Descamps v. United States

Justice KAGAN delivered the opinion of the Court.

*257The Armed Career Criminal Act (ACCA or Act), 18 U.S.C. § 924(e), increases the sentences of certain federal defendants who have three prior convictions "for a violent felony," including "burglary, arson, or extortion." To determine whether a past conviction is for one of those crimes, courts use what has become known as the "categorical approach": They compare the elements of the statute forming the basis of the defendant's conviction with the elements of the "generic" crime-i.e., the offense as commonly understood. The prior conviction qualifies as an ACCA predicate only if the statute's elements are the same as, or narrower than, those of the generic offense.

We have previously approved a variant of this method-labeled (not very inventively) the "modified categorical approach"-when a prior conviction is for violating a so-called "divisible statute." That kind of statute sets out one or more elements of the offense in the alternative-for example, stating that burglary involves entry into a building or an automobile. If one alternative (say, a building) matches an element in the generic offense, but the other (say, an automobile) does not, the modified categorical approach permits sentencing courts to consult a limited class of documents, such as indictments and jury instructions, to determine which alternative formed the basis of the defendant's prior conviction. The court can then do what the categorical approach demands: compare the elements of the crime of conviction (including the alternative element used in the case) with the elements of the generic crime.

*258This case presents the question whether sentencing courts may also consult those additional documents when a defendant was convicted under an "indivisible" statute-i.e., one not containing alternative elements-that criminalizes a broader swath of conduct than the relevant generic offense. That would enable a court to decide, based on information about a case's underlying facts, that the defendant's prior conviction qualifies as an ACCA predicate even though the elements *2282of the crime fail to satisfy our categorical test. Because that result would contravene our prior decisions and the principles underlying them, we hold that sentencing courts may not apply the modified categorical approach when the crime of which the defendant was convicted has a single, indivisible set of elements.

I

Petitioner Michael Descamps was convicted of being a felon in possession of a firearm, in violation of 18 U.S.C. § 922(g). That unadorned offense carries a maximum penalty of 10 years in prison. The Government, however, sought an enhanced sentence under ACCA, based on Descamps' prior state convictions for burglary, robbery, and felony harassment.

ACCA prescribes a mandatory minimum sentence of 15 years for a person who violates § 922(g) and "has three previous convictions ... for a violent felony or a serious drug offense." § 924(e)(1). The Act defines a "violent felony" to mean any felony, whether state or federal, that "has as an element the use, attempted use, or threatened use of physical force against the person of another," or that "is burglary, arson, or extortion, involves use of explosives, or otherwise involves conduct that presents a serious potential risk of physical injury to another." § 924(e)(2)(B).

Descamps argued that his prior burglary conviction could not count as an ACCA predicate offense under our categorical approach. He had pleaded guilty to violating California Penal Code Ann. § 459 (West 2010), which provides that a "person *259who enters" certain locations "with intent to commit grand or petit larceny or any felony is guilty of burglary." That statute does not require the entry to have been unlawful in the way most burglary laws do. Whereas burglary statutes generally demand breaking and entering or similar conduct, California's does not: It covers, for example, a shoplifter who enters a store, like any customer, during normal business hours. See People v. Barry, 94 Cal. 481, 483-484, 29 P. 1026, 1026-1027 (1892). In sweeping so widely, the state law goes beyond the normal, "generic" definition of burglary. According to Descamps, that asymmetry of offense elements precluded his conviction under § 459 from serving as an ACCA predicate, whether or not his own burglary involved an unlawful entry that could have satisfied the requirements of the generic crime.

The District Court disagreed. According to the court, our modified categorical approach permitted it to examine certain documents, including the record of the plea colloquy, to discover whether Descamps had "admitted the elements of a generic burglary" when entering his plea. App. 50a. And that transcript, the court ruled, showed that Descamps had done so. At the plea hearing, the prosecutor proffered that the crime " ' involve[d] the breaking and entering of a grocery store,' " and Descamps failed to object to that statement. Ibid. The plea proceedings, the District Court thought, thus established that Descamps' prior conviction qualified as a generic burglary (and so as a "violent felony") under ACCA. Applying the requisite penalty enhancement, the court sentenced Descamps to 262 months in prison-more than twice the term he would otherwise have received.

The Court of Appeals for the Ninth Circuit affirmed, relying on its recently issued decision in United States v. Aguila-Montes de Oca, 655 F.3d 915 (2011) (en banc) (per curiam ). There, a divided en banc court took much the same view of the modified categorical approach as had the District Court in this case. The en banc court held that when a sentencing *260court considers a conviction under § 459 -or *2283any other statute that is "categorically broader than the generic offense"-the court may scrutinize certain documents to determine the factual basis of the conviction. See id., at 940. Applying that approach, the Court of Appeals here found that Descamps' plea, as revealed in the colloquy, "rested on facts that satisfy the elements of the generic definition of burglary." 466 Fed.Appx. 563, 565 (2012).

