United States v. Joseph Simms

PUBLISHED

UNITED STATES COURT OF APPEALS FOR THE FOURTH CIRCUIT

No. 15-4640

UNITED STATES OF AMERICA,

Plaintiff - Appellee,

v.

JOSEPH DECORE SIMMS,

Defendant - Appellant. ------------------------------------------

FOURTH CIRCUIT FEDERAL DEFENDER OFFICES,

Amicus Supporting Appellant.

Appeal from the United States District Court for the Eastern District of North Carolina, at Greenville. Terrence W. Boyle, Chief District Judge. (4:15-cr-00010-BO-1)

Argued: September 26, 2018 Decided: January 24, 2019

Before GREGORY, Chief Judge, and WILKINSON, NIEMEYER, MOTZ, KING, DUNCAN, AGEE, KEENAN, WYNN, DIAZ, FLOYD, THACKER, HARRIS, RICHARDSON, and QUATTLEBAUM, Circuit Judges.

Reversed and remanded by published opinion. Judge Motz wrote the opinion, in which Chief Judge Gregory and Judges King, Wynn, Diaz, Floyd, Thacker, and Harris joined. Judge Wynn wrote a concurring opinion, in which Judge Harris joined. Judge Wilkinson wrote a dissenting opinion. Judge Niemeyer wrote a dissenting opinion, in which Judges

Wilkinson, Duncan, Agee, Keenan, and Quattlebaum joined. Judge Richardson wrote a dissenting opinion, in which Judge Quattlebaum joined.

ARGUED: Dhamian Blue, BLUE LLP, Raleigh, North Carolina, for Appellant. Phillip Anthony Rubin, OFFICE OF THE UNITED STATES ATTORNEY, Raleigh, North Carolina, for Appellee. Joshua B. Carpenter, FEDERAL DEFENDERS OF WESTERN NORTH CAROLINA, INC., Asheville, North Carolina, for Amici Curiae. ON BRIEF: Brian A. Benczkowski, Assistant Attorney General, Matthew S. Miner, Deputy Assistant Attorney General, John P. Taddei, Appellate Section, Criminal Division, UNITED STATES DEPARTMENT OF JUSTICE, Washington, D.C.; John Stuart Bruce, Acting United States Attorney, Robert J. Higdon, Jr., United States Attorney, Jennifer P. May- Parker, Assistant United States Attorney, OFFICE OF THE UNITED STATES ATTORNEY, Raleigh, North Carolina, for Appellee. Anthony Martinez, Federal Public Defender, FEDERAL PUBLIC DEFENDERS OF WESTERN NORTH CAROLINA, INC., Charlotte, North Carolina; Parks N. Small, Federal Public Defender, Columbia, South Carolina; Louis C. Allen, Federal Public Defender, Greensboro, North Carolina; Fred Heblich, Interim Federal Public Defender, Roanoke, Virginia; Christian M. Capece, Federal Public Defender, Charleston, West Virginia; James Wyda, Federal Public Defender, Baltimore, Maryland; Paresh S. Patel, Assistant Federal Public Defender, Greenbelt, Maryland; G. Alan DuBois, Federal Public Defender, Raleigh, North Carolina; Geremy Kamens, Federal Public Defender, Alexandria, Virginia; Brian J. Kornbrath, Federal Public Defender, OFFICE OF THE FEDERAL PUBLIC DEFENDER, Clarksburg, West Virginia, for Amici Curiae.

DIANA GRIBBON MOTZ, Circuit Judge:

Joseph Decore Simms was convicted of brandishing a firearm in connection with a “crime of violence,” as defined in 18 U.S.C. § 924(c)(3)(B). He appeals, contending that § 924(c)(3)(B), as long understood, is unconstitutionally vague. The Government concedes this point but urges us to abandon the settled meaning of the statute and employ a new definition of “crime of violence.”

We cannot do so. Neither the statutory language nor controlling precedent offer any support for the Government’s proposed reinterpretation. Rather, the text and structure of § 924(c)(3)(B) plainly set forth a definition of “crime of violence” that fails to comport with due process. Accordingly, we reverse and remand for further proceedings consistent with this opinion.

I.

This case arises from an April 2014 conspiracy to rob a McDonald’s in Goldsboro, North Carolina. Shortly after 1:00 a.m., Simms and a co-conspirator crawled into the McDonald’s through the drive-through window; a third robber served as a lookout. When inside, Simms pointed a gun at the manager, attempted to strike another employee, and demanded money. The manager complied and opened the restaurant’s safe. After removing the contents, Simms struck the manager with the gun, threw a cash drawer at the other employee, and fled with his two co-conspirators and $1,100.

After his arrest and indictment, Simms pleaded guilty to Count I, conspiracy to commit Hobbs Act robbery in violation of 18 U.S.C. § 1951, and Count II, brandishing a *4 firearm during and in relation to a “crime of violence” — that is, the Hobbs Act conspiracy in Count I — in violation of 18 U.S.C. § 924(c)(1)(A). But at sentencing, Simms argued that his conviction under Count II was unconstitutional in light of Johnson v. United States, 135 S. Ct. 2551 (2015). He contended that Hobbs Act conspiracy was not a “crime of violence” because the definition of this term in 18 U.S.C. § 924(c)(3)(B) was unconstitutionally vague, like the similar definition of “violent felony” that the Supreme Court struck down in Johnson. The district court rejected this argument and sentenced Simms to 115 months’ incarceration on Count I and 84 months on Count II, for a total consecutive sentence of 199 months’ imprisonment.

Simms appealed, again contending that his conviction under Count II could not stand because § 924(c)(3)(B) was unconstitutional. 1 After the parties briefed and argued the appeal before a panel of this court, the Supreme Court struck down as unconstitutionally vague a statute containing language materially identical to that challenged by Simms. See Sessions v. Dimaya, 138 S. Ct. 1204, 1223 (2018). Given the exceptional importance and recurring nature of the question presented here, we agreed to rehear the case en banc. For the reasons that follow, we now reverse.

1 The Government does not contend that Simms’s plea renders him unable to challenge his § 924(c) conviction. It has expressly disavowed any reliance on the appellate waiver provision in Simms’s plea agreement. See Gov. Supp. Br. at 4 n.2.

II.

We must determine whether the definition of “crime of violence” in § 924(c)(3)(B) satisfies the requirements of due process.

In resolving this question, we first set forth the statutory framework and examine Supreme Court precedent interpreting text materially identical to that at issue here. We then address the contours of Simms’s constitutional challenge, drawing on the Supreme Court’s consideration of identical challenges to similar statutory language. Finally, we explain why, in light of the plain text and binding Supreme Court precedent, we must hold § 924(c)(3)(B) unconstitutional.

A.

Federal law, as codified at 18 U.S.C. § 924(c)(1)(A), provides that a person who uses or carries a firearm “during and in relation to any crime of violence” or who “possesses a firearm” “in furtherance of any such crime” may be convicted of both the underlying crime (here, Hobbs Act conspiracy) and the additional, distinct crime of utilizing a firearm in connection with a “crime of violence,” with the latter punishable by at least five consecutive years of imprisonment.

Section 924(c)(3) defines “crime of violence” as “an offense that is a felony” and

(A) has as an element the use, attempted use, or threatened use of physical force against the person or property of another, or (B) that by its nature, involves a substantial risk that physical force against the person or property of another may be used in the course of committing the offense.

18 U.S.C. § 924(c)(3). Courts commonly refer to § 924(c)(3)(A) as the “force clause” and to § 924(c)(3)(B), the provision at issue here, as the “residual clause.” For Simms’s *6 § 924(c) conviction to stand, his Hobbs Act conspiracy offense must constitute a “crime of violence” under one of these two definitions.

Our analysis begins with the force clause, § 924(c)(3)(A). To determine whether an offense is a crime of violence under that clause, courts use an inquiry known as the “categorical” approach. They look to whether the statutory elements of the offense necessarily require the use, attempted use, or threatened use of physical force. See, e.g., Leocal v. Ashcroft, 543 U.S. 1, 7–10 (2004) (interpreting materially identical text in 18 U.S.C. § 16(a)); United States v. McNeal, 818 F.3d 141, 151–52 (4th Cir. 2016) (interpreting § 924(c)(3)(A)). This approach is “categorical” because courts consider only the crime as defined, not the particular facts in the case. See, e.g., McNeal, 818 F.3d at 152. To be more precise, we will refer to the force clause inquiry as the elements-based categorical approach, because it begins and ends with the offense’s elements. When a statute defines an offense in a way that allows for both violent and nonviolent means of commission, that offense is not “categorically” a crime of violence under the force clause.

Simms’s offense — conspiracy to commit Hobbs Act robbery — does not categorically qualify as a crime of violence under the elements-based categorical approach, as the United States now concedes. Gov. 28(j) Letter at 1, ECF No. 44 (Oct. 19, 2016); Simms Suppl. Br. at 1. This is so because to convict a defendant of this offense, the Government must prove only that the defendant agreed with another to commit actions that, if realized, would violate the Hobbs Act. Such an agreement does not invariably require the actual, attempted, or threatened use of physical force.

Accordingly, the only way we can sustain Simms’s conviction on Count II is if his commission of Hobbs Act conspiracy constitutes a crime of violence under the residual clause — that is, if we determine that he committed a felony offense “that by its nature[] involves a substantial risk that physical force against the person or property of another may be used in the course of committing the offense.” 18 U.S.C. § 924(c)(3)(B). Interpreting a materially identical clause in another statute, the Supreme Court has directed courts to employ a categorical approach that — as with the force clause — “look[s] to the elements and the nature of the offense of conviction, rather than to the particular facts.” Leocal, 543 U.S. at 7 (interpreting 18 U.S.C. § 16(b) to “require[]” categorical analysis); see also United States v. Aragon, 983 F.2d 1306, 1312–13 (4th Cir. 1993) (“conclud[ing] that the plain language of § 16(b) mandates that the court embark upon a categorical approach”). 2

Importantly, however, the analysis applicable to the residual clause constitutes a “distinctive form” of the categorical approach. Dimaya, 138 S. Ct. at 1211 (interpreting § 16(b)). Unlike the elements-based categorical approach of the force clause, the residual clause inquiry “goes beyond deciding whether creation of risk is an element of the

2 Because § 16(b) and § 924(c)(3)(B) are materially identical, we and other courts have treated “precedent respecting one as controlling analysis of the other.” In re Hubbard, 825 F.3d 225, 230 n.3 (4th Cir. 2016); see also, e.g., Evans v. Zych, 644 F.3d 447, 452 n.2 (6th Cir. 2011) (“recogniz[ing] essentially identical nature of § 924(c) and § 16” and applying § 16(b) case law to § 924(c)(3)(B) analysis); United States v. Serafin, 562 F.3d 1105, 1108 n.4 (10th Cir. 2009) (“cases interpreting [§ 16(b)] inform our analysis” of § 924(c)(3)(B)). The Government urges us to abandon this approach and reinterpret § 924(c)(3)(B); we address its argument in Parts IV-VI of this opinion.

crime.” Johnson, 135 S. Ct. at 2557 (interpreting 18 U.S.C. § 924(e)(2)(B)(ii)). Under the residual clause, courts must use the statutory definition of an offense to imagine its “ordinary case,” and then consider whether this imagined ordinary case entails a “substantial risk” of force. Dimaya, 138 S. Ct. at 1211 (internal quotation marks omitted); see also 18 U.S.C. § 924(c)(3)(B).

We call the analysis applicable to the residual clause the ordinary-case categorical approach. This ordinary-case categorical approach is “broader than” the elements-based categorical approach applicable to the force clause, “in the sense that force need not actually be applied” in a residual clause offense. Leocal, 543 U.S. at 11.

B.

Simms contends that conspiracy to commit Hobbs Act robbery does not constitute a crime of violence under the residual clause, § 924(c)(3)(B), because that clause — like the similarly worded residual clauses analyzed in Johnson, 135 S. Ct. 2551, and Dimaya, 138 S. Ct. 1204 — is void for vagueness.

As Justice Kagan explained in Dimaya, “‘[t]he prohibition of vagueness in criminal statutes’ . . . is an ‘essential’ of due process, required by both ‘ordinary notions of fair play and the settled rules of law.’” 138 S. Ct. at 1212 (plurality opinion) (quoting Johnson, 135 S. Ct. at 2557). Vague criminal statutes “invite the exercise of arbitrary power . . . by leaving people in the dark about what the law demands and allowing prosecutors and courts to make it up.” Id. at 1223–24 (Gorsuch, J., concurring in part and concurring in the judgment). The void-for-vagueness doctrine thus ensures citizens have fair notice of prohibited conduct, guards against discriminatory enforcement of *9 ambiguous laws, and respects the foundational principle that only Congress — not the executive or the courts — is empowered to establish the bounds of proscribed conduct. Id. at 1212 (plurality opinion).

In Johnson, the Supreme Court struck down the residual clause of the Armed Career Criminal Act (“ACCA”) as void for vagueness. 135 S. Ct. at 2557. ACCA enhances the sentence of those convicted of possessing a firearm in violation of 18 U.S.C. § 922(g) if they have three or more prior convictions for a “serious drug offense” or for a “violent felony.” It defines “violent felony” as any felony that:

(i) has as an element the use, attempted use, or threatened use of physical force against the person of another; or (ii) is burglary, arson, or extortion, involves use of explosives, or otherwise involves conduct that presents a serious potential risk of physical injury to another . . . .

18 U.S.C. § 924(e)(2)(B) (emphasis added). The italicized portion is the ACCA residual clause. Johnson, 135 S. Ct. at 2555–56.

The Johnson Court explained that “[t]wo features of the [ACCA] residual clause conspire[d] to make it unconstitutionally vague.” Id. at 2557. The clause left uncertainty about both (1) “how to estimate the risk posed by a crime,” and (2) “how much risk it takes for a crime to qualify as a violent felony.” Id. at 2557–58.

The first problem arose because the residual clause “tie[d] the judicial assessment of risk to a judicially imagined ‘ordinary case’ of a crime, not to real-world facts or statutory elements,” but “offer[ed] no reliable way” for judges to ascertain what the “ordinary case” involved. Id. Regarding the second problem, the Court clarified that it was “one thing to apply an imprecise ‘serious potential risk’ standard to real-world facts,” *10 but “quite another to apply it to a judge-imagined abstraction,” as the ACCA residual clause required. Id. at 2558. The Court held that the combination of these two flaws rendered the ACCA residual clause unconstitutional: “Each of the uncertainties in the residual clause may be tolerable in isolation, but ‘their sum makes a task for us which at best could be only guesswork.’” Id. at 2560 (quoting United States v. Evans, 333 U.S. 483, 495 (1948)).

The Supreme Court reiterated this logic in Dimaya, when it held 18 U.S.C. § 16(b) unconstitutionally vague. 138 S. Ct. at 1215–16. That statute defines “crime of violence,” a term incorporated into many criminal and immigration statutes, as:

(a) an offense that has as an element the use, attempted use, or threatened use of physical force against the person or property of another, or (b) any other offense that is a felony and that, by its nature, involves a substantial risk that physical force against the person or property of another may be used in the course of committing the offense.

18 U.S.C. § 16. Section 16(b) is similarly known as the “residual clause.” Dimaya, 138 S. Ct. at 1212.

In Dimaya, the Court held that a “straightforward application” of Johnson made clear that § 16(b), like the ACCA residual clause, was void for vagueness. Id. at 1213. In doing so, the Court carefully explained that the proper ordinary-case categorical analysis did not ask “whether ‘the particular facts’ underlying a conviction posed the substantial risk that § 16(b) demands,” nor whether “the statutory elements of a crime require (or entail) the creation of such a risk in each case that the crime covers.” Id. at 1211 (quoting Leocal, 543 U.S. at 7). Rather, a majority of the Court agreed that “§ 16(b) requires a court to ask whether the ordinary case of an offense poses the *11 requisite risk.” Id. (internal quotation marks omitted); see also id. at 1235 (Roberts, C.J., dissenting) (identifying ordinary-case categorical approach as the way “courts should assess whether a particular crime ‘by its nature’ involves a risk of the use of physical force”).

As a result, § 16(b) possessed “the same two features that conspired to make ACCA’s residual clause unconstitutionally vague.” Id. at 1216 (majority opinion) (alterations and internal quotation marks omitted) (quoting Johnson, 135 S. Ct. at 2557). As with the ACCA residual clause, § 16(b) required courts to “identify a crime’s ‘ordinary case’ in order to measure the crime’s risk.” Id. at 1215. But the statute failed to provide any guidance on how to “divin[e] the conduct entailed in a crime’s ordinary case.” Id. “And § 16(b) also possesse[d] the second fatal feature of ACCA’s residual clause: uncertainty about the level of risk that ma[de] a crime ‘violent.’” Id. Thus, like the ACCA residual clause, the Court held that § 16(b) violated due process. Id. at 1223.

C.

With these precedents in mind, we turn to the question at hand: is the materially identical statute at issue here also unconstitutionally vague?

Like the statutes examined in Johnson and Dimaya, § 924(c)(3)(B) requires a court to imagine the idealized ordinary case of a crime while providing no guidance on how to do so, rendering any judicial account of the ordinary case wholly speculative. Dimaya, 138 S. Ct. at 1213–14; Johnson, 135 S. Ct. at 2557–58. After conceiving of this judicial abstraction, a court must then assess its speculation using the same vague standard of “substantial risk” as § 16(b) required in Dimaya. 138 S. Ct. at 1214. This *12 conjectural exercise suffers from the same two fundamental flaws that, in combination, rendered the statutory provisions in Johnson and Dimaya void for vagueness. Dimaya, 138 S. Ct. at 1213–16; Johnson, 137 S. Ct. at 2557–58.

Section 924(c)(3)(B) is therefore unconstitutional. Three other circuits have reached this conclusion. United States v. Davis, 903 F.3d 483, 485–86 (5th Cir. 2018) (per curiam), cert. granted, 2019 WL 98544 (U.S. Jan. 4, 2019) (No. 18-431); United States v. Eshetu, 898 F.3d 36, 37 (D.C. Cir. 2018) (per curiam), petition for reh’g en banc filed, No. 15-3020 (Aug. 31, 2018); United States v. Salas, 889 F.3d 681, 684–86 (10th Cir. 2018), petition for cert. filed, No. 18-428 (U.S. Oct. 3, 2018). 3

To understand why, consider a situation where a defendant is charged with possessing a gun in conjunction with witness tampering. See 18 U.S.C. § 1512(b); Dimaya, 138 S. Ct. at 1232 (Gorsuch, J., concurring in part and concurring in the judgment). As the Supreme Court has explained, § 924(c) “requires the prosecution to make two showings”: the commission of an underlying crime and the use of a firearm. Smith v. United States, 508 U.S. 223, 227–28 (1993); see also Rosemond v. United States,

3 Before Dimaya was decided, a panel of the Eleventh Circuit concluded that the § 924(c)(3)(B) “determination [is] more precise” than that under the ACCA residual clause, because § 924(c) requires a “‘nexus’ between the . . . firearm offense and the predicate crime of violence.” Ovalles v. United States, 861 F.3d 1257, 1265–66 (11th Cir. 2017), reh’g en banc granted, opinion vacated, 889 F.3d 1259 (11th Cir. 2018), and on reh’g en banc, 905 F.3d 1231 (11th Cir. 2018), and opinion reinstated in part, 905 F.3d 1300 (11th Cir. 2018). But this conclusion is premised on a misunderstanding of § 924(c)(3)(B). In defining “crime of violence,” the statute directs us to consider what the underlying offense entails “by its nature,” not what it entails when committed using a firearm. 18 U.S.C. § 924(c)(3)(B). Therefore, the firearm “nexus” offers no meaningful guidance on what satisfies the distinct crime of violence element. See Salas, 889 F.3d at 685–86.

