the jury trial right, and thus the magnitude of its abridgement by this sort of judicial fact-finding. In particular, the majority noted that efficiency concerns are of no moment when such a Fifth and Sixth Amendment violation occurs. Id. at 849 (“The jury trial may have ‘never been efficient.’” (quoting Apprendi, 530 U.S., at 498, (Scalia, J., concurring)); see also id. at 842 (“There is no efficiency exception to the Fifth and Sixth Amendments.”). Nor did the majority seem moved by arguments that the occasions inquiry will often be so easy to determine that who decides will not make a difference: “Often, a defendant’s past offenses will be different enough and separated by enough time and space that there is little question he committed them on separate occasions. But none of that means a judge rather than a jury should make the call.” Id. at 842. The majority No. 22-5603 United States v. Cogdill Page 15

emphasized this point, noting that “[i]n a free society respectful of the individual, a criminal defendant enjoys the right to hold the government to the burden of proving its case beyond a reasonable doubt to a unanimous jury of his peers regardless of how overwhelming the evidence may seem to a judge.” Id. (citing Rose v. Clark, 478 U.S. 570, 578 (1986)). Most importantly for our purposes today, the Supreme Court stated that “a jury might have concluded that some or all [of the offenses] occurred on different occasions. Or it might not have done so. All we can say for certain is that the sentencing court erred in taking that decision from a jury of Mr. Erlinger’s peers.” Id. at 835.

III

In applying Erlinger to this case, the outcome seems easy to resolve at first blush. Cogdill was sentenced to a mandatory minimum of fifteen years based on a factual determination by a judge. The Supreme Court has since held that such factual determinations must be made by a jury or else the sentence runs afoul of the Fifth and Sixth Amendments. Therefore, we must vacate Cogdill’s sentence and remand his case to the district court for resentencing without the unconstitutionally-imposed mandatory minimum sentence. Yet there is an added wrinkle that we must address: whether and how harmless error review applies to cases where a sentencing judge has found a fact that the Fifth and Sixth Amendments require that only a jury determine the outcome. The majority in this case does not examine whether harmless error applies. Instead, it assumes a harmless error review based on this Court’s recent decision in United States v. Campbell, 122 F.4th 624 (6th Cir. 2024).

Campbell considered a nearly identical set of facts to this case. Like Defendant, Gerald Campbell pled guilty to being a felon in possession of a firearm. Id. at 627. And like the district court in this case, the Campbell district court applied the ACCA mandatory minimum by conducting its own different occasions analysis. Id. at 628. At the start of its analysis, the Campbell Court noted that the district court’s application of the ACCA enhancement plainly violated Erlinger. Id. at 629–30. Yet the Court also found that it must apply a harmless error review. Id. at 630–31. In finding that a harmless error review applies to Erlinger errors, the Court stated that “most constitutional errors are subject to a harmless error analysis, leaving us to ask whether the government has made it clear beyond a reasonable doubt that the outcome would No. 22-5603 United States v. Cogdill Page 16

not have been different without the constitutional violation.” Id. at 630 (internal quotation marks omitted). The Court then drew a comparison between Erlinger errors and the errors described in Apprendi and Alleyne. Id. Those two cases, according to the Campbell Court, “collectively bar a judge from increasing the prescribed range of penalties to which a criminal defendant is exposed under a preponderance of the evidence standard.” Id. The Court emphasized that these errors are functionally identical to Erlinger errors; and under the Supreme Court’s precedent in Washington v. Recuenco, 548 U.S. 212 (2006), Apprendi and Alleyne errors are subject to harmless error review. Id. at 630–31. According to the Campbell Court, this demonstrates that Erlinger errors are also subject to harmless error review. Id.

The Campbell Court also explained that Erlinger errors are not structural. The Court stated that the Supreme Court “utilizes a harmless error analysis, even when, it bears emphasizing, a jury does not consider an entire element of an offense or a sentencing enhancement altogether.” Id. at 630. It is thus “at odds with Supreme Court precedent to equate the failure to allow a jury to conduct fact finding as to a particular sentencing enhancement as on par with the wholesale denial of a jury trial right.” Id. at 630–31.

