United States v. Davis

JUSTICE GORSUCH delivered the opinion of the Court.¶

In our constitutional order, a vague law is no law at all. Only the people’s elected representatives in Congress have the power to write new federal criminal laws. And when Congress exercises that power, it has to write statutes that give ordinary people fair warning about what the law demands of them. Vague laws transgress both of those constitutional requirements. They hand off the legislature’s responsibility for defining criminal behavior to unelected prosecutors and judges, and they leave people with no sure way to know what consequences will attach to their conduct. When Congress passes a vague law, the role of courts under our Constitution is not to fashion a new, clearer law to take its place, but to treat the law as a nullity and invite Congress to try again.¶

Today we apply these principles to 18 U. S. C. §924(c). That statute threatens long prison sentences for anyone who uses a firearm in connection with certain other federal crimes. But which other federal crimes? The statute’s residual clause points to those felonies “that by [their] nature, involv[e] a substantial risk that physical force against the person or property of another may be used in the course of committing the offense.” §924(c)(3)(B). Even the government admits that this language, read in the way nearly everyone (including the government) has long understood it, provides no reliable way to determine which offenses qualify as crimes of violence and thus is unconstitutionally vague. So today the government attempts a new and alternative reading designed to save the residual clause. But this reading, it turns out, cannot be squared with the statute’s text, context, and history. Were we to adopt it, we would be effectively stepping outside our role as judges and writing a new law rather than applying the one Congress adopted.¶

I¶

After Maurice Davis and Andre Glover committed a string of gas station robberies in Texas, a federal prosecutor charged both men with multiple counts of robbery affecting interstate commerce in violation of the Hobbs Act, 18 U. S. C. §1951(a), and one count of conspiracy to commit Hobbs Act robbery. The prosecutor also charged Mr. Davis with being a felon in possession of a firearm. In the end, a jury acquitted Mr. Davis of one robbery charge and otherwise found the men guilty on all counts. And these convictions, none of which are challenged here, authorized the court to impose prison sentences of up to 70 years for Mr. Davis and up to 100 years for Mr. Glover.¶

But that was not all. This appeal concerns additional charges the government pursued against the men under §924(c). That statute authorizes heightened criminal penalties for using or carrying a firearm “during and in relation to,” or possessing a firearm “in furtherance of,” any federal “crime of violence or drug trafficking crime.” §924(c)(1)(A). The statute proceeds to define the term “crime of violence” in two subparts—the first known as the elements clause, and the second the residual clause.¶

According to §924(c)(3), a crime of violence is “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.”¶

Violators of §924(c) face a mandatory minimum sentence of five years in prison, over and above any sentence they receive for the underlying crime of violence or drug trafficking crime. The minimum sentence rises to 7 years if the defendant brandishes the firearm and 10 years if he discharges it. Certain types of weapons also trigger enhanced penalties—for example, a defendant who uses a short-barreled shotgun faces a minimum sentence of 10 years. And repeat violations of §924(c) carry a minimum sentence of 25 years. 1¶

At trial, the government argued that Mr. Davis and Mr. Glover had each committed two separate §924(c) violations by brandishing a short-barreled shotgun in connection with their crimes. Here, too, the jury agreed. These convictions yielded a mandatory minimum sentence for each man of 35 years, which had to run consecutively to their other sentences. Adding the §924(c) mandatory minimums to its discretionary sentences for their other crimes, the district court ultimately sentenced Mr. Glover to more¶

------¶

1 When this case was tried, a defendant convicted of two §924(c) violations in a single prosecution faced a 25-year minimum for the second violation. See Deal v. United States , 508 U. S. 129, 132 (1993); §1(a)(1), 112 Stat. 3469. In 2018, Congress changed the law so that, going forward, only a second §924(c) violation committed “after a prior [§924(c)] conviction . . . has become final” will trigger the 25-year minimum. Pub. L. 115–391, §403(a), 132 Stat. 5221.¶

than 41 years in prison and Mr. Davis to more than 50 years.¶

On appeal, both defendants argued that §924(c)’s residual clause is unconstitutionally vague. At first, the Fifth Circuit rejected the argument. United States v. Davis , 677 Fed. Appx. 933, 936 (2017) ( per curiam ). But after we vacated its judgment and remanded for further consideration in light of our decision in Sessions v. Dimaya , 584 U. S. ___ (2018), striking down a different, almost identically worded statute, the court reversed course and held §924(c)(3)(B) unconstitutional. 903 F. 3d 483, 486 (2018) ( per curiam ). It then held that Mr. Davis’s and Mr. Glover’s convictions on one of the two §924(c) counts, the one that charged robbery as a predicate crime of violence, could be sustained under the elements clause. But it held that the other count, which charged conspiracy as a predicate crime of violence, depended on the residual clause; and so it vacated the men’s convictions and sentences on that count.¶

Because the Fifth Circuit’s ruling deepened a dispute among the lower courts about the constitutionality of §924(c)’s residual clause, we granted certiorari to resolve the question. 586 U. S. ___ (2018). 2¶