572 U.S. 65, 75 (2014) (“§ 924(c) . . . punishes the temporal and relational conjunction of two separate acts, on the ground that together they pose an extreme risk of harm.”).

Thus, to evaluate whether witness tampering satisfies the “crime of violence” element of a § 924(c) violation, a court must assess whether the ordinary case of witness tampering, with or without a firearm, “involves a substantial risk [of] physical force.” 18 U.S.C. § 924(c)(3)(B). And when “divining the conduct entailed in [the] crime’s ordinary case,” the court must utilize some as-yet-unspecified method — “Surveys? Experts? Google? Gut instinct?” Dimaya, 138 S. Ct. at 1215 (majority opinion) (citing Johnson, 135 S. Ct. at 2557). The statute provides no guidance for this “abstract inquiry.” See Dimaya, 138 S. Ct. at 1215 (quoting Johnson, 135 S. Ct. at 2561). Instead, § 924(c)(3)(B) effectively requires judges to define the scope of criminal liability, and it directs them to do so using an unmoored, subjective abstraction that deprives the public of fair notice.

Just as in Dimaya, a “straightforward application” of controlling precedent to the challenged statute requires us to strike it down. Dimaya, 138 S. Ct. at 1213. The Supreme Court has made clear that where a statute requires courts to assess “both an ordinary-case requirement and an ill-defined risk threshold,” it is unconstitutional. Id. at 1223; see also Johnson, 135 S. Ct. at 2557–60. That principle controls here. Any conviction involving § 924(c)(3)(B) is subject to the same “guesswork,” “intuition,” “arbitrary enforcement,” and lack of “fair notice” that plagued both § 16(b) and the ACCA residual clause. Dimaya, 138 S. Ct. at 1223 (quoting Johnson, 135 S. Ct. at 2557–59). Because this “produces more unpredictability and arbitrariness than the Due *14 Process Clause tolerates,” id. (quoting Johnson, 135 S. Ct. at 2558), we must conclude that § 924(c)(3)(B) is unconstitutionally vague.

III.

The United States concedes that if we adhere to the ordinary-case categorical approach applied by the Supreme Court in Leocal, Johnson, and Dimaya, we must invalidate § 924(c)(3)(B). The Government, however, urges us to jettison this established interpretation and adopt a new reading of the challenged statutory language that employs a conduct-specific approach to the crime of violence analysis. This conduct-specific approach would consider the facts of each individual case, rather than the statutory definition of the underlying offense.

Before addressing the merits of the Government’s claim, we must determine whether to allow it to raise an argument that it previously abandoned. Although the Government initially argued in the alternative that a conduct-specific interpretation of § 924(c)(3)(B) was tenable, Gov. Br. at 28–31, it later expressly disclaimed this reading of § 924(c)(3)(B). The Government did so by submitting to us a written statement that “the position of the United States [is] that the categorical approach is the proper interpretation of the statute, and it is wholly unaffected by Johnson.” Gov. 28(j) Letter at 2 (Oct. 19, 2016). After the Supreme Court decided Dimaya and we ordered supplemental briefing, the Government again reversed course, deeming a conduct-specific reading of § 924(c)(3)(B) freshly controlling. Given these changing stances, *15 Simms asserts that the Government has forfeited any argument for a conduct-specific approach. 4

The Government offers two reasons why we should excuse its voluntary withdrawal of a conduct-specific interpretation of § 924(c)(3)(B). First, it argues that Dimaya represented “an intervening change in the law recognizing an issue that was not previously available.” United States v. Chittenden, 896 F.3d 633, 639 (4th Cir. 2018) (internal quotation marks omitted). At oral argument, the Government elaborated that the longstanding position of the United States that § 924(c)(3)(B) required an ordinary-case categorical approach came “more by way of assumption” than reason, and that Dimaya “caused a lot of searching in the Department [of Justice] to find the right answer.” Oral Arg. at 1:27:41, 1:29:00.

This claim does not hold water. Dimaya was not the first case to indicate that the ordinary-case categorical approach generated serious constitutional doubts. At the very least, the Government was on notice as to these vagueness concerns after Johnson was decided in 2015. Indeed, the Ninth Circuit held § 16(b) unconstitutionally vague four months later. See Dimaya v. Lynch, 803 F.3d 1110, 1120 (9th Cir. 2015), aff’d sub nom.

4 Simms frames this as “waiver,” which we have defined as “identif[ying] an issue, and then explicitly withdraw[ing] it.” United States v. Robinson, 744 F.3d 293, 298 (4th Cir. 2014) (internal quotation marks omitted). But waiver, in a technical sense, concerns a party’s relinquishment of rights before a district court; where a party raises and then knowingly withdraws a claim before the district court, there is no error for us to review. United States v. Olano, 507 U.S. 725, 733 (1993). Because the Government did not raise a conduct-specific argument before the district court, its double-turnabout before us is better treated as abandonment or forfeiture of a claim that we retain discretion to review. Id. at 733–34.

Sessions v. Dimaya, 138 S. Ct. 1204. But a full year thereafter, the Government asserted to us that “the categorical approach [was] the proper interpretation” of § 924(c)(3)(B) and that this conclusion was “wholly unaffected by Johnson.” Gov. 28(j) Letter at 2. The United States has yet to offer a coherent justification for its shifts in position, and its invocation of Dimaya offers no answer.

The Government next asserts that excusing its abandonment would not prejudice Simms, who has had a full opportunity to respond to the Government’s claims. That is certainly enough to convince us that we can look past the Government’s change of heart, e.g., United States v. Ramos-Cruz, 667 F.3d 487, 496 n.5 (4th Cir. 2012), but not necessarily that we should do so. As the Government is aware, we routinely exercise our discretion in favor of a strict reading of forfeiture. See, e.g., United States v. Brown, 742 F. App’x 742, 745 n.3 (4th Cir. 2018) (“[W]e conclude that Brown waived the argument that 18 U.S.C. § 924(c)(3)(B) (2012) is unconstitutional by failing to raise it in his opening brief.”); United States v. Khoa Dang Hoang, 737 F. App’x 136, 138 n.* (4th Cir. 2018) (deeming forfeited defendant’s challenge to validity of Miranda waiver); see also United States v. Cannon, 740 F. App’x 785, 791 n.2 (4th Cir. 2018) (deeming forfeited, at Government’s suggestion, defendant’s undeveloped sentencing challenges); United States v. Ballard, 727 F. App’x 757, 760 n.* (4th Cir. 2018) (same regarding defendant’s statutory speedy trial claim).

Nevertheless, in this case, we opt to proceed to the merits in view of the exceptional importance and recurring nature of the question presented. Particularly given *17 the markedly effective presentation by the parties and the amici before the en banc court, we see no reason to defer adjudication of the Government’s current argument.

IV.

The Government now contends that we must apply a different mode of analysis to § 924(c)(3)(B), a mode that the Supreme Court expressly rejected in both Dimaya and Johnson.

Specifically, the Government insists that the best reading of the statute — which, recall, asks whether “an offense that is a felony . . . by its nature[] involves a substantial risk that physical force . . . may be used in the course of committing the offense” — directs a court to analyze an offender’s specific conduct by diving into the facts before it instead of limiting the analysis to the offense’s ordinary case. In support of this contention, the Government argues that “the Supreme Court has been very clear” that the categorical approach is “essentially a saving construction” designed only to avoid the risk of unfairness that comes with reviewing conduct that underlies long-past convictions. Oral Arg. at 1:19:05. Alternatively, the Government claims that even if a conduct-specific approach is not the best reading of § 924(c)(3)(B), we still must adopt it to avoid striking down the statute. Three of our sister circuits have embraced the Government’s suggestion. 5

5 See United States v. Douglas, 907 F.3d 1, 16 (1st Cir. 2018), petition for cert. filed, No. 18-7331 (U.S. Jan. 7, 2019); Ovalles v. United States, 905 F.3d 1231, 1253 (11th Cir. 2018) (en banc); United States v. Barrett, 903 F.3d 166, 184 (2d Cir. 2018), (Continued)

We cannot do so. The Supreme Court did not invent the categorical approach out of whole cloth, as the Government would have us believe. While some other statutes invoking categorical analysis have been less than clear, the text and structure of § 924(c)(3)(B) unambiguously require courts to analyze the attributes of an “offense that is a felony . . . by its nature” — that is, categorically. And the Government’s comparisons to cases involving very different statutes, rather than bolstering its preference for a new mode of analysis, support adherence to the established ordinary-case categorical approach.

A.

As a preliminary matter, the Government presents a flawed historical premise. It claims that the categorical approach is nothing more than a “saving construction,” Oral Arg. at 1:19:05, “purely [a] judge-made doctrine” that was “first endorsed” less than thirty years ago, Gov. Supp. Br. at 17 (internal quotation marks omitted). Further, the Government asserts, this doctrine is grounded entirely in external considerations far afield from congressional language or intent.

petition for cert. filed, No. 18-6985 (U.S. Dec. 3, 2018). But see Davis, 903 F.3d at 485 (finding “a suggestion by a minority of justices in [Dimaya]” insufficient to overrule circuit precedent requiring ordinary-case categorical approach); Eshetu, 898 F.3d at 37 (emphasizing that “Dimaya nowise calls into question” circuit precedent requiring “a categorical approach”); Salas, 889 F.3d at 685 (holding § 924(c)(3)(B) unconstitutionally vague), reh’g and reh’g en banc denied, No. 16-2170 (May 23, 2018) (denying Government’s petition for rehearing on constitutional avoidance grounds).

This is simply not so. Although categorical analysis may be complicated, the rationale for it is simple and long-established: if Congress has conditioned a statutory penalty on commission of an offense generally — rather than on specific acts — courts must consider the crime as defined, rather than the offender’s conduct. See, e.g., Shepard v. United States, 544 U.S. 13, 19 (2005) (emphasizing Congress’s “language imposing the categorical approach” in ACCA); Leocal, 543 U.S. at 7 (highlighting Congress’s usage of “offense . . . by its nature” in § 16(b)). Such analysis became more frequent in the mid-1980s as Congress added general terms like “violent felony,” “crime of violence,” and “aggravated felony” to the United States Code, but the underlying principle is far from novel. Indeed, it has been an important part of American jurisprudence for more than a century. See, e.g., U.S. ex rel. Guarino v. Uhl, 107 F.2d 399, 400 (2d Cir. 1939) (L. Hand, J.) (explaining that “deporting officials may not consider the particular conduct for which the alien has been convicted” in determining whether a crime involved moral turpitude meriting deportation (citing U.S. ex rel. Mylius v. Uhl, 210 F. 860, 863 (2d Cir. 1914))).

Moreover, the Supreme Court has always rooted the categorical approach in the statutory language chosen by Congress and consistently defended this approach as a means of effectuating congressional intent. Thus, when analyzing ACCA four years after its passage, the Court looked first to “the language of” the statute, and only then to legislative history and practical concerns, to conclude that a categorical approach was “the only plausible interpretation of § 924(e)(2)(B)(ii).” Taylor v. United States, 495 U.S. 575, 600–02 (1990). Similarly, when defending application of the ordinary-case *20 categorical approach as it relates to ACCA’s residual clause in 2015, the Court again began with the statutory text before considering other arguments. Johnson, 135 S. Ct. at 2562.

The Supreme Court’s interpretation of ACCA in Taylor and Johnson was, of course, secondarily informed by considerations beyond the statutory text. But tellingly, the Court has deemed the text of § 16(b) — a statute far clearer than the ACCA residual clause and materially identical to the statute at issue here — so plain as to speak for itself. Thus, in 2004, when the Court first interpreted § 16(b) to require the ordinary-case categorical approach, it relied only on the text of the statute, and it did not invoke legislative history or practical concerns. Leocal, 543 U.S. at 7; accord Dimaya, 138 S. Ct. at 1204 (plurality opinion) (noting four-Justice plurality, Chief Justice, and Government all “accept[ed] that § 16(b), as long interpreted, demands a categorical approach”). The Government’s attempt to rewrite this history is utterly unpersuasive.

B.

Setting aside the origins of the categorical approach, we have reviewed § 924(c)(3)(B) on a clean slate and still find no reasonable construction of its text that supports the Government’s conduct-specific approach. The statutory structure, as well as Congress’s use of materially identical language to implement an ordinary-case categorical approach in § 16(b), render our conclusion inescapable. Whatever a judge’s personal feelings as to what does or does not constitute a crime of violence, we are bound to apply the definition that Congress has prescribed. And Congress could hardly have written a clearer call for the ordinary-case categorical approach than § 924(c)(3)(B).

The text of a statute is a court’s first and foremost guide to its meaning. See, e.g., Esquivel-Quintana v. Sessions, 137 S. Ct. 1562, 1568 (2017) (noting that when interpreting a statute, courts “begin, as always, with the text”). Here, the plain text of § 924(c)(3)(B) requires application of the ordinary-case categorical approach. The combination in § 924(c)(3)(B) of the phrase “offense that is a felony” with the qualifier “by its nature” makes Congress’s intent apparent.

As the Government itself admits, the definition of “nature” is “the basic or inherent features, character, or quality of something.” Oxford Dictionary of English 1183 (3d ed. 2010). Thus, § 924(c)(3)(B) directs courts to consider only the basic or inherent features of “an offense that is a felony.” 18 U.S.C. § 924(c)(3)(B); see also Dimaya, 138 S. Ct. at 1217–18 (plurality opinion) (explaining that phrase “by its nature” directs courts “to figure out what an offense normally . . . entails, not what happened to occur on one occasion”). Had Congress intended a conduct-specific analysis instead, “it presumably would have said so; other statutes, in other contexts, speak in just that way.” Descamps v. United States, 570 U.S. 254, 267–68 (2013); see, e.g., 18 U.S.C. § 1031(b)(2) (imposing heightened penalties for fraud offenses that “involve[] a conscious or reckless risk of serious personal injury”); 18 U.S.C. § 2332b(a)(1)(B) (punishing certain “conduct” that “creates a substantial risk of serious bodily injury”).

Moreover, we cannot adopt a reading of § 924(c)(3)(B) that renders part of the statute superfluous over one that gives effect to its “every clause and word.” United States v. Menasche, 348 U.S. 528, 538–39 (1955) (quoting Inhabitants of Montclair Twp. v. Ramsdell, 107 U.S. 147, 152 (1883)). This well-established rule against surplusage *22 further cuts against the Government’s conduct-specific reading, which would empty the phrase “by its nature” of meaning. Indeed, even under the Government’s interpretation, giving “by its nature” meaning would shift the § 924(c)(3)(B) inquiry away from conduct-specific facts and back towards a subjective consideration of that conduct’s “inherent features” — that is to say, another version of ordinary-case analysis.

The Government does not even attempt to address this problem. 6 Instead, it simply quotes a dissent in Dimaya to contend that in the context of § 924(c)(3)(B), “the phrase ‘by its nature’ is reasonably understood to ‘mean only that an offender who engages in risky conduct cannot benefit from the fortuitous fact that physical force was not actually used during his offense.’” Gov. Supp. Br. at 14 (quoting Dimaya, 138 S. Ct. at 1254 (Thomas, J., dissenting)). One of the dissenters here similarly insists that “by its nature” “broad[ens]” the scope of § 924(c)(3)(B), while another argues (seemingly to the contrary) that the phrase actually “limits the residual clause” by excluding conduct that risks force “merely incidentally or by happenstance.”

But the Dimaya Court did not embrace any of these readings of this text. And for good reason: each would still leave “by its nature” wholly superfluous. If we strike that phrase, § 924(c)(3)(B)’s remaining language would require a finding that an offense “involves a substantial risk that physical force . . . may be used in the course of committing the offense.” 18 U.S.C. § 924(c)(3)(B) (emphasis added). This text read on

6 Nor do any of the courts of appeals that have embraced a conduct-specific approach. For example, while the First Circuit found it “plausible that ‘by its nature’ refers to the real-world conduct of a particular offense,” it did not consider the surplusage problem that this interpretation creates. Douglas, 907 F.3d at 8.

its own would still encompass both actual uses of physical force and otherwise “risky” conduct in which force was not used. And the requirement that any risk be “substantial” would independently exclude “incidental” uses of force. In other words, under the Government’s conduct-specific reinterpretation, § 924(c)(3)(B) would have the same meaning without the phrase “by its nature” as it would with this phrase. This would drain the phrase “by its nature” of any effect, violating a cardinal rule of statutory construction.

The conclusion that “crime of violence” must be defined categorically is made even plainer when we consider the statutory context, as we must. See Davis v. Mich. Dep’t of Treasury, 489 U.S. 803, 809 (1989) (“[W]ords of a statute must be read in their context and with a view to their place in the overall statutory scheme.”). The Government concedes that § 924(c)(3)(A), which covers one half of the “crime of violence” definition in § 924(c), mandates a categorical approach. Gov. Supp. Br. at 4. It would stand to reason that § 924(c)(3)(B), the other half of this definition, also requires a categorical approach rather than analysis of “the facts of each defendant’s conduct.” Taylor, 495 U.S. at 601.

This conclusion is confirmed by the Supreme Court’s refusal to give statutory text variable meanings “depending on the presence or absence of constitutional concerns in each individual case,” explaining that such a “novel interpretive approach . . . would render every statute a chameleon.” Clark v. Martinez, 543 U.S. 371, 382 (2005). Under the longstanding interpretation of § 924(c)(3), the statute’s single reference to an “offense that is a felony” has a single meaning: it refers to a crime as defined by statute. Accepting the Government’s interpretation would require us instead to give this phrase *24 two contradictory meanings, depending on whether the force clause or the residual clause is in play. Specifically, the Government would have us read “offense that is a felony” to refer to an offense as defined by statute in prosecutions under the force clause, but to case-specific conduct in prosecutions under the residual clause. We refuse to so distort the statutory text.

Furthermore, § 924(c)(1)(A), which outlines the elements of a § 924(c) violation, expressly refers to “a crime of violence . . . that provides for an enhanced punishment if committed by the use of a deadly or dangerous weapon or device.” This phrasing would make no sense under a conduct-specific definition of “crime of violence,” as only statutes, not conduct-specific facts, can “provide[] for” an amount of punishment. 7

7 The Government responds by citing the phrase “during and in relation to any crime of violence,” 18 U.S.C. § 924(c)(1)(A) (emphasis added), to argue for a conduct- specific reading of “crime of violence.” The argument fails. In fact, it is not unusual to describe specific conduct or circumstances (like use of a firearm) “in relation to” a categorically defined generic offense (like a crime of violence). For example, in prosecutions under the force clause, § 924(c)(1)(A) itself refers to specific conduct (use of a firearm) “during and in relation to” a generic offense (a categorically defined crime of violence). In § 924(c) prosecutions where the predicate offense is a “drug trafficking crime” rather than a “crime of violence,” § 924(c)(1)(A) again refers to specific conduct “during and in relation to” a categorically defined crime. Congress has utilized an analogous approach in other statutes that incorporate the categorical “crime of violence” definition in § 16. See, e.g., 18 U.S.C. § 929(a)(2) (criminalizing use of armor-piercing ammunition “during and in relation to the commission of a crime of violence”); 18 U.S.C. § 1956(a)(3) (imposing criminal penalties for “conduct[ing] a financial transaction involving property . . . used to conduct or facilitate” a crime of violence); 18 U.S.C. § 25(b) (multiplying maximum sentence for adult who “intentionally uses a minor to commit a [federal] crime of violence”). And the Supreme Court has similarly interpreted a misdemeanor domestic violence statute to require a circumstance-specific analysis of whether a crime was “committed by” an offender with a “specified domestic relationship” to the victim, and a categorical analysis as to whether the crime requires force as an element. United States v. Hayes, 555 U.S. 415, 426 (2009).