It should be noted that the Campbell Court’s analysis mirrors Justice Kavanaugh’s dissent in Erlinger. Justice Kavanaugh argued that Erlinger errors are subject to harmless error review. Erlinger, 602 U.S. at 859 (Kavanaugh, J., dissenting). He stated that “[i]n cases like this where undisputed facts establish that the defendant’s prior crimes occurred a day or more apart, it will usually be evident beyond a reasonable doubt that the failure to submit the different-occasions question to the jury had no effect on the defendant’s sentence.” Id. at 860. What is noteworthy about this dissent is that while Justice Kavanaugh explicitly argued that harmless error review should apply to Erlinger errors, the Supreme Court’s majority very clearly did not adopt such a view. Nowhere in the majority opinion does the Erlinger Court invoke or apply a harmless error review. See id. at 825–49. Therefore, both the majority opinion in this case as well as the majority in Campbell attempt to advance Justice Kavanaugh’s erroneous dissent in opposition to the Supreme Court’s holding in Erlinger.

The Court’s decision in Campbell, and the majority’s subsequent reliance on that outcome in this case, directly contravenes the Supreme Court’s holding in Erlinger. Typically, No. 22-5603 United States v. Cogdill Page 17

harmless error analysis asks us to hypothesize whether “it is ‘clear beyond a reasonable doubt that a rational jury would have found the defendant guilty despite the lack of an actual jury finding on that element.’” United States v. Kuehne, 547 F.3d 667, 681 (6th Cir. 2008) (quoting Neder v. United States, 527 U.S. 1, 3 (1999)); see also Fed. R. Crim. P. 52 (“Any error, defect, irregularity, or variance that does not affect substantial rights must be disregarded.”). We apply harmless error to “small errors or defects that have little, if any, likelihood of having changed the result of the trial,” such as the erroneous admission of certain evidence. Chapman v. California, 386 U.S. 18, 22 (1967).

Other errors, however, so undermine “certain basic, constitutional guarantees that should define the framework of any criminal trial” that they are “structural.” Weaver v. Massachusetts, 582 U.S. 286, 295 (2017). Structural errors “defy analysis by ‘harmless-error’ standards.” Arizona v. Fulminante, 499 U.S. 279, 309 (1991). Rather, these errors “require reversal without regard to the evidence in the particular case” because “some errors necessarily render a trial fundamentally unfair.” Rose, 478 U.S. at 577. And structural errors “affect[] the framework within which the trial proceeds, rather than simply [reflect] an error in the trial process itself.” Fulminante, 499 U.S. at 310. “Denial of the right to a jury verdict of guilt beyond a reasonable doubt is certainly an error of the former sort, the jury guarantee being a ‘basic protectio[n]’ whose precise effects are unmeasurable, but without which a criminal trial cannot reliably serve its function.” Sullivan v. Louisiana, 508 U.S. 275, 281 (1993) (quoting Rose, 478 U.S. at 577).

Against this backdrop, a plain reading of Erlinger leads to the conclusion that a judge- made occasions inquiry which exposes the defendant to ACCA’s mandatory minimum sentence constitutes structural error.1 Erlinger held that “the Fifth and Sixth Amendments do not tolerate the denial of” the right to have a jury find beyond a reasonable doubt a fact that increases the minimum sentence, like the occasions inquiry. 602 U.S. at 849. “The right to trial by jury reflects . . . a profound judgment about the way in which law should be enforced and justice administered. The deprivation of that right, with consequences that are necessarily unquantifiable

1 It should be noted that the majority opinion states that “we and the parties agree that such error is not structural.” Yet that position does not represent the full panel of this Court, as this dissent explicitly argues that the error is structural. No. 22-5603 United States v. Cogdill Page 18

and indeterminate, unquestionably qualifies as structural error.” Sullivan, 508 U.S. at 281–82 (cleaned up). “Where [the Fifth and Sixth Amendments] right[s] [are] altogether denied, the State cannot contend that the deprivation was harmless because the evidence established the defendant’s guilt; the error in such a case is that the wrong entity judged the defendant guilty.” Rose, 478 U.S. at 578. In other words, some rights are so fundamental to the fairness of the trial process that whether the result was correct pales in comparison to whether the right was denied. “The very premise of structural-error review is that even convictions reflecting the ‘right’ result are reversed for the sake of protecting a basic right.” Neder, 527 U.S. at 34 (Scalia, J., dissenting).

Erlinger is in line with prior Supreme Court cases that unequivocally hold that the deprivation of the jury trial right typically constitutes structural error. Erlinger repeatedly explained the Fifth and Sixth Amendments jury trial right is fundamental—and how extreme a violation it is when judges take severe sentencing decisions out of the jury’s hands. 602 U.S. at 832 (holding that the Fifth and Sixth Amendments “represent not procedural formalities but fundamental reservations of power to the American people”); id. at 849 (“[T]he right to a jury trial has always been an important part of what keeps this Nation free.” (internal quotations omitted)). Erlinger’s reasoning, coupled with Supreme Court case law, indicate that the error Erlinger describes, and which caused Cogdill at least five additional years in prison, is structural. See Sullivan, 508 U.S. at 281–82; Rose, 478 U.S. at 578.