II¶

Our doctrine prohibiting the enforcement of vague laws rests on the twin constitutional pillars of due process and separation of powers. See Dimaya , 584 U. S., at ___–___ (plurality opinion) (slip op., at 4–5); id. , at ___–___¶

------¶

2 Compare United States v. Simms , 914 F. 3d 229, 236–246 (CA4 2019) (en banc), United States v. Salas , 889 F. 3d 681, 685–686 (CA10 2018), and United States v. Eshetu , 898 F. 3d 36, 37–38 (CADC 2018) (holding that §924(c)(3)(B) is vague), with United States v. Douglas , 907 F. 3d 1, 11–16 (CA1 2018), Ovalles v. United States , 905 F. 3d 1231, 1240–1252 (CA11 2018) (en banc), and United States v. Barrett , 903 F. 3d 166, 178–184 (CA2 2018) (taking the opposite view).¶

(GORSUCH, J., concurring in part and concurring in judgment) (slip op., at 2–9). Vague laws contravene the “first essential of due process of law” that statutes must give people “of common intelligence” fair notice of what the law demands of them. Connally v. General Constr. Co. , 269 U. S. 385, 391 (1926); see Collins v. Kentucky , 234 U. S. 634, 638 (1914). Vague laws also undermine the Constitution’s separation of powers and the democratic selfgovernance it aims to protect. Only the people’s elected representatives in the legislature are authorized to “make an act a crime.” United States v. Hudson , 7 Cranch 32, 34 (1812). Vague statutes threaten to hand responsibility for defining crimes to relatively unaccountable police, prosecutors, and judges, eroding the people’s ability to oversee the creation of the laws they are expected to abide. See Kolender v. Lawson , 461 U. S. 352, 357–358, and n. 7 (1983); United States v. L. Cohen Grocery Co. , 255 U. S. 81, 89–91 (1921); United States v. Reese , 92 U. S. 214, 221 (1876).¶

In recent years, this Court has applied these principles to two statutes that bear more than a passing resemblance to §924(c)(3)(B)’s residual clause. In Johnson v. United States , 576 U. S. ___ (2015), the Court addressed the residual clause of the Armed Career Criminal Act (ACCA), which defined a “violent felony” to include offenses that presented a “serious potential risk of physical injury to another.” §924(e)(2)(B)(ii). The ACCA’s residual clause required judges to use a form of what we’ve called the “categorical approach” to determine whether an offense qualified as a violent felony. Following the categorical approach, judges had to disregard how the defendant actually committed his crime. Instead, they were required to imagine the idealized ''ordinary case'' of the defendant’s crime and then guess whether a ''serious potential risk of physical injury to another'' would attend its commission. Id. , at ___ (slip op., at 4). Johnson held this judicial inquiry produced “more unpredictability and arbitrariness” when it comes to specifying unlawful conduct than the Constitution allows. Id. , at ___–___ (slip op., at 5–6).¶

Next, in Sessions v. Dimaya , we considered the residual clause of 18 U. S. C. §16, which defines a “crime of violence” for purposes of many federal statutes. Like §924(c)(3), §16 contains an elements clause and a residual clause. The only difference is that §16’s elements clause, unlike §924(c)(3)’s elements clause, isn’t limited to felonies; but there’s no material difference in the language or scope of the statutes’ residual clauses. 3 As with the ACCA, our precedent under §16’s residual clause required courts to use the categorical approach to determine whether an offense qualified as a crime of violence. Dimaya , 584 U. S., at ___–___ (slip op., at 2–3); see Leocal v. Ashcroft , 543 U. S. 1, 7, 10 (2004). And, again as with the ACCA, we held that §16’s residual clause was unconstitutionally vague because it required courts “to picture the kind of conduct that the crime involves in the ordinary case, and to judge whether that abstraction presents some not-well-specified-yet-sufficiently-large degree of risk.” Dimaya , 584 U. S., at ___ (slip op., at 11) (internal quotation marks omitted).¶

What do Johnson and Dimaya have to say about the statute before us? Those decisions teach that the imposition of criminal punishment can’t be made to depend on a judge’s estimation of the degree of risk posed by a crime’s imagined “ordinary case.” But does §924(c)(3)(B) require that sort of inquiry? The government and lower courts¶

------¶

3 Section 16 provides that the term “crime of violence” means “(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.”¶

have long thought so. For years, almost everyone understood §924(c)(3)(B) to require exactly the same categorical approach that this Court found problematic in the residual clauses of the ACCA and §16. 4 Today, the government acknowledges that, if this understanding is correct, then §924(c)(3)(B) must be held unconstitutional too.¶

But the government thinks it has now found a way around the problem. In the aftermath of our decisions holding the residual clauses of the ACCA and §16(b) unconstitutionally vague, the government “abandon[ed] its longstanding position” that §924(c)(3)(B) requires a categorical analysis and began urging lower courts to “adopt a new ‘case specific’ method” that would look to “the ‘defendant’s actual conduct’ in the predicate offense.” 903 F. 3d, at 485. Now, the government tries the same strategy in this Court, asking us to abandon the traditional categorical approach and hold that the statute actually commands the government’s new case-specific approach. So, while the consequences in this case may be of constitutional dimension, the real question before us turns out to be one of pure statutory interpretation.¶