Finally, the Government would have us interpret the materially identical 34-word phrase in § 924(c)(3)(B) and § 16(b) in entirely different ways. 8 This argument flies in the face of the traditional rule that “a legislative body generally uses a particular word with a consistent meaning in a given context.” Erlenbaugh v. United States, 409 U.S. 239, 243 (1972). Whatever force this interpretive presumption may have as to one “particular word,” it must carry more as applied to the 34-word phrase replicated in § 924(c)(3)(B) and § 16(b). See supra n.8. The presumption has more force still because Congress initially added “crime of violence” to § 924(c) and created § 16(b) in the same legislative enactment. 9 See Gustafson v. Alloyd Co., Inc., 513 U.S. 561, 570 (1995) (“[T]he normal rule of statutory construction” is that “identical words used in different parts of the same act are intended to have the same meaning.” (internal quotation marks omitted)); Erlenbaugh, 409 U.S. at 244 (noting that presumption of consistent meaning “certainly makes the most sense when the statutes were enacted by the same legislative body at the same time”). Thus, it is unsurprising that the Government has been unable to

8 Compare 18 U.S.C. § 924(c)(3) (“offense that is a felony and[] . . . that by its nature, involves a substantial risk that physical force against the person or property of another may be used in the course of committing the offense”), with 18 U.S.C. § 16(b) (“offense that is a felony and that, by its nature, involves a substantial risk that physical force against the person or property of another may be used in the course of committing the offense”). 9 Comprehensive Crime Control Act, Pub. L. No. 98–473, tit. II, § 1005(a), 98 Stat. 1976, 2138–39 (1984) (amending § 924(c) to reference “crime[s] of violence,” as defined in new § 16); id. § 1001, 98 Stat. at 2136 (enacting § 16); see also id. §§ 1801– 03, 98 Stat. at 2185 (enacting ACCA). Congress enacted § 924(c)(3)(B) two years later, adopting wording materially identical to § 16(b). Firearm Owners’ Protection Act, Pub. L. No. 99–308, § 104, 100 Stat. 449, 456–59 (1986).

cite even one case in which the Supreme Court or this court have interpreted two materially identical statutes differently, as it urges us to do with § 924(c)(3)(B) and § 16(b).

In sum, our de novo application of ordinary textual analysis yields a mountain of “textual evidence” that § 924(c)(3)’s residual clause — “like ACCA’s, except still more plainly” — “has no ‘plausible’ fact-based reading.” Dimaya, 138 S. Ct. at 1218 (plurality opinion) (quoting Johnson, 135 S. Ct. at 2562); see also id. at 1235 (Roberts, C.J., dissenting) (explaining that the ordinary-case categorical approach outlined in Leocal “provides a model for how courts should assess whether a particular crime ‘by its nature’ involves a risk of the use of physical force” (quoting § 16(b))). 10

C.

Resisting this conclusion, the Government contends that the phrases “offense,” “felony,” “by its nature,” “involves,” and “committing the offense” in § 924(c)(3)(B) “compel” a conduct-specific approach. Gov. Supp. Br. at 10.

Each of these terms can be susceptible to a conduct-specific analysis in isolation. See Nijhawan v. Holder, 557 U.S. 29, 33–34 (2009) (construing statute containing the words “offense” and “felony”); Hayes, 555 U.S. at 420–29 (construing statute containing the word “offense”); Taylor, 495 U.S. at 600–01 (construing statute containing the word

10 Despite the foregoing analysis, Judge Richardson maintains that we “rewrite[] history” to “claim[] that the Supreme Court has already resolved this issue.” Of course, we claim no such thing. Perhaps our colleague only means to assert that our conclusions accord with the reasoning of four Justices interpreting materially identical text mere months ago, as well as the indication of the Chief Justice; to that assertion, we readily agree.

“involves”); United States v. Price, 777 F.3d 700, 708–09 (4th Cir. 2015) (construing statute containing the terms “by its nature” and “involves”). But that hardly establishes that these five terms are susceptible to such a reading when appearing together in a single short statute. Rather, each of the cases on which the Government relies involves only fragments of the language in § 924(c)(3). Just as importantly, in none of these cases did the court actually rely on these fragments in reaching its holding. In fact, the reasoning that underlies each of the Government’s cases buttresses our conclusion that the text of § 924(c)(3)(B) requires an ordinary-case categorical approach.

Consider Nijhawan and Hayes. In both opinions, the Supreme Court interpreted statutes that paired a reference to “offense” with detailed qualifiers far too specific to refer to generic crimes. Nijhawan concerned an immigration statute that included only offenses “involv[ing] fraud or deceit in which the loss to the victim or victims exceed[ed] $10,000.” 8 U.S.C. § 1101(a)(43)(M)(i). In Hayes, the qualifier was even more specific: the relevant criminal statute only covered offenses “committed by a current or former spouse, parent, or guardian of the victim, by a person with whom the victim shares a child in common, by a person who is cohabiting with or has cohabited with the victim as a spouse, parent, or guardian, or by a person similarly situated to a spouse, parent, or guardian of the victim.” 18 U.S.C § 921(a)(33)(A)(ii).

Faced with these precise qualifiers, the Supreme Court concluded that Congress must have intended a conduct-specific approach for these statutes to have any force. See Nijhawan, 557 U.S. at 37–38 (holding that “to have any meaning at all,” qualifier “must refer to the particular circumstances in which an offender committed the crime”); Hayes, *28 555 U.S. at 427 (explaining that categorical reading of qualifier would render statute “a dead letter in some two-thirds of the States from the very moment of its enactment” (internal quotation marks omitted)). The specificity of each statute — not any talismanic use of “offense” or “involves” — was the crux of both rulings.

The “substantial risk of physical force” proviso in § 924(c)(3)(B) is a far cry from the specific qualifiers in Nijhawan and Hayes. Rather, the proviso in § 924(c)(3)(B) is materially identical to that in § 16(b), and it is similar to the force requirement in statutes that employ the elements-based categorical approach, like § 924(c)(3)(A) and § 16(a). Nijhawan and Hayes thus highlight the relative “absence of terms alluding to a crime’s circumstances” in § 924(c)(3)(B) beyond those also present in § 16(b). Dimaya, 138 S. Ct. at 1218 (plurality opinion); see also id. at 1217 (“Simple references to a ‘conviction,’ ‘felony,’ or ‘offense,’ we have stated, are ‘read naturally’ to denote the ‘crime as generally committed.’” (quoting Nijhawan, 557 U.S. at 34)). The differences between these statutes and § 924(c)(3)(B) reinforce rather than defeat a categorical reading here.

Moreover, despite seeking to rely on Nijhawan, the Government overlooks one of Nijhawan’s key teachings. For there, the Court expressly recognized that where Congress “uses similar statutory language and similar statutory structure in two adjoining provisions, it normally intends similar interpretations.” Nijhawan, 557 U.S. at 39. Here, Congress did not just use “similar” language in the “two adjoining provisions” of § 924(c)(3)(A) and § 924(c)(3)(B): it tied both definitions to the same use of the introductory phrase “offense that is a felony.” As we have already explained, the *29 Government’s reading would give that phrase a conduct-specific interpretation under one provision and a categorical interpretation under the adjoining provision.

As to Taylor, the Government’s reliance is inexplicable. It cites that case alone for the proposition that “the Supreme Court has previously relied on the absence of the word ‘involves’ as indicating that a categorical approach is required.” Gov. Supp. Br. at 13 (citing Taylor, 495 U.S. at 600). But Taylor did not rely on the absence of the word “involves” to require use of the categorical approach. The Taylor Court simply recognized that the ACCA force clause defined a “violent felony” as one that “‘has as an element’ — not any crime that, in a particular case, involves — the use or threat of force.” 495 U.S. at 600 (quoting § 924(e)(2)(B)(i)). The Government’s contrary argument is especially unconvincing given that the Supreme Court has held that the ACCA residual clause and § 16(b), both of which use the word “involves,” mandate the ordinary-case categorical approach. E.g., Johnson, 135 S. Ct. at 2561–62; Leocal, 543 U.S. at 7.

The Government’s reliance on Taylor, like its reliance on Nijhawan, also ignores a critical element of the Supreme Court’s analysis. The Taylor Court reasoned that the term “burglary” as used in ACCA “most likely refer[red] to the elements of the statute of conviction, not to the facts of each defendant’s conduct,” in part because of its proximity to the ACCA force clause. 495 U.S. at 600–01. In this case, the proximity of § 924(c)(3)(B) to the force clause in § 924(c)(3)(A) warrants a similar inference. Thus, Taylor offers no more support to the Government’s novel argument than Nijhawan or Hayes.

Finally, the Government emphasizes that we have applied a conduct-specific analysis to a provision of the Sex Offender Registration and Notification Act (“SORNA”) that uses the phrase “by its nature.” Price, 777 F.3d at 708–09. But in fact, each step of our textual analysis in that case supports a contrary result here.

In Price, we followed Nijhawan and Hayes to hold that “where a statute contains language that . . . refers to specific circumstances or conduct,” “Congress meant to allow the circumstance-specific approach’s more searching factual inquiry.” Id. at 708 (alteration in original) (internal quotation marks omitted). Here, Congress’s omission of this conduct-specific language in § 924(c)(3)(B) warrants the opposite inference. In Price, we noted that because Congress referenced “elements” in one subsection of SORNA, its contrasting terminology of “a criminal offense” in the next required a broader reading under the canon of meaningful variation. Id. at 708–09. Here, the ordinary-case categorical approach of § 924(c)(3)(B) is already broader than the elements-based approach of § 924(c)(3)(A), and both clauses refer to the same usage of the phrase “offense that is a felony.” The canon of meaningful variation thus offers the Government no support. And in Price, we recognized that because SORNA made “explicit reference to the ‘conduct’ underlying a prior offense, as well as the ‘nature’ of that conduct,” it “refer[red] to how an offense was committed — not a generic offense.” Id. at 709 (quoting Nijhawan, 557 U.S. at 37–39). Here, Congress wrote § 924(c)(3)(B) (unlike SORNA) to refer not to “conduct . . . by its nature” but “an offense that is a felony . . . by its nature.” 18 U.S.C. § 924(c)(3)(B). Price supports adherence to the *31 ordinary-case categorical approach, rather than adoption of the Government’s new construction.

To summarize, the text, structure, and context of § 924(c)(3)(B) clearly mandate use of the ordinary-case categorical approach, as do all relevant precedents. Given the Government’s concession that use of this mode of analysis renders the statute unconstitutionally vague, this dooms the Government’s defense of the statute.

V.

As the Supreme Court has repeatedly instructed, when statutory language is clear, “judicial inquiry is complete.” Conn. Nat’l Bank v. Germain, 503 U.S. 249, 254 (1992) (internal quotation marks omitted). Refusing to abide by this instruction, the Government heavily relies on considerations unrelated to the unambiguous text to support its new conduct-specific interpretation of § 924(c)(3)(B). We doubt such external considerations are relevant. But to the extent that they are, they confirm our decision to adhere to the ordinary-case categorical approach rather than adopting the Government’s new conduct-specific reinterpretation.

A.

We begin with practical considerations. The Government contends that notwithstanding the statutory text, we should apply a conduct-specific approach to § 924(c)(3)(B) because it, unlike the ACCA residual clause or some applications of § 16(b), always involves charges prosecuted contemporaneously with any underlying crimes of violence. According to the Government, this means “the categorical approach *32 serves no purpose” in the context of § 924(c)(3)(B). Gov. Supp. Br. at 20. Even setting aside the plain text of the statute, the argument fails for two fundamental reasons.

First, § 924(c)(3)(B)’s failure to require a court to assess past conduct hardly compels rejection of the categorical approach. Rather, the categorical approach applies to a number of statutes that require no assessment of past conduct. Most notably, the Government itself concedes that Congress has “require[d] a categorical approach” in defining a “crime of violence” under the force clause, § 924(c)(3)(A), the only other subsection of § 924(c)(3). Gov. Supp. Br. at 4. And the force clause, like the neighboring residual clause challenged here, applies only to contemporaneous offenses, not prior convictions.

Congress, through § 16, has also applied the categorical approach to many other statutes that similarly involve contemporaneous prosecutions. See, e.g., 18 U.S.C. § 1952(a)(2) (criminalizing travel or use of mail with intent to commit any “crime of violence”); 18 U.S.C. § 842(p)(2)(A) (prohibiting teaching, demonstration, or dissemination of information intended for use in committing a “crime of violence”); 18 U.S.C. § 3663A(c)(1)(A)(i) (mandating restitution to victims of certain “crime[s] of violence,” citing § 16); 18 U.S.C. § 5032 (permitting federal trials of juveniles in certain cases where “the offense charged is a crime of violence that is a felony”). 11

11 The Government also suggests that “Sixth Amendment concerns . . . led to the categorical approach” in ACCA and § 16(b) because conduct-specific interpretations of those statutes would require judicial fact-finding at sentencing, an issue that § 924(c)(3)(B) does not present. Gov. Supp. Br. at 20. In fact, non-sentencing applications of § 16(b) — including its incorporation into the immigration statute (Continued)

The second reason the Government’s practical argument fails is that, contrary to its suggestion otherwise, use of the categorical approach in § 924(c) does serve important purposes. Congress acted with good reasons in mandating a categorical approach here, even if the real-world benefits differ from those that arise in the context of prior convictions. Specifically, § 924(c)’s categorical “crime of violence” analysis serves the purpose of limiting the statute’s penalties to specific classes of federal crimes where the use of a firearm is especially dangerous. See Rosemond, 572 U.S. at 75 (“§ 924(c) . . . punishes the temporal and relational conjunction of two separate acts, on the ground that together they pose an extreme risk of harm.”); cf. Solem v. Helm, 463 U.S. 277, 292 (1983) (stating relevance of “gravity of the offense” to proportionality of sentence).

Congress amended § 924(c) over time to accomplish this very purpose. Before 1984, § 924(c) criminalized the use or unlawful carrying of a firearm in conjunction with “any [federal] felony.” 18 U.S.C. § 924(c) (1982). In 1984, Congress narrowed the statute to “crime[s] of violence,” defined in relevant part as they are today. Comprehensive Crime Control Act, § 1005(a), 98 Stat. at 2138–39. Two years later, Congress limited these “crimes of violence” to felonies while adding “drug trafficking crimes” to the statute’s scope. Firearm Owners’ Protection Act, § 104, 100 Stat. at 456–

reviewed in both Leocal and Dimaya — similarly present no Sixth Amendment concerns. Moreover, even more fundamental constitutional values militate against a conduct- specific approach here. For where Congress has spoken so plainly, judicially revising its work would threaten the basic structural principle of separation of powers. See, e.g., Lewis v. City of Chicago, 560 U.S. 205, 215 (2010) (“It is not for us to rewrite the statute . . . to achieve what we think Congress really intended.”).

59. Thus, Congress has amended § 924(c) to make it crystal clear that not every federal felony involving use of a gun constitutes a “crime of violence.” Rather, the modern statute requires the prosecution to make two distinct showings: first, that the defendant utilized a firearm, and second, that the defendant did so in conjunction with not just any felony, but specifically a crime of violence or drug trafficking crime. Smith, 508 U.S. at 227–28.

The Government’s interpretation of § 924(c)(3)(B) would eliminate Congress’s intentional choice to narrow the application of § 924(c) from “any felon[ies]” to only “crime[s] of violence” and “drug trafficking crime[s].” As the Government acknowledges, its reading of the statute would require juries to consider at trial whether an offender’s specific conduct involved a substantial risk of physical force. But the use, carrying, or possession of a firearm, standing alone, will always suffice to generate such a risk in a conduct-specific analysis, regardless of the nature of the underlying offense. If we were to adopt a conduct-specific approach, we would in effect judicially repeal the 1984 and 1986 congressional amendments to § 924(c).

Attempting to address this problem, the Government claims that jurors could avoid collapsing the firearm and crime of violence elements if they were instructed not to find a crime of violence based solely on the presence of a firearm. See Ovalles, 905 F.3d at 1250 n.8 (adopting Government’s proposed instruction). But this solution offers no affirmative principle to guide jury decision-making. How, exactly, are jurors to keep these two showings apart? Should they cross out facts involving the firearm? Imagine the firearm wasn’t there? Pretend it was inoperable? Instead of condemning jurors to *35 such an ill-defined inquiry, categorical analysis limits § 924(c)’s additional sanctions to the discrete and particularly serious classes of felonies selected by Congress — drug trafficking crimes and crimes of violence. 12

In sum, even if practical considerations could influence our construction of the clear text of § 924(c)(3)(B), they do not offer the Government any refuge.

B.

Having opened the door to external considerations, it is noteworthy that the Government does not ask us to examine the legislative history of § 924(c). Perhaps that is because the legislative history suggests that Congress intended a categorical approach. Recall that before 1984, § 924(c) penalized the use of a firearm in conjunction with any federal felony. A 1979 Senate bill unsuccessfully proposed amending § 924(c) to narrow its application to crimes of violence, largely as defined today. Criminal Code Reform Act of 1979, S. 1722, 96th Cong. § 1823 (1979). The Senate Judiciary Committee stated it was “doubtful” that “felonies involving the possession of narcotics with intent to distribute . . . would be considered by their nature to involve a substantial risk of the use of physical force against another.” S. Rep. No. 96–553, at 849 n.43 (1980). If § 924(c)(3)(B) required a conduct-specific analysis, this statement would make no sense,

12 Judge Wilkinson contends that we “derogat[e] the fact-finding function” of trial courts in a purported “arrogation of authority.” Far from it. We honor the difficult work of our colleagues on the trial courts by eliminating the need to waste time and energy interpreting an unconstitutional statute. Our holding rests not on a lack of respect for the excellent hard work of district judges, but on the inability of this court to rewrite an unambiguous congressional directive.

because using a firearm during and in relation to a narcotics felony would almost certainly generate a substantial risk of force.

Congress did not enact this specific bill. But it did subsequently limit § 924(c) to crimes of violence in 1984, and it defined the term “crime of violence” in § 16 at the same time. Comprehensive Crime Control Act, §§ 1001, 1005(a), 98 Stat. at 2136, 2138– 39. The Senate Judiciary Committee explained that it intended the revised statute to include “all persons who commit Federal crimes of violence, including those crimes set forth in statutes which already provide for enhanced sentences for their commission with a dangerous weapon.” S. Rep. No. 98–225, at 313 (1983) (emphasis added). The Committee additionally noted that the “crime of violence” limitation “expand[ed] the scope of predicate offenses . . . by including some violent misdemeanors, but restrict[ed] it by excluding non-violent felonies.” Id. at 313 n.9 (emphasis added). Though not dispositive, the Committee’s language again supports a categorical interpretation.

But, like practical considerations, quotes from committee reports cannot and do not control our reading of § 924(c)(3)(B). We note, however, that the Government has produced nothing from the legislative record to support its conduct-specific interpretation. And if Congress truly intended the materially identical text in § 16(b) and § 924(c)(3)(B) to embody such divergent approaches, we would expect to find some explanation of that divergence. Instead, the legislative history of § 924(c)(3)(B) offers the Government no assistance.

C.

The bedrock doctrine of stare decisis further weakens the Government’s position. Adherence to precedent is the preferred course for courts “because it promotes the evenhanded, predictable, and consistent development of legal principles, fosters reliance on judicial decisions, and contributes to the actual and perceived integrity of the judicial process.” Payne v. Tennessee, 501 U.S. 808, 827 (1991).

Of course, stare decisis “is not an inexorable command.” Id. at 828. But precedent exerts a particularly “powerful” pull in the context of “settled statutory meaning” because “Congress remains free to alter what [the courts] have done.” Shepard, 544 U.S. at 23 (internal quotation marks omitted).

