Harvard Law Review Harvard Law Review Harvard Law Review

Criminal Law

Recidivism Reformation: Eliminating Drug Predicates

The Armed Career Criminal Act (ACCA) imposes a minimum fifteen-year sentence for violating 18 U.S.C. § 922(g) after three “violent felony” or “serious drug offense” convictions. The ACCA disproportionately impacts people of color and imposes significant costs on the federal judiciary and the criminal justice system overall. This Essay contributes to ACCA recidivism analysis by studying, for the first time, all people sentenced under ACCA and released in 2009, 2010, and 2011 during an eight-year follow-up period. This Essay uses nonpublicly available criminal history information from the Federal Bureau of Investigation and recidivism data from the United States Sentencing Commission. People sentenced under the ACCA with fewer than three “violent felony” predicates (and one or more “serious drug offense” predicates) reoffended less than those sentenced under the ACCA with at least three “violent felony” predicates. This Essay therefore recommends that Congress eliminate drug offenses as ACCA predicates because people sentenced under ACCA with fewer than three “violent felon[ies]” recidivate at a rate no greater than the federal offender population as a whole.

Introduction

For the last thirty years, Congress has shown a desire to consider criminal history in sentencing. This desire can be seen in the Sentencing Reform Act of 1984,1×1. Pub. L. No. 98-473, tit. II, ch. II, 98 Stat. 1987 (codified as amended in scattered sections of 18 and 28 U.S.C.). as well as in every statute prescribing enhanced penalties for recidivism.2×2. See, e.g., Comprehensive Crime Control Act of 1984, Pub. L. No. 98-473, tit. II, 98 Stat. 1976 (codified as amended in scattered sections of 18 and 28 U.S.C.). For purposes of this Essay, recidivism “refers to a person’s relapse into criminal behavior, often after the person receives sanctions or undergoes intervention for a previous crime.” Recidivism, Nat’l Inst. of Just., https://nij.ojp.gov/topics/corrections/recidivism [https://perma.cc/QB6T-3L9J]. One of these statutes is the Armed Career Criminal Act3×3. 18 U.S.C. § 924(e). (ACCA), which imposes a minimum fifteen-year sentence for unlawfully possessing a weapon or ammunition under 18 U.S.C. § 922(g)4×4. An 18 U.S.C. § 922(g) offender, almost always, is a “felon-in-possession” (that is, a person previously convicted of a felony who is then found in possession of firearm). The text of the statute reads as follows:
It shall be unlawful for any person —
(1) who has been convicted in any court of, a crime punishable by imprisonment for a term exceeding one year;
(2) who is a fugitive from justice;
(3) who is an unlawful user of or addicted to any controlled substance (as defined in section 102 of the Controlled Substances Act (21 U.S.C. 802));
(4) who has been adjudicated as a mental defective or who has been committed to a mental institution;
(5) who, being an alien —
(A) is illegally or unlawfully in the United States; or
(B) except as provided in subsection (y)(2), has been admitted to the United States under a nonimmigrant visa (as that term is defined in section 101(a)(26) of the Immigration and Nationality Act (8 U.S.C. 1101(a)(26)));
(6) who has been discharged from the Armed Forces under dishonorable conditions;
(7) who, having been a citizen of the United States, has renounced his citizenship;
(8) who is subject to a court order that —
(A) was issued after a hearing of which such person received actual notice, and at which such person had an opportunity to participate;
(B) restrains such person from harassing, stalking, or threatening an intimate partner of such person or child of such intimate partner or person, or engaging in other conduct that would place an intimate partner in reasonable fear of bodily injury to the partner or child; and
(C)(i) includes a finding that such person represents a credible threat to the physical safety of such intimate partner or child; or
(ii) by its terms explicitly prohibits the use, attempted use, or threatened use of physical force against such intimate partner or child that would reasonably be expected to cause bodily injury; or
(9) who has been convicted in any court of a misdemeanor crime of domestic violence, to ship or transport in interstate or foreign commerce, or possess in or affecting commerce, any firearm or ammunition; or to receive any firearm or ammunition which has been shipped or transported in interstate or foreign commerce. Id. § 922(g).
after three prior convictions, committed on different occasions for (1) “violent felon[ies],” (2) “serious[footnote-short-crop]Under the ACCA, a drug offense is “serious” if it carries a sentence of ten years or more. See id. § 924(e)(2)(A).[/footnote-short-crop] drug offense[s],” or (3) some combination of both.5×5. Id. § 924(e). To examine recidivism rates for various categories of people sentenced under the ACCA, the author, as part of her Supreme Court Fellowship, designed a study and entered into an agreement with the Federal Bureau of Investigation (FBI) and the United States Sentencing Commission to obtain access to nonpublicly available criminal history data.6×6. See U.S. Sent’g Comm’n, Federal Armed Career Criminals: Prevalence, Patterns, and Pathways 81 n.83 (2021) [hereinafter Armed Career Criminals].

