UNITED STATES COURT OF APPEALS FOR THE THIRD CIRCUIT
No. 17-1861
UNITED STATES OF AMERICA
v.
MARC JAMES HARRIS,
Appellant
________________________
ORDER SUR PETITION
FOR REHEARING EN BANC
________________________
Present: CHAGARES, Chief Judge, JORDAN, HARDIMAN, KRAUSE, RESTREPO,
BIBAS, PORTER, MATEY, PHIPPS, MONTGOMERY-REEVES, CHUNG, AMBRO *,
FUENTES *, Circuit Judges
The petition for rehearing en banc filed by appellee in the above-entitled case
having been submitted to the judges who participated in the decision of this Court and to
all the other available circuit judges of the circuit in regular active service, and no judge
who concurred in the decision having asked for rehearing, and a majority of the judges of
the circuit in regular service not having voted for rehearing, the petition for rehearing en
banc by the Court, is denied. Judge Jordan, joined by Chagares, C.J., and Hardiman,
Krause, Bibas, Porter, & Matey, JJ., files the attached concurrence.
BY THE COURT,
s/ L. Felipe Restrepo
Circuit Judge
Date: November 27, 2023
cc: All Counsel of Record
*
Judge Ambro and Judge Fuentes’ votes are limited to panel rehearing.
United States v. Marc Harris (No. 17-1861)
Jordan, J., amended concurring in denial of rehearing en banc, joined by Chagares, C.J.,
and Hardiman, Krause, Bibas, Porter, & Matey, JJ.
We recognize that our decision today declining en banc reconsideration of this
matter will be a source of great frustration for the government. Frustration is the gift that
the “categorical approach” keeps on giving. This peculiar analytical construct has forced
us and other courts to reach perverse outcomes in many, many cases, this one being only
the latest. 1 And even when the result of applying the categorical approach sometimes
makes sense, time and effort is often wasted because a more obvious route to the sensible
result is readily available. Even worse is the difficulty of justifying the categorical
approach and its outcomes to the citizenry we serve. The public may not care whether
anyone finds the categorical approach frustrating, but they do care about justice, and we
are unable to explain how our holding in this case satisfies basic notions of right and
wrong. Despairing of that, we write to describe why the outcome here is compelled by
precedent and to highlight why changes in the categorical approach are needed.
For those who may not be familiar with the categorical approach, we provide a
brief overview of its origin and development, with particular focus on the Armed Career
Criminal Act (“ACCA”), 18 U.S.C. § 924(e), the legislation that, along with another
firearms statute, 18 U.S.C. § 924(c), and the Immigration and Nationality Act (“INA”), 8
U.S.C. § 1101, et seq., has been the primary seedbed for this extraordinary doctrine. We
then explain how application of the categorical approach regularly generates unjust
1
See infra Section III.
1
results and taxes judicial resources. We conclude by explaining our support for a more
fact-based approach that would maintain key aspects of Supreme Court precedent while
allowing courts to take account of an individual’s actual conduct and, hence, provide real
justice. 2
First, however, we provide a summary of the facts in this case and a defense of our
decision to decline en banc review.
I. Background
A. Harris’s Convictions
Marc Harris has a long rap sheet. In 2010, he pled guilty in federal court to yet
another crime: being a felon in unlawful possession of a firearm, a violation of 18 U.S.C.
§ 922(g). His earlier convictions in Pennsylvania state courts became relevant in that
federal proceeding because he was sentenced to enhanced penalties under ACCA. 3 To be
subject to such penalties, a defendant must have at least three prior convictions for crimes
that qualify as violent felonies or serious drug offenses. A “violent felony” is defined in
ACCA to include “any crime punishable by imprisonment for a term exceeding one year
… that … has as an element the use, attempted use, or threatened use of physical force
2
See United States v. Taylor, 142 S. Ct. 2015, 2033 (2022) (Thomas, J.,
dissenting) (“In light of the mischief that the categorical approach has caused, we should
welcome briefing on whether a conduct-based approach tacks closer to statutory text and
common sense … .”).
The maximum term of imprisonment for a violation of § 922(g) is ten years, 18
3
U.S.C. § 924(a)(2), but, with an ACCA enhancement, the punishment increases to a
minimum term of fifteen years, id. at § 924(e)(1).
2
against the person of another[.]” 18 U.S.C. § 924(e)(2)(B)(i). That particular definition is
contained in what has come to be called the “elements clause” or the “force clause” of
ACCA. 4
Among Harris’s prior convictions is one for first-degree aggravated assault under
§ 2702(a)(1) of title 18 of Pennsylvania’s consolidated statutes, and the District Court
relied on that conviction when it sentenced him under ACCA on the federal gun-
possession charge. During the hearing at which Harris pled guilty, the government
described the facts behind his § 2702(a)(1) conviction. He had snatched a woman’s
purse, and, shortly thereafter, when the woman recognized him on the street and fled into
her home, he shot a gun at the residence, narrowly missing a neighbor who had been with
4
Section 924(e)(2)(B) also contains an enumerated-offenses clause, which
specifically sets forth certain crimes that will qualify as violent felonies for purposes of
ACCA, including, “burglary, arson, or extortion, [or crimes] involv[ing] use of
explosives … .” Id. at § 924(e)(2)(B)(ii). And there is, or at least was, a so-called
“residual clause,” which Congress evidently meant to capture crimes committed with
violence but which may not be subject to easy classification in advance as being violent.
The residual clause provides that the term “violent felony” includes crimes “involv[ing]
conduct that presents a serious potential risk of physical injury to another[.]” Id.
Apropos our discussion today, the Supreme Court struck down that clause as being
unconstitutionally vague, not because there is anything vague about it when considering
an actual record of violent behavior but because, when forced to ignore the facts of a
crime and to instead apply the categorical approach, where hypotheticals replace reality,
the Court discovered that “[d]ecisions under the residual clause have proved to be
anything but evenhanded, predictable, or consistent.” Johnson v. United States, 576 U.S.
591, 606 (2015). Consequently, it held “that imposing an increased sentence under the
residual clause of the Armed Career Criminal Act violates the Constitution’s guarantee of
due process.” Id.
3
the woman and was following her inside. (Revised Answering Br. at 6.) When the
District Court asked Harris if those facts were true, he said yes. (Id.)
The Pennsylvania statute at issue specifies that “[a] person is guilty of aggravated
assault if he … attempts to cause serious bodily injury to another, or causes such injury
intentionally, knowingly or recklessly under circumstances manifesting extreme
indifference to the value of human life[.]” 18 Pa. Cons. Stat. § 2702(a)(1). By its terms,
the statute is aimed at punishing violent crimes. If some additional indication of
legislative intent were needed, the title “Aggravated assault” should suffice. 5 The
District Court decided that Harris’s first-degree aggravated assault conviction was a
qualifying ACCA predicate and, with that and his other offenses, he should receive an
enhanced sentence. Later, when Supreme Court precedent called into question the
contours of ACCA, 6 Harris filed a motion under 28 U.S.C. § 2255, 7 which the District
Court denied. We reversed that denial, and the government’s present petition for
rehearing en banc (the “Petition”) seeks to overturn our ruling.
5
A violation of § 2702(a)(1) is “a felony of the first degree.” 18 Pa. Cons. Stat.
§ 2702(b).
6
See Johnson, 576 U.S. at 601-02, 606; supra note 4 and accompanying text.
7
Section 2255(a) provides, in relevant part, “[a] prisoner in custody under
sentence of a court established by Act of Congress claiming the right to be released upon
the ground that the sentence was imposed in violation of the Constitution or laws of the
United States, or … that the sentence was in excess of the maximum authorized by law,
… may move the court which imposed the sentence to vacate, set aside or correct the
sentence.”
4
B. The Categorical Approach Applied to Harris
If one not versed in the intricacies of the categorical approach read the foregoing
and thought that shooting a gun at someone is obviously enough to support a conviction
for aggravated assault under § 2702(a)(1), which, again, is the pertinent state statute, he’d
be right. Gunplay will indeed get you convicted of first-degree aggravated assault in
Pennsylvania. But if, in turn, the reader thought a conviction for that shooting would
constitute a violent felony under ACCA, he would be dead wrong, even if the victim was
literally dead. Thanks to the categorical approach, that felony is not “violent” for ACCA
purposes, because the categorical approach does not deal well with the obvious. It
attends only to theoretical possibilities, taking no account of the defendant’s actual
conduct and asking only what other conduct in some other case might be prosecuted
under the statute of conviction.
Here’s how it works: one examines the text of the statute under which the earlier
conviction in question was obtained – in this case, § 2702(a)(1), the proposed ACCA
predicate conviction – and then one hypothesizes the least culpable conduct with the least
culpable mens rea that could qualify as a violation of that statute. As a plurality of the
Supreme Court recently put it, “[t]he focus is … on whether … a state offense necessarily
involves the defendant’s ‘use, attempted use, or threatened use of physical force against
the person of another.’ If any – even the least culpable – of the acts criminalized do not
entail that kind of force,” then “the statute of conviction does not categorically match the
federal standard, and so cannot serve as an ACCA predicate.” Borden v. United States,
141 S. Ct. 1817, 1822 (2021) (quoting 18 U.S.C. § 924(e)(2)(B)(i)). Thus, “the
5
categorical approach requires courts not only to ignore the actual manner in which the
defendant committed the prior offense, but also to presume that the defendant did so by
engaging in no more than ‘the minimum conduct criminalized by the state statute[,]’”
United States v. Ramos, 892 F.3d 599, 606 (3d Cir. 2018) (quoting Moncrieffe v. Holder,
569 U.S. 184, 191 (2013)), and with the least culpable mindset, Borden, 141 S.Ct. at
1825. It is that set of presumptions that often leads to perverse results in categorical-
approach cases.