Here, the considered circuit consensus, pre-Dimaya, that § 924(c)(3)(B) requires an ordinary-case categorical approach strongly favors adherence to such an interpretation. 13 “In this case, time has enhanced even the usual precedential force,”

13 See, e.g., United States v. Taylor, 814 F.3d 340, 377 (6th Cir. 2016) (“Section 924(c)(3)(B) . . . does not allow a court to consider risk-related conduct beyond that which is an element of the predicate crime,” because “[t]he phrase ‘by its nature’ indicates that a court’s analysis . . . is confined to the offense itself.”); United States v. Fuertes, 805 F.3d 485, 498–99 (4th Cir. 2015) (applying categorical approach under § 924(c)(3)(B)); United States v. McGuire, 706 F.3d 1333, 1336–37 (11th Cir. 2013) (“We employ this categorical approach because of the statute’s terms: It asks whether [the defendant] committed ‘an offense’ that . . . ‘by its nature, involves a substantial risk that physical force against the person or property of another may be used.’”), overruled in Ovalles, 905 F.3d at 1253; United States v. Acosta, 470 F.3d 132, 134–37 (2d Cir. 2006) (applying categorical approach under § 924(c)(3)(B)), overruled in Barrett, 903 F.3d at 178; United States v. Munro, 394 F.3d 865, 870–71 (10th Cir. 2005) (explaining that § 924(c)(3)(B) determination is a “legal conclusion” and applying categorical analysis); United States v. Jennings, 195 F.3d 795, 797–98 (5th Cir. 1999) (stating that use of “by its nature” in § 924(c)(3)(B) “requires courts to determine whether an offense constitutes (Continued)

Shepard, 544 U.S. at 23, because the federal courts of appeals have interpreted § 924(c)(3)(B) to preclude conduct-specific analysis for decades. And although Congress amended § 924(c) repeatedly as this consensus built, it never materially changed the definition of “crime of violence” after 1986. See, e.g., Pub. L. No. 105–386, § 1, 112 Stat. 3469, 3469 (1998) (adding “possess[ing] a firearm” “in furtherance of” specified crimes to § 924(c)); Pub. L. No. 109–92, § 6, 119 Stat. 2095, 2102 (2005) (creating § 924(c)(5) to impose additional penalties for armor-piercing ammunition).

Circuit case law (even if unanimous, as here) carries less weight than Supreme Court jurisprudence, and we are hesitant to over-analyze congressional inaction. But we would be remiss to discount the fact that between at least 1994 and April 2018, every § 924(c) conviction of which we are aware relied on precedents treating crime of violence determinations under § 924(c)(3)(B) (not to mention those under § 16(b)) 14 as questions

a crime of violence without examining the underlying facts surrounding the conviction” (internal quotation marks omitted)); United States v. Kennedy, 133 F.3d 53, 56–57 (D.C Cir. 1998) (under § 924(c)(3)(B), “the question is whether the crime with which [appellant] was charged constituted a . . . crime of violence” under a “categorical approach” (internal quotation marks omitted)); United States v. Amparo, 68 F.3d 1222, 1225–26 (9th Cir. 1995) (“[A]pplication of a categorical approach to section 924(c)(3)(B) is required by our cases.”); United States v. Moore, 38 F.3d 977, 979 (8th Cir. 1994) (emphasizing reference to “by its nature” in § 924(c)(3)(B) and holding that “[t]o determine the nature of a crime requires an examination of the elements which compose it, . . . not . . . an exploration of the underlying facts”), abrogation on other grounds recognized, United States v. Torres-Villalobos, 487 F.3d 607, 616 (8th Cir. 2007). 14 See, e.g., Jobson v. Ashcroft, 326 F.3d 367, 372 (2d Cir. 2003) (deeming the categorical approach “not only consistent with precedent and sound policy,” but “also necessary in view of the language of” § 16(b) (internal quotation marks omitted)); Aragon, 983 F.2d at 1313 (“[T]he plain language of § 16(b) mandates that the court (Continued)

of law requiring categorical analysis. Stability, consistency, and coherence therefore counsel against the Government’s rejection of this approach.

D.

Finally, a trio of remaining considerations further confirm our adherence to the ordinary-case categorical approach or otherwise offer no support to the Government.

Although we have chosen not to treat the Government’s withdrawal of its conduct-specific interpretation as abandonment, it remains notable that the United States embraced an ordinary-case categorical reading of this language for years, and even disclaimed a conduct-specific reading after Johnson. In Dimaya, for example, the Government “accept[ed]” that materially identical text in § 16(b), as incorporated into the Immigration and Nationality Act, “demand[ed] a categorical approach.” Dimaya, 138 S. Ct. at 1216. It did so even though a conduct-specific inquiry in that case, which involved a deportation determination, raised none of the Sixth Amendment concerns implicated in the context of criminal sentencing. The Government was right to take this position: a majority of Justices in both Johnson and Dimaya accepted this reading and declined to reinterpret the ACCA residual clause and § 16(b) to employ a conduct-specific approach. Dimaya, 138 S. Ct. at 1216–18 (plurality opinion); id. at 1235 (Roberts, C.J., dissenting) (citing Leocal, 543 U.S. at 10–11); Johnson, 135 S. Ct. at 2561–62. Regardless of the

embark upon a categorical approach to determine whether a particular crime, ‘by its nature,’ qualifies as a ‘crime of violence.’”).

Government’s rationale, its most recent change of position can only be read as a latebreaking effort to resuscitate a twice-rejected argument.

Moreover, if the Government were correct and the text at issue here were truly so ambiguous as to lend itself to two very different forms of analysis — an ordinary-case categorical approach in § 16(b) and a conduct-specific approach in § 924(c)(3)(B) — the rule of lenity might nonetheless yield the same result that we reach now. For in the case at hand, this doctrine would tilt in favor of Simms. See Hayes, 555 U.S. at 436 (Roberts, C.J., dissenting) (describing statute that was ambiguous about categorical approach as “textbook case for application of the rule of lenity”).

Further, we are not persuaded by the Government’s contention that “the strong societal interest in finality,” particularly as it pertains to guilty pleas, favors saving § 924(c)(3)(B) rather than striking it down. Lee v. United States, 137 S. Ct. 1958, 1967 (2017). Under the unusual circumstances presented here, finality cuts both ways. The parties agree that the longstanding interpretation of § 924(c)(3)(B) is unconstitutional. Accordingly, past convictions employing the ordinary-case categorical approach will be called into question regardless of whether we invalidate the statute or adopt a new conduct-specific reading that all rejected before Dimaya. If concerns about opening the floodgates are relevant to our interpretation of § 924(c)(3)(B) — and we doubt that they are — they do not favor either reading.

VI.

Perhaps recognizing that its textual analysis of § 924(c)(3)(B) is untenable and its reliance on external considerations unconvincing, the Government urges us to embrace its new reading of the statute as a matter of constitutional avoidance. We will not “lightly ascribe . . . an unconstitutional intent” to Congress. Ward v. Dixie Nat’l Life Ins. Co., 595 F.3d 164, 177 (4th Cir. 2010). But given the clear language of § 924(c)(3)(B), we lack the power to avoid its constitutional infirmity.

We are “obligated to construe [a] statute to avoid [constitutional] problems.” INS v. St. Cyr, 533 U.S. 289, 300 (2001). However, we may do this only if such a reading is “fairly possible,” id. (internal quotation marks omitted), “after the application of ordinary textual analysis,” Jennings v. Rodriguez, 138 S. Ct. 830, 842 (2018) (internal quotation marks omitted). Where, as here, there is an “absence of more than one plausible construction,” the canon of constitutional avoidance “simply has no application.” Jennings, 138 S. Ct. at 842 (internal quotation marks omitted).

This limitation is an important one. As the Supreme Court recently explained, “[s]potting a constitutional issue does not give a court the authority to rewrite a statute as it pleases.” Id. at 843. Rather, constitutional avoidance serves the “basic democratic function of maintaining a set of statutes that reflect, rather than distort, the policy choices that elected representatives have made.” Almendarez-Torres v. United States, 523 U.S. 224, 238 (1998). The doctrine is thus “a means of giving effect to congressional intent, not of subverting it.” Clark, 543 U.S. at 382. In the words of the Chief Justice, “rewrit[ing] a law to conform it to constitutional requirements . . . would constitute a *42 serious invasion of the legislative domain.” United States v. Stevens, 559 U.S. 460, 481 (2010) (alterations and internal quotation marks omitted).

In this regard, the seemingly disparate doctrines of vagueness and constitutional avoidance unite. Both demand respect for the distinct functions that Congress and the judiciary fulfill in our constitutional republic. Due process requires Congress to speak in definite terms, particularly where the consequences for individual liberties are steep, in part because Congress alone — not the executive or the judiciary — is equipped to balance competing policy priorities and to define the boundaries of criminal law. For similar reasons, although courts must interpret statutes under the presumption that legislators do not intend to violate the Constitution, judges cannot revise invalid statutes. To the contrary, while the grave remedy of striking down a statute as unconstitutional lies within the judicial province, rewriting it is a task solely for the elected legislature.

These principles compel us to reject the Government’s constitutional avoidance argument. Given the text and context of § 924(c)(3)(B), accepting the Government’s new interpretation would amount to judicially rewriting the statute.

The United States does not merely seek to narrow an ambiguous statute, as the Supreme Court did in Skilling v. United States, 561 U.S. 358, 404 (2010), and Zadvydas v. Davis, 533 U.S. 678, 682 (2001) — the latter being “a notably generous application of the constitutional-avoidance canon.” Jennings, 138 S. Ct. at 843. Here, the Government asks us to go “much further,” id., by adopting a conduct-specific reading of § 924(c)(3)(B) that directly conflicts with how courts and the United States itself have thoughtfully interpreted this statute (and materially identical text in § 16(b)) since its *43 enactment three decades ago. Tellingly, the Government has yet to identify any case in which the Supreme Court has done anything comparable in the name of constitutional avoidance. 15

We cannot usurp the legislative role to edit out the constitutional flaw in § 924(c)(3)(B). The plain text, structure, and context of § 924(c)(3) are more than enough to convince us that its residual clause has no plausible conduct-specific interpretation. We therefore refuse to rewrite the statute, just as the Supreme Court refused to rewrite § 16(b) and the ACCA residual clause, even in the face of vigorous dissents urging the employment of constitutional avoidance. See Dimaya, 138 S. Ct. at 1216–18 (plurality opinion); id. at 1232–33 (Gorsuch, J., concurring in part and concurring in the judgment); Johnson, 135 S. Ct. at 2561–62.

VII.

We are mindful of the consequences of holding § 924(c)(3)(B) unconstitutional. The provision is part of a widely used criminal statute enacted by Congress, and like § 16(b), it has been incorporated into other parts of the federal criminal code. See 18

15 Judge Richardson’s puzzling invocation of Justice Kavanaugh’s scholarship only underscores this point. Far from endorsing the unprecedented application of constitutional avoidance that Judge Richardson champions, then-Judge Kavanaugh warned against overreliance on this canon and others triggered by a threshold finding of ambiguity. Brett Kavanaugh, Fixing Statutory Interpretation, 129 HARV. L. REV. 2118, 2144 (2016) (book review). Thus, he proposed “jettison[ing]” constitutional avoidance and instead allowing judges to first “determine the best reading of the statute” and then, “[i]f that reading turn[s] out to be unconstitutional,” “say as much.” Id. at 2148. Of course, we lack the power to follow this approach and, in any event, a faithful employment of the constitutional avoidance canon compels the result we reach.

U.S.C. §§ 844(o), 1028(b)(3)(B). We do, however, note that our ruling is limited in important respects.

First, § 924(c)(3)(B) appears to be the last federal statute that directs courts to impose penalties based on a drifting ordinary-case categorical inquiry and an indeterminate risk threshold. The Government has not pointed to any other statute that our decision places in jeopardy, and there is no colorable argument that the elements-based categorical approach of § 924(c)(3)(A) suffers from any similar indeterminacy. Section 924(c)(3)(B) is, in other words, the last Johnson domino to fall. In striking it down, we leave intact the balance of the definition of “crime of violence” and the entirety of the definition of “drug trafficking crime” in § 924(c).

Second, nothing in our holding restricts the broad discretion of district judges to make case-by-case sentencing determinations. Here, for example, the district court was authorized to sentence Simms to as much as 240 months’ incarceration on his conviction for Hobbs Act conspiracy alone — that is, more than three years longer than the total sentence it actually imposed on both counts. Although judges must consult the advisory Sentencing Guidelines, they may vary from these recommendations in light of the factors listed in 18 U.S.C. § 3553(a). See, e.g., Peugh v. United States, 569 U.S. 530, 536 (2013). Foremost among these is “the nature and circumstances of the offense,” which will necessarily include whether the offender used a firearm. 18 U.S.C. § 3553(a)(1).

Last, Congress is fully equipped to revise the statute here. As Justice Gorsuch noted in Dimaya:

Vagueness doctrine represents a procedural, not a substantive, demand. It does not forbid the legislature from acting toward any end it wishes, but only requires it to act with enough clarity that reasonable people can know what it is required of them and judges can apply the law consistent with their limited office.

Dimaya, 138 S. Ct. at 1233 (Gorsuch, J., concurring in part and concurring in the judgment). These insights apply here. If Congress deems it appropriate to replace § 924(c)(3)(B) with a statute that achieves similar aims with sufficiently definite terms, the Due Process Clause poses no obstacle.

VIII.

For the foregoing reasons, the judgment of the district court is

REVERSED AND REMANDED.

WYNN, Circuit Judge, with whom HARRIS, Circuit Judge, joins, concurring:

In this case we must decide whether to jettison the “categorical” approach for determining whether an offense is a “crime of violence” for the purpose of 18 U.S.C. § 924(c)(3)(B)—an approach that the executive and judicial branches uniformly have embraced for several decades, and which the government advanced to this court as recently as October 2016 1—and replace it with a “case-specific” approach. The dissenting opinions argue that, in refusing to abandon the categorical approach, the majority opinion—with which I concur in full—is at odds with the doctrine of constitutional avoidance. E.g., Post at 85 (Richardson, J.). In particular, my dissenting colleagues maintain that constitutional avoidance compels adoption of the case-specific approach in order to “reflect, rather than distort, the policy choices that the elected representatives have made.” Post at 84 (Niemeyer, J.) (quoting Almendarez-Torres v. United States, 523 U.S. 224, 238 (1998)).

Yet it is my dissenting colleagues’ misguided application of the doctrine of constitutional avoidance—not the majority opinion’s adherence to the long-standing understanding of Section 924(c)(3)(B)—that has the potential to “distort” the Framers’ carefully crafted allocation of powers for defining crimes and punishments. Relying on

1 Gov. 28(j) letter at 2, ECF No. 44 (Oct. 19, 2016); see also, e.g., United States v. Prickett, 839 F.3d 697 (8th Cir. 2016); United States v. Taylor, 814 F.3d 340, 378, 397 n.19 (6th Cir. 2016); United States v. Fuertes, 805 F.3d 485, 498 (4th Cir. 2015); United States v. McGuire, 706 F.3d 1333, 1336-37 (11th Cir. 2013); United States v. Serafin, 562 F.3d 1105, 1107 (10th Cir. 2009); United States v. Jennings, 195 F.3d 795, 797-98 (5th Cir. 1999); United States v. Kennedy, 133 F.3d 53, 56-57 (D.C. Cir. 1998); United States v. Amparo, 68 F.3d 1222, 1224 (9th Cir. 1995).

constitutional avoidance as a basis for replacing the categorical approach with a case-specific approach, as my dissenting colleagues suggest, would broaden the universe of defendants subject to Section 924(c)(3)(B). But, to date, the Supreme Court has applied the doctrine only to narrow a criminal statute’s breadth and thereby forestall constitutional concerns. By relying on constitutional avoidance to expand a criminal statute’s reach, my dissenting colleagues embrace an unprecedented application of the doctrine of constitutional avoidance that empowers the judiciary to usurp Congress’s exclusive authority to establish crimes and punishments. I write separately to explain why I do not believe the Constitution permits judicial encroachment into such a well-defined legislative province.

I.

Section 924(c)(1)(A) provides that a person who “uses or carries a firearm” “during and in relation to any crime of violence” or who “possesses a firearm” “in furtherance of any such crime” may be convicted of both the underlying crime and the additional crime of using a firearm in connection with a “crime of violence.” Section 924(c)(3)(B) defines “crime of violence” as a felony offense “that by its nature, involves a substantial risk that physical force against the person or property of another may be used in the course of committing the offense.” Accordingly, to convict a defendant of violating Section 924(c), a jury must find both that the defendant possessed or used a firearm and that the defendant did so in connection with a “crime of violence.”

As the majority explains, prior to the Supreme Court’s decision in Sessions v. Dimaya, 138 S. Ct. 1204 (2018), federal courts—at the government’s urging and in *48 accordance with the Supreme Court’s construction of virtually identical language in other statutes, see, e.g., Taylor v. United States, 495 U.S. 595, 601 (1990)—universally held that courts must apply the ordinary-case categorical approach in determining whether a predicate offense constitutes a crime of violence for purposes of Section 924(c)(3)(B). Under the ordinary-case categorical approach, a court must determine whether the hypothetical “ordinary case” of an offense—i.e., not under the particular facts giving rise to the defendant’s prosecution, Dimaya, 138 S. Ct. at 1211—“involves a substantial risk that physical force against the person or property of another may be used in the course of committing the offense.” By contrast, under the case-specific, or “circumstance-specific,” approach now advanced by the government, and embraced by my colleagues in dissent, a jury would decide whether the specific conduct giving rise to the defendant’s prosecution involves a “substantial risk that physical force against the person or property of another may be used.” See United States v. Price, 777 F.3d 700, 704 (4th Cir. 2015).

If this Court were to replace the ordinary-case categorical approach with the case-specific approach, then offenses that courts have found do not constitute crimes of violence under the ordinary-case categorical approach will amount to crimes of violence under the case-specific approach. The Ninth Circuit has held, for example, that involuntary manslaughter, 18 U.S.C. § 1112, is not categorically a crime of violence because it “requires a mental state of only gross negligence.” United States v. Benally, 843 F.3d 350, 354 (9th Cir. 2016). Yet it takes little effort to imagine specific cases of involuntary manslaughter—particularly cases involving guns, which must be present to implicate Section 924(c)(3)(B)—that a jury would find constitute a “crime of violence” *49 under the case-specific approach. In Benally, for example, the defendant shot and killed the victim with a rifle. Id. at 352. At trial, the government presented that the “shooting was accidental and part of a drunken game.” Id. at 352. Accordingly, the Benally defendant, who was not amenable to prosecution under Section 924(c)(3)(B) under the ordinary-case categorical approach, would face a high probability of conviction under the case-specific approach because a jury would likely find that the defendant’s conduct posed “a substantial risk that physical force against the person . . . of another may be used,” notwithstanding that the crime requires “a mental state of only gross negligence.” Id. at 354.

Benally demonstrates that if this Court abandons use of the ordinary-case categorical approach to determine whether an offense constitutes a “crime of violence” for purposes of Section 924(c)(3)(B) and adopts the case-specific approach the government belatedly advances—and our dissenting colleagues embrace—then the universe of defendants subject to prosecution under Section 924(c)(3)(B) will be substantially greater. Put differently, the adoption of the case-specific approach would expand the reach of Section 924(c)(3)(B) to a new class of offenders—namely those offenders who commit offenses that do not “ordinar[ily]” pose a “substantial risk” of application of physical force against another, but which pose such a risk under the *50 specific factual circumstances of the offender’s case (perhaps because the offender was carrying a gun when he committed the offense). 2

II.