Using that nonpublic data, this Essay compares the recidivism rate of people sentenced under the ACCA with fewer than three “violent felony” predicates (and one or more “serious drug offense” predicates) with those sentenced under the ACCA with at least three “violent felony” predicates. Based on this comparison, this Essay recommends that Congress eliminate drug offenses as ACCA predicates because people sentenced under the ACCA who committed fewer than three “violent felon[ies]” recidivate at a rate no greater than the federal population as a whole.

Many scholars have criticized the ACCA for reasons ranging from its “randomized draconianism”7×7. Mark Kleiman, How to Have Less Crime and Less Punishment, The Atlantic (Aug. 9, 2010), https://www.theatlantic.com/national/archive/2010/08/how-to-have-less-crime-and-less-punishment/61123 [https://perma.cc/FK29-KS76]. to its disproportionate impact on people of color.8×8. See, e.g., Leah M. Litman, Essay, Residual Impact: Resentencing Implications of Johnson’s Potential Ruling on ACCA’s Constitutionality, 115 Colum. L. Rev. Sidebar 55, 56 (2015) (“[The] ACCA is a flashpoint for many of the most pressing issues facing criminal law today — Hispanic and [B]lack offenders receive the ACCA enhancement at higher rates than [W]hite offenders do . . . .”). This Essay is the first to provide an empirical basis upon which to make necessary changes. Former Sentencing Commissioner Professor Rachel Barkow recently argued that “Congress should have allowed the Sentencing Commission to use data and evidence to guide sentencing policy . . . instead of trying to take that task on for itself.”9×9. Rachel E. Barkow, The Supreme Court, 2018 Term — Comment: Categorical Mistakes: The Flawed Framework of the Armed Career Criminal Act and Mandatory Minimum Sentencing, 133 Harv. L. Rev. 200, 209 (2019). This Essay does just that — it recommends changes to the scope of the ACCA’s penalties by providing empirical analysis regarding the scope, impact, and effectiveness of the current penalty structure.

This Essay concentrates on the utilitarian purposes of punishment: specific deterrence,10×10. Many authors argue that punishment must deter crime in order “to justify its extreme institutional costs.” Jacob Bronsther, Two Theories of Deterrent Punishment, 53 Tulsa L. Rev. 461, 462 (2018) (citing, inter alia, Douglas Husak, Holistic Retributivism, 88 Calif. L. Rev. 991, 996 (2000); Victor Tadros, The Ends of Harm: The Moral Foundations of Criminal Law 88–110 (2011); Michael T. Cahill, Punishment Pluralism, in Retributivism: Essays on Theory and Policy 25, 39 (Mark D. White ed., 2011)). rehabilitation, and incapacitation.11×11. See Michele Cotton, Back with a Vengeance: The Resilience of Retribution as an Articulated Purpose of Criminal Punishment, 37 Am. Crim. L. Rev. 1313, 1313 (2000). Generally, criminal punishment seeks to serve four purposes: retribution, deterrence, rehabilitation, and incapacitation.12×12. See id. Retribution “supposes that crime inherently merits punishment”13×13. Id. at 1315. and focuses on giving the offender what he deserves. Deterrence seeks to prevent future illegal conduct, both by the offender, through specific deterrence, and by the rest of the population, through general deterrence.14×14. See id. at 1316. Rehabilitation attempts to assist the offender, perhaps through psychological counseling or job training, to improve himself to prevent his future misconduct.15×15. See id. at 1316–17. Incapacitation protects society from the offender through imprisonment.16×16. See id. at 1316.