We granted certiorari, 567 U.S. ----, 133 S.Ct. 90, 183 L.Ed.2d 730 (2012), to resolve a Circuit split on whether the modified categorical approach applies to statutes like § 459 that contain a single, "indivisible" set of elements sweeping more broadly than the corresponding generic offense.1 We hold that it does not, and so reverse.

II

Our caselaw explaining the categorical approach and its "modified" counterpart all but resolves this case. In those decisions, as shown below, the modified approach serves a limited function: It helps effectuate the categorical analysis when a divisible statute, listing potential offense elements in the alternative, renders opaque which element played a part in the defendant's conviction. So understood, the modified approach cannot convert Descamps' conviction under § 459 into an ACCA predicate, because that state law defines burglary not alternatively, but only more broadly than the generic offense.

We begin with Taylor v. United States, 495 U.S. 575, 110 S.Ct. 2143, 109 L.Ed.2d 607 (1990), which established the rule for determining when a defendant's prior conviction counts as one of ACCA's enumerated *261predicate offenses (e.g., burglary). Taylor adopted a "formal categorical approach": Sentencing courts may "look only to the statutory definitions"-i.e., the elements-of a defendant's prior offenses, and not "to the particular facts underlying those convictions." Id., at 600, 110 S.Ct. 2143. If the relevant statute has the same elements as the "generic" ACCA crime, then the prior conviction can serve as an ACCA predicate; so too if the statute defines the crime more narrowly, because anyone convicted under that law is "necessarily ... guilty of all the [generic crime's] elements." Id., at 599, 110 S.Ct. 2143. But if the statute sweeps more broadly than the generic crime, a conviction under that law cannot count as an ACCA predicate, even if the defendant actually committed the offense in its generic form. The key, we emphasized, is elements, not facts. So, for example, we held that a defendant can receive an ACCA enhancement for burglary only if he was convicted of a crime having "the basic elements" of generic burglary-i.e., "unlawful or unprivileged entry into, or remaining in, a building or structure, with intent to commit a crime." Ibid . And indeed, we indicated that the very statute at issue here, § 459, does not fit that bill because "California defines 'burglary' so broadly as to include shoplifting." Id., at 591, 110 S.Ct. 2143.

At the same time, Taylor recognized a "narrow range of cases" in which sentencing courts-applying what we would later dub the "modified categorical approach"-*2284may look beyond the statutory elements to "the charging paper and jury instructions" used in a case. Id., at 602, 110 S.Ct. 2143. To explain when courts should resort to that approach, we hypothesized a statute with alternative elements-more particularly, a burglary statute (otherwise conforming to the generic crime) that prohibits "entry of an automobile as well as a building." Ibid. One of those alternatives (a building) corresponds to an element in generic burglary, whereas the other (an automobile) does not. In a typical case brought under the statute, the prosecutor charges one of those two alternatives, *262and the judge instructs the jury accordingly. So if the case involves entry into a building, the jury is "actually required to find all the elements of generic burglary," as the categorical approach demands. Ibid. But the statute alone does not disclose whether that has occurred. Because the statute is "divisible"-i.e., comprises multiple, alternative versions of the crime-a later sentencing court cannot tell, without reviewing something more, if the defendant's conviction was for the generic (building) or non-generic (automobile) form of burglary. Hence Taylor permitted sentencing courts, as a tool for implementing the categorical approach, to examine a limited class of documents to determine which of a statute's alternative elements formed the basis of the defendant's prior conviction.

In Shepard v. United States, 544 U.S. 13, 125 S.Ct. 1254, 161 L.Ed.2d 205 (2005), the hypothetical we posited in Taylor became real: We confronted a Massachusetts burglary statute covering entries into "boats and cars" as well as buildings. 544 U.S., at 17, 125 S.Ct. 1254. The defendant there pleaded guilty to violating the statute, and we first confirmed that Taylor 's categorical approach applies not just to jury verdicts, but also to plea agreements. That meant, we held, that a conviction based on a guilty plea can qualify as an ACCA predicate only if the defendant "necessarily admitted [the] elements of the generic offense." Id., at 26, 125 S.Ct. 1254. But as we had anticipated in Taylor, the divisible nature of the Massachusetts burglary statute confounded that inquiry: No one could know, just from looking at the statute, which version of the offense Shepard was convicted of. Accordingly, we again authorized sentencing courts to scrutinize a restricted set of materials-here, "the terms of a plea agreement or transcript of colloquy between judge and defendant"-to determine if the defendant had pleaded guilty to entering a building or, alternatively, a car or boat. Ibid. Yet we again underscored the narrow scope of that review: It was not to determine "what the defendant and state judge must have understood as the factual basis of the *263prior plea," but only to assess whether the plea was to the version of the crime in the Massachusetts statute (burglary of a building) corresponding to the generic offense. Id., at 25-26, 125 S.Ct. 1254 (plurality opinion).