In approaching the parties’ dispute over the statute’s meaning, we begin by acknowledging that the government¶

------¶

4 See, e.g. , United States v. Acosta , 470 F. 3d 132, 134–135 (CA2 2006); United States v. Butler , 496 Fed. Appx. 158, 161 (CA3 2012); United States v. Fuertes , 805 F. 3d 485, 498 (CA4 2015); United States v. Williams , 343 F. 3d 423, 431 (CA5 2003); Evans v. Zych , 644 F. 3d 447, 453 (CA6 2011); United States v. Jackson , 865 F. 3d 946, 952 (CA7 2017), vacated and remanded, 584 U. S. ___ (2018); United States v. Moore , 38 F. 3d 977, 979–980 (CA8 1994); United States v. Amparo , 68 F. 3d 1222, 1225–1226 (CA9 1995); United States v. Munro , 394 F. 3d 865, 870 (CA10 2005); United States v. McGuire , 706 F. 3d 1333, 13361337 (CA11 2013); United States v. Kennedy , 133 F. 3d 53, 56 (CADC 1998); see also Ovalles v. United States , 905 F. 3d 1231, 1295 (CA11 2018) (en banc) (J. Pryor, J., dissenting) (“For years, and even after Johnson , the government consistently has urged that we apply a categorical approach to §924(c)”).¶

is right about at least two things. First, a case-specific approach would avoid the vagueness problems that doomed the statutes in Johnson and Dimaya . In those cases, we recognized that there would be no vagueness problem with asking a jury to decide whether a defendant’s ''real-world conduct'' created a substantial risk of physical violence. Dimaya , 584 U. S., at ___–___ (slip op., at 10–11); see Johnson , 576 U. S., at ___, ___ (slip op., at 6, 12). Second, a case-specific approach wouldn’t yield the same practical and Sixth Amendment complications under §924(c) that it would have under the ACCA or §16. Those other statutes, in at least some of their applications, required a judge to determine whether a defendant’s prior conviction was for a “crime of violence” or “violent felony.” In that context, a case-specific approach would have entailed “reconstruct[ing], long after the original conviction, the conduct underlying that conviction.” Id. , at ___ (slip op., at 13). And having a judge, not a jury, make findings about that underlying conduct would have “raise[d] serious Sixth Amendment concerns.” Descamps v. United States , 570 U. S. 254, 269–270 (2013). By contrast, a §924(c) prosecution focuses on the conduct with which the defendant is currently charged . The government already has to prove to a jury that the defendant committed all the acts necessary to punish him for the underlying crime of violence or drug trafficking crime. So it wouldn’t be that difficult to ask the jury to make an additional finding about whether the defendant’s conduct also created a substantial risk that force would be used.¶

But all this just tells us that it might have been a good idea for Congress to have written a residual clause for §924(c) using a case-specific approach. It doesn’t tell us whether Congress actually wrote such a clause. To answer that question, we need to examine the statute’s text, context, and history. And when we do that, it becomes clear that the statute simply cannot support the government’s newly minted case-specific theory.¶

III A¶

Right out of the gate, the government faces a challenge. This Court, in a unanimous opinion, has already read the nearly identical language of 18 U. S. C. §16(b) to mandate a categorical approach. And, importantly, the Court did so without so much as mentioning the practical and constitutional concerns described above. Instead, the Court got there based entirely on the text. In Leocal , the Court wrote:¶

“In determining whether petitioner’s conviction falls within the ambit of §16, the statute directs our focus to the ‘offense’ of conviction. See §16(a) (defining a crime of violence as ‘ an offense that has as an element the use . . . of physical force against the person or property of another’ (emphasis added)); §16(b) (defining the term as ‘ 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’ (emphasis added)). This 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., at 7.¶

Leocal went on to suggest that burglary would always be a crime of violence under §16(b) “because burglary, by its nature , involves a substantial risk that the burglar will use force against a victim in completing the crime,” regardless of how any particular burglar might act on a specific occasion. Id. , at 10 (emphasis added); see also Dimaya , 584 U. S., at ___ (slip op., at 14) (plurality opinion) (reaffirming that “§16(b)’s text . . . demands a categorical approach”). And what was true of §16(b) seems to us at least as true of §924(c)(3)(B): It’s not even close; the statutory text commands the categorical approach.¶

Consider the word “offense.” It’s true that “in ordinary speech,” this word can carry at least two possible meanings. It can refer to “a generic crime, say, the crime of fraud or theft in general,” or it can refer to “the specific acts in which an offender engaged on a specific occasion.” Nijhawan v. Holder , 557 U. S. 29, 33–34 (2009). But the word “offense” appears just once in §924(c)(3), in the statute’s prefatory language. And everyone agrees that, in connection with the elements clause, the term “offense” carries the first, “generic” meaning. Cf. id. , at 36 (similar language of the ACCA’s elements clause “refers directly to generic crimes”). So reading this statute most naturally, we would expect “offense” to retain that same meaning in connection with the residual clause. After all, “[i]n all but the most unusual situations, a single use of a statutory phrase must have a fixed meaning.” Cochise Consultancy, Inc. v. United States ex rel. Hunt , 587 U. S. ___, ___ (2019) (slip op., at 5).¶