At the time of the ACCA’s passing, Congress intended that the ACCA would “substantial[ly] deter[]” crime and “incapacitat[e] career criminals.”17×17. S. Rep. No. 97-585, at 8 (1982). Accordingly, this Essay uses recidivism as one of the key measures to determine whether it is necessary to incarcerate all people eligible for ACCA sentencing for fifteen years in order to achieve these utilitarian purposes.

Congress enacted the ACCA to combat violence,18×18. Id.; see also infra p. 427. yet as currently structured, the ACCA punishes some people who have never committed a violent offense. The triggering offense, 18 U.S.C. § 922(g), simply requires possession, receipt, transport, or shipment of a firearm or ammunition, not use. Additionally, the predicate convictions may be exclusively drug offenses.

The ACCA also imposes significant administrative costs on the judiciary. In Taylor v. United States,19×19. 495 U.S. 575 (1990). the Supreme Court mandated that courts use the “categorical approach” to determine if a prior offense serves as an ACCA predicate.20×20. Id. at 602. Sentencing decisions applying the categorical approach are notoriously complicated and extensively litigated. In 2011, Ninth Circuit Judge Bybee wrote that “over the past decade, perhaps no other area of the law has demanded more of our resources.”21×21. United States v. Aguila-Montes de Oca, 655 F.3d 915, 917 (9th Cir. 2011) (en banc) (opinion of Bybee, J.) (collecting cases). Fourth Circuit Judge Agee lamented: “The dockets of our court and all federal courts are now clogged with [ACCA] cases.”22×22. United States v. Vann, 660 F.3d 771, 787 (4th Cir. 2011) (en banc) (Agee, J., concurring) (per curiam). Justice Alito opined that “only Congress can rescue the federal courts from the mire into which [the] ACCA’s draftsmanship and Taylor’s ‘categorical approach’ have pushed us.”23×23. Chambers v. United States, 555 U.S. 122, 132 (2009) (Alito, J., concurring in the judgment). Federal court dockets are filled with ACCA cases, not only relating to the ACCA’s “violent felony” provision but also regarding the “serious drug offense” clause.24×24. See, e.g., Shular v. United States, 140 S. Ct. 779, 782 (2020); United States v. Cantu, 964 F.3d 924, 934 (10th Cir. 2020) (holding that convictions under Oklahoma statute are not “serious drug offenses” because the statute includes the distribution of drugs that are not federally controlled).

Relatedly, the ACCA also imposes notable costs on the criminal justice system overall. Federal public defenders have commented that “[t]he ACCA is not only poorly drafted, but its irrational harshness has become one of the engines driving mass over-incarceration in America.”25×25. Stephen R. Sady & Gillian R. Schroff, Johnson: Remembrance of Illegal Sentences Past, 28 Fed. Sent’g Rep. 58, 63 (2015). In fiscal year 2021, 264 people (0.5% of all people sentenced in federal court) were sentenced under the ACCA.26×26. U.S. Sent’g Comm’n, 2021 Annual Report and Sourcebook of Federal Sentencing Statistics 77 tbl.23 (2022) [hereinafter 2021 Sourcebook of Federal Sentencing Statistics], https://www.ussc.gov/sites/default/files/pdf/research-and-publications/annual-reports-and-sourcebooks/2021/2021_Annual_Report_and_Sourcebook.pdf [https://perma.cc/6GSC-YD3W]. Because of its lengthy minimum sentence, people convicted under the ACCA represent a larger percentage of the federal prison population than the percentage of the annual federal docket they comprise. As of June 27, 2020, 3572 incarcerated people were sentenced under the ACCA, comprising 2.5% of the federal prison population.27×27. Armed Career Criminals, supra note 7, at 29. In the fiscal years 2017, 2018, 2019, and 2020, respectively, 268, 301, 312, and 261 people were sentenced under the ACCA. Id. at 19 fig.1; U.S. Sent’g Comm’n, 2020 Annual Report and Sourcebook of Federal Sentencing Statistics 77 tbl.23 (2021), https://www.ussc.gov/sites/default/files/pdf/research-and-publications/annual-reports-and-sourcebooks/2020/2020-Annual-Report-and-Sourcebook.pdf [https://perma.cc/L3K3-G65Z].