Two more recent decisions have further emphasized the elements-based rationale-applicable only to divisible statutes-for examining documents like an indictment or plea agreement. In Nijhawan v. Holder, 557 U.S. 29, 129 S.Ct. 2294, 174 L.Ed.2d 22 (2009), we discussed another Massachusetts statute, this one prohibiting " 'Breaking and Entering at Night' " in any of four alternative places: a "building, ship, vessel, or vehicle." Id., at 35, 129 S.Ct. 2294. We recognized that when a statute so "refer[s] to several different crimes," not all of which qualify as an ACCA predicate, a court must determine which crime formed the basis of the defendant's conviction. Ibid. That is why, we explained, Taylor and Shepard developed the modified categorical *2285approach. By reviewing the extra-statutory materials approved in those cases, courts could discover "which statutory phrase," contained within a statute listing "several different" crimes, "covered a prior conviction." 557 U.S., at 41, 129 S.Ct. 2294. And a year later, we repeated that understanding of when and why courts can resort to those documents: "[T]he 'modified categorical approach' that we have approved permits a court to determine which statutory phrase was the basis for the conviction." Johnson v. United States, 559 U.S. 133, 144, 130 S.Ct. 1265, 176 L.Ed.2d 1 (2010) (citation omitted).

Applied in that way-which is the only way we have ever allowed-the modified approach merely helps implement the categorical approach when a defendant was convicted of violating a divisible statute. The modified approach thus acts not as an exception, but instead as a tool. It retains the categorical approach's central feature: a focus on the elements, rather than the facts, of a crime. And it preserves the categorical approach's basic method: comparing those elements with the generic offense's. All the modified approach adds is a mechanism for making that comparison *264when a statute lists multiple, alternative elements, and so effectively creates "several different ... crimes." Nijhawan, 557 U.S., at 41, 129 S.Ct. 2294. If at least one, but not all of those crimes matches the generic version, a court needs a way to find out which the defendant was convicted of. That is the job, as we have always understood it, of the modified approach: to identify, from among several alternatives, the crime of conviction so that the court can compare it to the generic offense.2

The modified approach thus has no role to play in this case. The dispute here does not concern any list of alternative elements. Rather, it involves a simple discrepancy between generic burglary and the crime established in § 459. The former requires an unlawful entry along the lines of breaking and entering. See 3 W. LaFave, Substantive Criminal Law § 21.1(a) (2d ed. 2003) (hereinafter LaFave). The latter does not, and indeed covers simple shoplifting, as even the Government *265acknowledges. See Brief for United States 38; Barry, 94 Cal., at 483-484, 29 P., at 1026-1027. In Taylor 's words, then, § 459"define[s] burglary more broadly" than the generic offense. 495 U.S., at 599, 110 S.Ct. 2143. And because that is true-because California, *2286to get a conviction, need not prove that Descamps broke and entered-a § 459 violation cannot serve as an ACCA predicate. Whether Descamps did break and enter makes no difference. And likewise, whether he ever admitted to breaking and entering is irrelevant. Our decisions authorize review of the plea colloquy or other approved extra-statutory documents only when a statute defines burglary not (as here) overbroadly, but instead alternatively, with one statutory phrase corresponding to the generic crime and another not. In that circumstance, a court may look to the additional documents to determine which of the statutory offenses (generic or non-generic) formed the basis of the defendant's conviction. But here no uncertainty of that kind exists, and so the categorical approach needs no help from its modified partner. We know Descamps' crime of conviction, and it does not correspond to the relevant generic offense. Under our prior decisions, the inquiry is over.

III

The Court of Appeals took a different view. Dismissing everything we have said on the subject as "lack[ing] conclusive weight," the Ninth Circuit held in Aguila-Montes that the modified categorical approach could turn a conviction under any statute into an ACCA predicate offense. 655 F.3d, at 931. The statute, like § 459, could contain a single, indivisible set of elements covering far more conduct than the generic crime-and still, a sentencing court could "conside[r] to some degree the factual basis for the defendant's conviction" or, otherwise stated, "the particular acts the defendant committed." Id., at 935-936. More specifically, the court could look to reliable materials (the charging document, jury instructions, plea colloquy, and so forth) to determine *266"what facts" can "confident[ly]" be thought to underlie the defendant's conviction in light of the " prosecutorial theory of the case" and the "facts put forward by the government." Id., at 936-937. It makes no difference, in the Ninth Circuit's view, whether "specific words in the statute" of conviction " ' actually required' " the jury (or judge accepting a plea) "to find a particular generic element." Id., at 936 (quoting Taylor, 495 U.S., at 602, 110 S.Ct. 2143; internal quotation marks omitted).3 *2287That approach-which an objecting judge aptly called "modified factual," 655 F.3d, at 948 (Berzon, J., concurring in judgment)-turns an elements-based inquiry into an *267evidence-based one. It asks not whether "statutory definitions" necessarily require an adjudicator to find the generic offense, but instead whether the prosecutor's case realistically led the adjudicator to make that determination. And it makes examination of extra-statutory documents not a tool used in a "narrow range of cases" to identify the relevant element from a statute with multiple alternatives, but rather a device employed in every case to evaluate the facts that the judge or jury found. By this point, it should be clear that the Ninth Circuit's new way of identifying ACCA predicates has no roots in our precedents. But more: Aguila-Montes subverts those decisions, conflicting with each of the rationales supporting the categorical approach and threatening to undo all its benefits.