So how did application of the categorical approach work out for Harris in this
case? Really well. He ended up winning. See United States v. Harris, 68 F.4th 140, 141
(3d Cir. 2023) (“Harris’ aggravated assault conviction is stricken as a predicate, and he
no longer has the three violent felony convictions necessary to justify the
enhancement.”). And he won because, to his good fortune, we had already applied the
categorical approach in a similar case, United States v. Mayo, 901 F.3d 218 (3d Cir.
2018), with a result that was ideal for him. The very same Pennsylvania statute and the
very same ACCA question were presented in Mayo as in Harris’s case, namely, whether a
conviction under 18 Pa. Cons. Stat. § 2702(a)(1) constitutes a qualifying ACCA predicate
offense for a defendant charged with violating 18 U.S.C. § 922(g). We said in Mayo that
it did not, although it was undisputed that the defendant there was convicted of first-
degree aggravated assault because he had “hit [the victim] on the head with a brick,
punched and kicked [the victim] … , and hit [the victim] with a glass bottle[.]” Mayo,
901 F.3d at 222 (alterations in original). None of that real-world violence mattered. It
never does with the categorical approach. What mattered in Mayo, and ultimately for
6
Harris, is that aggravated assault under Pennsylvania’s § 2702(a)(1) does not provide a
categorical match with ACCA’s elements clause, which calls for the use of physical force
against another. Id. at 224.
There is not a categorical match for two reasons. First, § 2702(a)(1) allows
conviction for criminal acts committed not only intentionally and knowingly, but also
“recklessly under circumstances manifesting extreme indifference to the value of human
life,” 18 Pa. Cons. Stat. § 2702(a)(1), and the Supreme Court has held that the definition
of “violent felony” in ACCA’s elements clause does not cover recklessly committed
crimes. See Borden, 141 S. Ct. at 1825 (“We must decide whether the elements clause’s
definition of ‘violent felony’ … includes offenses criminalizing reckless conduct. We
hold that it does not.”). While some Courts of Appeals have held a mens rea of “extreme
recklessness” is distinguishable and can be a categorical match, e.g., United States v.
Manley, 521 F.4th 143, 151 (4th Cir. 2022), we have not yet addressed that question. 8 And
if it is not distinguishable, then when viewed through the lens of the categorical approach,
ACCA cannot count a violation of § 2702(a)(1) as a violent felony. Being forced to
assume, counterfactually, that Harris acted only recklessly, we would be required to let
him evade the enhanced penalty ACCA was designed to impose.
The second reason there is no categorical match is more definitive: § 2702(a)(1)
encompasses a wider range of possible actus rei than does ACCA. That is, Pennsylvania
courts have upheld convictions under § 2702(a)(1) when the crime at issue involved no
8
We leave that question for another day.
7
use of force at all. So, in that respect, too, the state statute is broader in application than
ACCA’s elements clause, which clearly does require “the use, attempted use, or
threatened use of physical force against the person of another[,]” 18 U.S.C.
§ 924(e)(2)(B)(i). Mayo, 901 F.3d at 227-28. There is thus no categorical match. See
Commonwealth v. Thomas, 867 A.2d 594, 597 (Pa. Super. Ct. 2005) (upholding
conviction of a mother for starving her four-year-old son, and declaring that “evidence of
the use of force or the threat of force is not an element of the crime of aggravated assault”
(emphasis added)); Commonwealth v. Taylor, No. 1641 WDA 2013, 2015 WL 7576457,
at *1, *6 (Pa. Super. Ct. Feb. 9, 2015) (affirming conviction under § 2702(a)(1) for
defendant’s “criminal neglect” of her twin six-year-old children, which included failing
to feed and clothe them).
C. The Pushback on Mayo
Attorneys for the United States Department of Justice have, quite understandably,
been trying to get the result in Mayo overturned ever since. We knew at the time it was a
“wholly unsatisfying and counterintuitive” outcome. Mayo, 901 F.3d at 230. But we
believed then, and still do, that it is an outcome dictated by the categorical approach. For
its part, however, “[t]he [g]overnment has consistently argued that Mayo was wrongly
decided and should be overturned, in part because it improperly relied on a ‘single
intermediate appellate court’ decision [– namely, the just-referenced opinion in
Commonwealth v. Thomas –] in concluding that Pennsylvania first-degree aggravated
assault can be committed without the use of physical force.” Harris, 68 F.4th at 143
(quoting the government’s brief).
8
To address that concern, we turned to the Supreme Court of Pennsylvania and
asked it to accept certification of this question: “Whether the Pennsylvania First-Degree
Aggravated Assault provision, codified at 18 Pa. Cons. Stat. § 2702(a)(1), requires some
use of physical force, as the [United States] contends, or, instead, as the Pennsylvania
Superior [Court] said in … Commonwealth v. Thomas, 867 A.2d … [at] 597 …, the
statute means that ‘the use of force or threat of force is not an element of the crime[.]’”
United States v. Harris, 289 A.3d 1060, 1064 (Pa. 2023).
The court granted our petition for certification and confirmed that “there is no
express element in Section 2702(a)(1) requiring the use or attempted use of physical
force, or any reference to force at all.” Id. at 1070. The court went on to observe that
“the General Assembly was cognizant of how to codify the manner of causing a
particular bodily injury as an element of the crime. The legislature did not restrict the
manner of causing or attempting to cause serious bodily injury in subsection (a)(1) [of
§ 2702], and,” the court added, “we decline the invitation to do so by judicial fiat.” Id. at
1070-71 (citations omitted). Given the Pennsylvania Supreme Court’s definitive
statement that § 2702(a)(1) does not include the use of force as an element, and, in light
of our own binding precedent in Mayo, we concluded that Harris was entitled to relief on
his § 2255 motion.
Nevertheless, the government continues to argue that we should take this case en
banc and overrule Mayo. By the government’s lights, it does not really matter what the
Pennsylvania Supreme Court has to say about Pennsylvania law. ACCA, it argues, is a
federal statute and, therefore, “the question presented is whether each [of Harris’s]
9
statutory offense[s] presents as an element … ‘the use, attempted use, or threatened use
of physical force against the person of another.’” (Petition at 2 (quoting 18 U.S.C.
§ 924(e)(2)(B)(i)).)
The government is of course correct that the definition of “physical force,” as that
term is used in ACCA, is a matter of federal law, not state law. But the government is
wrong to imply that state law has nothing to do with whether a hypothetical state
conviction – i.e., a conviction for the least culpable conduct with the least culpable
mindset that could be prosecuted under the state statute in question – is relevant to the
“physical force” question under ACCA. State law has everything to do with it. It is what
the categorical approach was designed to take account of, however poorly it does so.
Here, as already discussed, the outcome of that analytical process is plain.
Pennsylvania’s aggravated assault statute, per the authoritative opinion of the
Pennsylvania Supreme Court, does not include the use of force as an element. Period.
That should be the end of it.
But the government presses on with its argument about the use of force, saying,
“[a]n element that requires proof of bodily injury [, as does §2702(a)(1),] equates to
‘physical force’ as defined by federal law.” (Petition at 3.) That must be so, the
government says, because the Supreme Court in Johnson v. United States, 559 U.S. 133,
134 (2010), stated that “‘physical force’ means violent force – i.e., force capable of
causing physical pain or injury to another person.” (Petition at 3.) This same argument is
pressed in a variety of ways in the government’s Petition, but the import is always the
same: “‘serious bodily injury’ … cannot result except through the application of ‘physical
10
force’ as stated in ACCA.” (Petition at 3 n.1.) And therein lies the fundamental flaw in
the government’s argument: the quoted statement is false, as a matter of fact and logic.
As a matter of fact, it is wrong because serious bodily injury can and does result
from inaction, or, in other words, from the absence of any force at all. That was the point
of the Superior Court’s analysis in Commonwealth v. Thomas. The mother in that case
starved her four-year-old child to death, and there was no proof of restraint or force of
any kind. See Thomas, 867 A.2d at 597 (“Mother argues that the evidence was not
sufficient to sustain her conviction because the Commonwealth failed to demonstrate
either the use of force or the threat of force.”). Even under Thomas’s facts, though, some
force had to be used against the child, says the government; the child must have been
restrained in some way. Assuming that were true, it still does not answer the point. One
can imagine an infant instead of a four-year-old being the victim, and no restraint would
be required to effect the crime. Or one can imagine a bed-ridden invalid or an
incapacitated elderly person, and again no force at all would be required. A heartless
person with the duty to care for a victim could do nothing, exert no physical force at all
on the victim, and simply watch the serious bodily harm occur as the victim starved or
got gangrenous bedsores and died as a result. That would be a crime under the
Pennsylvania aggravated assault statute, because it focuses on bodily injury, not the use
of force. Even the government now acknowledges that “the state Supreme Court
correctly confirmed: ‘a person may commit first-degree aggravated assault by starving a
person to death by willfully not performing an act required by law, as a parent or
guardian.’” (Petition at 8 (quoting Harris, 289 A.3d at 1074).) So factually, serious
11
bodily injury can be caused by a failure to act – the complete absence of any force –
rather than through the application of force.