However, rolling back mandatory minimum sentences can be politically problematic. Congressional representatives likely fear that released prisoners who recidivate will pose a political liability. Yet now may be the opportune moment for restructuring the ACCA. Congress has showed its willingness to revisit mandatory minimums, as in the First Step Act,28×28. Pub. L. No. 115-391, 132 Stat. 5194 (2018) (codified as amended in scattered sections of 18, 21, 34, and 42 U.S.C.). for example. Moreover, because COVID-19 spreads rapidly in prisons and jails, the current pandemic has further motivated policymakers to consider ways to reduce incarceration, not just for the sake of inmates but also for the protection of prison employees and surrounding communities. Most recently, President Joe Biden issued an executive order directing the Department of Justice not to renew contracts with privately operated criminal detention facilities,29×29. Exec. Order No. 14,006, 86 Fed. Reg. 7483, 7483 (Jan. 26, 2021). creating further need to reduce the federal prison population.

This Essay makes a novel contribution to ACCA recidivism analysis by studying all people sentenced under the ACCA30×30. That is, all ACCA offenders whose ACCA sentence was not vacated; were U.S. citizens; possessed valid FBI numbers; were not reported dead, escaped, or detained; and whose presentence investigation report was submitted to the U.S. Sentencing Commission. Armed Career Criminals, supra note 7, at 41. and released in 2009, 2010, and 2011 during an eight-year follow-up period. Most prior recidivism studies involved only people released during one year31×31. See, e.g., U.S. Sent’g Comm’n, Recidivism Among Federal Firearms Offenders (2019) [hereinafter Firearms Recidivism Report for 2005 Releasees], https://www.ussc.gov/sites/default/files/pdf/research-and-publications/research-publications/2019/20190627_Recidivism_Firearms.pdf [https://perma.cc/LSZ2-E5ZS]. or follow-up periods of no more than a few years.32×32. See Megan C. Kurlychek et al., Long-Term Crime Desistance and Recidivism PatternsEvidence from the Essex County Convicted Felon Study, 50 Criminology 71, 79 (2012); see also, e.g., J.J. Prescott et al., Understanding Violent-Crime Recidivism, 95 Notre Dame L. Rev. 1643, 1672, 1684 (2020) (detailing the article’s study using a three-year follow-up period and describing a separate study by the Bureau of Justice Statistics that also used a three-year follow-up period). Generally, recidivism studies are only as accurate as the public information upon which they are based,33×33. See, e.g., Prescott et al., supra note 33, at 1683. but this study utilizes nonpublicly available criminal history information from the FBI and recidivism data from the U.S. Sentencing Commission.34×34. Armed Career Criminals, supra note 7, at 17, 81 n.83.

This study compares the recidivism rates of three categories of people sentenced under the ACCA: (1) 616 people (83.4% of the total) with at least three “violent felony” predicates; (2) 33 people (4.5% of the total) with no “violent felony” predicates (who committed only “serious drug offense[s]”); and (3) 90 people (12.2% of the total) with one or two “violent felony” predicates (and one or more “serious drug offense[s]”).35×35. Id. at 47, 53 fig.17. People with at least three “violent felony” predicates recidivated at the highest rate, followed by those with one or two “violent felony” predicates.36×36. See id. at 54 tbl.8, 57. People with no “violent felony” predicates recidivated at the lowest rate.37×37. See id. The study report notes that “the 33 individual offenders in the drug trafficking pathway were insufficient in number to conduct a meaningful analysis to support any reliable conclusions about recidivism regarding those offenders.” Id. at 57. No demographics appear to explain the recidivism disparities between the categories.

Part I explains how the ACCA works and how it relates to the firearms-possession statute, 18 U.S.C. § 922(g). Part II provides a detailed history of the ACCA, including its expansion at the same time the federal sentencing guidelines were coming into effect. Additionally, Part II details sentencing statistics under the ACCA for the last ten years. Part II also highlights the problems associated with the ACCA, such as its failure to achieve its intended goals of preventing violent crime and reducing sentencing inconsistencies; its harshness in comparison to the average sentences for various crimes; its costs to the judiciary and the criminal justice system overall; and the racial and geographic disparities in its application. Part III explains the methodology of this Essay’s study and the demographics of the groups studied, comparing them to the demographics of people more recently sentenced under the ACCA. Part III then presents the overall recidivism rates for people in this Essay’s study; the difference between the recidivism rates of those who committed drug and violent offenses in this study; and an analysis of the results. Finally, Part IV explains what the author recommends in light of these findings and why now is the opportune time for Congress, already focused on criminal justice reform, to implement these recommendations.