A

This Court offered three grounds for establishing our elements-centric, "formal categorical approach." Taylor, 495 U.S., at 600, 110 S.Ct. 2143. First, it comports with ACCA's text and history. Second, it avoids the Sixth Amendment concerns that would arise from sentencing courts' making findings of fact that properly belong to juries. And third, it averts "the practical difficulties and potential unfairness of a factual approach." Id., at 601, 110 S.Ct. 2143. When assessed in light of those three reasons, the Ninth Circuit's ruling strikes out swinging.

Start with the statutory text and history. As we have long recognized, ACCA increases the sentence of a defendant who has three "previous convictions" for a violent felony-not a defendant who has thrice committed such a crime. 18 U.S.C. § 924(e)(1) ; see Taylor, 495 U.S., at 600, 110 S.Ct. 2143. That language shows, as Taylor explained, 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 not to the facts underlying the prior convictions." Ibid. ; see Shepard, 544 U.S., at 19, 125 S.Ct. 1254. If Congress had wanted to increase a sentence based on the facts of a prior offense, it presumably would have said so; other statutes, in other *268contexts, speak in just that way. See Nijhawan, 557 U.S., at 36, 129 S.Ct. 2294 (construing an immigration statute as requiring a " 'circumstance-specific,' not a 'categorical,' " approach). But in ACCA, Taylor found, Congress made a deliberate decision to treat every conviction of a crime in the same manner: During the lengthy debate preceding the statute's enactment, "no one suggested that a particular crime might sometimes count towards enhancement and sometimes not, depending on the facts of the case." 495 U.S., at 601, 110 S.Ct. 2143. Congress instead meant ACCA to function as an on-off switch, directing that a prior crime would qualify as a predicate offense in all cases or in none.

The Ninth Circuit's approach runs headlong into that congressional choice. Instead of reviewing documents like an indictment or plea colloquy only to determine "which statutory phrase was the basis for the conviction," the Ninth Circuit looks to those materials to discover what the defendant actually did. Johnson, 559 U.S., at 144, 130 S.Ct. 1265. This case demonstrates the point. Descamps was not convicted of generic burglary, because (as the Government agrees) § 459 does not contain that crime's required unlawful-*2288entry element. See Brief for United States 38, 43-44. At most, the colloquy showed that Descamps committed generic burglary, and so hypothetically could have been convicted under a law criminalizing that conduct. But that is just what we said, in Taylor and elsewhere, is not enough. See 495 U.S., at 600, 110 S.Ct. 2143 ; Carachuri-Rosendo v. Holder, 560 U.S. ----, ----, 130 S.Ct. 2577, 2586, 177 L.Ed.2d 68 (2010) (rejecting such a " 'hypothetical approach' " given a similar statute's directive to "look to the conviction itself," rather than "to what might have or could have been charged"). And the necessary result of the Ninth Circuit's method is exactly the differential treatment we thought Congress, in enacting ACCA, took care to prevent. In the two years since Aguila-Montes, the Ninth Circuit has treated some, but not other, convictions under § 459 as ACCA predicates, based on minor variations in the cases' plea documents. Compare, *269e.g., 466 Fed.Appx., at 565 (Descamps' § 459 conviction counts as generic burglary), with 655 F.3d, at 946 (Aguila-Montes' does not).

Similarly, consider (though Aguila-Montes did not) the categorical approach's Sixth Amendment underpinnings. We have held that "[o]ther than the fact of a prior conviction, any fact that increases the penalty for a crime beyond the prescribed statutory maximum must be submitted to a jury, and proved beyond a reasonable doubt." Apprendi v. New Jersey, 530 U.S. 466, 490, 120 S.Ct. 2348, 147 L.Ed.2d 435 (2000). Under ACCA, the court's finding of a predicate offense indisputably increases the maximum penalty. Accordingly, that finding would (at the least) raise serious Sixth Amendment concerns if it went beyond merely identifying a prior conviction. Those concerns, we recognized in Shepard, counsel against allowing a sentencing court to "make a disputed" determination "about what the defendant and state judge must have understood as the factual basis of the prior plea," or what the jury in a prior trial must have accepted as the theory of the crime. 544 U.S., at 25, 125 S.Ct. 1254 (plurality opinion); see id., at 28, 125 S.Ct. 1254 (THOMAS, J., concurring in part and concurring in judgment) (stating that such a finding would "giv[e] rise to constitutional error, not doubt"). Hence our insistence on the categorical approach.