Still, insists the government, “whether the state court believes that this omission to
act by a parent involves ‘physical force’ is irrelevant, at least with regard to the
application of ACCA.” (Petition at 8-9.) To overcome the obvious contradiction of
proclaiming that total inaction is the same as physical force, the government asserts that
“[t]he U.S. Supreme Court has made clear that the reference to ‘physical force’ in the
elements clause [of ACCA] simply distinguishes ‘force exerted by and through concrete
bodies,’ as opposed to ‘intellectual force or emotional force.’” (Petition at 9 (quoting
United States v. Castleman, 572 U.S. 157, 170 (2014) and Johnson, 559 U.S. at 138).)
Never mind that, earlier in the Petition, the government defined “physical force” as
“violent force.” Now, shifting gears, and trying to make the absence of force equate to
the use of force, it simply declares that, “[w]ithout question, death by starvation involves
‘physical force’ under this binding definition.” (Petition at 9.) But assertion is not
reasoning, and saying something is so “without question” does not make it so.
The government’s argument leaves logic behind. To say that bodily injury results
from physical force does not demonstrate that bodily injury results only from physical
force, as Harris’s lawyers rightly note. Making their point stronger still, defense counsel
assume the truth of the government’s premise (which we reject as a matter of plain
English) that one can rightly say a failure to act is an application of force, and they then
proceed to show that the government still loses because calling something “force” does
not make it ACCA-qualifying force. Here’s the example they use to highlight the
12
government’s error: “[S]uffice it to say that just as ‘all pickpockets are criminals’ does
not validly imply that ‘all criminals are pickpockets,’ ‘ACCA force is capable of causing
physical pain or injury’ does not validly imply ‘all force causing pain or injury is ACCA
force.’” (Petition Response at 9.) All true, and the government has no answer for it.
So, we are left – unless the Supreme Court takes this matter up – with yet another
absurd result dictated by the categorical approach. A violent criminal who shot a gun at
someone is saved from facing the ACCA penalties Congress designed for just such
behavior because we are commanded to ignore readily provable, indeed admitted,
conduct that should be accounted for. How on Earth did we end up here?
II. How We Got Here
A. The Armed Career Criminal Act
When Congress passed ACCA as part of the Comprehensive Crime Control Act of
1984, it subjected violent recidivist offenders to more serious penalties than those
received by other criminals. To that end, it identified burglaries and robberies as “violent
felonies,” singling out those crimes because of their frequency and perceived harm to
society. See S. Rep. No. 98-190, at 5 (1983) (“[I]n terms of the likelihood of being
victimized, burglary is the primary threat, followed by robbery.”); H.R. Rep. No. 98-
1073, at 3 (1984) (“Robberies and burglaries are the most damaging crimes to society.”).
The legislation provided definitions for burglary and robbery, 9 with the dual aims of
9
It defined burglary as “any offense involving entering or remaining
surreptitiously within a building that is the property of another with intent to engage in
conduct constituting a Federal or State offense.” S. Rep. No. 98-190, at 2 (1983). It
defined robbery as “any offense involving the taking of the property of another from the
13
respecting “the prerogatives of the States in defining their own offenses” while ensuring
“that the same type of conduct is punishable on the Federal level in all cases.” S. Rep.
No. 98-190, at 20 (emphasis added). Congress wanted to avoid “put[ting] Federal courts
in a position of having to interpret and apply State laws on robbery and burglary in
Federal criminal trials.” H.R. Rep. No. 98-1073, at 5-6. According to the legislative
history, the intent was to avoid a situation in which “culpable offenders might escape
punishment on a technicality” due to “the wide variation among states and localities in
the ways that offenses are labeled[.]” S. Rep. No. 98-190, at 20.
Congress decided to expand the scope of ACCA in 1986. 10 It replaced burglary
and robbery with three categories of offenses: (1) “violent felonies,” which included
burglary and robbery, as well as other enumerated offenses; 11 (2) “serious drug
person or presence of another by force or violence, or by threatening or placing another
person in fear that any person will imminently be subjected to bodily injury[.]” Id.
10
See 132 Cong. Rec. 7697 (1986) (“[T]he time has come to broaden that
definition so that we may have a greater sweep and more effective use of this important
statute.”).
11
As noted in part earlier, supra note 4 and accompanying text, the term “violent
felony” is defined in ACCA as:
[a] crime punishable by imprisonment for a term exceeding one year …
involving the use or carrying of a firearm, knife, or destructive device that
would be punishable by imprisonment for such term if committed by an
adult, that – (i) has as an element the use, attempted use, or threatened use of
physical force against the person of another; or (ii) is burglary, arson, or
extortion, involves use of explosives, or otherwise involves conduct that
presents a serious potential risk of physical injury to another[.]”
18 U.S.C. § 924(e)(2)(B).
14
trafficking offenses”; and (3) crimes that involved the use, attempted use, or threatened
use of violence against a person. Career Criminal Amendments Act of 1986, H.R. 4885,
99th Cong. (1986); H.R. Rep. No. 99-849, at 3 (1986). 12
B. The Advent of the Categorical Approach
The categorical approach made its first appearance in Supreme Court case law in
Taylor v. United States, 495 U.S. 575 (1990). Like Harris in the case before us now,
Arthur Taylor pled guilty to being a felon in unlawful possession of a firearm. Id. at 578.
He had previously been convicted of robbery, assault, and two second-degree burglaries
in Missouri. Id. at 578 & n.1. Applying ACCA, the district court adjudged Taylor to be
a career offender and sentenced him to imprisonment for the mandatory minimum of
fifteen years. Id. at 579. The Eighth Circuit affirmed, reasoning that “the word
12
In addition to broadening ACCA’s application, Congress deleted the statutory
definitions of burglary and robbery. Why it did so is a mystery. The Supreme Court has
observed, “[t]he legislative history as a whole suggests that the deletion of the 1984
definition of burglary may have been an inadvertent casualty of a complex drafting
process.” Taylor v. United States, 495 U.S. 575, 590-91 (1990).
In 1989, then-Senator Joseph Biden introduced a bill that would have
“reenact[ed]” those definitions, explaining that his bill “corrects an error that occurred
inadvertently when the definition of burglary was deleted from the Armed Career
Criminal statute in 1986.” Id. at 590 n.5 (quoting 135 Cong. Rec. 23519 (1989)). The
apparently uncontroversial bill passed in the Senate by a vote of 100 to 0 but was never
taken up in the House of Representatives. See A Bill to Implement the President’s 1989
Drug Control Strategy, S. 1711, 101st Cong. (1989),
https://www.congress.gov/bill/101st-congress/senate-bill/1711/all-actions-without-
amendments.
15
‘burglary’ in § 924(e)(2)(B)(ii) ‘means “burglary” however a state chooses to define
it[.]’” 13 Id.
The Supreme Court, in contrast, thought it “implausible that Congress intended the
meaning of ‘burglary’ ... to depend on the definition adopted by the State of conviction.”
Id. at 590. Were that the case, the Court reasoned, whether a defendant “receive[d] a
sentence enhancement” would not depend on the defendant’s conduct, but rather, on
“whether the State of his prior conviction happened to call that conduct ‘burglary.”’ Id.
at 590-91. That construction of ACCA would conflict with Congress’s intent that “the
same type of conduct” be “punishable on the Federal level in all cases.” Id. at 582
(emphasis added) (quoting S. Rep. No. 98-190, at 20 (1983)). 14
But, in spite of that entirely correct emphasis on offender conduct, the Supreme
Court decision in Taylor then took a fateful turn. The conduct of the defendant was hard
to determine based on the available record, so, instead of looking to his conduct to
determine the applicability of ACCA’s enhanced sentencing scheme, the Court compared
the elements of the burglary statutes he violated with the elements of “generic
burglary[,]” which the Court defined as “having the basic elements of unlawful or
13
Section 924(e)(2)(B)(ii) is the enumerated-offense clause of ACCA and, as
noted earlier, defines a “violent felony” as “any crime punishable by imprisonment for a
term exceeding one year” 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[.]” See supra note 10.
Cf. United States v. Booker, 543 U.S. 220, 253 (2005) (“Congress’ basic goal in
14
passing the [new federal sentencing regime] was to move the sentencing system in the
direction of increased uniformity.”).
16
unprivileged entry into, or remaining in, a building or structure, with intent to commit a
crime.” 15 Id. at 599. Extrapolating from the case before it, with an indeterminate record
on the offense conduct, the Court decided that, going forward, courts had to apply “a
formal categorical approach, looking only to the statutory definitions of the prior
offenses, and not to the particular facts underlying those convictions.” Id. at 600.
The Court justified its new categorical approach with three reasons. First, looking
at statutory text, the Court observed that § 924(e)(1) “refers to ‘a person who …has three
previous convictions’ for – not a person who has committed – three previous violent
felonies or drug offenses.” Id. (quoting § 924(e)(1)). Then too, “[s]ection
924(e)(2)(B)(i) defines ‘violent felony’ as any crime punishable by imprisonment for
more than a year that ‘has as an element’ – not any crime that, in a particular case,
involves – the use or threat of force.” Id. Thus, “[r]ead in this context,” the Court
decided, ACCA requires focus on “the elements of the statute of conviction, not … the
facts of each defendant’s conduct.” Id. at 600-01.