Because a shift to the case-specific approach will broaden the class of offenders subject to prosecution under Section 924(c)(3)(B), I believe constitutional avoidance is an improper basis for a court to choose the case-specific approach over the ordinary-case categorical approach. As the majority opinion correctly explains, the canon of

2 One of my dissenting colleagues posits in a separate opinion that the ordinary- case approach “is not inherently ‘narrower’ than the case-specific approach” because “a judicially-imagined ‘ordinary case’ can ensnare a defendant whose actual conduct created no risk of violence, a result barred by the case-specific approach.” Post, at 100–01 n.6 (Richardson, J.). Whatever the merits of that contention in other contexts, it makes little sense in the context of Section 924(c)(1)(A).

Put simply, my dissenting colleague’s argument rests on the premise that a jury will find that a defendant did not, as a matter of fact, engage in conduct that posed a substantial risk of violence, even though the defendant necessarily (1) committed a crime that “ordinar[il]y” creates a substantial risk of violence and (2) committed that crime while carrying a firearm—“an article that is typically and characteristically dangerous,” McLaughlin v. United States, 476 U.S. 16, 17 (1986), and therefore “inherently violent,” Pelissero v. Thompson, 470 F.3d 442, 447 (4th Cir. 1999). The prospect that a jury would render such a verdict is, indeed, vanishingly small. Cf. Smith v. United States, 508 U.S. 223, 240 (1993) (“We therefore see no reason why Congress would have intended courts and juries applying § 924(c)(1) to draw a fine metaphysical distinction between a gun’s role in a drug offense as a weapon and its role as an item of barter; it creates a grave possibility of violence and death in either capacity.”).

Rather, it is far more likely that a jury would find that an offense that is not “ordinar[il]y” a crime of violence amounts to a crime of violence when, in a particular case, the defendant committed the offense while carrying a firearm. In other words, the number of additional defendants convicted of violating Section 924(c)(1)(A) under the case-specific approach would likely substantially exceed the number of defendants—if any—who would be relieved of liability under that approach.

constitutional avoidance is a “tool for choosing between competing plausible interpretations of a statutory text, resting on the reasonable presumption that Congress did not intend the alternative which raises serious constitutional doubts.” Clark v. Martinez, 543 U.S. 371, 381 (2005); ante at 41–43.

In the context of constitutional challenges to criminal statutes, the Supreme Court and lower courts have applied the doctrine of constitutional avoidance to adopt a “narrow[ing] constru[ction]” that avoids constitutional concerns that would arise were the statute construed more broadly. Skilling v. United States, 561 U.S. 358, 407 n.40, 409 n.43 (2010) (emphasis added) (quoting United States v. Lanier, 520 U.S. 259, 257–68 n.6 (1997)). For example, in Skilling, the Supreme Court considered a vagueness challenge to the honest-services wire fraud statute. Id. at 402–03. Recognizing that the defendant’s “vagueness challenge ha[d] force,” the Court applied the doctrine of constitutional avoidance to “par[e] down” the statute to “only [its] bribe-and-kickback core.” Id. at 409 & n.43. In reaching this conclusion, the Supreme Court emphasized that, in view of the statute’s legislative history, as well as consistent judicial interpretation, there was “no doubt that Congress intended § 1346 to reach at least bribes and kickbacks.” Id. at 405, 408–09 (emphasis in original). Put differently, the Supreme Court applied the doctrine of constitutional avoidance to narrow the statute’s scope to a range of conduct that Congress unambiguously intended to render unlawful.

Skilling recognized that “cases ‘paring down’ federal statutes to avoid constitutional shoals are legion.” Id. at 406 n.40, 409 n.43 (emphasis added) (collecting cases). By contrast, the government has not identified a single case—nor have I—in *52 which the Supreme Court has applied the doctrine of constitutional avoidance to expand the reach of a criminal statute, as my colleagues in dissent suggest we do in construing Section 924(c)(3)(B).

That no court appears to have applied constitutional avoidance to construe a statute to “proscribe a wider range of offensive conduct,” id. at 408 (emphasis added), is unsurprising. The Constitution vests the “power of punishment . . . in the legislative, not in the judicial department.” Dowling v. United States, 473 U.S. 207, 214 (1985) (quoting United States v. Wiltberger, 5 Wheat. 76, 95 (1820)). Indeed, it “is the legislature, not the Court, which is to define a crime, and ordain its punishment.” Id.

The Supreme Court has held that applying the doctrine of constitutional avoidance to “par[e] down” the scope of a constitutionally suspect criminal statute does not amount to usurpation of Congress’s singular authority to define crimes and punishments. See Skilling, 561 U.S. at 409 n.43 (explaining that the Court does not usurp Congress’s authority to define crimes and punishment when it “par[es] down federal statutes to avoid constitutional shoals” because, in such circumstances, “the Court does not legislate, but instead respects the legislature, by preserving a statute through a limiting interpretation”). But see id. at 422 (Scalia, J., dissenting) (“I know of no precedent for such ‘paring down,’ and it seems to me clearly beyond judicial power.”). But the Supreme Court has sanctioned the use of constitutional avoidance to narrow a potentially unconstitutional criminal statute’s scope only when there is “no doubt that Congress intended [the statute] to reach” the conduct proscribed by the limiting construction. Id. at 408 (majority op.) (emphasis added). In Skilling, for example, the Court’s limiting *53 construction preserved the honest-services fraud statute’s “solid core.” Id. Put differently, a court may apply constitutional avoidance to narrow the scope of a constitutionally suspect statute without infringing on Congress’s exclusive authority to define crimes and punishments because the conduct proscribed by the limiting construction necessarily falls within the scope of the conduct Congress intended to proscribe.

The same cannot be said when a court applies the doctrine of constitutional avoidance to expand a criminal statute’s reach. That is because the doctrine of constitutional avoidance presupposes that there is at least some doubt as to the scope of conduct a statute proscribes—to apply the doctrine there must be at least two “competing plausible interpretations of a statutory text.” See Clark, 543 U.S. at 381. Accordingly, applying the doctrine of constitutional avoidance to adopt the more expansive of “competing plausible interpretations” of a criminal statute’s reach—as my colleagues in dissent would have us do here—entails the judiciary holding unlawful conduct for which there necessarily is some doubt as to whether Congress intended to make the conduct a crime. The Supreme Court, however, has long recognized that the judiciary lacks “the power to define new federal crimes,” Skilling, 561 U.S.at 415 (Scalia, J., dissenting) (citing United States v. Hudson, 7 Cranch 32, 34 (1812))—precisely what would occur if a judicial construction of a criminal statute encompasses conduct Congress did not intend to make unlawful.

That certainly is the case here. Given (1) the Supreme Court’s long-settled construction of several key terms in Section 924(c)(3)(B) as a compelling application of *54 the ordinary-case categorical approach, (2) the government’s decades-long position that the language of Section 924(c)(3)(B) demands application of the ordinary-case categorical approach, and (3) lower federal courts’ unanimous determination, prior to Dimaya, that the ordinary-case categorical approach applies, there is, at the very least, a strong possibility—indeed, likelihood, as the majority opinion persuasively explains— that Congress did not intend for the case-specific approach to apply. Accordingly, my dissenting colleagues’ novel application of the doctrine of constitutional avoidance to expand Section 924(c)(3)(B)’s reach by requiring application of the case-specific approach would “arrogat[e],” post at 62 (Wilkinson, J.), to the judiciary the authority “to define [a] new federal crime,” Skilling, 561 U.S at 415 (Scalia, J., dissenting).

Not only does applying the doctrine of constitutional avoidance to expand the reach of a criminal statute conflict with Congress’s exclusive authority to define crimes and punishments, but it also stands in tension with the rule of lenity, which requires that “any ambiguity over the . . . scope [of a criminal statute] be resolved in [a criminal defendant’s] favor.” Crandon v. United States, 494 U.S. 152, 168 (1990). When a court resorts to the doctrine of constitutional avoidance to construe a statute—as my dissenting colleagues would have us do here—it necessarily does so because there is some “ambiguity” as to the statute’s reach. Applying the doctrine of constitutional avoidance to narrow a criminal statute’s scope advances the interests served by the rule of lenity because it resolves the statutory ambiguity in the criminal defendant’s favor.

By contrast, applying constitutional avoidance to widen a statute’s reach fails to keep faith with the rule of lenity because it resolves a statutory ambiguity in a manner *55 contrary to the interests of criminal defendants. Indeed, were the judiciary to rely on constitutional avoidance to interpret a statute’s breadth in a manner that extends beyond what the text “clearly warrants”—as necessarily occurs when a court adopts a broad reading of an ambiguous statute—it would violate the due process principle of “fair warning” undergirding the rule of lenity. United States v. Hayes, 555 U.S. 415, 436–37 (2009) (Roberts, C.J., dissenting) (quoting Crandon, 494 U.S. at 160).

III.

My good colleague in dissent asserts that it is the majority’s refusal to abandon the ordinary-case categorical approach—rather than the dissenting opinions’ unprecedented application of constitutional avoidance to usurp Congress’s exclusive authority to define crimes and punishments—that amounts to an “arrogation of authority . . . all too prevalent in appellate ranks throughout the country.” Post at 62 (Wilkinson, J.). As my colleague sees it, this Court’s decision to continue to construe Section 924(c)(3)(B) as requiring use of the ordinary-case categorical approach reflects a “troubling trend” of “appellate micro-management”—that by applying the ordinary-case categorical approach this Court is “displac[ing]” the historical roles of “trial courts, juries, and indeed Congress.” Post at 60, 63-64, 67 (Wilkinson, J.).

Whatever the merits of my dissenting colleague’s concern about the “appellate regency” subverting the traditional role of trial judges, juries, and Congress in criminal adjudication, post at 65 (Wilkinson, J.), such concerns have no place in this case. There is no question that this Court’s resolution keeps faith with the Supreme Court’s long-standing conclusion that when Congress used identical terms in other statutes, it intended *56 for the categorical approach to apply, see Taylor, 495 U.S. at 601, and with this Court’s conclusion that those terms demand application of the categorical approach in present-offense, as well as past-offense, cases, see Fuertes, 805 F.3d at 498; United States v. Aragon, 983 F.2d 1306, 1312 (4th Cir. 1993). There also can be no question that this Court’s resolution of this case adheres to the—until recently unquestioned—decades-long judicial understanding of Section 924(c)(3)(B). See supra note 1. And there is no question that this Court’s resolution of this case mirrors the construction of Section 924(c)(3)(B) that the government advanced from the statute’s adoption until Dimaya.

Accordingly, this Court’s resolution of this case in no way breaks with the historical understanding of Section 924(c)(3)(B) and the roles of juries and trial courts in applying that statute. Rather, it is the government’s post-Dimaya about-face as to the proper construction of Section 924(c)(3)(B), and my dissenting colleagues’ embrace of that construction, that turns away from the historical understanding and application of that statute. Likewise, my dissenting colleagues’ proposed expansion of the doctrine of constitutional avoidance in a manner that empowers the judiciary to “trench upon the legislature’s [exclusive] powers” to define crimes and punishments, Salinas v. United States, 522 U.S. 52, 60–61 (1997), is the only evidence of a “troubling trend” of “appellate micro-management” that this case provides.

My colleagues in dissent rejoin that surely Congress intended to punish individuals like Defendant—and, possibly, like the defendant in Benally—who committed, under the specific facts of their case, what are undisputedly “violent” crimes, as that term is used colloquially. Post at 63-64 (Wilkinson, J.); post at 85 (Richardson, *57 J.). And they are right. That is why Congress has proscribed conspiracy to commit Hobbs Act robbery and involuntary manslaughter, and why Congress has imposed substantial criminal penalties for individuals who commit those crimes. But that is not the question we must address. Rather, we are tasked with deciding whether Congress intended to hold Defendant criminally liable for the additional offense of using a firearm in commission of an “offense” that “by its nature, involves a substantial risk that physical force against the person or property of another may be used.” § 924(c)(3)(B) (emphases added). And by expressly choosing to render unlawful only those “offenses” that “by [their] nature” involve application of violent force, Congress sought to limit the universe of defendants prosecutable under Section 924(c)(3)(B) to those defendants who commit crimes the ordinary manifestation of which “involve[] a substantial risk that physical force against the person or property of another may be used.”

As the majority opinion ably explains, the Supreme Court long has held that when Congress uses that specific language, it intends for courts to apply the ordinary-case categorical approach. And in accordance with the Supreme Court’s construction of that language, the government long has argued—including earlier in this case—that Congress intended for courts to apply the ordinary-case categorical approach in determining whether a defendant is amenable to prosecution under Section 924(c)(3)(B). Put simply, until Dimaya, the judiciary and the government agreed that Congress intended for Section 924(c)(3)(B) to require application of the ordinary-case categorical approach, and therefore that Congress did not contemplate prosecuting individuals like Defendant under *58 that provision. Since Congress enacted Section 924(c)(3)(B) in 1986, Congress never has disturbed that settled understanding.

Of course, Congress is free to amend Section 924(c)(3)(B) to permit further punishment of Defendant under that provision. But this Court is not free to expand the statute’s reach through an unprecedented application of the doctrine of constitutional avoidance, 3 see Clark, 543 U.S. at 382 (2005) (“[Constitutional avoidance] is thus a means of giving effect to congressional intent, not of subverting it.”), particularly in light of the Supreme Court’s settled construction of the meaning of Section 924(c)(3)(B)’s key terms, see Rodriguez de Quijas v. Shearson/Am. Express, Inc., 490 U.S. 477, 484 (1989)

3 Another one of my dissenting colleagues claims in his separate opinion that I “seek[] to replace the doctrine of constitutional avoidance with a search and destroy mission to strike down federal statutes.” Post at 66 (Wilkinson, J.). Not true. My analysis leaves ample room for courts to rely on constitutional avoidance to adopt narrowing constructions of criminal statutes some applications of which pose constitutional concerns—that simply is not how my dissenting colleagues propose to apply the doctrine here. My dissenting colleague further asserts that my understanding of constitutional avoidance “embrace[s] the odd proposition that upholding a statute is somehow a usurpation of Congress’ authority.” Post at 66. Again, not true. I am not the first to recognize that—even when a limiting, as opposed to a widening, construction is proposed to save a potentially unconstitutional statute—courts do not “have the power, in order to uphold an enactment, to rewrite it.” Skilling, 561 U.S. at 423 (Scalia, J., dissenting). That is particularly true given that, in the decades leading up to Dimaya, neither the government nor any court believed the case-specific approach applied. Cf. id. (explaining that “[i]f it were a ‘fairly possible’ or ‘reasonable’ construction, not ‘contrary to the intent of Congress’”—and therefore amenable to application of the doctrine of constitutional avoidance—“one would think that some court would have adopted it”). In such circumstances, judicial “restraint” demands “resist[ing] the temptation to make all things right with the stroke of [the judicial] pen”—as my colleagues in dissent would have this Court do. Id. at 423–24.

(‘‘[T]he Court of Appeals should . . . leav[e] to th[e Supreme] Court the prerogative of overruling its own decisions.’’).

IV.

In sum, neither my dissenting colleagues nor the government points to a single case in which the Supreme Court has sanctioned the use of constitutional avoidance in a manner that expands the scope of a criminal statute, as it would if we applied the case-specific approach to Section 924(c)(3)(B). And, for decades, courts—at the government’s express invitation—have applied the ordinary-case categorical approach in accordance with the settled understanding of that statute’s text. Constitutional avoidance is “not a license for the judiciary to rewrite language enacted by the legislature.” Salinas, 522 U.S. at 60–61. But that is precisely what would occur if this Court followed my dissenting colleagues’ suggestion and relied on constitutional avoidance to expand Section 924(c)(3)(B)’s reach. Accordingly, any justification for the case-specific approach grounded in constitutional avoidance radically departs from the established scope and application of the doctrine. *60 WILKINSON, Circuit Judge, dissenting:

Once upon a time, now seemingly a geologic age ago, the federal judiciary appeared sold on the inherent advantages that trial courts and trial juries bring to fact-finding in our criminal justice system. No longer. My colleagues in the majority ably demonstrate that application of the categorical approach to 18 U.S.C. § 924(c)(3)(B) saddles that statute with a fatal constitutional infirmity. My colleagues in dissent—whom I join—ably demonstrate why that infirmity need not exist; the better reading of the statute avoids it by applying the case-specific approach in place of the categorical. I write separately to further explain how application of the categorical approach here is part of a troubling trend: the gratuitous conversion of issues of fact into questions of law; the usurpation of authority by appellate courts and the resultant atrophy of trial courts’ fact-finding function.

****

Appellate courts have taken the Sentencing Guidelines and the categorical approach to crimes of violence as an invitation to wade deep into sentencing. This despite the faith that Gall v. United States placed in district judges with respect to the sentencing function. 552 U.S. 38 (2007). There is no mistaking Gall’s thrust: district judges are “in a superior position to find facts,” and, because of this, appellate judges must give them “due deference.” Id. at 51. The district judge has “greater familiarity with[] the individual case and the individual defendant before him than the . . . appeals court.” Id. at 51-52. Unlike the appeals court, the district “judge sees and hears the evidence, makes credibility determinations, has full knowledge of the facts and gains insight not conveyed *61 by the record.” Id. at 51. In short, “district courts have an institutional advantage over appellate courts” in setting sentences. Id. at 52.

Because of this institutional advantage—that district judges have the eyes and ears that appellate judges lack—Gall commanded appeals courts to apply a “deferential abuse-of-discretion” standard to critical sentencing determinations. Id. This standard, when combined with the clearly-erroneous standard for findings of fact, should have instilled in us a proper respect for the trial courts’ basic judgment calls. Id. at 51. The Gall Court underscored this point when it wrote that “[t]he fact that the appellate court might reasonably have concluded that a different sentence was appropriate is insufficient to justify reversal of the district court.” Id. at 51.

One would think that Gall fixed principal responsibility for sentencing where it belonged: with the trial courts. After all, that case simply reflected the basic principle that sentencing is a fact-bound matter which requires the exercise of wide discretion by trial judges. Landmark cases such as Gall and Anderson v. City of Bessemer City, 470 U.S. 564 (1985), inhabit that ironic junction where the role of facts becomes the essence of the rule of law. But Gall’s promise was never fulfilled. By adopting, expanding, and rigidly applying the categorical approach to the characterization of crimes of violence by a defendant, appellate courts have subjected too many discretionary trial-court sentencing determinations to de novo review. This appellate approach has not only led courts to “reach counterintuitive results, and ones which are not what Congress intended.” United States v. Faust, 853 F.3d 39, 61 (1st Cir. 2017) (Lynch J., concurring). It has further *62 undermined the authority of district courts as finders of fact, even where the facts are clearly ascertainable and readily discerned.

My fine colleague, Judge Motz, has written thoughtfully. And my friends in the majority are certainly not alone in derogating the fact-finding function, but part of an overreach all too prevalent in appellate ranks throughout the country. The majority cloaks its latest arrogation of authority in dulcet reassurances: “[N]othing in our holding restricts the broad discretion of district judges to make case-by-case sentencing determinations … Although judges must consult the advisory Sentencing Guidelines, they may vary from these recommendations in light of the factors listed in 18 U.S.C. § 3553(a).” Maj. Op. at 44. It is true that district judges retain the ability to vary from the Sentencing Guidelines’ recommendations. But variant sentences can shed their presumption of reasonableness and are more easily slapped down. United States v. Johnson, 242 Fed. Appx. 7, 10 (4th Cir. 2007). What is more, the time and effort we are spending on making categorical determinations unmistakably sends this message: categorical holdings may, in the context of the Guidelines, be “advisory,” Maj. Op. at 44, but they carry the peremptory bearing of commands. Finally, despite the majority’s reassurances, it is openly skeptical of the ability of triers of fact to make the kind of determinations a case-specific approach here would require. Maj. Op. at 34. Plainly the majority has far more faith in the ability of appellate judges to make legal determinations than it has in triers-of-fact to make factual ones.