To prevail, the government admits it must persuade us that the singular term “offense” bears a split personality in §924(c), carrying the “generic” meaning in connection with the elements clause but then taking on the “specific act” meaning in connection with the residual clause. And, the government suggests, this isn’t quite as implausible as it may sound; sometimes the term “offense” can carry both meanings simultaneously. To illustrate its point, the government posits a statute defining a “youthful gun crime” as “an offense that has as an element the use of a gun and is committed by someone under the age of 21.” Tr. of Oral Arg. 16. This statute, the government suggests, would leave us little choice but to understand the single word “offense” as encompassing both the generic crime and the manner of its commission on a specific occasion. To which we say: Fair enough. It’s possible for surrounding text to make clear that “offense” carries a double meaning. But absent evidence to the contrary, we presume the term is being used consistently. And nothing in §924(c)(3)(B) comes close to rebutting that presumption.¶

Just the opposite. The language of the residual clause itself reinforces the conclusion that the term “offense” carries the same “generic” meaning throughout the statute. Section 924(c)(3)(B), just like §16(b), speaks of an offense that, “by its nature,” involves a certain type of risk. And that would be an exceedingly strange way of referring to the circumstances of a specific offender’s conduct. As both sides agree, the “nature” of a thing typically denotes its ''normal and characteristic quality,'' Dimaya , 584 U. S., at ___ (slip op., at 14) (quoting Webster’s Third New International Dictionary 1507 (2002)), or its ''basic or inherent features,'' United States v. Barrett , 903 F. 3d 166, 182 (CA2 2018) (quoting Oxford Dictionary of English 1183 (A. Stevenson ed., 3d ed. 2010)). So in plain English, when we speak of the nature of an offense, we’re talking about “what an offense normally—or, as we have repeatedly said, ‘ordinarily’—entails, not what happened to occur on one occasion.” Dimaya , 584 U. S., at ___ (slip op., at 14); see Leocal , 543 U. S., at 7 (contrasting the “nature of the offense” with “the particular facts [of] petitioner’s crime”). 5¶

Once again, the government asks us to overlook this obvious reading of the text in favor of a strained one. It suggests that the statute might be referring to the “na-¶

------¶

5 The government’s own regulations reflect this understanding of the ordinary meaning of “by its nature.” A Department of Justice regulation provides that an inmate is not eligible for early release if he was convicted of an offense “that, by its nature or conduct , presents a serious potential risk of physical force.” 28 CFR §550.55(b)(5)(iii) (2017) (emphasis added); see Bush v. Pitzer , 133 F. 3d 455, 458 (CA7 1997) (denying early release because “[c]onspiracy does not by its ‘nature’ present a serious risk; but Bush’s ‘conduct’ did so”).¶

ture” of the defendant’s conduct on a particular occasion. But while this reading may be linguistically feasible, we struggle to see why, if it had intended this meaning, Congress would have used the phrase “by its nature” at all. The government suggests that “by its nature” keeps the focus on the offender’s conduct and excludes evidence about his personality , such as whether he has violent tendencies. But even without the words “by its nature,” nothing in the statute remotely suggests that courts are allowed to consider character evidence—a type of evidence usually off-limits during the guilt phase of a criminal trial. Cf. Fed. Rule Evid. 404.¶

B¶

Things become clearer yet when we consider §924(c)(3)(B)’s role in the broader context of the federal criminal code. As we’ve explained, the language of §924(c)(3)(B) is almost identical to the language of §16(b), which this Court has read to mandate a categorical approach. And we normally presume that the same language in related statutes carries a consistent meaning. See, e.g. , Sullivan v. Stroop , 496 U. S. 478, 484 (1990).¶

This case perfectly illustrates why we do that. There are dozens of federal statutes that use the phrase “crime of violence” to refer to presently charged conduct rather than a past conviction. Some of those statutes cross-reference the definition of “crime of violence” in §924(c)(3), while others are governed by the virtually identical definition in §16. The choice appears completely random. Reading the similar language in §924(c)(3)(B) and §16(b) similarly yields sensibly congruent applications across all these other statutes. But if we accepted the government’s invitation to reinterpret §924(c)(3)(B) as alone endorsing a case-specific approach, we would produce a series of seemingly inexplicable results.¶

Take just a few examples. If the government were right,¶

Congress would have mandated the case-specific approach in a prosecution for providing explosives to facilitate a crime of violence, 18 U. S. C. §844( o ), but the (nowinvalidated) categorical approach in a prosecution for providing information about explosives to facilitate a crime of violence, §842(p)(2). It would have mandated the case-specific approach in a prosecution for using false identification documents in connection with a crime of violence, §1028(b)(3)(B), but the categorical approach in a prosecution for using confidential phone records in connection with a crime of violence, §1039(e)(1). It would have mandated the case-specific approach in a prosecution for giving someone a firearm to use in a crime of violence, §924(h), but the categorical approach in a prosecution for giving a minor a handgun to use in a crime of violence, §924(a)(6)(B)(ii). It would have mandated the case-specific approach in a prosecution for traveling to another State to acquire a firearm for use in a crime of violence, §924(g), but the categorical approach in a prosecution for traveling to another State to commit a crime of violence, §1952(a)(2). And it would have mandated the case-specific approach in a prosecution for carrying armor-piercing ammunition in connection with a crime of violence, §924(c)(5), but the categorical approach in a prosecution for carrying a firearm while “in possession of armor piercing ammunition capable of being fired in that firearm” in connection with a crime of violence, §929(a)(1).¶