Yet again, the Ninth Circuit's ruling flouts our reasoning-here, by extending judicial factfinding beyond the recognition of a prior conviction. Our modified categorical approach merely assists the sentencing court in identifying the defendant's crime of conviction, as we have held the Sixth Amendment permits. But the Ninth Circuit's reworking authorizes the court to try to discern what a trial showed, or a plea proceeding revealed, about the defendant's underlying conduct. See Aguila-Montes, 655 F.3d, at 937. And there's the constitutional rub. The Sixth Amendment contemplates that a jury-not a sentencing court-will find such facts, unanimously and beyond a reasonable doubt. And the only facts the court can be sure the jury so found are those *270constituting elements of the offense-as distinct from amplifying but legally extraneous circumstances. See, e.g., Richardson v. United States, 526 U.S. 813, 817, 119 S.Ct. 1707, 143 L.Ed.2d 985 (1999). Similarly, as Shepard indicated, when a defendant pleads guilty to a crime, he waives his right to a jury determination of only that offense's elements; whatever he says, or fails to say, about superfluous facts cannot license a later sentencing court to impose extra punishment. See 544 U.S., at 24-26, 125 S.Ct. 1254 (plurality opinion). So when the District Court here enhanced Descamps' sentence, based on *2289his supposed acquiescence to a prosecutorial statement (that he "broke and entered") irrelevant to the crime charged, the court did just what we have said it cannot: rely on its own finding about a non-elemental fact to increase a defendant's maximum sentence.

Finally, the Ninth Circuit's decision creates the same "daunting" difficulties and inequities that first encouraged us to adopt the categorical approach. Taylor, 495 U.S., at 601-602, 110 S.Ct. 2143. In case after case, sentencing courts following Aguila-Montes would have to expend resources examining (often aged) documents for evidence that a defendant admitted in a plea colloquy, or a prosecutor showed at trial, facts that, although unnecessary to the crime of conviction, satisfy an element of the relevant generic offense. The meaning of those documents will often be uncertain. And the statements of fact in them may be downright wrong. A defendant, after all, often has little incentive to contest facts that are not elements of the charged offense-and may have good reason not to. At trial, extraneous facts and arguments may confuse the jury. (Indeed, the court may prohibit them for that reason.) And during plea hearings, the defendant may not wish to irk the prosecutor or court by squabbling about superfluous factual allegations. In this case, for example, Descamps may have let the prosecutor's statement go by because it was irrelevant to the proceedings. He likely was not thinking about the possibility that his silence could come *271back to haunt him in an ACCA sentencing 30 years in the future. (Actually, he could not have been thinking that thought: ACCA was not even on the books at the time of Descamps' burglary conviction.)

Still worse, the Aguila-Montes approach will deprive some defendants of the benefits of their negotiated plea deals. Assume (as happens every day) that a defendant surrenders his right to trial in exchange for the government's agreement that he plead guilty to a less serious crime, whose elements do not match an ACCA offense. Under the Ninth Circuit's view, a later sentencing court could still treat the defendant as though he had pleaded to an ACCA predicate, based on legally extraneous statements found in the old record. Taylor recognized the problem: "[I]f a guilty plea to a lesser, nonburglary offense was the result of a plea bargain," the Court stated, "it would seem unfair to impose a sentence enhancement as if the defendant had pleaded guilty" to generic burglary. 495 U.S., at 601-602, 110 S.Ct. 2143. That way of proceeding, on top of everything else, would allow a later sentencing court to rewrite the parties' bargain.

B

The Ninth Circuit defended its (excessively) modified approach by denying any real distinction between divisible and indivisible statutes extending further than the generic offense. "The only conceptual difference," the court reasoned, "is that [a divisible statute] creates an explicitly finite list of possible means of commission, while [an indivisible one] creates an implied list of every means of commission that otherwise fits the definition of a given crime." Aguila-Montes, 655 F.3d, at 927. For example, an indivisible statute "requir[ing] use of a 'weapon' is not meaningfully different"-or so says the Ninth Circuit-"from a statute that simply lists every kind of weapon in existence ... ('gun, axe, sword, baton, slingshot, knife, machete, bat,' and so on)." Ibid. In a similar way, every indivisible statute can be imaginatively *272reconstructed as a divisible one. And if that is true, the Ninth Circuit asks, why limit the modified categorical *2290approach only to explicitly divisible statutes?