Both Justice Kavanaugh’s Erlinger dissent and this Court’s decision in Campbell resisted this conclusion by relying heavily on the Supreme Court’s decision in Recuenco. 602 U.S. at 859 (Kavanaugh, J., dissenting); Campbell, 122 F.4th at 630. Recuenco’s reasoning was based on a previous Supreme Court case, Neder, which held that the omission of an essential element from a jury instruction constituted harmless error. Recuenco, 548 U.S. at 219–220 (citing Neder, 527 U.S. at 20). Recuenco extended Neder’s holding to sentencing factors that the government fails to charge to the jury, concluding that harmless error applies to such omissions. Id. at 220. Recuenco based its reasoning on Apprendi’s analogization of essential elements of a crime and sentencing factors. Id. Apprendi counseled that sentencing factors were just as important as criminal elements for purposes of the jury trial right. 530 U.S. at 478. “Accordingly, we have No. 22-5603 United States v. Cogdill Page 19

treated sentencing factors, like elements, as facts that have to be tried to the jury and proved beyond a reasonable doubt.” Recuenco, 548 U.S. at 220. The Recuenco majority then held that this equivalence required the Court to treat failures to charge the jury on required elements of a crime the same way it treats failures to charge the jury on sentencing factors: with harmless error analysis. Id. at 222. But Apprendi’s exhortation to treat sentencing factors as essential elements of a crime was to underscore the importance of the sentencing factors and the Sixth Amendment requirement that juries—not judges—decide them. Apprendi, 530 U.S. at 478. It distorts Apprendi’s logic—which, again, repeatedly emphasizes the necessity of a jury deciding certain sentencing factors—to hold that where a jury does not do so, such an error is harmless.2

Recuenco, however, does not control this case. In Recuenco, the defendant was charged with the required sentencing factor and both sides introduced evidence related to the factor at trial—the prosecution just failed to properly instruct the jury on that factor. Recuenco, 548 U.S. at 215. Not so in this case, where Cogdill pleaded guilty without admitting any of the necessary facts for ACCA’s mandatory minimum to apply. We need not extend Recuenco to these facts, where there is no trial, evidence, or record for the appellate court to even conduct a harmless error analysis, and where the error is clearly ill-suited for such analysis. Rose, 478 U.S. at 578 (“Harmless-error analysis thus presupposes a trial, at which the defendant, represented by counsel, may present evidence and argument before an impartial judge and jury.”).

That harmless error should not apply to Erlinger errors is particularly clear in light of the Supreme Court’s more recent statements about sentencing errors and the jury trial right in Erlinger. Erlinger echoes Apprendi’s exaltation of the importance of the jury trial right and the requirement, under that right, that a jury decide a fact that increases a defendant’s exposure to punishment. It therefore seems unlikely that the Erlinger majority anticipated that harmless error analysis would apply to its decision. First, it is impossible to ignore that the Erlinger majority, despite every opportunity, failed to mention harmless error in its opinion, instead electing to

2 Further, though Recuenco postdates Apprendi and Alleyne, it is difficult to square its reliance on Neder with the Supreme Court’s evolution in its thinking on sentencing jurisprudence, particularly in cases like Erlinger. See Freeze v. State, 827 N.E.2d 600, 605 (Ind. Ct. App. 2005) (“We believe the validity of Neder might be short- lived, in light of the seismic shift in the Supreme Court’s Sixth Amendment jurisprudence since 1999.”). No. 22-5603 United States v. Cogdill Page 20

emphasize the importance of the jury trial right and of juries finding certain sentencing factors beyond a reasonable doubt. 602 U.S. at 828, 833, 835, 842, 849.