There would be no rhyme or reason to any of this. Nor does the government offer any plausible account why Congress would have wanted courts to take such dramatically different approaches to classifying offenses as crimes of violence in these various provisions. To hold, as the government urges, that §16(b) requires the categorical approach while §924(c)(3)(B) requires the case-specific approach would make a hash of the federal criminal code.¶

C¶

Section 924(c)(3)(B)’s history provides still further evidence that it carries the same categorical-approach command as §16(b). It’s no accident that the language of the two laws is almost exactly the same. The statutory term “crime of violence” traces its origins to the Comprehensive Crime Control Act of 1984. There, Congress enacted the definition of “crime of violence” in §16. §1001(a), 98 Stat. 2136. It also “employed the term ‘crime of violence’ in numerous places in the Act,” Leocal , 543 U. S., at 6, including in §924(c). §1005(a), 98 Stat. 2138. At that time, Congress didn’t provide a separate definition of “crime of violence” in §924(c) but relied on §16’s general definition. The two statutes, thus, were originally designed to be read together.¶

Admittedly, things changed a bit over time. Eventually, Congress expanded §924(c)’s predicate offenses to include drug trafficking crimes as well as crimes of violence. §§104(a)(2)(B)–(C), 100 Stat. 457. When it did so, Congress added a subsection-specific definition of “drug trafficking crime” in §924(c)(2)—and, perhaps thinking that both terms should be defined in the same place, it also added a subsection-specific definition of “crime of violence” in §924(c)(3). §104(a)(2)(F), id. , at 457. But even then, Congress didn’t write a new definition of that term. Instead, it copied and pasted the definition from §16 without making any material changes to the language of the residual clause. The government suggests that, in doing so, Congress “intentionally separated” and “decoupled” the two definitions. Brief for United States 34, 37. But importing the residual clause from §16 into §924(c)(3) almost word for word would have been a bizarre way of suggesting that the two clauses should bear drastically different meanings. Usually when statutory language ''is obviously transplanted from . . . other legislation,'' we have reason to think ''it brings the old soil with it.'' Sekhar v. United¶

States , 570 U. S. 729, 733 (2013).¶

What’s more, when Congress copied §16(b)’s language into §924(c) in 1986, it proceeded on the premise that the language required a categorical approach. By then courts had, as the government puts it, “beg[u]n to settle” on the view that §16(b) demanded a categorical analysis. Brief for United States 36–37. Of particular significance, the Second Circuit, along with a number of district courts, had relied on the categorical approach to hold that selling drugs could never qualify as a crime of violence because “[w]hile the traffic in drugs is often accompanied by violence,” it can also be carried out through consensual sales and thus “does not by its nature involve substantial risk that physical violence will be used.” United States v. Diaz , 778 F. 2d 86, 88 (1985). Congress moved quickly to abrogate those decisions. But, notably, it didn’t do so by directing a case-specific approach or changing the language courts had read to require the categorical approach. Instead, it accepted the categorical approach as given and simply declared that certain drug trafficking crimes automatically trigger §924 penalties, regardless of the risk of violence that attends them. §§104(a)(2)(B)–(C), 100 Stat. 457.¶

The government’s reply to this development misses the mark. The government argues that §16(b) had not acquired such a well-settled judicial construction by 1986 that the reenactment of its language in §924(c)(3)(B) should be presumed to have incorporated the same construction. We agree. See Jerman v. Carlisle, McNellie, Rini, Kramer & Ulrich, L. P. A. , 559 U. S. 573, 590 (2010) (interpretations of three courts of appeals “may not have ‘settled’ the meaning” of a statute for purposes of the reenactment canon). But Congress in 1986 did more than just reenact language that a handful of courts had interpreted to require the categorical approach. It amended §924(c) specifically to abrogate the results of those decisions, without making any attempt to overturn the categorical reading on which they were based. And that would have been an odd way of proceeding if Congress had thought the categorical reading erroneous.¶

There’s yet one further and distinct way in which §924(c)’s history undermines the government’s casespecific reading of the residual clause. As originally enacted in 1968, §924(c) prohibited the use of a firearm in connection with any federal felony. §102, 82 Stat. 1224. The 1984 amendments narrowed §924(c) by limiting its predicate offenses to “crimes of violence.” But the casespecific reading would go a long way toward nullifying that limitation and restoring the statute’s original breadth. After all, how many felonies don’t involve a substantial risk of physical force when they’re committed using a firearm—let alone when the defendant brandishes or discharges the firearm?¶