I. The Armed Career Criminal Act

The Armed Career Criminal Act (ACCA) imposes a minimum38×38. A “mandatory minimum penalty” is “a federal criminal statute requiring, upon conviction of a federal criminal offense and the satisfaction of criteria set forth in that statute, the imposition of a specified minimum term of imprisonment.” U.S. Sent’g Comm’n, Report to the Congress: Overview of Mandatory Minimum Penalties in the Federal Criminal Justice System 4 (2011) [hereinafter 2011 Report to Congress]. fifteen-year sentence for violating 18 U.S.C. § 922(g) after committing, on different occasions, three prior “violent felon[ies],” “serious drug offense[s],” or both.39×39. 18 U.S.C. § 924(e). Specifically, the statute provides:

In the case of a person who violates section 922(g) of this title and has three previous convictions by any court referred to in section 922(g)(1) of this title for a violent felony or a serious drug offense, or both, committed on occasions different from one another, such person shall be fined under this title and imprisoned not less than fifteen years . . . .40×40. Id. § 924(e)(1).

18 U.S.C. § 922(g) bans specific classes of people from transporting, possessing, receiving, or shipping firearms or ammunition and ordinarily carries a maximum ten-year penalty and no minimum penalty.41×41. Firearms Recidivism Report for 2005 Releasees, supra note 32, at 8.

The punishment for violating § 922(g) is found at § 924. The statute generally provides a maximum imprisonment of ten years. However, an enhanced penalty, a fifteen-year minimum sentence,42×42. A court may not grant probation to a person eligible for sentencing under the ACCA for his § 922(g) offense. is provided where a person violates § 922(g) and possesses three previous convictions, committed on different occasions, for “violent felon[ies],” “serious drug offense[s],” or both.43×43. 18 U.S.C. § 924(e). “Violent felony” means:

any crime punishable by imprisonment for a term exceeding one year, or any act of juvenile delinquency 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 . . . .44×44. Id. § 924(e)(2)(B). The statute also provides a catch-all provision, “or otherwise involves conduct that presents a serious potential risk of physical injury to another,” id., but in Johnson v. United States, 576 U.S. 591 (2015), the Supreme Court held that the quoted language above, commonly referred to as the “residual clause,” was unconstitutionally vague, id. at 594, 597. The Johnson holding applies retroactively for cases under collateral review. Welch v. United States, 578 U.S. 120, 135 (2016).
There are individuals in this Essay’s study sample who were sentenced under the “residual clause” and otherwise for crimes that today are no longer considered ACCA predicates. The author removed offenders sentenced under the “residual clause” only if their sentences were vacated.

A “serious drug offense” is a federal45×45. A federal controlled-substance offense is defined as “an offense under the Controlled Substance Act (21 U.S.C. 801 et seq.), the Controlled Substances Import and Export Act (21 U.S.C. 951 et seq.), or chapter 705 of title 46.” 18 U.S.C. § 924(e)(2)(A)(i). or state controlled-substance offense46×46. A state controlled-substance offense is defined as “an offense under State law, involving manufacturing, distributing, or possessing with intent to manufacture or distribute, a controlled substance (as defined in section 102 of the Controlled Substances Act (21 U.S.C. 802)).” Id. § 924(e)(2)(A)(ii). with a maximum term of imprisonment of ten years or more. The ACCA places no time or age limitation on predicate offenses other than requiring that the prior convictions occur before the 18 U.S.C. § 922(g) offense.47×47. See, e.g., United States v. Moody, 770 F.3d 577, 580 (7th Cir. 2014) (“The district court . . . had no authority to ignore the [ACCA predicate] conviction because of its age or its underlying circumstances. Such considerations are irrelevant in determining predicate offenses under the Act.” (citing, inter alia, Shepard v. United States, 544 U.S. 13, 15–17 (2005))); United States v. Keesee, 358 F.3d 1217, 1221 (9th Cir. 2004) (“The only time limitation supported by the language of the Armed Career Criminal Act is that the predicate convictions be ‘previous.’” (quoting 18 U.S.C. § 924(e)(1))).