Second, looking at legislative history, the Court said, “Congress generally took a
categorical approach to predicate offenses[,]” and “no one suggested that a particular
crime might sometimes count towards enhancement and sometimes not, depending on the
facts of the case.” Id. at 601. Surely Congress would have said something in the
15
The Court noted that this definition was “practically identical to the
[accidentally deleted] 1984 definition” and was therefore consistent with Congress’s
legislative intent. Taylor, 495 U.S. at 598; see also id. (“As we have seen, there simply is
no plausible alternative that Congress could have had in mind.”).
17
legislative history of ACCA if it “had meant to adopt an approach that would require the
sentencing court to engage in an elaborate factfinding process regarding the defendant’s
prior offenses[.]” Id.
Finally, the Court postulated that a conduct-based approach would burden federal
courts in numerous ways. Id. at 601-02. Chief among these was that, as was the case in
Taylor, a defendant’s record of conviction might not clearly identify the defendant’s
criminal conduct for past convictions. Id. at 601. Absent a categorical approach, the
Court was concerned that trial courts would get caught up in examining witnesses from
past criminal proceedings to ascertain the facts underlying earlier convictions. Id. And if
that were to happen and a judge were to make a ruling as to those facts, it could very well
implicate the defendant’s Sixth Amendment right to a jury trial. Id.
In addition to creating the categorical approach, the Court also briefly
foreshadowed what we now call the “modified categorical approach.” It observed that
many states’ “burglary statutes … define burglary more broadly” than the generic
definition “by including places, such as automobiles and vending machines, other than
buildings.” Id. at 599. In cases in which such statutes are implicated, and “in a narrow
range of cases where a jury was actually required to find all the elements of generic
burglary[,]” then, “if the indictment or information and jury instructions show that the
defendant was charged only with a burglary of a building, … the Government [would] be
allowed to use the conviction for enhancement.” Id. at 602.
So, while the Court’s reasoning in Taylor began with an emphasis on Congress’s
desire that recidivists who have engaged in the same type of criminal conduct in their
18
pasts should be exposed to the same enhanced federal penalties, regardless of differences
in the state statutes criminalizing such conduct, id. at 581-83, 587-89, the Court ended up
in a very different place. It concluded that “the only plausible interpretation of … the
enhancement statute … [is that] it generally requires the trial court to look only to the fact
of conviction and the statutory definition of the prior offense.” Id. at 602.
C. The Modified Categorical Approach – Allowing a Peek at the Record
When the Supreme Court tried to apply the categorical approach in Shepard v.
United States, 544 U.S. 13 (2005), it faced the problem it predicted in Taylor. As the
Court had anticipated, the case presented a defendant who was convicted under a state
statute that, although containing all the elements of generic burglary, contained additional
elements that rendered it overbroad when compared with the generic version. The
Massachusetts statute at issue provided that “[w]hoever, in the night time, breaks and
enters a building, ship, vessel or vehicle, with intent to commit a felony, ... shall be
punished by imprisonment[.]” Mass. Gen. Laws Ann. ch. 266, § 16 (West 2000). The
statute therefore covered burglary of boats and cars, as well as buildings. The problem
was thus plainly presented: “[n]o one could know, just from looking at the statute, which
version of the offense Shepard was convicted of. Accordingly, [the Court] again
authorized sentencing courts to scrutinize a restricted set of materials ... to determine if
the defendant had pleaded guilty to entering a building or, alternatively, a car or boat.”
Descamps v. United States, 570 U.S. 254, 262 (2013) (citing Shepard, 544 U.S. at 26).
The Shepard Court endorsed Taylor’s suggestion that trial courts could look to the
record of conviction when the statute contained alternative elements, but it limited the
19
permissible record to the “charging document, written plea agreement, transcript of plea
colloquy, and any explicit factual finding by the trial judge to which the defendant
assented.” 16 Shepard, 544 U.S. at 16.
Next, in Descamps v. United States, the Court instructed sentencing judges to
apply this peek-at-the-record method, or “modified categorical approach[,]” only when
dealing with a state statute that is “divisible.” 570 U.S. at 257-58. The difficulty in
Descamps was not that the California burglary statute at issue embodied more than one
version of the crime; the problem was that it set out just one version but lacked an
element of generic burglary. It did not include “an unlawful entry” requirement “along
the lines of breaking and entering.” Id. at 264. In other words, it was an indivisible
statute that was not a categorical match for generic burglary, so it could not serve as a
predicate conviction for an ACCA-enhanced sentence, no matter what the defendant’s
actual conduct had been. The Court declared “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.” Id. at 258.17
In so ruling, the Court overturned the trial court’s application of ACCA
16
enhancement to the defendant because that court had relied on a police report in applying
the modified categorical approach. Shepard v. United States, 544 U.S. 13, 18, 25-26
(2005).
17
See also Moncrieffe v. Holder, 569 U.S. 184, 190-91 (2013) (“Because we
examine what the state conviction necessarily involved, not the facts underlying the case,
we must presume that the conviction ‘rested upon [nothing] more than the least of th[e]
acts’ criminalized, and then determine whether even those acts are encompassed by the
generic federal offense.”).
20
The Court went on to explain that the modified categorical approach applies only
when a statute presents alternative sets of elements, as opposed to alternative means of
committing an element. Id. at 257, 271-74. To illustrate the distinction, the Court gave
an example of a divisible statute that would be amenable to the modified categorical
approach:
[A]ssume … that a statute criminalizes assault with any of eight specified
weapons; and suppose further … that only assault with a gun counts as an
ACCA offense. A later sentencing court need only check the charging
documents and instructions … to determine whether in convicting a
defendant under that divisible statute, the jury necessarily found that he
committed the ACCA-qualifying crime.
Id. at 272. 18 Thus, when considering convictions under divisible statutes, a court can turn
to Shepard documents to determine which alternative statutory prohibition is implicated
by the criminal conduct at issue in the case, and then the court must apply the categorical
approach using the set of elements in that prohibition. 19 When considering convictions
under indivisible statutes, however, conduct is irrelevant, even when it is known and
would readily establish that the crime as committed meets the definition of “violent
felony” under ACCA.
18
Justice Kennedy appeared to be less than convinced by this hypothetical, saying,
“the dichotomy between divisible and indivisible state criminal statutes is not all that
clear.” Descamps v. United States, 570 U.S. 254, 279 (2013) (Kennedy, J., concurring).
19
This gloss on the categorical approach thus involves the added oddity of telling
a judge to pay attention to the record but only briefly. After ascertaining what the
defendant did to get convicted of a predicate offense, the judge is supposed to
immediately forget all about that and contemplate only purely theoretical possibilities.
21
In dissent, Justice Alito expressed skepticism about the means-versus-elements
dichotomy, and he questioned whether the statutes previously considered by the Court
contained alternative elements as opposed to mere factual means. Id. at 283-87. The
majority demurred, saying, “if, as the dissent claims, the state laws at issue in [previous]
cases set out ‘merely alternative means, not alternative elements’ of an offense, that is
news to us.” Id. at 264 n.2 (citation omitted). The Court further observed that it could
“see no real-world reason to worry” about the distinction. Id.
In Mathis v. United States, 579 U.S. 500 (2016), the Court sharpened the means-
versus-elements distinction. The defendant in that case had five burglary convictions
under Iowa law. Id. Iowa defined burglary to include entering an “occupied structure,”
Iowa Code §§ 713.1 (2013), which in turn was defined as “any building, structure,
appurtenances to buildings and structures, land, water or air vehicle, or similar place
adapted for overnight accommodation of persons, or occupied by persons for the purpose
of carrying on business or other activity therein, or for the storage or safekeeping of
anything of value[,]” id. § 702.12. Although the statute strongly resembled the divisible
statutes at issue in Taylor and Shepard – as well as the example hypothesized in
Descamps – the Court held that the different structures listed in § 702.12 were means, not
elements, and accordingly, that the statute was indivisible. Mathis, 579 U.S. at 506.
The Court explained that “‘[e]lements’ are the ‘constituent parts’ of a crime’s legal
definition – the things the ‘prosecution must prove to sustain a conviction[,]”’ id. at 504
(citing Black’s Law Dictionary (10th ed. 2014)), and “what the jury must find beyond a
reasonable doubt to convict the defendant,” id. By contrast “facts” or “means” are “real-
22
world things – extraneous to the crime’s legal requirements[.] … In particular, they need
neither be found by a jury nor admitted by a defendant.” Id.
Mathis also includes a vigorous defense of the categorical approach, with the
Justices in the majority saying, “[o]ur decisions have given three basic reasons for
adhering to an elements-only inquiry. First, ACCA’s text favors that approach. …
Second, a construction of ACCA allowing a sentencing judge to go any further would
raise serious Sixth Amendment concerns. … And third, an elements-focus avoids
unfairness to defendants. Statements of ‘non-elemental fact’ in the records of prior
convictions are prone to error precisely because their proof is unnecessary.” Id. at 510-
12.