Recognizing this difficulty, the government assures us that a jury wouldn’t be allowed to find a felony to be a crime of violence solely because the defendant used a firearm, although it could consider the firearm as a “factor.” Tr. of Oral Arg. 8. But the government identifies no textual basis for this rule, and exactly how it would work in practice is anyone’s guess. The government says, for example, that “selling counterfeit handbags” while carrying a gun wouldn’t be a crime of violence under its approach. Id. , at 9. But why not? Because the counterfeithandbag trade is so inherently peaceful that there’s no substantial risk of a violent confrontation with dissatisfied customers, territorial competitors, or dogged police officers? And how are jurors supposed to determine that? The defendant presumably knew the risks of his trade, and he chose to arm himself. See United States v. Simms , 914 F. 3d 229, 247–248 (CA4 2019) (en banc) (refusing to “condem[n] jurors to such an ill-defined inquiry”). Even granting the government its handbag example, we suspect its approach would result in the vast majority of federal felonies becoming potential predicates for §924(c) charges, contrary to the limitation Congress deliberately imposed when it restricted the statute’s application to crimes of violence.¶

D¶

With all this statutory evidence now arrayed against it, the government answers that it should prevail anyway because of the canon of constitutional avoidance. Maybe the case-specific approach doesn’t represent the best reading of the statute—but, the government insists, it is our duty to adopt any ''fairly possible'' reading of a statute to save it from being held unconstitutional. Brief for United States 45. 6¶

We doubt, however, the canon could play a proper role in this case even if the government’s reading were “possible.” True, when presented with two “fair alternatives,” this Court has sometimes adopted the narrower construction of a criminal statute to avoid having to hold it unconstitutional if it were construed more broadly. United States v. Rumely , 345 U. S. 41, 45, 47 (1953); see, e.g. , Skilling v. United States , 561 U. S. 358, 405–406, and n. 40 (2010); United States v. Lanier , 520 U. S. 259, 265267, and n. 6 (1997). But no one before us has identified a case in which this Court has invoked the canon to expand the reach of a criminal statute in order to save it. Yet that¶

------¶

6 There are at least two different canons of construction that sometimes go by the name “constitutional avoidance.” The one the government invokes here is perhaps better termed the presumption of constitutionality. Of long lineage, it holds that courts should, if possible, interpret ambiguous statutes to avoid rendering them unconstitutional, see, e.g. , Parsons v. Bedford , 3 Pet. 433, 448–449 (1830) (Story, J.), and it is distinct from the more modern (and more debated) constitutional doubt canon, which suggests courts should construe ambiguous statutes to avoid the need even to address serious questions about their constitutionality, see Rust v. Sullivan , 500 U. S. 173, 190–191 (1991).¶

is exactly what the government seeks here. Its casespecific reading would cause §924(c)(3)(B)’s penalties to apply to conduct they have not previously been understood to reach: categorically nonviolent felonies committed in violent ways. See Simms , 914 F. 3d, at 256–257 (Wynn, J., concurring). 7¶

Employing the avoidance canon to expand a criminal statute’s scope would risk offending the very same due process and separation-of-powers principles on which the vagueness doctrine itself rests. See supra , at 4–5. Everyone agrees that Mr. Davis and Mr. Glover did many things that Congress had declared to be crimes; and no matter how we rule today, they will face substantial prison sentences for those offenses. But does §924(c)(3)(B) require them to suffer additional punishment, on top of everything else? Even if you think it’s possible to read the statute to impose such additional punishment, it’s impossible to say that Congress surely intended that result, or that the law gave Mr. Davis and Mr. Glover fair warning that §924(c)’s mandatory penalties would apply to their conduct. Respect for due process and the separation of powers suggests a court may not, in order to save Congress the trou-¶

------¶

7 The government claims to have found cases invoking the canon to expand a statute’s reach, but none actually stands for that proposition. Each simply remarks in passing that a construction the Court arrived at for other reasons had the additional benefit of avoiding vagueness concerns; none suggests that a narrower construction was available. See United States v. Grace , 461 U. S. 171, 176 (1983) (accepting government’s construction, which was “not contested by appellees”); United States v. Culbert , 435 U. S. 371, 379 (1978) (finding statute clear and refusing to “manufacture ambiguity where none exists”); United States v. Shreveport Grain & Elevator Co. , 287 U. S. 77, 82–83 (1932) (finding statute unambiguous and construing it according to “the natural import of its terms”). And the dissent, despite compiling a page-long list of constitutional avoidance cases spanning “more than 200 years,” post , at 25–26, has been unable to find any better examples. See post , at 29–30 (opinion of KAVANAUGH, J.).¶

ble of having to write a new law, construe a criminal statute to penalize conduct it does not clearly proscribe.¶

Employing the canon as the government wishes would also sit uneasily with the rule of lenity’s teaching that ambiguities about the breadth of a criminal statute should be resolved in the defendant’s favor. That rule is “perhaps not much less old than” the task of statutory “construction itself.” United States v. Wiltberger , 5 Wheat. 76, 95 (1820) (Marshall, C. J.). And much like the vagueness doctrine, it is founded on “the tenderness of the law for the rights of individuals” to fair notice of the law “and on the plain principle that the power of punishment is vested in the legislative, not in the judicial department.” Ibid. ; see Lanier , 520 U. S., at 265–266, and n. 5. Applying constitutional avoidance to narrow a criminal statute, as this Court has historically done, accords with the rule of lenity. By contrast, using the avoidance canon instead to adopt a more expansive reading of a criminal statute would place these traditionally sympathetic doctrines at war with one another. 8¶