People who provide “substantial assistance” may be sentenced below a statutory mandatory minimum.48×48. See 18 U.S.C. § 3553(e). See generally 28 U.S.C. § 994(n) (directing the U.S. Sentencing Commission to propose guidelines that “reflect the general appropriateness” of sentencing ranges below statutory mandatory minimums when a defendant provides “substantial assistance”); U.S. Sent’g Guidelines Manual § 5K1.1 (U.S. Sent’g Comm’n 2016). Through the last ten years, approximately one-fifth of people sentenced under the ACCA were relieved from the mandatory minimum for providing “substantial assistance.”49×49. Armed Career Criminals, supra note 7, at 27. That percentage is generally higher than the percentage of people overall who receive “substantial assistance” departures from the Sentencing Guidelines. For example, 9.6% of all people convicted in federal court received such departures in fiscal year 2021.50×50. 2021 Sourcebook of Federal Sentencing Statistics, supra note 27, at 84 tbl.29.

The ACCA sentence length during the last ten years for a person who received “substantial assistance” relief was generally ten years or less, ranging from 106 months to 122 months.51×51. Armed Career Criminals, supra note 7, at 28. Sentences for people who remained subject to the ACCA’s mandatory minimum ranged from 197 months to 214 months.52×52. Id.

II. The ACCA’s History and Problems

A. The ACCA’s History

Crime rates in the United States continuously increased beginning in the 1960s until they peaked in 1991.53×53. See Lauren-Brooke Eisen & Oliver Roeder, America’s Faulty Perception of Crime Rates, Brennan Ctr. for Just. (Mar. 16, 2015), https://www.brennancenter.org/our-work/analysis-opinion/americas-faulty-perception-crime-rates [https://perma.cc/GW4M-VEMF]. Homicides doubled between 1960 and 1980.54×54. Barry Latzer, Michelle Alexander Is Wrong About Mass Incarceration, Nat’l Rev. (Apr. 4, 2019, 11:41 AM), https://www.nationalreview.com/magazine/2019/04/22/michelle-alexander-is-wrong-about-mass-incarceration [https://perma.cc/3FPE-Z83B]. During this time period, many experts predicted that the worst was yet to come.55×55. Glenn C. Loury, Race, Incarceration, and American Values 4 (2008). The prevalence and fear of future crime motivated President Lyndon Johnson to announce a national “War on Crime” on March 8, 1965, and President Richard Nixon to announce the “War on Drugs” in June 1971.56×56. See A History of the Drug War, Drug Pol’y All., https://drugpolicy.org/issues/brief-history-drug-war [https://perma.cc/QUG4-7UEH], Elizabeth Hinton, Why We Should Reconsider the War on Crime, Time (Mar. 20, 2015, 7:00 AM), https://time.com/3746059/war-on-crime-history [https://perma.cc/U3HM-JWE7]; Matthew D. Lassiter et al., Detroit Under Fire: Police Violence, Crime Politics, and the Struggle for Racial Justice in the Civil Rights Era, Univ. of Mich. Carceral State Project (Mar. 2021), https://policing.umhistorylabs.lsa.umich.edu/s/detroitunderfire/page/national-and-local-war-on-crime [https://perma.cc/UTW5-HSE9]. It was against this backdrop that Congress began the road to the ACCA.