Now triers of fact have suffered a double whammy. Just as application of the categorical approach to crime-of-violence sentencing determinations displaces the district *63 judge, application of the categorical approach to the residual clause of 18 U.S.C. § 924(c)(3) displaces both the trial judge and jury. The majority converts the question before us—whether Simms committed a crime of violence—from an ordinary one of fact into a lofty one of constitutional law that has nothing at all to do with Simms as an individual. It makes the standard of review de novo. It discards the limited sufficiency review normally reserved for jury verdicts. See Musacchio v. United States, 136 S.Ct. 706, 715 (2016). It jettisons the clear error review normally accorded district courts’ sentencing findings. In so doing it runs the role of juries and trial judges further into the ground.

To this, the majority offers only an adjectival denial that begs the question. Ante at n. 12. Its whole exercise is systemically damaging. In certain circumstances, the categorical approach is useful because it relieves district courts of the institutional burden involved in reaching back through time to discover the undiscoverable. Moncrieffe v. Holder, 569 U.S. 184, 200–01 (2013). Those circumstances are not present here. Joseph Simms pleaded guilty, not because of some legal misunderstanding but because he robbed a McDonald’s at gunpoint. Simms and his conspirator snuck into the occupied restaurant in the middle of the night through the drive-through window. Once inside, Simms used his gun to menace the employees and demand money. He even tried to strike one of them before ordering the workers to gather in the back office. He thereupon forced the manager at gunpoint to open the safe and hand over the money. With the money in hand, Simms struck the manager on the top of his head with the gun, threw the cash drawer at another employee, and fled.

Those are the facts. That is the actuality of it all. The trial court had this entire narrative at its disposal because, like all cases brought under §924(c), this conduct was part of instant offense. Simms’ actions are amply detailed in the pre-sentencing report which was not contested. Thus, there is not even an arguable institutional deficit at work here; the trier of fact’s unique competence should be more than sufficient to decide whether Simms committed a crime of violence.

Unlike the categorical approach, the case-specific approach would allow the trial judge and jury to look at what is before them and do what they do best. Under this approach—approved by the First Circuit in United States v. Douglas, 907 F.3d 1 (1st Cir. 2018), the Eleventh Circuit in Ovalles v. United States, 905 F.3d 1231 (11th Cir. 2018) (en banc), and the Second Circuit in United States v. Barrett, 903 F.3d 166 (2d Cir. 2018)—the trier of fact would determine whether robberies like Simms’ “involved a substantial risk that physical force against the person or property of another may be used,” resolving the question no differently than it would the application of any other element to the facts of the offense. 18 U.S.C. §924(c)(3)(B). Our review would be appropriately deferential and strictly limited, stepping in only if no reasonable trier of fact could have found a crime of violence. See Jackson v. Virginia, 443 U.S. 307, 318-19 (1979). Thus would the case-specific approach begin to right the indefensible imbalance occasioned by appellate micro-management.

Ah, but in abjuring the case-specific approach, the majority is able to say, as a matter of law naturally, that the statute is too vague and that Simms, in conspiring to rob a fast food restaurant at gunpoint, committed no crime of violence. And as it stands, I am *65 left scratching my head and attempting to find some even remotely plausible explanation for the appellate assumption of what should be a routine fact-finding role. I suppose it could be an example of the old Barnum-and-Bailey urge to move our act to center ring. Or perhaps we simply do all this because we can—an exhibition of appellate muscularity profoundly at odds with the restraint belonging to the rule of law.

This appellate regency is not consonant with common sense, and is at cross-purposes with the thrust of humane sentencing reforms. Those reforms have concentrated in major part on reducing mandatory sentences for drug offenses. See, e.g., First Step Act of 2018, S. 756, 115th Cong. §401 (2018) (reducing mandatory minimums for prior drug offenders but expanding them to prior violent felons under the Controlled Substances Act). Here, by contrast, the majority is mounting a transparent attack upon Congress’ intent to punish crimes of violence—particularly recidivist crimes of violence—more severely, an intent grounded in the basic proposition that violent offenders, 18 U.S.C. § 924(c), and repeat offenders, 18 U.S.C. § 924(e), pose a greater threat to public wellbeing.

To repeat, the majority’s approach is even at odds with the asserted rationale of the categorical approach itself, which is to spare district courts the burden of relitigating difficult-to-ascertain facts to determine whether or not past offenses are crimes of violence. Moncrieffe, 569 U.S. at 200–01. But this is a present-offense case. The trier of fact here has all the tools it needs to do the job. Rather than use these tools, the majority would have us take yet another gratuitous flight from reality by throwing out the entire residual clause of §924(c)(3) and potentially releasing a large and untold number of the *66 most violent cohort of criminal offenders prematurely back on to the streets. This approach floats impossibly above the reality that the trier of fact could capably bring to the task. The hallmarks of the appellate ascendancy are once again painfully apparent, including an obliviousness to the fates of future victims, unknown and anonymous to judges, but whom Congress, in its wisdom, sought to shelter.

The credo now seems to be to inflict more, not less, damage upon the whole edifice of law. The majority is so wedded to the categorical approach that it would rather strike down the residual clause of § 924(c)(3) in its entirety than recognize trial courts’ inherent capabilities. Constitutional assault now takes precedence over constitutional avoidance. *

Only by adopting the categorical approach does the majority create constitutional problems for itself. The vagueness question that plagued the Supreme Court in Sessions v. Dimaya, 138 S. Ct. 1204 (2018), disappears altogether upon the adoption of the case-specific approach. The text of 18 U.S.C. 924(c)(3)(B) supports it: as Judge Niemeyer indicates, language like “involves conduct” and “in the course of” and “by its nature” can easily and fairly be read to mandate a case-specific approach. The elements clause of 18

* My friend Judge Wynn seeks to replace the doctrine of constitutional avoidance with a search and destroy mission to strike down federal statutes. He then proceeds to embrace the odd proposition that upholding a statute is somehow a usurpation of Congress’ authority, and that interpreting a statute with due respect for the prerogatives of a coordinate branch is somehow an impermissible judicial exercise. It is further odd to suggest that an interpretation of a statute that allows triers of fact to apply its precise language to specific conduct represents a judicial “expansion” of an offense or indeed anything other than the most textually faithful rendering of the legislature’s command.

U.S.C. § 924(c)(3)(A) and the residual clause of 18 U.S.C. § 924(c)(3)(B) were written very differently for a reason. Determinations of “substantial risk” are familiar territory to district judges and juries, Johnson v. United States, 135 S. Ct. 2551, 2561 (2015), and we have interpreted similar language to require a case-specific approach before, United States v. Price, 777 F.3d 700 (4th Cir. 2015). What, if anything, the text leaves ambiguous, Congress’ purpose makes manifest: this statute was plainly intended to punish people like Joseph Simms who use firearms to inspire fear and visit lawless and unforgettable trauma upon their fellow citizens.

This whole enterprise will not end well. Territorial grabs are seldom becoming to the rule of law, and the ouster of trial courts, juries, and indeed Congress from their rightful place in criminal justice is becoming one of the worst. This case does not even mount a frontal assault upon the categorical approach, only to the most extreme manifestations of it. Lacking any rational explanation for the present state of affairs, I can only conclude that the very words “categorical approach” exert some irresistible pull on the appellate psyche. Like the children of medieval Hamelin, robbed of their reason, courts are left only to follow the pied piper’s song off the cliff. But that does not mean we cannot wish a fond and affectionate farewell to those tumbling into the seas. *68 NIEMEYER, Circuit Judge, with whom Judges WILKINSON, DUNCAN, AGEE, KEENAN, and QUATTLEBAUM join, dissenting:

In holding that the definition of “crime of violence” in 18 U.S.C. § 924(c)(3)(B) is unconstitutionally vague and that § 924(c)(1)’s prohibition on using a firearm in connection with a crime of violence can therefore never be enforced contemporaneously with a Hobbs Act robbery conspiracy under 18 U.S.C. § 1951, the majority insists that the “categorical approach” must be applied to assess whether the charged Hobbs Act conspiracy is a crime of violence. But this insistence both (1) distorts the basic reasons for and limited application of the categorical approach and (2) rejects a sensible reading of § 924(c) that would avoid holding a key part of a major criminal law unconstitutional. Consequently, thousands of § 924(c)(1) convictions will unnecessarily be challenged as premised on what the majority today concludes is an unconstitutionally vague provision, even though the parties in those cases had little difficulty understanding, enforcing, or defending the § 924(c)(1) charges at issue.

In reaching its conclusion, the majority relies on the Supreme Court’s decisions in Johnson v. United States, 135 S. Ct. 2551 (2015), and Sessions v. Dimaya, 138 S. Ct. 1204 (2018), where the Court found unconstitutionally vague the “residual clauses” defining “violent felony” and “crime of violence” in 18 U.S.C. § 924(e) and § 16(b), which were used to categorically characterize prior convictions. While § 924(c) employs a similar residual clause to define “crime of violence” in describing proscribed conduct, the definition in § 924(c) does not require one to imagine in the abstract the “ordinary case” of a particular crime and then to assess its risk of violence, as required by § 924(e) *69 and § 16(b) to determine whether prior convictions were for violent crimes. Johnson, 135 S. Ct. at 2557–58; Dimaya, 138 S. Ct. at 1215–16. Instead, as relevant here, § 924(c)(1) requires rather straightforwardly that the government prove that the defendant (1) used, carried, or possessed a firearm (2) in connection with (3) a crime of violence, which, in this case, was alleged to be a Hobbs Act conspiracy. In this distinct context, § 924(c) can readily be read as requiring the government to prove to the jury, as an element of the offense, that a particular Hobbs Act conspiracy charged in an indictment qualifies as a crime of violence by showing that that offense, “by its nature,” posed “a substantial risk” of physical force being used. 18 U.S.C. § 924(c)(3)(B). And in both Johnson and Dimaya, the Supreme Court could not have been clearer that there are no vagueness problems when a crime-of-violence definition like § 924(c)(3)(B)’s is applied to “real-world conduct” in precisely this manner — that is, to assess “the riskiness of conduct in which an individual defendant engage[d] on a particular occasion.” Johnson, 135 S. Ct. at 2561; see also Dimaya, 138 S. Ct. at 1215–16.

In this case, Joseph Simms pleaded guilty to two counts — Count I charging him with conspiracy to commit Hobbs Act robbery, in violation of 18 U.S.C. § 1951, and Count II charging him with brandishing a firearm during the commission of a “crime of violence” — namely, the Hobbs Act conspiracy — in violation of § 924(c)(1)(A). The crime-of-violence element in the § 924(c)(1) charge, even though defined in the same terms as the crime-of-violence definition in § 16(b), did not refer to a prior conviction in the abstract but to the real-world conduct alleged in charging the Hobbs Act violation. Specifically, Count I charged that Simms and others conspired to, and did in fact, rob *70 McDonald’s Restaurant # 31619 in Goldsboro, North Carolina, on April 14, 2014. To prove that Simms committed that crime, the government would have been required to show that Simms conspired unlawfully to obstruct, delay, or affect commerce by “robbery,” defined to mean:

the unlawful taking or obtaining of personal property from the person or in the presence of another, against his will, by means of actual or threatened force, or violence, or fear of injury, immediate or future, to his person or property, or property in his custody or possession, or the person or property of a relative or member of his family or of anyone in his company at the time of the taking or obtaining.

Id. § 1951(b)(1) (emphasis added). And then to prove further that Simms’s particular Hobbs Act conspiracy offense — which, again, was alleged to have culminated in an actual robbery — qualified as a crime of violence for purposes of the § 924(c)(1)(A) charge, the government would have been required to show that his conspiracy, “by its nature, involve[d] a substantial risk” of physical force being used “against the person or property of another.” Id. § 924(c)(3)(B).

No one has asserted or can assert that this charge against Simms is unconstitutionally vague, as defined in Kolender v. Lawson, 461 U.S. 352, 357–58 (1983) (stating that the Due Process Clause “requires that a penal statute define the criminal offense with sufficient definiteness that ordinary people can understand what conduct is prohibited and in a manner that . . . establish[es] minimal guidelines to govern law enforcement”). Instead, the majority purports to make it vague by applying a “categorical approach” designed for completely distinct and limited circumstances, even though nothing in the statute so requires. Simms’s conviction should be affirmed.

I

Simms and two others conspired to rob the McDonald’s Restaurant on April 14, 2014. Shortly after 1:00 a.m., Simms and one co-conspirator entered the restaurant by crawling through the drive-through window while the third person served as a lookout. Simms then pointed a gun at the restaurant’s manager and demanded money. At gunpoint, the manager opened the restaurant’s safe, after which Simms struck the manager on the top of his head with the gun and threw a cash drawer at another employee. The robbers fled with $1,100.

After Simms was arrested and indicted, he pleaded guilty to Counts I and II of the indictment. Count I charged him with conspiracy to commit Hobbs Act robbery, in violation of 18 U.S.C. § 1951, and it specifically alleged, as an overt act taken in furtherance of the conspiracy, that Simms and others committed the agreed-upon robbery. And Count II charged that Simms had brandished a firearm during and in relation to a crime of violence — namely, the Hobbs Act violation alleged in Count I — in violation of 18 U.S.C. § 924(c)(1)(A). In exchange for his guilty plea on these two counts, the government agreed to dismiss a third count charging that Simms had possessed a firearm after having been convicted of a felony, in violation of 18 U.S.C. §§ 922(g)(1) and 924(a)(2).

During sentencing, Simms objected to his § 924(c)(1)(A) conviction under Count II in light of the Supreme Court’s decision in Johnson, contending that the required element that he use a firearm during and in relation to a “crime of violence” was not satisfied because the residual clause used to define “crime of violence” in *72 § 924(c)(3)(B) was unconstitutionally vague. The district court overruled Simms’s objection and sentenced him to 199 months’ imprisonment — 115 months for the Hobbs Act conspiracy and a mandatory consecutive 84 months for the § 924(c)(1)(A) offense.

Simms filed this appeal, maintaining that his § 924(c) conviction cannot stand because it is premised on a definition of “crime of violence” that is unconstitutionally vague under the reasoning of Johnson and Dimaya, the latter of which was decided after Simms filed his notice of appeal.

II

Simms contends that because there is “[n]o meaningful difference” between the definition of “violent felony” in § 924(e)(2)(B)(ii) that was found unconstitutionally vague in Johnson and the definition of “crime of violence” that was used to support his conviction under § 924(c)(1)(A), we must find the latter definition unconstitutionally vague and invalidate his conviction. And Dimaya, he maintains, confirms this conclusion since the text of “Section 924(c)’s residual clause [defining crime of violence] is identical to Section 16(b).” According to Simms, § 924(c)(3)(B) compels the use of “the same ordinary case approach and risk inquiry” that doomed the provisions in Johnson and Dimaya, and the same result must therefore also follow here. The majority adopts the same basic approach. Such an argument, however, is too formulaic, overlooking the role and context of the definitions, as well as the limited role of the categorical approach.

In both Johnson and Dimaya, the Court addressed the question of applying the “categorical approach” to assess whether prior convictions were for violent crimes when *73 such crimes were defined to include any felony that, in Johnson, “involves conduct that presents a serious potential risk of physical injury,” 18 U.S.C. § 924(e)(2)(B)(ii) or, in Dimaya, that “by its nature, involves a substantial risk that physical force against the person or property of another may be used in the course of committing the offense,” id. § 16(b). The difficulty the Supreme Court found in making that determination arose not from the inability to assess violent conduct as defined by those “residual clauses,” but from assessing whether the prior conviction, considered under the categorical approach, was for a violent crime under those definitions. In short, it was the limitation imposed by applying the categorical approach — where facts and conduct may not be considered — to those residual clauses that created the problem of vagueness for the Court.

The “categorical approach” is a specific form of analysis adopted in light of the need to determine whether prior convictions were for “violent felonies” or “crimes of violence” under federal provisions — chief among them 18 U.S.C. § 924(e) — that impose sentencing or other consequences on individuals who have those types of convictions in their criminal histories. In Taylor v. United States, 495 U.S. 575, 600–02 (1990), the Supreme Court recognized that, when applying § 924(e)’s sentencing enhancement, the question of whether a defendant’s prior conviction was for a “violent felony” must focus exclusively on (1) the elements of the statute under which the defendant was convicted and (2) the fact of conviction, and exclude any evaluation of the particular facts or conduct underlying the prior conviction. In reaching this conclusion, the Court started with the statutory text but hardly found it conclusive. See id. at 600–01 (1990) (“First, the language of § 924(e) generally supports the inference that Congress *74 intended the sentencing court to look only to the fact that the defendant had been convicted of crimes falling within certain categories, and not to the facts underlying the prior convictions. . . . Read in . . . context, [the text] most likely refers to the elements of the statute of conviction, not to the facts of each defendant’s conduct” (emphasis added)). Rather, what appeared to seal the deal in favor of the “categorical approach” was the Court’s serious concern that, in the context of prior convictions, “the practical difficulties and potential unfairness of a factual approach [would be] daunting.” Id. at 601. In this regard, the Court stressed both the impracticality of attempting to relitigate potentially long-ago events, as well as the Sixth Amendment problem that would be created if sentencing judges were to make factual findings about the nature of the conduct underlying a defendant’s prior conviction. Id. at 601–02; see also Descamps v. United States, 570 U.S. 254, 269–70 (2013) (emphasizing “the categorical approach’s Sixth Amendment underpinnings,” as well as the “‘daunting’ difficulties and inequities that first encouraged [the Taylor Court] to adopt the categorical approach”).

While the precise question resolved by Taylor was whether a defendant who had in fact committed “burglary,” as federal law defines that term, but was convicted under a broader state statute should be treated as having a “burglary” conviction when applying § 924(e), the Supreme Court subsequently confirmed that the “categorical approach” also applied to the “residual clause” portion of the “violent felony” definition, see James v. United States, 550 U.S. 192, 201–02 (2007). It did so even though it acknowledged that the residual clause’s text — covering “crime[s]” that “involve[] conduct that presents a serious potential risk of physical injury to another,” § 924(e)(2)(B)(ii) — was “more *75 ambiguous” and “pose[d] greater interpretive difficulty” as to whether “it too refers to crimes as generically defined.” Nijhawan v. Holder, 557 U.S. 29, 36 (2009) (emphasis added).

Unique problems arose, however, as courts attempted to apply the “categorical approach” to determine whether the offense of conviction — as defined by its elements, rather than its underlying conduct — was one that “involves conduct that presents a serious potential risk of physical injury.” 18 U.S.C. § 924(e)(2)(B)(ii) (emphasis added). Initially, in attempting to provide clarity, the Court instructed that “the proper inquiry is whether the conduct encompassed by the elements of the offense, in the ordinary case, presents a serious potential risk of injury to another.” James, 550 U.S. at 208 (emphasis added); see also Leocal v. Ashcroft, 543 U.S. 1, 7, 10 (2004) (similarly holding that, “[i]n determining whether [an individual’s prior] conviction falls within the ambit of § 16[’s definition of crime of violence], the statute directs our focus to the ‘offense’ of conviction,” “rather than to the particular facts relating to [the individual’s] crime,” and interpreting § 16(b) in particular as “cover[ing] offenses that naturally involve a person acting in disregard of the risk that physical force might be used against another in committing an offense”). But in both Johnson and Dimaya, the Court ultimately concluded that this ordinary-case form of the “categorical approach” entailed too much speculation to be consistent with the Due Process Clause’s prohibition on enforcing vague laws.

In particular, the Court in Johnson addressed the residual clause portion of the definition of “violent felony” in § 924(e) — a statute that provides a 15-year mandatory *76 minimum sentence for certain firearm offenses if the defendant has three prior convictions for a “violent felony.” Addressing how § 924(e)’s residual clause failed in the context of the categorical approach, the Court explained:

Two features of the residual clause conspire to make it unconstitutionally vague. In the first place, the residual clause leaves grave uncertainty about how to estimate the risk posed by a crime. It ties the judicial assessment of risk to a judicially imagined “ordinary case” of a crime [as required by the categorical approach], not to real-world facts or statutory elements.