IV¶

What does the dissent have to say about all this? It starts by emphasizing that §924(c)(3)(B) has been used in “tens of thousands of federal prosecutions” since its enactment 33 years ago. Post , at 2 (opinion of KAVANAUGH, J.). And the dissent finds it “surprising” and “extraordinary” that, after all those prosecutions over all that time,¶

------¶

8 Admittedly, abandoning the categorical approach in favor of the case-specific approach would also have the effect of excluding from the statute’s coverage defendants who commit categorically violent felonies in nonviolent ways, and in that respect would be more “lenient” for some defendants. Regardless, the constitutional principles underlying the rule of lenity counsel caution before invoking constitutional avoidance to construe the statute to punish conduct that it does not unambiguously proscribe.¶

the statute could “suddenly” be deemed unconstitutional. Post , at 2–3. But the government concedes that §924(c)(3)(B) is unconstitutional if it means what everyone has understood it to mean in nearly all of those prosecutions over all those years. So the only way the statute can be saved is if we were “suddenly” to give it a new meaning different from the one it has borne for the last three decades. And if we could do that , it would indeed be “surprising” and “extraordinary.”¶

The dissent defends giving this old law a new meaning by appealing to intuition. It suggests that a categorical reading of §924(c)(3)(B) is “unnatural” because “[i]f you were to ask John Q. Public whether a particular crime posed a substantial risk of violence, surely he would respond, ‘Well, tell me how it went downwhat happened ?’ ” Post , at 13 (some internal quotation marks omitted). Maybe so. But the language in the statute before us isn’t the language posited in the dissent’s push poll. Section 924(c)(3)(B) doesn’t ask about the risk that “a particular crime posed” but about the risk that an “offense . . . by its nature, involves.” And a categorical reading of this categorical language seemed anything but “unnatural” to the unanimous Court in Leocal or the plurality in Dimaya . 9 Nor did the government think the categorical reading of §924(c)(3)(B) “unnatural” when it embraced that reading for decades. The dissent asks us to overlook the government’s prior view, explaining that the government only defended a categorical reading of the statute “when it did not matter for constitutional vagueness purposes”—that is, before Johnson and Dimaya identified constitutional problems with the categorical approach. Post , at 34. But¶

------¶

9 To be sure, the dissent suggests that Leocal and Dimaya adopted a categorical reading simply to avoid practical and constitutional problems. Post , at 15–16, 23, and n. 23. But, as we have seen, this too is mistaken. Leocal did not even mention those problems, and Dimaya held that the text demanded a categorical approach. See supra , at 9.¶

isn’t that exactly the point? Isn’t it at least a little revealing that, when the government had no motive to concoct an alternative reading, even it thought the best reading of §924(c)(3)(B) demanded a categorical analysis?¶

If this line of attack won’t work, the dissent tries another by telling us that we have “not fully account[ed] for the long tradition of substantial-risk criminal statutes.” Post , at 34. The dissent proceeds to offer a lengthy bill of particulars, citing dozens of state and federal laws that do not use the categorical approach. Post , at 7–10, and nn. 4–17. But what does this prove? Most of the statutes the dissent cites impose penalties on whoever “creates,” or “engages in conduct that creates,” or acts under “circumstances that create” a substantial risk of harm; others employ similar language. Not a single one imposes penalties for committing certain acts during “an offense . . . that by its nature, involves” a substantial risk, or anything similar. Marching through the dissent’s own catalog thus only winds up confirming that legislatures know how to write risk-based statutes that require a case-specific analysis—and that §924(c)(3)(B) is not a statute like that.¶

When the dissent finally turns to address the words Congress actually wrote in §924(c)(3)(B), its main argument seems to be that a categorical reading violates the canon against superfluity. On this account, reading “offense” generically in connection with the residual clause makes the residual clause “duplicate” the elements clause and leaves it with “virtually nothing” to do. Post , at 20. But that is a surprising assertion coming from the dissent, which devotes several pages to describing the “many” offenders who have been convicted under the residual clause using the categorical approach but who “might not” be prosecutable under the elements clause. Post , at 30–33. It is also wrong. As this Court has long understood, the residual clause, read categorically, “sweeps more broadly” than the elements clause—potentially reaching offenses, like burglary, that do not have violence as an element but that arguably create a substantial risk of violence. Leocal , 543 U. S., at 10. So even under the categorical reading, the residual clause is far from superfluous.¶