Second, the Erlinger majority’s discussion of the jury trial right underscores that harmless error is incompatible with the type of error the Erlinger majority described. As described above, Erlinger discussed the importance of the Sixth Amendment requirement that juries find facts that increase criminal penalties explicitly without regard for judicial efficiency— the typical justification for harmless error analysis. Compare United States v. Hasting, 461 U.S. 499, 509 (1983) (“The goal [of harmless error analysis] is to conserve judicial resources by enabling appellate courts to cleanse the judicial process of prejudicial error without becoming mired in harmless error.” (citation omitted)) with Erlinger, 602 U.S. at 842 (“There is no efficiency exception to the Fifth and Sixth Amendments.”). And Erlinger’s repeated statements that the ease with which the occasions inquiry might be decided does not change the fact that a jury must determine that inquiry further counsel against the application of the harmless error analysis. Id. (“Often, a defendant’s past offenses will be different enough and separated by enough time and space that there is little question he committed them on separate occasions. But none of that means a judge rather than a jury should make the call.”).

Third and relatedly, Erlinger held that, “a criminal defendant enjoys the right to hold the government to the burden of proving its case beyond a reasonable doubt to a unanimous jury of his peers ‘regardless of how overwhelming’ the evidence may seem to a judge.” Id. (quoting Rose, 478 U.S. 570, 578 (1986)). The Court’s reliance on Rose also indicates harmless error’s inapplicability to an Erlinger error. In Rose, the Supreme Court held that harmless error analysis cannot apply where there is no “trial, at which the defendant, represented by counsel, may present evidence and argument before an impartial judge and jury.” 478 U.S. at 578. But in cases such as this one where the defendant elected to plead guilty, there has been no trial at which the defendant has presented evidence on whether his prior offenses were committed on multiple occasions. Therefore the preconditions for harmless error—a trial of a jury’s peers and a record for the reviewing court to analyze—have not been satisfied. As Rose aptly phrased it: “Where [the jury trial] right is altogether denied, the State cannot contend that the deprivation was harmless because the evidence established the defendant’s guilt; the error in such a case is No. 22-5603 United States v. Cogdill Page 21

that the wrong entity judged the defendant guilty.” Id. Erlinger effectively echoed this reasoning by noting that “[p]resented with evidence about the times, locations, purpose, and character of those crimes, a jury might have concluded that some or all occurred on different occasions. Or it might not have done so. All we can say for certain is that the sentencing court erred in taking that decision from a jury of Mr. Erlinger’s peers.” 602 U.S. at 835.

All of these reasons point in the same direction: that a judge’s determination, rather than a jury’s, of a sentencing factor that increases the minimum or maximum of a defendant’s sentence is structural error. Such an error is not subject to harmless error review. Our holding is supported not just by the text and reasoning of Erlinger, but also by logic. If the issue is who should decide a question, then it should not matter whether the answer to the question remains the same. Rather, the very act of the judge, and not the jury, deciding this question is what violates the Sixth Amendment jury trial right. Id.; see also United States v. Legins, 34 F.4th 304, 323 (4th Cir. 2022) (“[I]t is bizarre that a deprivation of the jury right, which reflects a distrust of judges to adjudicate criminal guilt, can be set aside as harmless when we judges find the result sufficiently clear.”). It is no cure to say that the jury would have reached the same result—it is enough to say that they were never given the chance. In fact, the Supreme Court has held that even conducting the harmless error analysis infringes on the Fifth and Sixth Amendments. “[T]o hypothesize a guilty verdict that was never in fact rendered—no matter how inescapable the findings to support that verdict might be—would violate the jury-trial guarantee.” Sullivan, 508 U.S. at 279. In any event, “[h]armless-error review applies only when the jury actually renders a verdict—that is, when it has found the defendant guilty of all the elements of the crime.” Neder, 527 U.S. at 38 (Scalia, J., dissenting in part). In this case, a jury did not render the decision necessary to justify Cogdill’s mandatory minimum sentence. Therefore, harmless error cannot apply.

The majority in this case argues that we must apply a harmless error analysis because we are bound by Sixth Circuit precedent. It asserts that this Circuit’s interpretation of Supreme Court precedent has to be followed by our Circuit even if that interpretation is actually at odds with the holding of the Supreme Court. Contrary to the majority’s argument, this Court is not beholden to precedent that directly contravenes Supreme Court holdings. Instead, where prior No. 22-5603 United States v. Cogdill Page 22

cases in this Circuit depart from the Supreme Court’s case law, we are bound to disregard the Circuit’s erroneous holdings and to correct the error. See, e.g., Smith v. Cook, 956 F.3d 377, 391 n.3 (6th Cir. 2020). The majority contends that its argument is favored by the fact that the Supreme Court has issued no further opinions since Campbell “that would permit us to disregard Campbell.”3 There are two flaws with this argument. First, Campbell was on decided on November 22, 2024, just three months ago, giving the Supreme Court virtually no time to correct the Campbell Court’s error. Second, Campbell directly conflicts with the Supreme Court’s holding in Erlinger. Where Circuit precedent clearly violates established Supreme Court precedent, it is our duty to disregard the Circuit’s erroneous holding, and as soon as possible, to alter the Circuit’s law to faithfully reflect the Supreme Court’s jurisprudence. The majority would have us do otherwise.