• • * *

At the same time, the residual clause leaves uncertainty about how much risk it takes for a crime to qualify as a violent felony. It is one thing to apply an imprecise “serious potential risk” standard to real-world facts; it is quite another to apply it to a judge-imagined abstraction [as required by the categorical approach].

Johnson, 135 S. Ct. at 2557–58 (emphases added). The Court noted further that its own “repeated attempts and repeated failures to craft a principled and objective standard out of the residual clause confirm[ed] its hopeless indeterminacy.” Id. at 2558–59 (referring to the Court’s prior decisions in James, 550 U.S. 192; Begay v. United States, 553 U.S. 137 (2008); Chambers v. United States, 555 U.S. 122 (2009); and Sykes v. United States, 564 U.S. 1 (2011)). The Court concluded that “[i]nvoking so shapeless a provision to condemn someone to prison for 15 years to life does not comport with the Constitution’s guarantee of due process.” Id. at 2560. In so concluding, however, the Court was quick to note that numerous statutes using terms like “substantial risk” — which the government had warned would be affected by the Court’s holding — were not, in fact, in constitutional jeopardy. The Court explained:

[A]lmost all of the cited laws require gauging the riskiness of conduct in which an individual defendant engages on a particular occasion. As a general matter, we do not doubt the constitutionality of laws that call for the application of a qualitative standard such as “substantial risk” to real- world conduct; “the law is full of instances where a man’s fate depends on his estimating rightly . . . some matter of degree.”

Id. at 2561 (first emphasis in original; second emphasis added) (quoting Nash v. United States, 229 U.S. 373, 377 (1913)).

In Dimaya, the Court followed precisely the Johnson analysis, finding the residual clause used to define “crime of violence” in § 16(b) unconstitutionally vague. At issue in Dimaya were provisions of the Immigration and Nationality Act that rendered deportable any alien convicted of an “aggravated felony,” 8 U.S.C. § 1227(a)(2)(A)(iii), which was defined to include “crime[s] of violence,” as defined in § 16, see id. § 1101(a)(43)(F). Section 16, in turn, defined “crime of violence” in part by a residual clause similar to the one used in § 924(e), which was addressed in Johnson. Specifically, the § 16 residual clause defined crime of violence to include any felony offense “that, by its nature, involves a substantial risk that physical force against the person or property of another may be used in the course of committing the offense.” 18 U.S.C. § 16(b).

After recalling in extensive detail the holding and reasoning of Johnson, the Dimaya Court applied Johnson to find the residual clause in § 16(b) also unconstitutionally vague. It explained:

To begin where Johnson did, § 16(b) also calls for a court to identify a crime’s “ordinary case” in order to measure the crime’s risk. . . . Nothing in § 16(b) helps courts to perform that task, just as nothing in [§ 924(e)] did. . . . [T]he “ordinary case” remains, as Johnson described it, an excessively “speculative,” essentially inscrutable thing.

And § 16(b) also possesses the second fatal feature of [§ 924(e)’s] residual clause: uncertainty about the level of risk that makes a crime “violent.” In [§ 924(e)], that threshold was “serious potential risk”; in § 16(b), it is “substantial risk.” . . . The difficulty comes, in § 16’s residual clause just as in [§ 924(e)’s], from applying such a standard to “a judge- imagined abstraction” — i.e., “an idealized ordinary case of the crime.” It is then that the standard ceases to work in a way consistent with due process.

138 S. Ct. at 1215–16 (citations omitted). But again, as in Johnson, the Court repeated that it did “‘not doubt’ the constitutionality of applying § 16(b)’s ‘substantial risk [standard] to real-world conduct.’” Id. at 1215 (quoting Johnson, 135 S. Ct. at 2561).

Thus, in both Johnson and Dimaya, the constitutional vagueness problem arose from having to imagine the ordinary case of an offense in the abstract — as required by the categorical approach — and then to assess that imagined crime’s risk of violence.

But the two features identified in Johnson and Dimaya that rendered the residual clauses there unconstitutional — the requirements of (1) imagining the ordinary case of a crime and (2) applying to it a qualitative risk standard — do not arise when enforcing § 924(c) because that statute does not require judicial assessment of an imagined ordinary case. Instead, § 924(c) uses “crime of violence” as an element that is charged in the indictment in the context of real-world conduct on a particular occasion. See United States v. Rodriguez-Moreno, 526 U.S. 275, 280 (1999) (noting that the commission of a crime of violence is “an essential conduct element of the § 924(c)(1) offense”). In essence, rather than creating sentencing or immigration consequences for past convictions, § 924(c)(1) “punishes the temporal and relational conjunction of two separate acts, on the *79 ground that together they pose an extreme risk of harm.” Rosemond v. United States, 572

U.S. 65, 75 (2014).

Because real-world conduct was charged in Simms’s indictment, including the Hobbs Act violation alleged to be a crime of violence, we are not left to speculate about an “ordinary case” — a “judge-imagined abstraction.” Johnson, 135 S. Ct. at 2557–58. Instead, the indictment informs us of the conduct alleged to constitute a crime of violence, precluding the need for guesswork. In Count II, Simms was charged with brandishing a firearm on April 14, 2014, while conspiring to and actually robbing a specified McDonald’s restaurant and its employees in Goldsboro, with the robbery element being defined as “taking . . . personal property from [a] person . . . by means of actual or threatened force, or violence, or fear of injury . . . to his person or property.” 18 U.S.C. § 1951(b)(1). Simms was thus clearly notified of the government’s allegation that his Hobbs Act conspiracy was among the offenses that, “by [their] nature, involve[] a substantial risk that physical force against the person or property of another may be used in the course of committing the offense.” Id. § 924(c)(3)(B).

This indictment does not invite reliance on some vague, imagined, ordinary offense. Rather, it alleges specific offense conduct, which had to be proved with real-world facts in order to obtain a conviction. Simms was thus placed on full notice of the charges and of what had to be proved against him. The problem identified in Johnson and Dimaya with courts’ imagining “an ‘idealized ordinary case of the crime,’” such that “no one could tell how much risk the offense generally posed,” Dimaya, 138 S. Ct. at 1214 (quoting Johnson, 135 S. Ct. at 2557), is simply not a problem in enforcing § 924(c). The *80 prosecution of the offense does not involve divining the degree of risk from “a judge-imagined abstraction — i.e., an idealized ordinary case of the crime.” Id. at 1216 (internal quotation marks and citation omitted).

In short, in both Johnson and Dimaya, the Supreme Court was presented with the problem of applying the ordinary-case categorical approach to a prior conviction and thus imagining an abstract offense with a risk that is unknowable due to that very act of judicial imagination. Under § 924(c), however, no prior conviction must be imagined. To the contrary, the commission of a crime of violence must be alleged as an element of the § 924(c) offense and must be proved with real-world facts that occurred on a particular occasion. Thus, the definition of crime of violence, as used in § 924(c) as an element of the offense, does not implicate the unique context identified in both Johnson and Dimaya, where the analysis was limited by the severe restrictions imposed by the categorical approach as applied to assess prior convictions.

The majority argues that Congress adopted the “categorical approach” by including the language in § 924(c)(3)(B) that a crime of violence includes a felony “offense . . . that by its nature, involves a substantial risk that physical force against the person or property of another may be used in the course of committing the offense.” It asserts that the phrase “offense . . . that by its nature,” naturally construed, means “categorically” and thus incorporates the “categorical approach.” Ante at 18, 21. Indeed, the majority concludes that the only “plausible construction” of § 924(c)(3)(B) is that it “requires application of the ordinary-case categorical approach.” Ante at 21, 41.

But the majority’s insistence on interpreting § 924(c)(3)(B) as requiring the “categorical approach” represents the judicial equivalent of cramming a square peg into a round hole. Again, the lynchpin of the majority’s textual argument is § 924(c)(3)(B)’s provision that an offense qualifies as a “violent felony” if “by its nature” it involves the requisite risk of physical force, and it notes that “nature” is defined as “the basic or inherent features, character, or quality of something.” Ante at 21 (quoting Oxford Dictionary of English 1183 (3d ed. 2010)). But while the majority emphasizes that § 924(c)(3)(B) uses the phrase “by its nature” to modify “offense,” it overlooks that “offense” is a word that, as the Supreme Court has recognized, is sometimes used in ordinary speech to “refer to a generic crime, say, the crime of fraud or theft in general” and other times readily used to “refer to the specific acts in which an offender engaged on a specific occasion” — e.g., “the fraud that the defendant planned and executed last month.” Nijhawan, 557 U.S. at 33–34. And it is just as linguistically reasonable to speak of the “nature” of the defendant’s particular offense, as it is to refer to the “nature” of the general offense as defined by law. See United States v. Barrett, 903 F.3d 166, 182 (2d Cir. 2018) (noting that “nothing in the[] definitions [of the word ‘nature’] indicates whether the offense whose inherent characteristics are to be considered is the generic crime or the particular one charged”); accord United States v. Douglas, 907 F.3d 1, 11 (1st Cir. 2018); Ovalles v. United States, 905 F.3d 1231, 1248 (11th Cir. 2018) (en banc).

In addition, the majority’s argument is undermined by another feature of § 924(c)(3)’s text — namely, the textual difference between subsections (A) and (B). Subsection (A) provides that a “crime of violence” includes any felony offense that has *82 “as an element the use, attempted use, or threatened use of physical force,” whereas subsection (B) provides that a “crime of violence” includes any felony offense that “by its nature[] involves a substantial risk” of physical force being used in the course of its commission. The phrase “by its nature” in subsection (B) is thus distinct from the use of “elements” in subsection (A), suggesting that the analysis under subsection (B) should focus not on how an offense is defined by law, but on the nature of the defendant’s offense and the degree of risk posed by it. And this interpretation is only strengthened by subsection (B)’s focus on the risk that force may be “used in the course of committing the offense.” Moreover, such a distinction between subsections (A) and (B) makes good sense. Subsection (A) thus qualifies a felony as a “crime of violence” if its elements require proof that the defendant “use[d], attempt[ed] [to] use, or threaten[ed] [to] use . . . physical force,” while subsection (B) allows that even if an offense does not have such an element, it can still qualify as a crime of violence if, as committed on a particular occasion, it actually “involve[d] a substantial risk” of physical force being used. And Congress added further certain phrases to emphasize that this determination is intended to be relatively broad, covering even those particular violations that, “by [their] nature, involve a substantial risk that physical force . . . may be used.”

This last statutory feature in particular also precludes the majority’s additional argument that the rule against surplusage compels a categorical interpretation of § 924(c)(3)(B). See ante at 21–22.

To be sure, this reading of § 924(c)(3)(B) differs from the interpretation the Supreme Court in Johnson gave to the similar language defining “violent felony” in *83 § 924(e), as well as from the Dimaya plurality’s interpretation of the materially identical text in § 16(b). But this difference is readily explained by the fact that § 924(e) and § 16(b) are applied to characterize prior convictions, and the Court was therefore compelled to apply the categorical approach. See Johnson, 135 S. Ct. at 2562 (concluding that “‘[t]he only plausible interpretation’” of § 924(e)’s residual clause is that it “requires use of the categorical approach” (1) because the text’s “emphasis on convictions indicates that ‘Congress intended the sentencing court to look only to the fact that the defendant had been convicted of crimes falling within certain categories’” and (2) because of “the utter impracticality of requiring a sentencing court to reconstruct, long after the original conviction, the conduct underlying that conviction” (emphasis added) (quoting Taylor, 495 U.S. at 600–02)); see also Dimaya, 138 S. Ct. at 1217–18 (plurality opinion) (emphasizing, among other reasons for concluding that § 16(b) must be read as incorporating the categorical approach, (1) that the “Court adopted the categorical approach in part to avoid the Sixth Amendment concerns that would arise from sentencing courts’ making findings of fact that properly belong to juries” and (2) that a fact-based reading would create “the daunting difficulties of accurately reconstructing, often many years later, the conduct underlying a conviction’” (emphasis added) (internal quotation marks, brackets, and citations omitted)).

By contrast, the definition of crime of violence in § 924(c)(3) applies only to contemporaneously charged conduct that is made an element of the § 924(c)(1) violation and never to prior convictions. And, in that distinct context, the important practical considerations and Sixth Amendment concerns that led the Court to develop and adhere to *84 the categorical approach for similar residual clauses simply do not apply. It is therefore not surprising that § 924(c) has presented little difficulty to courts, prosecutors, and defense counsel in its enforcement, and the so-called problems with that statute have been created only by the effort to force the ordinary-case categorical approach, applicable in other circumstances, onto a statute that punishes the use of a firearm in connection with a crime of violence, in this case a Hobbs Act conspiracy.

But were there any doubt about the proper construction of § 924(c) we would nonetheless have to give the statute the benefit of the doubt when a constitutional reading of it is “fairly possible,” as I surely conclude. I.N.S. v. St. Cyr, 533 U.S. 289, 299–300 (2001). This constitutional avoidance doctrine serves the “basic democratic function of maintaining a set of statutes that reflect, rather than distort, the policy choices that the elected representatives have made.” Almendarez-Torres v. United States, 523 U.S. 224, 238 (1998).

I thus readily conclude that the residual clause in § 924(c) defining the “crime of violence” element of the offense is not unconstitutionally vague. Simms received ample notice of how to conform his conduct to the law, and the statute provides adequate standards to prevent “arbitrary and discriminatory enforcement.” Kolender, 461 U.S. at 357. I would accordingly affirm the district court’s judgment. *85 RICHARDSON, Circuit Judge, with whom Judge QUATTLEBAUM joins, dissenting:

While robbing a McDonalds, Simms struck the manager on the head with a gun. By statute, if the nature of Simms’s offense involved a “substantial risk of physical force,” then it was a “crime of violence.” 18 U.S.C. § 924(c)(3)(B). And if we look to the facts here, it seems obvious that Simms’s offense was a crime of violence: of course hitting someone over the head with a gun involves a substantial risk of physical force. And Simms agreed when he pleaded guilty.

My colleagues in the majority take a different tack. Refusing to read the statute to apply to the facts of this case, they insist that the statute must be applied to some hypothetical conduct that would be involved in an “ordinary case” of federal robbery. Yet, by the majority’s own account, the hypothetical, “ordinary case” interpretation is not workable here. In fact, even before adopting it, the majority accepts that its interpretation is unconstitutionally vague under the Fifth Amendment.

This is not the right approach. It is our duty as judges, if we can, to give statutes a reasonable interpretation that conforms to the Constitution. Because the statute permits looking directly to Simms’s conduct in pistol whipping a manager during a robbery, I respectfully dissent.

I. Avoiding an Unconstitutional Construction

Any statutory provision must be interpreted, whenever possible, to avoid unconstitutionality. Our duty to adopt saving constructions dates back more than two hundred years. See Mossman v. Higginson, 4 U.S. (4 Dall.) 12 (1800); Society for the *86 Propagation of the Gospel v. Wheeler, 22 F. Cas. 756, 769 (C.C.D.N.H. 1814). 1 At base, the rule reflects respect for the political branches, presuming they intend to comply with our Constitution. 2 So long as it does not require doing violence to unambiguous language, courts must act to save a statute with any plausible constitutional interpretation. Grenada Cty. Supervisors v. Brogden, 112 U.S. 261, 269 (1884).

Here, we face a binary choice. On the one hand, if Section 924(c)(3)(B) compels the “ordinary case” interpretation, then it is unconstitutionally vague and Simms escapes liability. On the other, if the statute permits a case-specific interpretation, then it is constitutional and Simms’s guilty plea stands.

1 Justice Story later explained that if a

section admits of two interpretations, one of which brings it within, and the other presses it beyond the constitutional authority of congress, it will become our duty to adopt the former construction; because a presumption never ought to be indulged, that congress meant to exercise or usurp any unconstitutional authority, unless that conclusion is forced upon the Court by language altogether unambiguous.

United States v. Coombs, 37 U.S. (12 Pet.) 72, 76 (1838); see also Blodgett v. Holden, 275 U.S. 142, 148 (1927) (Holmes, J., concurring) (“[A]s between two possible interpretations of a statute, by one of which it would be unconstitutional and by the other valid, [a court’s] plain duty is to adopt that which will save the Act.”). 2 The requirement that a court adopt a plausible interpretation to avoid striking down a statute as unconstitutional is sometimes referred to as the presumption of validity or the unconstitutionality canon. It is related to the more recent, and arguably weaker, “doubts” canon, which urges courts to adopt plausible interpretations of statutes to avoid having to contemplate difficult constitutional questions. See United States ex rel. Attorney General v. Delaware & Hudson Co., 213 U.S. 366, 408 (1909); see also Jennings v. Rodriguez, 138 S. Ct. 830, 836 (2018). For a discussion of these two principles, see Adrian Vermeule, Saving Constructions, 85 GEO. L.J. 1945, 1948–50 (1997).

While our choice is binary, statutory interpretation often is not. Typically, judges seek the “best reading of the statute by interpreting the words of the statute, taking account of the context of the whole statute, and applying the agreed-upon semantic canons.” Brett Kavanaugh, Fixing Statutory Interpretation, 129 HARV. L. REV. 2118, 2121 (2016) (book review). Along the way to the “best” interpretation, we discard other possibilities. Not all the rejects are flat-out wrong (although some are). An interpretation may have some support in the text, and even from a canon, but still get left on the cuttingroom floor in favor of another, better option. And we keep on discarding until we find the best one.

But in some circumstances, we must abandon the search for the best reading of the statute. When a statute’s best interpretation would render it unconstitutional, a court must identify and adopt any fairly possible interpretation that would save it. See Hooper v. California, 155 U.S. 648, 657 (1895) (“The elementary rule is that every reasonable construction must be resorted to, in order to save a statute from unconstitutionality.”). That means looking for a plausible constitutional option for Section 924(c)(3)(B). 3

3 I agree with much of Judge Niemeyer’s fine opinion finding that the proper construction of the residual clause requires looking to Simms’s actual conduct. In my view, however, the duty to adopt a saving construction obviates the need to identify the best reading, requiring us to determine only whether a fairly possible reading permits looking to Simms’s actual conduct. See Ante at 84 (Niemeyer, J., dissenting) (noting in the alternative that the constitutional avoidance doctrine would apply if there were any doubt about the proper construction).

II. The Residual Clause of Section 924(c)(3)

Having framed our endeavor, I begin by asking what readings of the statute are even conceivable. Once the conceivable interpretations are identified, I turn to evaluating which of those are plausible.

Section 924(c)(3) defines the term “crime of violence” to mean

an offense that is a felony and— (A) has as an element the use, attempted use, or threatened use of physical force against the person or property of another, or (B) that by its nature, involves a substantial risk that physical force against the person or property of another may be used in the course of committing the offense.

(emphasis added). This statute provides two definitions of what constitutes a “crime of violence.” The first, sometimes called the “elements clause,” focuses on whether the “elements” of an offense include the “use, attempted use, or threatened use of physical force.” The second, called the “residual clause,” focuses on whether the offense’s “nature” involves “a substantial risk that physical force” may be “used in the course of committing the offense.” As its name conveys, the second clause is residual, meaning it operates as a catch-all for those offenders engaged in this type of risky conduct that might otherwise fail to meet the elements clause.

A. Three Conceivable Interpretations of the Residual Clause

The language and context of the residual clause point to three conceivable readings, each asking us to look at the offense through a different lens. First, an elements-based one: Does the offense legally require proof of a substantial risk of physical force? Second, a hypothetical one: Does a judicially imagined “ordinary” *89 version of the offense involve conduct that presents a substantial risk of force? Third, a factual one: Does the specific offense involve actual conduct that presents a substantial risk of physical force? See generally Johnson v. United States, 135 S. Ct. 2551, 2557 (2015) (noting three possibilities for consideration: the statutory “elements,” a “judicially imagined ordinary case,” or “real-world facts”). These readings have been fleshed out in the context of other, similar statutes, although there is no binding Supreme Court precedent governing this particular statute.