The simple answer is: Because only divisible statutes enable a sentencing court to conclude that a jury (or judge at a plea hearing) has convicted the defendant of every element of the generic crime. A prosecutor charging a violation of a divisible statute must generally select the relevant element from its list of alternatives. See, e.g., The Confiscation Cases, 20 Wall. 92, 104, 22 L.Ed. 320 (1874) ("[A]n indictment or a criminal information which charges the person accused, in the disjunctive, with being guilty of one or of another of several offences, would be destitute of the necessary certainty, and would be wholly insufficient").4 And the jury, as instructions in the case will make clear, must then find that element, unanimously and beyond a reasonable doubt. So assume, along the lines of the Ninth Circuit's example, that a statute criminalizes assault with any of eight specified weapons; and suppose further, as the Ninth Circuit did, that only assault with a gun counts as an ACCA offense. A later sentencing court need only check the charging documents and instructions ("Do they refer to a gun or something else?") to determine whether in convicting a defendant under that divisible statute, the jury necessarily found that he committed the ACCA-qualifying crime.

None of that is true of an overbroad, indivisible statute. A sentencing court, to be sure, can hypothetically reconceive such a statute in divisible terms. So, as Aguila-Montes reveals, *273a court blessed with sufficient time and imagination could devise a laundry list of potential "weapons"-not just the eight the Ninth Circuit mentioned, but also (for starters) grenades, pipe bombs, spears, tire irons, BB guns, nunchucks, and crossbows. But the thing about hypothetical lists is that they are, well, hypothetical. As long as the statute itself requires only an indeterminate "weapon," that is all the indictment must (or is likely to) allege and all the jury instructions must (or are likely to) mention. And most important, that is all the jury must find to convict the defendant. The jurors need not all agree on whether the defendant used a gun or a knife or a tire iron (or any other particular weapon that might appear in an imagined divisible statute), because the actual statute requires the jury to find only a "weapon." And even if in many cases, the jury could have readily reached consensus on the weapon used, a later sentencing court cannot supply that missing judgment. Whatever the underlying facts or the evidence presented, the defendant still would not have been convicted, in the deliberate and considered way the Constitution guarantees, of an offense with the same (or narrower) elements as the supposed generic crime (assault with a gun).

Indeed, accepting the Ninth Circuit's contrary reasoning would altogether collapse the distinction between a categorical and a fact-specific approach. After all, the Ninth Circuit's "weapons" example is just the tip of the iceberg: Courts can go much further in reconceiving indivisible statutes as impliedly divisible ones. In fact, every element of every statute can be imaginatively transformed as the Ninth *2291Circuit suggests-so that every crime is seen as containing an infinite number of sub-crimes corresponding to "all the possible ways an individual can commit" it. Aguila-Montes, 655 F.3d, at 927. (Think: Professor Plum, in the ballroom, with the candlestick?; Colonel Mustard, in the conservatory, with the rope, on a snowy day, to cover up his affair with Mrs. Peacock?) If a sentencing court, as the Ninth Circuit *274holds, can compare each of those "implied ... means of commission" to the generic ACCA offense, ibid. (emphasis deleted), then the categorical approach is at an end. At that point, the court is merely asking whether a particular set of facts leading to a conviction conforms to a generic ACCA offense. And that is what we have expressly and repeatedly forbidden. Courts may modify the categorical approach to accommodate alternative "statutory definitions." Ibid. ; cf. MCI Telecommunications Corp. v. American Telephone & Telegraph Co., 512 U.S. 218, 225, 114 S.Ct. 2223, 129 L.Ed.2d 182 (1994) (" ' [T]o modify' means to change moderately or in minor fashion"). They may not, by pretending that every fact pattern is an "implied" statutory definition, Aguila-Montes, 655 F.3d, at 927, convert that approach into its opposite.

IV

The Government tries to distance itself from the Ninth Circuit by offering a purportedly narrower theory-that although an indivisible statute that is "truly missing" an element of the generic offense cannot give rise to an ACCA conviction, California's burglary law can do so because it merely "contains a broader version of the [generic] element of unlawfulness of entry." Brief for United States 11-12. The Government's argument proceeds in three steps. It begins from the premise that sentencing courts applying ACCA should consider not only the statute defining a prior crime but also any judicial interpretations of it. Next, the Government points to a California decision holding (not surprisingly) that a defendant cannot "burglariz[e] his own home"; the case's reasoning, the Government notes, is that § 459 (though not saying so explicitly) requires "an entry which invades a possessory right." People v. Gauze, 15 Cal.3d 709, 713-716, 125 Cal.Rptr. 773, 542 P.2d 1365, 1367-1368 (1975). Given that precedent, the Government contends, § 459 includes a kind of "unlawful entry" element, although it is broader than the generic crime's analogous requirement. Finally, *275the Government asserts that sentencing courts may use the modified approach "to determine whether a particular defendant's conviction under" such an overbroad statute actually "was for [the] generic" crime. Brief for United States 11.