But the categorical approach, full or modified, has never garnered the Justices’
unanimous approval. In Mathis, for example, Justice Kennedy’s concurrence described
the Court’s decision as “a stark illustration of the arbitrary and inequitable results
produced by applying an elements based approach” and observed that “[it] could not have
been Congress’ intent” either “for a career offender to escape his statutorily mandated
punishment when the record makes it clear beyond any possible doubt that [he]
committed generic burglary” or to create “vast sentencing disparities for defendants
convicted of identical criminal conduct in different jurisdictions.” Id. at 520-21
(Kennedy, J., concurring) (internal quotation marks omitted). And in his dissent, Justice
Alito sympathized with the district and appellate courts tasked with ascertaining when to
apply the categorical or modified categorical approach, and he wished us all “good luck.”
23
Id. at 539 (Alito, J., dissenting). Good luck wishes, however, have not been sufficient to
stave off a multitude of unintended but truly damaging consequences.
III. Some Consequences of the Categorical Approach
The vitality of the judicial branch is rooted in public confidence that judges will
apply the law justly, making every reasonable effort to carry out the will of the elected
branches of government. Our authority “depends in large measure on the public’s
willingness to respect … [our] decisions.” Williams-Yulee v. Fla. Bar, 575 U.S. 433,
445-46 (2015). And that, in turn, depends on our decisions making sense. As currently
administered, the categorical approach undermines public trust in our justice system
precisely because it produces decisions that cannot be squared with common sense. This
is so not only when ACCA is involved but in other kinds of cases too, including those
dealing with sentencing enhancements for other firearms offenses, 18 U.S.C. § 924(c),
and the aggravated felony provision of the INA, 8 U.S.C. § 1227(a)(2)(A)(iii).
A. The Armed Career Criminal Act
Time and again, federal courts have been required to hold that state law felony
convictions for conduct that plainly involved the use of force – including convictions for
voluntary manslaughter, aggravated assault, assault with a deadly weapon with intent to
kill, attempted rape, first-degree sexual abuse, sexual abuse by forcible compulsion,
taking indecent liberties with a child, maliciously damaging or destroying property by
means of an explosive, first-degree robbery, second-degree robbery, first-degree
burglary, and second-degree burglary – do not qualify as “violent felonies” under
24
ACCA. 20 Today, we are likewise compelled to reiterate, bizarre though it sounds, that
shooting at a fleeing victim is not a “violent felony,” Harris, 68 F.4th at 141, while in
20
See, e.g., United States v. Jenkins, 68 F.4th 148, 155 (3d Cir. 2023) (holding that
second-degree aggravated assault of a police officer was not a violent felony under
ACCA); United States v. Mayo, 901 F.3d 218, 222 (3d Cir. 2018) (holding that an
aggravated assault during which the victim was hit in the head with a brick was not a
violent felony under ACCA); United States v. Al-Muwwakkil, 983 F.3d 748, 760-61 (4th
Cir. 2020) (holding that attempted rape and burglary under Virginia law did not constitute
violent felonies under ACCA); United States v. Vann, 660 F.3d 771, 776 (4th Cir. 2011)
(holding that a conviction for taking indecent liberties with a child under North Carolina
law did not constitute a violent felony under ACCA); Dunlap v. United States, 784 F.
App’x 379, 381 (6th Cir. 2019) (holding that voluntary manslaughter and aggravated
assault were not violent felonies under ACCA, even though the district court recounted
that the underlying offenses involved killing a person with a handgun and forcefully
inserting a glass crack pipe into a victim’s genitals); Lofton v. United States, 920 F.3d
572, 576 (8th Cir. 2019) (holding that a conviction for criminal sexual abuse did not
constitute a violent felony under ACCA); Brown v. United States, 929 F.3d 554, 560 (8th
Cir. 2019) (concluding that conviction for second-degree burglary did not count as a
violent felony under ACCA); United States v. Mathews, 37 F.4th 622, 624, 626 (9th Cir.
2022) (holding that the crime of property damage under 18 U.S.C. § 844(i), which
involved placing a “bomb packed with steel balls (to increase the risk of personal injury)”
in an alley by the victim’s house, was not categorically a violent felony under ACCA);
United States v. Walton, 881 F.3d 768, 770 (9th Cir. 2018) (holding that neither first-
degree robbery under Alabama law nor second-degree robbery under California law was
a violent felony under ACCA); United States v. Davis, 875 F.3d 592, 600, 604 (11th Cir.
2017) (concluding that Alabama’s first degree sexual abuse statute was not divisible, and
that defendant’s conviction for sexual abuse by forcible compulsion was not a violent
felony under ACCA); United States v. Parrott, 585 F. Supp. 3d 661, 662-63 (E.D. Pa.
2022) (holding that second-degree robbery under Pennsylvania law did not categorically
qualify as a predicate ACCA offense); United States v. Singleton, 252 F. Supp. 3d 423,
426-27, 433 (E.D. Pa. 2017) (holding that defendant’s convictions under Pennsylvania’s
first-degree robbery statute for four prior robberies performed at gunpoint did not qualify
as violent felonies under ACCA); United States v. Brown, 249 F. Supp. 3d 287, 295
(D.D.C. 2017) (holding that a conviction for assault with a deadly weapon with intent to
kill under North Carolina law did not qualify as a violent offense under ACCA); United
States v. McNeal, No. 13-cr-16, 2017 WL 5186385, at *3-5 (E.D. Pa. July 14, 2017)
(holding that first-degree robbery under Pennsylvania law did not qualify as a violent
felony under ACCA even though “there was little doubt the defendant’s robberies were
violent,” as one of them involved shocking a victim with a taser and the other involved
holding a victim at gunpoint); United States v. Bayya, No. 13-cr-00558, 2015 WL
25
Mayo, we were forced to say that bashing a victim in the head with a brick was not
“violent” under ACCA. Mayo, 901 F.3d at 222, 230. Such outcomes are completely
confounding, and the sense of injustice they ignite was foreseen.
As noted earlier, supra at Section I.B., a plurality of the Supreme Court in Borden
held that a prior state law conviction for aggravated assault was not a “violent felony”
under ACCA because the statute of conviction could be satisfied by a reckless use of
force and thus was not a categorical match for the generic federal standard, which
required intentional use of force. 141 S. Ct. at 1822, 1825. The dissent, meanwhile,
objected that the Court’s holding would reverberate across the nation, relieving
defendants who had committed indisputably violent acts from ACCA’s enhanced
penalties. Id. at 1855-56 (Kavanaugh, J., dissenting). Those consequences, the dissent
urged, would “override[] Congress’s policy judgment about the risk posed by serial
violent felons who unlawfully possess firearms.” Id. at 1857. Descamps and Mathis only
strengthened the categorical approach’s grip on ACCA, despite other pointed dissents.
Mathis, 579 U.S. at 536 (Alito, J., dissenting); Descamps, 570 U.S. at 281 (Alito, J.,
dissenting).
The dissenters’ warnings have proved prescient. Take, for example, the Sixth
Circuit’s decision in Dunlap v. United States, 784 F. App’x 379 (6th Cir. 2019). The
defendant there was convicted on two counts of being a felon in possession of a firearm
8751795, at *3 (D. Or. Dec. 14, 2015) (holding that a conviction under Oregon’s first-
degree burglary statute did not qualify as a violent offense under ACCA).
26
and ammunition, in violation of 18 U.S.C. § 922(g), after he “shot [his victim] in the head
at close range, pistol-whipped him, and robbed him of his cellphone and $150 in cash.”
Id. at 381. The district court sentenced the defendant as a career offender under ACCA
because it determined that his extensive criminal history – which included convictions for
voluntary manslaughter for fatally shooting a man and aggravated assault for sexually
assaulting a woman with a glass pipe – demonstrated “a disregard for the rights of others”
and “a willingness to inflict whatever injury might flow … [without] the slightest
remorse.” Id. (internal quotation marks omitted). Applying the categorical approach,
however, the Sixth Circuit had no choice but to conclude that neither his conviction for
aggravated assault nor his conviction for voluntary manslaughter qualified as a “violent
felony” under ACCA because the state statutes could, in some hypothetical case, reach
less violent conduct. Id. at 387, 389.
Recall that Congress’s purpose in passing ACCA was to “address the special
danger created when … a violent criminal[] possesses a gun,” Borden, 141 S. Ct. at 1830
(internal quotation marks omitted), and then juxtapose that purpose with the outcome just
described. Consider further that the outcome is not some outlier but is a common
consequence of the categorical approach.
B. Using or Carrying a Firearm in Furtherance of a Crime of Violence,
Under 18 U.S.C. § 924(c)
Such consequences also regularly arise in firearms cases covered by 18 U.S.C.
§ 924(c), which, among other things, is meant to subject those who carry a gun “in
furtherance of” a “crime of violence” to a mandatory minimum of five years in prison,
27
consecutive to other sentences. 18 U.S.C. §§ 924(c)(1)(A), (D)(ii); see also U.S. Sent’g
Guidelines Manual § 2K2.1 (U.S. Sent’g Comm’n 2023). As with ACCA, the
Congressional intent behind § 924(c) has frequently been thwarted by the categorical
approach, since the mandate to apply that approach obligates federal courts to determine
if the predicate crime of conviction “has as an element the use, attempted use, or
threatened use of physical force,” as required by 18 U.S.C. § 924(c)(3)(A). See United
States v. Taylor, 142 S. Ct. 2015, 2020 (2022).