In the absence of a harmless error review, Cogdill would surely prevail in this appeal. Structural errors require an automatic reversal of the district court. See Weaver v. Massachusetts, 582 U.S. 286, 290 (2017). And considering that Erlinger errors are structural, this Court should automatically reverse the district court and require that a jury conduct the different occasions inquiry.

IV

Even if we were to adopt the majority’s approach and apply harmless error review, Cogdill would still prevail. In evaluating harmlessness, we should not ask, as the majority does, whether the jury would have reached the same result anyway. That approach is derived from the Campbell Court’s analysis.4 Campbell, 122 F.4th at 630. Campbell relied on Recuenco for applying a version of harmless error analysis typically applied to trial errors, but Recuenco does not demand such a result in this case. While Recuenco established the rule that failure to instruct

3 In support of this argument, the majority relies on two cases: United States v. Calvetti, 836 F.3d 654 (6th Cir. 2016), and United States v. Moody, 206 F.3d 609 (6th Cir. 2000). Those citations are misleading. Neither of those cases stand for the proposition that Circuit precedent can overrule Supreme Court authority, nor do they stand for the proposition that this Court is bound to follow law that contravenes the Supreme Court. 4 The Campbell Court specifically instructed that the proper analysis is to ask “whether the government has made it clear beyond a reasonable doubt that the outcome would not have been different without the constitutional violation.” Campbell, 122 F.4th at 630. No. 22-5603 United States v. Cogdill Page 23

a jury on a required sentencing element is reviewed for harmless error, that holding does not apply to this case because there were no faulty jury instructions because there was no jury or trial in the first instance. Further, Recuenco never specified what that harmless error review entails. See United States v. Guerrero-Jasso, 752 F.3d 1186, 1201 (9th Cir. 2014) (Berzon, J., concurring) (“[Recuenco] did not mandate any particular method for conducting that review.”); United States v. Lewis, 802 F.3d 449, 457 (3d Cir. 2015) (“Importantly, at no point did the Supreme Court explain what harmless-error review should consist of.”); Legins, 34 F.4th at 321 (“That top-line holding [of Recuenco] does not, of course, resolve how to perform the harmless- error analysis.”). Thus, even if we were to believe that Recuenco controls the question before us, it provided no guidance, let alone binding guidance, as to how to apply the harmless error review it required.

Further, it makes little sense to apply the more stringent harmless error analysis that asks whether the record indicates that the jury would have reached the same result in cases like Erlinger and Cogdill’s. In Owens v. Paris, this Court considered the habeas petition of a defendant who was convicted at trial and whose sentence was enhanced past his presumptive sentence, based on state law, in part by the judge-found fact that the offense was committed with “exceptional cruelty.” 932 F.3d 456, 458 (6th Cir. 2019). In performing harmless error analysis on the clear Apprendi violation, this Court held that the question was “whether the jury would have returned the same verdict absent the error.” Id. at 460.

But applying Owens to this case is akin to fitting a square peg in a round hole. For one thing, Owens actually involved a jury trial whose record the Court could analyze in determining whether the error in that case was harmless. We lack such a record in this case.5 And the Supreme Court has instructed that “[h]armless-error analysis thus presupposes a trial, at which the defendant, represented by counsel, may present evidence and argument before an impartial

5 There are further factual distinctions between Owens and this case. Owens involved judicial factfinding that enabled the district court to vary from a presumptive sentence, rather than mandated a minimum sentence. The Court therefore proceeded on the rationale that “the question whether the court would have imposed the same sentence on remand is itself incoherent in cases where—as here—the court’s factfinding liberated the court to impose the sentence that the court in fact imposed, rather than mandated that sentence.” Owens, F.3d at 460. But in this case, the district court’s factual finding required a mandatory minimum sentence, rendering Owens even further afield from the instant appeal. No. 22-5603 United States v. Cogdill Page 24

judge and jury.” Rose, 478 U.S. at 578. No such trial, evidence, or argument occurred here. Owens therefore may control cases where a defendant has been properly tried but the prosecution fails to charge the jury with a required sentencing factor, such that harmless error review applies. But given that no record exists here, Owens and its prescribed harmless error instruction to review the record and predict how the jury would have found has no bearing on this case.