Although elaborately developed in the Government's brief, this argument's first two steps turn out to be sideshows. We may reserve the question whether, in determining a crime's elements, a sentencing court should take account not only of the relevant statute's text, but of judicial rulings interpreting it. And we may assume, as the Government insists, that California caselaw treats § 459 as including an element of entry "invading a possessory right"-although, truth be told, we find the state decisions on that score contradictory and confusing.5 Even on those assumptions, *2292§ 459's elements do not come into line with generic burglary's. As the Government concedes, almost every entry onto another's property with intent to steal-including, for example, a shoplifter's walking into an open store-"invades a possessory right" under § 459. See Brief for United States 38; Gauze, 15 Cal.3d, at 714, 125 Cal.Rptr. 773, 542 P.2d, at 1367. By contrast, generic burglary's unlawful-entry element excludes any case in which a person enters premises open to the public, no matter his intent; the generic crime requires breaking and entering or similar unlawful activity. See Brief for United States 38; LaFave § 21.1(a). So everything rests on the Government's *276third point: that this mismatch does not preclude applying the modified categorical approach, because it results not from a missing element but instead from an element's overbreadth.

But for starters, we see no principled way to make that distinction. Most overbroad statutes can also be characterized as missing an element; and most statutes missing an element can also be labeled overbroad. Here is the only conclusion in Aguila-Montes we agree with: "[I]t is difficult, if not impossible" to determine which is which. 655 F.3d, at 925. The example that court gave was as follows: A statute of conviction punishes possession of pornography, but a federal law carries a sentence enhancement for possession of child pornography. Is the statute of conviction overbroad because it includes both adult and child pornography; or is that law instead missing the element of involvement of minors? The same name game can be played with § 459. The Government labors mightily to turn what it fears looks like a missing-element statute into an overbroad statute through the incorporation of judicial decisions. But even putting those decisions aside, the Government might have described § 459 as merely having an overbroad element because "entry" includes both the lawful and the unlawful kind. And conversely, Descamps could claim that even as judicially interpreted, § 459 is entirely missing generic burglary's element of breaking and entering or similar unlawful conduct. All is in the eye of the beholder, and prone to endless manipulation.

In any event, and more fundamentally, we see no reason why the Government's distinction should matter. Whether the statute of conviction has an overbroad or missing element, the problem is the same: Because of the mismatch in elements, a person convicted under that statute is never convicted of the generic crime. In this case, for example, Descamps was not convicted of generic burglary because § 459, whether viewed as missing an element or containing an overbroad *277one, does not require breaking and entering. So every reason we have given-textual, constitutional, and practical-for rejecting the Ninth Circuit's proposed approach applies to the Government's as well. See supra, at 2287 - 2290. At bottom, the Government wants the same thing as the Ninth Circuit (if nominally in a few fewer cases): It too wishes a sentencing court to look beyond the elements to the evidence or, otherwise said, to explore whether a person convicted of one crime could also have been convicted of another, more serious offense. But that circumstance-specific review is just what the categorical approach precludes. And as we have explained, we adopted the modified approach to help implement *2293the categorical inquiry, not to undermine it.

V

Descamps may (or may not) have broken and entered, and so committed generic burglary. But § 459 -the crime of which he was convicted-does not require the factfinder (whether jury or judge) to make that determination. Because generic unlawful entry is not an element, or an alternative element, of § 459, a conviction under that statute is never for generic burglary. And that decides this case in Descamps' favor; the District Court should not have enhanced his sentence under ACCA.6 That court and the Ninth Circuit erred in invoking the modified categorical approach to look behind Descamps' conviction in search of *278record evidence that he actually committed the generic offense. The modified approach does not authorize a sentencing court to substitute such a facts-based inquiry for an elements-based one. A court may use the modified approach only to determine which alternative element in a divisible statute formed the basis of the defendant's conviction. Accordingly, we reverse the judgment of the Court of Appeals.

It is so ordered.

Compare, e.g., 466 Fed.Appx. 563, 565 (C.A.9 2012) (case below) (applying the modified categorical approach to § 459 ); United States v. Armstead, 467 F.3d 943, 947-950 (C.A.6 2006) (applying that approach to a similar, indivisible statute), with, e.g., United States v. Beardsley, 691 F.3d 252, 268-274 (C.A.2 2012) (holding that the modified categorical approach applies only to divisible statutes); United States v. Giggey, 551 F.3d 27, 40 (C.A.1 2008) (en banc) (same).