When it was first suggested to our Court, in the context of an armed robbery case,
that the categorical approach should apply when evaluating a § 924(c) charge, we
rejected the notion, explaining that, “[w]hen the predicate offense, Hobbs Act robbery,
and the § 924(c) offense are contemporaneous and tried to the same jury, the record of all
necessary facts [is] before the district court.” United States v. Robinson, 844 F.3d 137,
141 (3d Cir. 2016), abrogated on other grounds by United States v. Davis, 139 S. Ct.
2319 (2019). In other words, it made no sense to apply the categorical approach because
the same jury was both hearing the facts pertaining to the predicate offense and deciding
whether that offense “was committed with ‘the use, attempted use, or threatened use of
physical force[.]’” Id. There was no work for the categorical approach to do. Or so we
thought.
It turns out we were wrong. In the course of declaring a portion of § 924(c) to be
unconstitutionally vague in a later case, the Supreme Court said that it had “already read
… nearly identical language” in another statute “to mandate a categorical approach.”
Davis, 139 S. Ct. at 2327. So, despite the reality that a single fact-finder is looking at all
28
of the information necessary to decide guilt or innocence on the charges for both the
predicate offense and the firearm offense, the categorical approach now rules in § 924(c)
cases too. We are pushed again into an alternative universe where egregious violence
must be ignored, despite clear Congressional intent to the contrary.
The results are thus not surprising. They regularly fail to achieve justice but do
compound the tragedy of the underlying crimes. In United States v. Capers, for instance,
a § 924(c) conviction was predicated on the defendant’s shooting and killing a man who
was walking down a street with his eleven-month old daughter in his arms. 20 F.4th 105,
112, 116-17 (2d Cir. 2021). Although the Second Circuit acknowledged the crime was
“violent, even murderous,” the court had to vacate the defendant’s sentence because the
crime was charged as part of a conspiracy to violate the Racketeer-Influenced and
Corrupt Organizations Act (RICO), and some RICO conspiracies – not this one, mind
you, but some others – could theoretically encompass acts that do not involve the use of
physical force. Id. at 118 (internal quotation marks and citations omitted). Many other
defendants have similarly escaped the enhanced sentences that Congress said should
apply under §924(c). 21
21
See, e.g., United States v. Eldridge, 63 F.4th 962, 963 (2d Cir. 2023) (holding
that second-degree kidnapping, in violation of New York state law, did not constitute a
crime of violence under §924(c)); United States v. Barrett, 937 F.3d 126, 128 (2d Cir.
2019) (holding that a “violent, even murderous” Hobbs Act robbery conspiracy was not a
“crime of violence”), abrogated on other grounds by Lora v. United States, 599 U.S. 453
(2023); United States v. Davis, 53 F.4th 168, 169-70, 173 (4th Cir. 2022) (holding that
conviction under federal arson statute did not qualify as a crime of violence under
§ 924(c)); United States v. Taylor, 979 F.3d 203, 205, 210 (4th Cir. 2020) (holding that
convictions for attempted Hobbs Act robbery and conspiracy to commit Hobbs Act
robbery did not qualify as crimes of violence under § 924(c), even though the underlying
29
C. The Immigration and Nationality Act
In the immigration context, too, the categorical approach has produced disturbing
results. Under the INA, an alien who commits an “aggravated felony” may be
immediately deported and is ineligible for discretionary relief. See Moncrieffe, 569 U.S.
at 187 (first citing 8 U.S.C. § 1227(a)(2)(A)(iii); and then citing id. §§ 1129b(a)(3),
1129b(b)(1)(C), 1158(b)(2)(A)(ii), 1158(b)(2)(B)(i)). Enumerated offenses include,
among others, “murder, rape, [and] sexual abuse of a minor.” 8 U.S.C. §1101(a)(43)(A).
The statute was meant to “identify categories of criminal conduct that evidence such a
high degree of societal danger that an alien found to have engaged in such conduct should
not be allowed to obtain permission to remain in this country.” Moncrieffe, 569 U.S. at
218 (Alito, J., dissenting). Here again, however, the categorical approach has frustrated
Congress’s “clear objective.” Id. at 218-19.
offense involved a fatal shooting); United States v. Walker, 934 F.3d 375, 379 (4th Cir.
2019) (holding that kidnapping, in violation of 18 U.S.C. § 1201(a), did not qualify as a
crime of violence under § 924(c)); United States v. Fuertes, 805 F.3d 485, 500 (4th Cir.
2015) (holding that sex trafficking by force, fraud, or coercion is not categorically a
crime of violence under § 924(c)); United States v. Brazier, 933 F.3d 796, 800-01 (7th
Cir. 2019) (holding that convictions for kidnapping and holding a victim for ransom did
not constitute crimes of violence under § 924(c)); United States v. Lung’aho, 72 F.4th
845, 851 (8th Cir. 2023) (holding that arson, in violation of 18 U.S.C. § 844(f)(1), was
not a crime of violence under § 924(c)); United States v. Gill, No. 07-cr-0149, 2023 WL
349844, at *12 (D. Md. Jan. 20, 2023) (“An indivisible offense that includes felony
murder – like the Maryland first- and second-degree statutes at issue here – is not
categorically a ‘crime of violence’ under § 924(c)[.]”); Hernandez v. United States, No.
16-CV-22657-HUCK, 2016 WL 8078310, at *3 (S.D. Fla. Dec. 7, 2016) (holding that
hostage-taking, in violation of 18 U.S.C. § 1203, “does not categorically qualify as a
‘crime of violence’ under 18 U.S.C. § 924(c)’s elements clause”).
30
In case after case, courts have confronted abhorrent conduct that qualifies as an
enumerated offense under the text of the INA but ceases to qualify after application of the
categorical approach. These include the case of the alien in Larios-Reyes v. Lynch, who
sexually molested a four-year-old child but succeeded in having his order of removal
vacated because the categorical approach required the court to “focus on the minimum
conduct necessary for a violation of the state statute” and did not allow consideration of
“whether Larios-Reyes’s actual conduct constitute[d] ‘sexual abuse of a minor[.]’” 843
F.3d 146, 152, 154-55, 159 (4th Cir. 2016) (internal quotation marks omitted).
We too recently found ourselves compelled to reach a result difficult to square
with the language of the INA. In Cabeda v. Attorney General, although “it [was]
indisputable … that [the petitioner] repeatedly had sex with a minor,” we had to conclude
under the “formalistic framework” of the categorical approach that, notwithstanding the
reality of the conduct or the harm to the victim, the petitioner had not committed “sexual
abuse of a minor” and therefore was not removable for an aggravated felony. 971 F.3d
165, 166-67, 174-76 (3d Cir. 2020).
Other Courts of Appeals have found themselves in similar predicaments. While
Congress intended that aliens committing such heinous acts should not be permitted to
remain here, the categorical approach has required us to reach the opposite result. 22
22
See, e.g., Quinteros v. Att’y Gen., 945 F.3d 772, 785-86 (3d Cir. 2019) (holding
that conviction for conspiracy to commit assault with a dangerous weapon did not
constitute an aggravated felony under the categorical approach); Omargharib v. Holder,
775 F.3d 192, 194 (4th Cir. 2014) (holding that a grand larceny conviction under Virginia
law did not constitute an aggravated felony under the categorical approach); Kerr v.
31
IV. Proposed Changes to the Categorical Approach
A. The First Law of Holes
No one is suggesting that anything but laudable goals, such as uniformity and
consistency, 23 underlie the categorial approach, and an elements-based analysis may be
justified in some contexts. 24 But good intentions ought not blind us to practical
consequences. The real-world problem with the categorical approach is not where it
begins – which is in allowing judges to look beyond the labeling of a state offense – but
rather where it ends – which is in prohibiting them from looking at the factual record.
It has been more than three decades since the Supreme Court decided in Taylor
that the “only plausible interpretation” of ACCA is that it requires the categorical
approach. Taylor, 495 U.S. at 602. With all respect, it is now time for the Court to give
Holder, 352 F. App’x 958, 963 (5th Cir. 2009) (holding that a conviction for false
imprisonment under Florida law did not constitute an aggravated felony under the
categorical approach); Keeley v. Whitaker, 910 F.3d 878, 881 (6th Cir. 2018) (holding
that a conviction under Ohio’s rape statute did not constitute an aggravated felony under
the categorical approach); United States v. Martinez, 786 F.3d 1227, 1233 (9th Cir. 2015)
(holding that conviction for third-degree child molestation under Washington law did not
constitute an “aggravated felony” under the INA); Nicanor-Romero v. Mukasey, 523 F.3d
992, 1008 (9th Cir. 2008) (holding that petitioner’s conviction for annoying or molesting
a child under 18 years of age did not constitute a “crime involving moral turpitude” or an
aggravated felony under the categorical approach); Gomez-Ponce v. Holder, 571 F.
App’x 528, 530 (9th Cir. 2014) (vacating removal of an alien and holding that oral
copulation with a minor was not “categorically a crime involving moral turpitude”).
23
Taylor, 495 U.S. at 582; see also United States v. Doctor, 842 F.3d 306, 313
(4th Cir. 2016) (Wilkinson, J., concurring) (“It surprises me that we have arrived at this
point, because in theory, the categorical approach makes a good deal of sense .... But
what was fine in theory has sometimes proven to be less so in practice.”).