Without its misplaced reliance on the superfluity canon, there is little left of the dissent’s textual analysis. The dissent asserts that the phrase “by its nature” must “focu[s] on the defendant’s actual conduct”—but only because this “follows” from the dissent’s earlier (and mistaken) superfluity argument. Post , at 21. Next, the dissent claims that “the word ‘involves'' and “the phrase ‘in the course of committing the offense'' both support a casespecific approach. Post , at 22. But these words do not favor either reading: It is just as natural to ask whether the offense of robbery ordinarily “involves” a substantial risk that violence will be used “in the course of committing the offense” as it is to ask whether a particular robbery “involved” a substantial risk that violence would be used “in the course of committing the offense.” If anything, the statute’s use of the present and not the past tense lends further support to the categorical reading. 10 The dissent thinks it significant, too, that the statute before us “does not use the term ‘conviction,'' post , at 23; but that word is hardly a prerequisite for the categorical approach, as Dimaya makes clear. Remarkably, the dissent has noth-¶

------¶

10 The dissent claims that Taylor v. United States , 495 U. S. 575 (1990), and Nijhawan v. Holder , 557 U. S. 29 (2009), pointed to “the absence of the word ‘involved’ ” as one reason to adopt a categorical approach. Post , at 22. Not true. Taylor explained that the ACCA’s elements clause requires a categorical approach in part because it refers to a crime “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. All the work in that sentence was being done by the phrase “in a particular case,” not by the word “involves.” And Nijhawan noted that the Court had construed the ACCA’s residual clause, which refers to crimes “that ‘ involv[e] conduct that presents a serious potential risk of physical injury,” to require the categorical approach. 557 U. S., at 36.¶

ing at all to say about §924(c)(3)’s history or its relationship with other criminal statutes; it just ignores those arguments. And when it comes to the constitutional avoidance canon, the dissent does not even try to explain how using that canon to criminalize conduct that isn’t criminal under the fairest reading of a statute might be reconciled with traditional principles of fair notice and separation of powers. Instead, the dissent seems willing to consign ''thousands'' of defendants to prison for “years—potentially decades,” not because it is certain or even likely that Congress ordained those penalties, but because it is merely “possible” Congress might have done so. Post , at 30, 33–34. In our republic, a speculative possibility that a man’s conduct violated the law should never be enough to justify taking his liberty.¶

In the end, the dissent is forced to argue that holding §924(c)(3)(B) unconstitutional would invite “bad” social policy consequences. Post , at 34. In fact, the dissent’s legal analysis only comes sandwiched between a lengthy paean to laws that impose severe punishments for gun crimes and a rogue’s gallery of offenses that may now be punished somewhat less severely. See post , at 1–2, 30–34. The dissent acknowledges that “the consequences cannot change our understanding of the law.” Post , at 34. But what’s the point of all this talk of “bad” consequences if not to suggest that judges should be tempted into reading the law to satisfy their policy goals? Even taken on their own terms, too, the dissent’s policy concerns are considerably overblown. While the dissent worries that our ruling may elicit challenges to past §924(c) convictions, post , at 33, the dissent’s preferred approach—saving §924(c)(3)(B) by changing its meaning—would also call into question countless convictions premised on the categorical reading. And defendants whose §924(c) convictions are overturned by virtue of today’s ruling will not even necessarily receive lighter sentences: As this Court has noted, when a defendant’s §924(c) conviction is invalidated, courts of appeals “routinely” vacate the defendant’s entire sentence on all counts “so that the district court may increase the sentences for any remaining counts” if such an increase is warranted. Dean v. United States , 581 U. S. ___, ___ (2017) (slip op., at 5).¶

Of course, too, Congress always remains free to adopt a case-specific approach to defining crimes of violence for purposes of §924(c)(3)(B) going forward. As Mr. Davis and Mr. Glover point out, one easy way of achieving that goal would be to amend the statute so it covers any felony that, “based on the facts underlying the offense, involved a substantial risk” that physical force against the person or property of another would be used in the course of committing the offense. Brief for Respondents 46 (quoting H. R. 7113, 115th Cong., 2d Sess. (2018); emphasis deleted); see also Tr. of Oral Arg. 19 (government’s counsel agreeing that this language would offer “clearer” support for the case-specific approach than the current version of the statute does). The dissent’s catalog of case-specific, riskbased criminal statutes supplies plenty of other models Congress could follow. Alternatively still, Congress might choose to retain the categorical approach but avoid vagueness in other ways, such as by defining crimes of violence to include certain enumerated offenses or offenses that carry certain minimum penalties. All these options and more are on the table. But these are options that belong to Congress to consider; no matter how tempting, this Court is not in the business of writing new statutes to right every social wrong it may perceive.¶

*¶

We agree with the court of appeals’ conclusion that §924(c)(3)(B) is unconstitutionally vague. At the same time, exactly what that holding means for Mr. Davis and Mr. Glover remains to be determined. After the Fifth¶

Circuit vacated their convictions and sentences on one of the two §924(c) counts at issue, both men sought rehearing and argued that the court should have vacated their sentences on all counts. In response, the government conceded that, if §924(c)(3)(B) is held to be vague, then the defendants are entitled to a full resentencing, not just the more limited remedy the court had granted them. The Fifth Circuit has deferred ruling on the rehearing petitions pending our decision, so we remand the case to allow the court to address those petitions. The judgment below is affirmed in part and vacated in part, and the case is remanded for further proceedings consistent with this opinion.¶

It is so ordered.¶