The dissent delves into the nuances of various States' laws in an effort to cast doubt on this understanding of our prior holdings, arguing that we used the modified categorical approach in cases like Taylor, Shepard , and Johnson "in relation to statutes that may not have been divisible" in the way that we have just described. Post, at 2297 (ALITO, J.). But if, as the dissent claims, the state laws at issue in those cases set out "merely alternative means, not alternative elements" of an offense, post, at 2298, that is news to us. And more important, it would have been news to the Taylor, Shepard , and Johnson Courts: All those decisions rested on the explicit premise that the laws "contain [ed] statutory phrases that cover several different ... crimes," not several different methods of committing one offense. Johnson, 559 U.S., at 144, 130 S.Ct. 1265 (citing Nijhawan, 557 U.S., at 41, 129 S.Ct. 2294). And if the dissent's real point is that distinguishing between "alternative elements" and "alternative means" is difficult, we can see no real-world reason to worry. Whatever a statute lists (whether elements or means), the documents we approved in Taylor and Shepard -i.e., indictment, jury instructions, plea colloquy, and plea agreement-would reflect the crime's elements. So a court need not parse state law in the way the dissent suggests: When a state law is drafted in the alternative, the court merely resorts to the approved documents and compares the elements revealed there to those of the generic offense.

The dissent, as we understand it, takes the same view as the Ninth Circuit; accordingly, each of the reasons-statutory, constitutional, and practical-that leads us to reject Aguila-Montes proves fatal to the dissent's position as well. The dissent several times obscures its call to explore facts with language from our categorical cases, asking whether "the relevant portions of the state record clearly show that the jury necessarily found, or the defendant necessarily admitted, the elements of [the] generic [offense]." Post, at 2302; see Shepard, 544 U.S., at 24, 125 S.Ct. 1254 (plurality opinion) (reiterating Taylor 's"demanding requirement that ... a prior conviction 'necessarily' involve[ ] " a jury finding on each element of the generic offense) (emphasis added). But the dissent nowhere explains how a factfinder can have "necessarily found" a non-element-that is, a fact that by definition is not necessary to support a conviction. The dissent's fundamental view is that a sentencing court should be able to make reasonable "inference[s]" about what the factfinder really (even though not necessarily) found. See post, at 2302 - 2303. That position accords with our dissenting colleague's previously expressed skepticism about the categorical approach. See Moncrieffe v. Holder, 569 U.S. ----, ----, 133 S.Ct. 1678, 1701, 185 L.Ed.2d 727 (2013) (ALITO, J., dissenting) ("I would hold that the categorical approach is not controlling where the state conviction at issue was based on a state statute that encompasses both a substantial number of cases that qualify under the federal standard and a substantial number that do not. In such situations, it is appropriate to look beyond the elements of the state offense and to rely as well on facts that were admitted in state court or that, taking a realistic view, were clearly proved"). But there are several decades of water over that dam, and the dissent offers no newly persuasive reasons for revisiting our precedents.

See also 1 C. Wright & A. Leipold, Federal Practice and Procedure: Criminal § 125, pp. 550-551 (4th ed. 2008) ("If a single statute sets forth several different offenses, [a] pleading ... that does not indicate which crime [the] defendant allegedly committed is insufficient"); 5 W. LaFave, J. Israel, N. King, & O. Kerr, Criminal Procedure § 19.3(a), p. 263 (3d ed. 2007) ("[W]here a statute specifies several different ways in which the crime can be committed, [courts often] hold that the pleading must refer to the particular alternative presented in the individual case").

Several decisions treat "invasion of a possessory right" as an aspect of § 459's entry element, see, e.g., People v. Waidla, 22 Cal.4th 690, 723, 94 Cal.Rptr.2d 396, 996 P.2d 46, 65 (2000) ; Fortes v. Sacramento Munic. Ct. Dist., 113 Cal.App.3d 704, 712-714, 170 Cal.Rptr. 292, 296-297 (1980), but others view the issue of possessory right as bearing only on the affirmative defense of consent, see, e.g., People v. Sherow, 196 Cal.App.4th 1296, 1303-1305, 1311, and n. 9, 128 Cal.Rptr.3d 255, 260-261, 266, and n. 9 (2011) ; People v. Felix, 23 Cal.App.4th 1385, 1397, 28 Cal.Rptr.2d 860, 867 (1994). And California's pattern jury instructions do not require the jury to find invasion of a possessory right before convicting a defendant of burglary. See 1 Cal. Jury Instr., Crim., No. 1700 (2012).

The Government here forfeited an alternative argument that § 459 qualifies as a predicate offense under ACCA's "residual clause," which covers statutes "involv[ing] conduct that presents a serious potential risk of physical injury to another." 18 U.S.C. § 924(e)(2)(B)(ii). We express no view on that argument's merits. Compare United States v. Mayer, 560 F.3d 948, 960-963 (C.A.9 2009) (holding that Oregon's burglary statute falls within the residual clause, even though it does not include all of generic burglary's elements), with id., at 951 (Kozinski, C.J., dissenting from denial of rehearing en banc) (arguing that the panel opinion "is a train wreck in the making").