See, e.g., Blockburger v. United States, 284 U.S. 299, 304 (1932) (applying an
24
elements-based test to prohibit successive prosecutions for the same criminal act.).
32
that conclusion another long, hard look. The time-consuming casuistry compelled by the
categorical approach dwarfs the concerns that were raised in Taylor and should call into
question the defense of the approach outlined in Mathis. Will there be cases in which the
record is unclear? Certainly. But those cases are the ones in which the categorical
approach may be resorted to as needed, and we can trust district judges to recognize when
that is so. Will a Sixth Amendment jury issue arise from time to time? Perhaps. And
again, the categorical approach may prove useful then, as district judges will recognize.
Such exceptions, however, do not justify the current rule. 25
The chorus of voices over the years calling for a release from the analytical fetters
of Taylor and its progeny could hardly be louder, reflecting a broad consensus that the
categorical approach has gone terribly awry and that serious corrections are in order. 26
25
A good argument can be made for scrapping the categorical approach entirely
and requiring sentencing courts to look at the true facts provable in each case when the
crime of conviction is not obviously a “crime of violence” or “serious drug offense.” If
the requisite facts cannot be proven in those outlier cases, then the prosecution will have
failed to carry its burden of showing that the prior conviction in question is a qualifying
predicate offense for the sentencing enhancement it seeks. Judge Hardiman advocates
this approach.
26
See, e.g., Taylor, 142 S. Ct. at 2032 (Thomas, J., dissenting) (“[C]ourts
attempting to apply the categorical approach waste time thinking up improbable
hypotheticals, making the approach very difficult to administer.” (internal quotation
marks omitted)); Mathis v. United States, 579 U.S. 500, 538 (2016) (Alito, J., dissenting)
(“The Court’s approach calls for sentencing judges to delve into pointless abstract
questions.”); Chambers v. United States, 555 U.S. 122, 133 (2009) (Alito, J., concurring)
(“[T]he ‘categorical approach’ to predicate offenses has created numerous splits among
the lower federal courts, the resolution of which could occupy this Court for years.”); De
Lima v. Sessions, 867 F.3d 260, 268 (1st Cir. 2017) (“Even a single such categorical
analysis is an arduous task” that “is often difficult and time consuming.”); United States
v. Scott, 990 F.3d 94, 126 (2d Cir. 2021) (Park, J., concurring on behalf of five judges,
and collecting cases demonstrating discontent) (“As a growing number of judges across
33
the country have explained, the categorical approach perverts the will of Congress, leads
to inconsistent results, wastes judicial resources, and undermines confidence in the
administration of justice.”); United States v. Chapman, 866 F.3d 129, 139 (3d Cir. 2017)
(Jordan, J., concurring) (“Forcing judges to close their eyes [when applying the
categorical approach] to what is obvious promotes inefficiency and guarantees difficult-
to-explain sentences.”); Doctor, 842 F.3d at 313 (Wilkinson, J., concurring, and
collecting cases demonstrating counterintuitive results of the categorical approach)
(“[The categorical approach] involves an exhaustive review of state law as courts search
for a non-violent needle in a haystack or conjure up some hypothetical situation to
demonstrate that the predicate state crime just might conceivably reach some presumably
less culpable behavior outside the federal generic.”); Vann, 660 F.3d at 787 (Agee, J.,
concurring) (“The dockets of ... all federal courts are now clogged with [ACCA] cases.”);
United States v. Reyes-Contreras, 910 F.3d 169, 186 (5th Cir. 2018), abrogated in part
by Borden v. United States, 141 S. Ct. 1817 (2021) (“The well-intentioned experiment
[the categorical approach] that launched fifteen years ago has crashed and burned.”);
United States v. Burris, 912 F.3d 386, 407 (6th Cir. 2019) (en banc) (Thapar, J.,
concurring, and collecting cases with similar sentiment) (“A casual reader … might
struggle to understand why we are even debating if ramming a vehicle into a police
officer is a crime of violence. The reader’s struggle would be understandable. The time
has come to dispose of the long-baffling categorical approach.”); Bridges v. United
States, 991 F.3d 793, 804 (7th Cir. 2021) (“[T]he categorical approach frequently
produces counterintuitive results and has been the subject of much judicial
handwringing.”); Brown, 929 F.3d at 561 (Loken, J., dissenting) (“[T]his [outcome]
‘demonstrates the absurdity of applying the categorical approach to the enumerated-
offenses clause.’ Whether to abandon the categorical approach is of course an issue for
the Supreme Court.” (quoting Quarles v. United States, 139 S.Ct. 1872, 1880 (2019)
(Thomas, J., concurring))); United States v. Aguila-Montes de Oca, 655 F.3d 915, 917
(9th Cir. 2011) (en banc), abrogated by Young v. Holder, 697 F.3d 976 (9th Cir. 2012),
and Descamps, 570 U.S. 254 (“In the twenty years since Taylor, we have struggled to
understand the contours of the Supreme Court’s framework. Indeed, … perhaps no other
area of the law has demanded more of our resources.”); United States v. Valdivia-Flores,
876 F.3d 1201, 1210 (9th Cir. 2017) (O’Scannlain, C.J., specially concurring to
“highlight how it illustrates the bizarre and arbitrary effects of the ever-spreading
categorical approach”); Davis, 875 F.3d at 595 (“So [in this categorical approach case]
we go down the rabbit hole again to a realm where we must close our eyes as judges to
what we know as men and women. It is a pretend place in which a crime that the
defendant committed violently is transformed into a non-violent one … Curiouser and
curiouser it has all become, as the holding we must enter in this case shows. Still we are
required to follow the rabbit.”).
34
That observation is not an expression of discontent with the work courts are
expected to do. It is instead a protest about one of the tools we are being required to use.
We have, for instance, come a very long way from the intent of Congress to apply ACCA
with an eye toward the even-handed evaluation of offender conduct. Indeed, we are in a
world that is exactly the opposite of what Congress wanted, a world now where “two
defendants who, in their past, independently committed identical criminal acts in two
different states and have essentially the same criminal history will find that the
applicability of ACCA to their current cases depends not on their past criminal conduct
but on the phrasing of the different state criminal statutes.” United States v. Chapman,
866 F.3d 129, 137 (3d Cir. 2017) (Jordan, J., concurring).
And it’s not just courts complaining about these inequities, or at least recognizing
that there is a major problem. Lawyers on both sides of the “v.” see it too. Several years
ago, Judge Wilkinson of the Fourth Circuit quoted defense counsel who described the
categorical approach as “a morass of jurisprudential goo” and a “particularly glorious
goo, because the confusion almost inevitably helps our clients.” United States v. Doctor,
842 F.3d 306, 316 (4th Cir. 2016) (Wilkinson, J., concurring). On the other side, an
experienced Assistant United States Attorney in our own Circuit has called the
categorical approach “a unique theory of statutory interpretation that warps the
application of federal criminal provisions,” and an “approach [that] subjects an offender
to criminal penalties not based on what he did, but on whether someone else could violate
the same statute he did but do it in a less violent way.” Robert A. Zauzmer, Fixing the
35
Categorical Approach “Mess”, 69 Dep’t Just. J. Fed. L. & Prac. 3, 5, 10 (2021)
(emphasis omitted).
No one is quite sure who first said it, but a wise proverb declares the “First Law of
Holes” to be, “when you are in one, stop digging.” 27 We are in one, and it is not getting
better with shouts of encouragement and more shovels.
27
If you’re inclined to believe Wikipedia, the origin of the adage is unknown but
an early example was found “on page six of The Washington Post dated 25 October 1911,
in the form: ‘Nor would a wise man, seeing that he was in a hole, go to work and blindly
dig it deeper[.]’” Law of Holes, Wikipedia,
https://en.wikipedia.org/wiki/Law_of_holes#:~:text=The%20law%20of%20holes%20or,s
top%20making%20the%20situation%20worse (last visited Oct. 13, 2023). Another apt
metaphor was provided by Justice Alito in his dissent in Mathis:
Sabine Moreau lives in Solre–sur–Sambre, a town in Belgium located 38 miles
south of Brussels. One day she set out in her car to pick up a friend at the Brussels
train station, a trip that should have taken under an hour. She programmed her
GPS and headed off. Although the GPS sent her south, not north, she apparently
thought nothing of it. She dutifully stayed on the prescribed course. Nor was she
deterred when she saw road signs in German for Cologne, Aachen, and Frankfurt.
“I asked myself no questions,” she later recounted. “I kept my foot down.”
Hours passed. After crossing through Germany, she entered Austria. Twice she
stopped to refuel her car. She was involved in a minor traffic accident. When she
tired, she pulled over and slept in her car. She crossed the Alps, drove through
Slovenia, entered Croatia, and finally arrived in Zagreb—two days and 900 miles
after leaving her home. Either she had not properly set her GPS or the device had
malfunctioned. But Moreau apparently refused to entertain that thought until she
arrived in the Croatian capital. Only then, she told reporters, did she realize that
she had gone off course, and she called home, where the police were investigating
her disappearance.
Twenty-six years ago, in Taylor v. United States, 495 U.S. 575, 602 (1990), this
Court set out on a journey like Moreau’s. Our task in Taylor, like Moreau’s short
trip to the train station, might not seem very difficult—determining when a
conviction for burglary counts as a prior conviction for burglary under the Armed
Career Criminal Act … . But things have not worked out that way.