The first and most formalistic reading requires looking at the elements of the “offense.” See Descamps v. United States, 570 U.S. 254, 257 (2013). That is, one would ask whether one element of the crime is a substantial risk that physical force may be used.

The second reading concentrates not on the legal elements of the crime but on the conduct thought to be ordinarily involved in violating those elements. This approach is similar to the first in that it is “categorical”: once the courts decide whether a particular criminal statute satisfies the residual clause, that holding applies to any conviction under the statute, without a factfinder looking at what that defendant did. Yet it differs from the first in that it looks to conduct: namely “whether the conduct encompassed by the elements of the offense, in the ordinary case, presents a serious potential risk.” James v. United States, 550 U.S. 192, 208 (2007) (emphasis added). This approach is an exercise in judicial imagination, requiring “a court to picture the kind of conduct that the crime involves in ‘the ordinary case,’ and to judge whether that abstraction presents a serious potential risk.” Johnson, 135 S. Ct. at 2557.

Real-world facts are the focus of the approach resulting from the third and final interpretation. This approach is like the second approach, and unlike the first, in that it looks at conduct. Yet unlike the second approach, it is not categorical and requires no judicial ingenuity: it asks only whether the actual conduct committed by the defendant presents a substantial risk on a case-by-case basis. The third approach has significant appeal, relieving judges of imagining hypotheticals and empowering factfinders (generally juries) to do their job. See Ante at 64 (Wilkinson, J., dissenting).

B. The Case-Specific Interpretation of the Residual Clause is Plausible

Thus, three conceivable interpretations are available. The first and third readings are constitutional; the second, “ordinary case” reading, would be unconstitutionally vague. We must therefore ask whether the first and third are fairly possible. If either is, then the “ordinary case” interpretation must be rejected to save the statute.

Neither party argues that the first interpretation is plausible, and for good reason. To be sure, the term “offense” can be read to encompass the formalistic concept of a crime as defined by its elements. Indeed, “offense” takes that meaning in subsection A’s elements clause. Yet there is little else good to say about this interpretation, and two considerations make it untenable. First, Congress knew how to specify an elements-based approach and did so in subsection A, but not in subsection B. Subsection B contemplates an understanding of “offense” that is broader than the formalistic subsection A. See United States v. Price, 777 F.3d 700, 708–09 (4th Cir. 2015). Second, I am aware of no other federal criminal statute with a “substantial risk” that “physical *91 force” “may be used” as an actual element. And we generally avoid interpreting statutes to render them a nullity.

The third interpretation is a different story. Consider again the term “offense.” The term is ambiguous: it is consistent with the elements-based approach (as already noted), but also with the others. The Supreme Court has found it may refer to (1) “a generic crime, say, the crime of fraud or theft in general,” or (2) “the specific acts in which an offender engaged on a specific occasion, say, the fraud that the defendant planned and executed last month.” Nijhawan v. Holder, 557 U.S. 29, 33–34 (2009); see also United States v. Hayes, 555 U.S. 415, 426 (2009). While definition (1) fits the elements-based and “ordinary case” approaches, definition (2) permits consideration of the real-world conduct committed by the defendant.

To decide which definition is appropriate, courts look to the statutory context. In some cases, context compels a case-specific approach based on real-world conduct. See Nijhawan, 557 U.S. at 40. In others, context compels the “generic” interpretation. See id. at 37. The term may even be used in both senses within the same sentence: to mean a generic crime and how the specific crime was committed. See Hayes, 555 U.S. at 421 (holding “offense” that “has, as an element, the use or attempted use of physical force . . . , committed by a current or former spouse,” requires a categorical determination of the elements but a fact-specific evaluation of the person who committed the offense).

Our own precedent confirms that “offense” may compel examining real-world conduct. In Price, the statute defined “sex offense” to mean a “criminal offense that is a specified offense against a minor.” 777 F.3d at 704 (quoting 42 U.S.C. *92 § 16911(5)(A)(ii)). The statute then defined “specified offense[s] against a minor” according to a list of enumerated offenses and a residual clause encompassing “[a]ny conduct that by its nature is a sex offense against a minor.” Id. (quoting 42 U.S.C. § 16911(7)(I)). In that context, we concluded, the statute “refers to how an offense was committed—not a generic offense.” Id. at 709. 4

Here, we need only decide whether the fact-specific approach—which again, the Supreme Court has held is a permissible reading of the word “offense”—is plausible. And the context here shows that it is.

Consider the phrase “by its nature.” In ordinary use, “nature” means a “normal and characteristic quality” of something. Webster’s Third New International Dictionary 1507–08 (1986); see also Oxford Dictionary of English 1183 (3d ed. 2010) (“basic or inherent features, character, or qualities of something”). This definition, however, merely raises the question of what the “something” is? See United States v. Barrett, 903 F.3d 166, 182 (2d Cir. 2018). It could be either the hypothetical, “ordinary case” conduct of the second interpretation or the case-specific, real-world conduct of the third. Compare

4 The majority attempts to distinguish Price on the basis that Section 924(c)(3)(B)’s residual clause lacks “conduct-specific language.” Majority Op. at 30. Yet the majority has found that the residual clause demands adopting the “ordinary case” approach, an interpretation that focuses exclusively on hypothetical conduct.

Even were the majority’s distinction convincing, this would miss the point. Cases like Nijhawan, Hayes, and Price reveal that the phrasing used here does not unavoidably refer to the “ordinary case” approach. It makes no difference whether the majority can find some linguistic distinction for each case, because a reading calling for the fact- specific approach need only be fairly possible.

Dugger v. Adams, 489 U.S. 401, 410 n.6 (1989) (using “by its nature” to refer to general characteristics and not individualized circumstances), with H.J. Inc. v. Nw. Bell Tel. Co., 492 U.S. 229, 241–42 (1989) (noting that RICO continuity may be shown by “past conduct that by its nature projects into the future with a threat of repetition,” and that this conduct “depends on the specific facts of each case” (emphasis added)), and Doolan v. Carr, 125 U.S. 618, 625 (1887) (“[E]ven a patent from the government of the United States, issued with all the forms of law, may be shown to be void by extrinsic evidence if it be such evidence as by its nature is capable of showing a want of authority for its issue.” (emphasis added)). Thus, the phrase “by its nature” does not resolve the meaning of “offense” one way or the other.

Nor does the rest of the residual clause: “involves a substantial risk that physical force against the person or property of another may be used in the course of committing the offense.” Either hypothetical conduct or actual conduct may “involve substantial risk.” Similarly, the predictive phrase “may be used” applies equally well to hypothetical or actual conduct. And the same is true of “in the course of committing the offense.”

Tellingly, the majority concedes that all the individual terms in the statute are susceptible to a case-specific interpretation. Somehow, it claims, the combination of words compels the “ordinary case” approach. See Majority Op. at 26–27. But a handful of maybe’s simply do not add up to a never.

III. Precedent Does Not Demand the “Ordinary Case” Approach

The majority also claims that the Supreme Court has already resolved this issue by holding that the language used in Section 924(c)(3)(B) permits only one reading: the *94 hypothetical conduct of an ordinary case. In so doing, the majority rewrites history. The Supreme Court has never held that the language like that in Section 924(c)(3)(B) necessarily requires the “ordinary case” analysis.

The origins of the modern “ordinary case” approach lie in Supreme Court decisions interpreting statutory sentencing enhancements based on prior convictions for “violent felonies.” Those decisions adopted a “categorical approach,” requiring courts to look at the elements of the offense and not the defendant’s conduct. See Shepard v. United States, 544 U.S. 13, 19 (2005); Taylor v. United States, 495 U.S. 575, 600–02 (1990). In addition to the text and practical considerations, one of the main motivators of the “categorical approach” was the need to avoid a potential Sixth Amendment problem: if a mandatory sentencing enhancement arose from the facts underlying a prior conviction, as opposed to the fact of conviction itself, then those facts could not be determined by a judge (absent a defendant’s wavier). Shepard, 544 U.S. at 24 (plurality opinion). The importance of the Sixth Amendment to the history of the categorical approach is hardly controversial. The Supreme Court itself has directed courts to consider “the categorical approach’s Sixth Amendment underpinnings” when applying it. Descamps, 570 U.S. at 269. And multiple panel decisions by this Court have acknowledged as much. See, e.g., Price, 777 F.3d at 709–10. 5

5 The majority downplays the importance of the Sixth Amendment concern, suggesting that these cases were only “secondarily informed” by nontextual considerations. Majority Op. at 20. That argument is ad hoc, running against uniform Supreme Court and circuit-level precedent to the contrary. Equally unconvincing is the majority’s suggestion that, because the Court started with the text in these cases, the text (Continued)

Over time, the categorical approach was extended to new contexts. There are a number of statutes that mix and match similar language to define “crimes of violence” or “violent felonies.” The language at issue in Taylor and Shepard, the classic cases, encompasses crimes that correspond to the “generic” definition of certain enumerated crimes. Other statutes include language resembling the “elements clause” of Section 924(c)(3)(A) and the “residual clause” of Section 924(c)(3)(B). Unsurprisingly, the Court transplanted the categorical approach into these similar contexts. Yet the Court continued to note the important Sixth Amendment problems that the categorical approach avoided. See James, 550 U.S. at 213–14.

The majority identifies only two cases suggesting that statutory text alone— unaccompanied by constitutional and other considerations—was sufficient to support the categorical approach. Neither is persuasive in interpreting the statute at hand. The first case, Leocal v. Ashcroft, only superficially supports the majority. After quoting language from the residual clause in 18 U.S.C. § 16(b) (textually similar to the residual clause of Section 924(c)(3)(B)), the Court stated: “[t]his language requires us to look to the elements and the nature of the offense of conviction, rather than to the particular facts relating to petitioner’s crime.” 543 U.S. 1, 7 (2004). Yet it is hardly surprising that the Supreme Court did not analyze the issue in detail: the parties, amici, and the Court

alone must have been decisive. Id. at 19. The Supreme Court, like all courts, always starts from the text in statutory interpretation cases. See, e.g., Sebelius v. Cloer, 569 U.S. 369, 376 (2013).

accepted the application of the categorical approach as an untested premise before addressing the issue presented. Cf. Sessions v. Dimaya, 138 S. Ct. 1204, 1232 (2018) (Gorsuch, J., concurring in part and concurring in judgment) (accepting the “ordinary case” approach to Section 16(b) as a premise).

As everyone accepted the categorical approach, Leocal had no reason to consider any saving constructions to avoid unconstitutionality (or constitutional doubts). So even were one to read Leocal as finding the “best” interpretation, that does not answer whether that is the only plausible interpretation. Moreover, Leocal addressed the application of Section 16 to prior convictions. It is in that context that applying a factual approach would raise the serious practical and Sixth Amendment concerns at issue in Taylor, Shepard, and James. Section 924(c)(3) presents none of those problems, as it applies only to contemporaneous conduct. Nor did Leocal foresee that applying the “ordinary case” approach would be unconstitutionally vague under the Fifth Amendment. In fact, Leocal did not even use the term “ordinary case.” In sum, Leocal’s acceptance of an undisputed premise while interpreting a different statute in a different context without discussing the plausibility of a saving construction does not dictate the interpretation of Section 924(c)(3)(B).

The second case, Dimaya, is even less supportive. It is true that Section IV(A) of Justice Kagan’s opinion in Dimaya held that similar language demands the use of the categorical approach, but that interpretation was only endorsed by a plurality of the Supreme Court. Justice Gorsuch chose not to join that section of the opinion. Under the Marks rule, the binding holding of the Court is the narrowest opinion from those *97 concurring in the judgment. Marks v. United States, 430 U.S. 188, 193 (1977). Thus, Justice Gorsuch’s narrower concurrence controls. He (like the parties) simply assumed that the “ordinary case” approach applied, and on that basis held that § 16(b) is impermissibly vague. Dimaya, 138 S. Ct. at 1232–33 (Gorsuch, J., concurring in part and concurring in judgment).

As for the debated language, Justice Gorsuch noted that “[p]lausibly, anyway, the word ‘nature’ might refer to an inevitable characteristic of the offense; one that would present itself automatically, whenever the statute is violated.” Dimaya, 138 S. Ct. at 1233. He continued, “[w]hile I remain open to different arguments about our precedent and the proper reading of language like this, I would address them in another case, whether involving the INA or a different statute, where the parties have a chance to be heard and we might benefit from their learning.” Id. Thus, Justice Gorsuch unambiguously stated that he is open to other interpretations, even involving the statute at issue in his own opinion. As his concurrence shows, the majority gets Dimaya backward. Dimaya makes clear that the interpretation of this language is a live issue before the Supreme Court, not one settled by precedent.

If anything, the history of the categorical approach shows that the “ordinary case” approach should be jettisoned wherever possible. In truth, there is not one categorical approach; there are two. When the purely elements-based categorical approach from Taylor and Shepard was applied to residual clauses, it evolved (or perhaps devolved) into the “ordinary case” approach. See James, 550 U.S. at 208. Those two approaches are only superficially similar. The elements-based approach looks only at the formal *98 definition of the offense; the “ordinary case” approach tries to construct a hypothetical model of conduct for the offense. In both logic and application, they are as different from each other as they are from the case-specific approach. History has shown that the “ordinary case” approach is unworkable, so much so that its application has been declared unconstitutionally vague twice. Johnson, 135 S. Ct. at 2563; Dimaya, 138 S. Ct. at 1232–33 (Gorsuch, J., concurring in part and concurring in judgment).

In contexts where the Sixth Amendment applies—in particular, mandatory sentencing enhancements for prior convictions—the “ordinary case” approach is the only way to interpret a residual clause. But in many contexts, those Sixth Amendment concerns do not apply. This is one of them: The consideration of contemporaneous conduct under Section 924(c)(3)(B) need not create any Sixth Amendment concerns. See Ante at 79 (Niemeyer, J., dissenting). And we have noted the importance of this distinction when rejecting specific applications of the categorical approach. Price, 777 F.3d at 709–10 (noting, in rejecting the categorical approach, that “Sixth Amendment concerns that compel the judicial use of the categorical approach in other contexts are not present here”). When Sixth Amendment concerns do not apply, but Fifth Amendment vagueness concerns do, we must adopt the fact-specific interpretation.

IV. The Majority’s Remaining Arguments

The majority advances a grab-bag of other considerations to support its “ordinary case” interpretation. Many of these are unpersuasive on their face, but I mention three briefly.

First, the majority argues that adopting the factual approach to the residual clause would lead to a contradiction: “offense that is a felony” would be read to have a categorical interpretation in the elements clause, but a case-specific interpretation in the residual clause. Majority Op. at 23–24. This reasoning is hollow because, as I have already explained, there is not only one categorical approach. The hypothetical conduct of the “ordinary case” approach adopted by the majority differs substantially from the legal elements of the elements-based approach. Thus, the interpretation the majority adopts is similarly contradictory. And in any event, neither “contradiction” is that concerning. See Nijhawan, 557 U.S. at 37 (The “statute lists several of its ‘offenses’ in language that must refer to generic crimes . . . [but] it lists certain other ‘offenses’ using language that almost certainly does not refer to generic crimes but refers to specific circumstances.”).

The majority also suggests that reading the residual clause to apply to real-world factual conduct violates the rule against surplusage by draining the phrase “by its nature” of any meaning. At the outset, the Supreme Court has explained that the rule against surplusage “is not absolute” and is itself to be avoided when it causes more problems than it solves. Lamie v. U.S. Trustee, 540 U.S. 526, 536 (2004). In any event, “by its nature” is not superfluous under the case-specific approach for two reasons. First, the phrase distinguishes subsection B’s conduct-based residual clause from subsection A’s elements clause. The latter addresses the legal elements of the offense (“has as an element”), while the former focuses more broadly on conduct (whether hypothetical or actual conduct). See Price, 777 F.3d at 708–09. Second, “by its nature” limits the *100 residual clause to conduct that involves risk inherently or naturally, not merely incidentally or by happenstance. In this sense, the phrase “by its nature” acts as an intensifier that limits the clause’s application to more dangerous crimes.

Finally, the majority makes passing reference to the rule of lenity. Majority Op. at 40. The majority cites no authority for the proposition that the rule of lenity, grounded in the need for fair warning, demands that this Court knowingly adopt an interpretation so unclear that it is unconstitutionally vague. Indeed, this makes no sense. Cf. United States v. Lanier, 520 U.S. 259, 266 (1997) (describing the “rule of lenity” as the “junior version of the vagueness doctrine” that “ensures fair warning by so resolving ambiguity in a criminal statute as to apply it only to conduct clearly covered”); United States v. Harriss, 347 U.S. 612, 617–18 (1954) (noting that while the Constitution “is violated by a criminal statute that fails to give a person of ordinary intelligence fair notice,” a court is under a “duty” to give a “reasonable construction of the statute” that makes the “class of offenses . . . constitutionally definite”). It may well be that no ordinary person would be fairly warned that their conduct might be implicated by the hypothetical, “ordinary case” of robbery, but most everyone would understand that pistol whipping someone during an armed robbery is a crime of violence. 6

6 Judge Wynn suggests that the avoidance doctrine should allow only those interpretations that “narrow” a criminal statute. See Ante at 47 (Wynn, J., concurring). Even if one accepts this novel theory, the “ordinary case” approach is not inherently “narrower” than the case-specific approach. Since the “ordinary case” approach focuses on judicially-imagined hypothetical conduct in place of a defendant’s own actions, that approach is both over- and under-inclusive. While some defendants would be guilty under the case-specific approach but not the “ordinary case” approach, other defendants (Continued)

• • * *

Occasionally we find ourselves so far down the rabbit hole that logic appears upside down. It is true that the faithful application of the law may lead to odd results. And I certainly do not fault the majority for following its analysis to this seemingly nonsensical result. For good judges must be willing to reach results that go against their preferences when those results are dictated by principles or precedent.

But judges are also required to exercise restraint. Voiding an act of Congress is an extraordinary exercise of judicial power, and accordingly, such power should only be invoked when it is incumbent upon us to do so. Here, the statute plausibly permits looking directly at Simms’s conduct in pistol whipping a manager during a robbery in determining whether his conduct qualified as a crime of violence under the residual

would find themselves in the exact reverse position. These latter defendants, whose actual conduct did not create a substantial risk of violence, would be guilty when evaluated under the guise of a hypothetical ordinary case. Assuming it matters, I cannot as an empirical matter reasonably identify whether the former group is larger than the latter. Cf. Dimaya, 138 S. Ct. at 1232 (Gorsuch, J., concurring in part and concurring in judgment) (rhetorically asking whether “a conviction for witness tampering ordinarily involve[s] a threat to the kneecaps or just the promise of a bribe”). Moreover, the concurrence’s characterization of what it means to “expand” a statute’s reach is perverse. A judicially-imagined “ordinary case” can ensnare a defendant whose actual conduct created no risk of violence, a result barred by the case- specific approach. It is the “ordinary case” approach that expands the statute to cover hypothetical conduct imagined by a court. And this is why the “ordinary case” approach is unconstitutionally vague. A far more precise approach is to let juries determine guilt based on the real actions taken by the defendant. See Harriss, 347 U.S. at 618 (noting the duty to make the “class of offenses . . . constitutionally definite”). It is entirely proper to replace an unworkable, unconstitutionally vague interpretation with a readily- comprehensible, textually-supported interpretation, even if that interpretation would “expand” liability by clarifying which conduct falls within the statute and which does not.

clause. We are therefore required to adopt that reading. For these reasons, I respectfully dissent.