The Career Criminal Life Sentence Act of 198157×57. S. 1688, 97th Cong. (1981). proposed a mandatory life sentence for a person with three robbery or burglary convictions if the third instance involved the use of a firearm.58×58. See Barkow, supra note 10, at 211. Congress consulted with the Department of Justice and modified the Act to provide a fifteen-year minimum sentence and a maximum life sentence for a person illegally possessing a firearm with three prior robbery or burglary convictions, retitling it the “Armed Career Criminal Act of 1984.”59×59. Pub. L. No. 98-473, tit. II, ch. XVIII, 98 Stat. 2185 (repealed 1986). The accompanying House Report cited recidivism statistics showing that a small number of repeat offenders commit a large percentage of crimes.60×60. H.R. Rep. No. 98-1073, at 1 (1984), as reprinted in 1984 U.S.C.C.A.N. 3661, 3661 (“It has . . . become increasingly clear that a large percentage of . . . crimes [involving theft or violence] are committed by a very small percentage of repeat offenders.”). The purpose of the statute was to incapacitate habitual criminals with longer sentences.61×61. Barkow, supra note 10, at 212 (“If several hundred of the worst career criminals are sentenced to 15-year Federal prison terms, that in itself will prevent tens of thousands of felonies.” Id. at 212 n.100 (quoting Armed Robbery and Burglary Prevention Act: Hearing on H.R. 6386 Before the Subcomm. on Crime of the H. Comm. on the Judiciary, 97th Cong. 11 (1982) (statement of Rep. Ron Wyden); and then citing id. at 63 (statement of Prof. Charles Wellford, Director, Inst. of Crim. Just. & Criminology at the Univ. of Maryland) (stating that because “the level of deterrence that we can probably achieve in our system will not be a factor that will slow them down . . . we look to . . . incapacitation”))).

In the Career Criminals Amendment Act,62×62. 18 U.S.C. § 924(e). part of the Anti-Drug Abuse Act of 1986,63×63. Pub. L. No. 99-570, 100 Stat. 3207 (codified as amended in scattered sections of the U.S. Code). Congress introduced the ACCA in its present form by expanding it to include “violent felon[ies]” and “serious drug offenses” as predicates.64×64. 18 U.S.C. § 924(e); see also Barkow, supra note 10, at 212; 2011 Report to Congress, supra note 39, at 26. A Representative who supported adding drug offenses to the ACCA argued: “All the evidence . . . we have seen in our investigation indicates that drugs and violent crime go hand-in-hand.”65×65. Armed Career Criminal Act Amendments: Hearing on S. 2312 Before the Subcomm. on Crim. L. of the S. Comm. on the Judiciary, 99th Cong. 5 (1986) [hereinafter Hearing on ACCA Amendments] (statement of Rep. Ron Wyden). The U.S. Sentencing Commission responded to the ACCA by promulgating section 4B1.4 (Armed Career Criminal) of the Federal Sentencing Guidelines to provide guidance about the enhanced penalties to courts sentencing people under the ACCA. In 1988, Congress amended the ACCA,66×66. Anti-Drug Abuse Amendments Act of 1988, Pub. L. No. 100-690, tit. VI, 102 Stat. 4312 (codified as amended in scattered sections of the U.S. Code); Minor and Technical Criminal Law Amendments Act of 1988, Pub. L. No. 100-690, tit. VII, subtit. B, 102 Stat. 4395 (codified as amended in scattered sections of the U.S. Code). clarifying that predicate offenses must be “committed on occasions different from one another”67×67. 18 U.S.C. § 924(e)(1). and including as “violent felon[ies]” “act[s] of juvenile delinquency involving the use or carrying of a firearm, knife, or destructive device that would be punishable by imprisonment . . . if committed by an adult.”68×68. Id. § 924(e)(2)(B).

The goals of incapacitation and deterrence motivated Congress to create the ACCA. Congress “anticipated that the entry of the Federal Government into the field of prosecuting violent street crime w[ould] have a substantial deterrent effect.”69×69. S. Rep. No. 97-585, at 8 (1982) (statement of Sen. Arlen Specter). In enacting the ACCA, legislators sought “to improve public safety and reduce violent crime by incapacitating career criminals, through lengthy incarceration”70×70. Id. and to target “the most dangerous, frequent[,] and hardened offenders.”71×71. Id. at 5. Congress seemingly enacted the ACCA to combat violence, yet as currently structured, the ACCA punishes some people who never committed a violent offense.