As if the factual distinctions and the unsatisfied preconditions for harmless error were not enough, Erlinger seems to expressly prohibit the Owens framework. The Supreme Court held that “[p]resented with evidence about the times, locations, purpose, and character of those crimes, a jury might have concluded that some or all occurred on different occasions. Or it might not have done so. All we can say for certain is that the sentencing court erred in taking that decision from a jury of Mr. Erlinger’s peers.” Erlinger, 602 U.S. at 835. To ask whether a jury would have concluded that Cogdill’s offenses occurred on different occasions—which is what Owens would require—would be in direct contravention of Erlinger’s holding and reasoning, not to mention other Supreme Court precedent. We cannot hypothesize about what an imaginary jury, if impaneled, would have found based on the evidence produced at sentencing. See Sullivan, 508 U.S. at 279 (holding that such imaginative exercises about a verdict that was never rendered actually violate the jury trial guarantee); Rose, 478 U.S. at 578; accord United States v. Guerrero-Jasso, 752 F.3d 1186, 1197 (9th Cir. 2014) (Berzon, J., concurring) (“The defendant waived a jury trial, so there is no pre-conviction factual record at all. To create one after the fact is not only to undermine Apprendi but to broaden the defendant’s waiver, by allowing courts to make factual determinations concerning a trial that never occurred.”).

Erlinger further forbids this approach because it constrains the documents we may look at on appeal. Erlinger prevents district courts from reviewing Shepard documents—such as judicial records, plea agreements, and colloquies between a judge and the defendant—in the context of the occasions inquiry. Id. at 839–41. “To sanction that practice would be to allow a sentencing court to do exactly what the Fifth and Sixth Amendments forbid.” Id. at 840. A three-judge panel of this Court cannot do what the Supreme Court has forbidden district courts themselves from doing. Otherwise we yield the bizarre result that “[t]he remedy for a constitutional violation by a trial judge (making the determination of criminal guilt reserved to No. 22-5603 United States v. Cogdill Page 25

the jury) is a repetition of the same constitutional violation by the appellate court (making the determination of criminal guilt reserved to the jury).” Neder v. United States, 527 U.S. 1, 32 (1999) (Scalia, J., dissenting in part).

Therefore, even if a harmless error analysis applied to this case, the question we should ask is whether the sentence would remain the same without the error. See United States v. Gillis, 592 F.3d 696, 699 (6th Cir. 2009) (“There can be no harmless error unless the appellate court can determine from the record that the same sentence would be imposed on remand.”). And we would not be alone in doing so. At least one of our sister circuits has interpreted Recuenco’s harmless error command to ask whether the error “would have made no difference to the sentence.” Lewis, 802 F.3d at 456 (quoting Parker v. Dugger, 498 U.S. 308, 319 (1991)). This is in fact how the Supreme Court framed sentencing errors for years before Neder and before Recuenco’s extension of Neder to sentencing errors. See, e.g., Sochor v. Fla., 504 U.S. 527, 540 (1992) (holding that a sentencing error is harmless only if “it did not contribute to the sentence obtained” (quoting Chapman v. California, 386 U.S. 18, 24 (1967)); Sullivan, 508 U.S. at 279; Parker, 498 U.S. at 319. Recuenco’s silence on the contours of the harmless error test in this context allows us to apply the correct version of harmless error. Asking whether the sentence would remain the same without the error focuses on the constitutional violation—the impermissible judge-found fact—and properly inquires whether the result would change without that error, as harmless error analysis is meant to do.

Based on the record, it is not the case that Cogdill’s sentence would remain the same without the judge-conducted occasions inquiry. See Gillis, 592 F.3d at 699 (“For the government to carry its burden, it must demonstrate to the Court with certainty that the error at sentencing did not cause the defendant to receive a more severe sentence.”). The error is therefore not harmless. Cogdill’s offense had a statutory maximum of 10 years at the time of sentencing. After finding that Cogdill’s prior offenses occurred on multiple occasions and therefore qualified him as an armed career criminal, the district court applied ACCA’s 15-year mandatory minimum. Without the district court’s erroneous finding, Cogdill could not have been sentenced to more than 10 years. The error was therefore harmful, and certainly affected No. 22-5603 United States v. Cogdill Page 26

Cogdill’s substantial rights. We should thus remand Cogdill’s case for resentencing without the erroneous 15-year mandatory minimum.

For these reasons, I respectfully dissent.