Mathis, 579 U.S. at 536-37 (Alito, J., dissenting).
36
B. A More Fact-Based Approach
There is a better way, and we are not the first to suggest it. 28 It is simply this:
courts should be allowed to look at the record when it is readily available. A return to the
facts, when the record allows it, would promote substantial justice.
The illogical results compelled by the categorical approach often derive from its
well-intentioned but ill-conceived under-inclusivity. Again using ACCA sentencing as
an example, it will often be the case that the federal offense under consideration and the
state predicate offense match factually but that the state statute is broader elementally.
Permitting judges to rely on the facts underlying past convictions, when those facts are
readily ascertainable from Shepard documents, would allow application of the
enhancement based on reality and thus cure that under-inclusivity. While the categorial
approach could be a “default inquiry,” one that would apply in the absence of a factual
record, judges could use discretion to consider a defendant’s conduct based on the
available record. Doctor, 842 F.3d at 315, 319 (Wilkinson, J., concurring) (discussing
the discretion that district judges already have in deviating from the Sentencing
Guidelines and arguing that “the district court may decide in the face of an inconclusive
record to apply the categorical approach to predicate offenses, but it also should enjoy the
28
See, e.g., Descamps, 570 U.S. at 288-89 (Alito, J., dissenting) (“[I]n Shepard,
we observed that the factual circumstances of a defendant’s prior conviction may be
relevant to determining whether it qualifies as a violent felony under ACCA. … And in
Nijhawan, we departed from the categorical approach altogether and instead applied a
“circumstance-specific” approach. See [Nijhawan v. Holder,] 557 U.S. [29,] 36, 38
[(2009)]. If anything, then, Nijhawan undermines the majority’s position that rigid
adherence to elements is always required.”).
37
discretion and the tools to craft a more individualized sentence when such would serve
the ends of justice”).
In such a fact-supplemented approach, courts would begin by applying the
categorical approach to determine if the state offense contains an unmatched element. If
so, and the state statute is broader than the ACCA element, courts could “peek” at the
actual conduct underlying the state conviction, as made evident (at the very least) by the
Shepard documents, to determine whether the relevant conduct falls within the portion
covered by the ACCA offense. If covered, the unmatched element would be disregarded
in the categorical analysis and a sentence enhancement would be imposed if the
remaining elements also matched and satisfied all elements of the ACCA-qualifying
offense. If not, the inquiry would end and the underlying conviction would not qualify as
a predicate offense for ACCA. This kind of analysis would ensure a defendant’s conduct
met the requirements of ACCA, notwithstanding the particular phrasing of any state
statute. Used in this way, the categorical approach would provide direction but not
dictate an unreasonable result.
This already occurs, to a degree, with the modified categorical approach. It allows
courts to reference Shepard documents, so that, instead of sifting through state codes and
state judicial precedent, a district court can be informed by readily ascertainable facts. It
will, for instance, often be immediately apparent that a prior conviction for first-degree
aggravated assault did involve violence (as in the case before us now). One can argue, as
the dissent in Shepard did, that the universe of documents the Supreme Court has
sanctioned for review is too narrow. See Shepard, 544 U.S. at 28 (O’Connor, J.,
38
dissenting) (“The Court today adopts a rule that is not compelled by statute or by this
Court’s precedent, that makes little sense as a practical matter, and that will substantially
frustrate Congress’ scheme for punishing repeat violent offenders who violate federal gun
laws.”). But just loosening the reins enough to permit resort to Shepard documents
would be a major step forward.
To assist in appellate review, district courts employing this fact-supplemented
approach could expressly note what record facts they are relying on and explain how
those facts are clear enough to support their conclusion that defaulting to the categorical
approach is not appropriate. An example of a record that is adequate to support the kind
of analysis we are suggesting is the one before us right now. The reason for Harris’s
first-degree aggravated assault conviction is, based on the plea colloquy alone,
abundantly clear, and the clarity of the record justifies using the approach we are
advocating.
This kind of analysis can and should be done without regard to the means-versus-
elements dichotomy that currently occupies so much time to so little purpose. If the facts
of a crime are readily available through Shepard documents, why should it matter
whether a statute is divisible along strictly elemental lines? That may be an interesting
intellectual exercise, but it does not have any apparent connection to Congressional will
or to the advancement of justice in any given case. Quite the contrary. As already
outlined, the current analytical construct frustrates rather than fulfills Congressional
intent and the just instinct that punishment should fit the crime.
39
Nor should Sixth Amendment concerns prevent a course correction. Judges are
permitted to review Shepard documents when the modified categorical approach applies.
Also, they must consider, inter alia, the “history and characteristics of the defendant,” 18
U.S.C. § 3553(a)(1), and the need to “protect the public from further crimes of the
defendant,” 18 U.S.C. § 3553(a)(2)(C). Those considerations – and the Shepard
documents – relate to the defendant’s past, not his current crime of conviction. So, the
Supreme Court’s decision in Apprendi v. New Jersey, 530 U.S. 466 (2000), which had
nothing to do with recidivism, does not apply. 29
29
Judge Matey joins this concurrence but notes two issues that, in his view,
animate the many problems produced by the categorical approach. First, assuming
statutes like 18 U.S.C. §§ 922(g) and 924(e) fall within Congress’s power “[t]o regulate
Commerce . . . among the several States,” U.S. const. art. I, § 8, cl. 3, then fixing any
problems posed should be a congressional, not judicial, challenge. Lewis v. City of
Chicago, 560 U.S. 205, 217 (2010) (“[I]t is not our task to assess the consequences of
each approach and adopt the one that produces the least mischief. Our charge is to give
effect to the law Congress enacted.”). But Judge Matey notes it is unclear whether
Congress can craft a uniform definition of “violent felony” since “[u]nder our federal
system, the ‘States possess primary authority for defining and enforcing the criminal
law.’” United States v. Lopez, 514 U.S. 549, 561 n.3 (1995) (quoting Brecht v.
Abrahamson, 507 U.S. 619, 635 (1993)). In his view, shoehorning state standards for
“violent felonies” into a single federal statute seems unlikely to succeed. Cf. Rachel E.
Barkow, Categorical Mistakes: The Flawed Framework of The Armed Career Criminal
Act and Mandatory Minimum Sentencing, 133 Harv. L. Rev. 200, 208, 237–38 (2019)
(“It is remarkable that Congress would so cavalierly disrupt what has traditionally been a
local matter and impose such a harsh punishment regime without pausing to think about
or analyze how its new regime would have to adjust to fifty-one different jurisdictions
and the ways they define crime.”).
Second, Judge Matey remains concerned that attempts to fix the categorical
approach may diverge from the historical tradition of the Sixth Amendment’s
requirement that “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 (2000), despite the Supreme Court’s
allowance for an exception for “the fact of a prior conviction,” id.; see Almendarez–
Torres v. United States, 523 U.S. 224 (1998); Shepard v. United States, 544 U.S. 13, 27
40
The Mathis majority suggested there would be “unfairness to defendants” because
“[s]tatements of non-elemental fact in the records of prior convictions are prone to error
precisely because their proof is unnecessary.” Mathis, 579 U.S. at 512 (internal quotation
marks and citation omitted). “At trial,” the majority said, “and still more at plea hearings,
a defendant may have no incentive to contest what does not matter under the law; to the
contrary, he ‘may have good reason not to’ – or even be precluded from doing so by the
court.” Id. at 512 (quoting Descamps, 570 U.S. at 270). Respectfully, we think that
reasoning wrongly calls into question much of our justice system. Guilty pleas are not
mere theater. The facts are carefully put on record to ensure that they satisfy the
elements of the crime and can support a conviction. A judge cannot accept a guilty plea
without hearing and finding those facts. That process is critical to the administration of
justice. It ensures that defendants willingly and knowingly admit to their actions and
accept guilt for them. If we cannot accept those pleaded facts as given but should view
them as suspect, we have bigger problems than the categorical approach. And what is
true of guilty pleas should be even more true of verdicts after a full trial.
The system does not always work perfectly, but it is the very best we have for
ascertaining truth. As such, we ought to rely on it and the outcomes it generates, as
reflected in Shepard documents, to determine whether a defendant’s prior convictions
warrant an ACCA enhancement or similar consequences dictated by other statutes.
(2005) (“Almendarez–Torres has been eroded by . . . subsequent Sixth Amendment
jurisprudence[.]”) (Thomas, J., concurring in part & concurring in judgment).
41
IV. Conclusion
Considering the well-intentioned provenance of the categorical approach, it is
ironic that it has come to be such an impediment to the sound administration of justice.
Last year, Justice Thomas chose a different metaphor than hole digging, but his point
seemed much the same when he memorably observed that the “‘categorical approach’ has
led the Federal Judiciary on a ‘journey Through the Looking Glass,’ during which we
have found many ‘strange things.’” Taylor, 142 S. Ct. at 2026 (Thomas, J., dissenting)
(quoting Lewis Carroll, Alice in Wonderland and Through the Looking Glass 227 (J.
Messner ed. 1982)). We hope the journey, and the digging, ends soon.
In the meantime, we have no choice but to deny the government’s petition for en
banc review.
42