The triggering offense, 18 U.S.C. § 922(g), simply requires possession, receipt, transport, or shipment — not use — of a firearm or ammunition,72×72. Only seventeen states criminalize the possession of ammunition by any nonintoxicated person. United States v. Young, 766 F.3d 621, 632 (6th Cir. 2014) (Stranch, J., concurring). and it may have been for any purpose, including self-defense or hunting.73×73. Some jurisdictions recognize an “innocent possessor” defense, see United States v. Mason, 233 F.3d 619, 623 (D.C. Cir. 2000) (collecting cases), which generally requires “(1) the firearm was attained innocently and held with no illicit purpose and (2) possession of the firearm was transitory — i.e., in light of the circumstances presented, there is a good basis to find that the defendant took adequate measures to rid himself of possession of the firearm as promptly as reasonably possible,” id. at 624; see also United States v. Wolak, 923 F.2d 1193, 1198 (6th Cir. 1991). The First Circuit “declined to create a ‘mandatory safe harbor’ for innocent possession, [but] it also acknowledged that ‘there are circumstances that arguably come within the letter of the law but in which conviction would be unjust,’ such as if a felon snatched away a loaded gun from his school-aged son and then called the police to retrieve it.” United States v. Baird, 712 F.3d 623, 629 (1st Cir. 2013) (quoting United States v. Teemer, 394 F.3d 59, 64–65 (1st Cir. 2005)). The Eighth Circuit recently rejected the possibility of an “innocent possessor” defense to 18 U.S.C. § 922(g). See United States v. Becerra, 958 F.3d 725, 730 (8th Cir. 2020). Additionally, the required predicates include controlled-substance offenses. Thus, an ACCA defendant’s predicates could consist entirely of drug offenses.

B. Failure to Target the Most Dangerous

As noted above, the ACCA originally included only robbery and burglary as predicates. One of the impetuses for adding drug predicates was the perceived relationship between drugs and violence.74×74. See, e.g., Hearing on ACCA Amendments, supra note 66, at 5 (statement of Rep. Ron Wyden) (“All the evidence . . . we have seen in our investigation indicates that drugs and violent crime go hand-in-hand.”). However, more recently, empirical analysis has weakened this perception. For example, a study found that only 1.06% of drug crimes committed in 2010 involved physical injury to a victim.75×75. Barkow, supra note 10, at 229 (citing Evan Tsen Lee et al., Which Felonies Pose a “Serious Potential Risk of Injury” for Federal Sentencing Purposes?, 26 Fed. Sent’g Rep. 118, 119 tbl.1 (2013)). Moreover, a Bureau of Justice Statistics study found that only 1.1% of people convicted of drug possession and 1.6% of people convicted of drug trafficking are rearrested for a violent felony, which is lower than the 1.9% average for all people with convictions.76×76. Id. at 230 (citing Shima Baradaran & Frank L. McIntyre, Predicting Violence, 90 Tex. L. Rev. 497, 561 tbl.3 (2012)).

Even the enumerated “violent felon[ies]” may not involve violence or even physical contact with anyone, as is usually the case for burglary.77×77. Id. at 231 (citing Richard F. Culp et al., Is Burglary a Crime of Violence? An Analysis of National Data 1998–2007, at 29–30 (2015)). Most burglars target empty locations to reduce the likelihood of discovery.78×78. Id. (citing Culp et al., supra note 78, at vii, 19). Between 1998 and 2007, burglary co-occurred with a violent crime only 7.6% of the time, and physical injury occurred in only 2.7% of all burglaries.79×79. Id. at 231 n.245 (citing Culp et al., supra note 78, at 29–30).

“Armed career criminals” may not be as dangerous as their name makes them sound. The most prevalent previous conviction for people sentenced under the ACCA in fiscal year 2019 was a public order offense, committed by 85.3% of them.80×80. Armed Career Criminals, supra note 7, at 31. As might be expected, the second-most-common prior offense for people sentenced under the ACCA that year was drug trafficking, a potential81×81. Only “serious” drug offenses are ACCA predicates, those imposing sentences of ten years or more. 18 U.S.C. § 924(e)(2)(A). ACCA predicate.82×82. In fiscal year 2019, 70.8% of people sentenced under the ACCA had committed the previous offense of drug trafficking. Armed Career Criminals, supra note 7, at 31.

Granted, 83.7% of people sentenced under the ACCA in fiscal year 2019 previously committed an offense characterized as “violent.”83×83. Id. at 32. “For purposes of this analysis, the [U.S. Sentencing] Commission identified offenses that are generally accepted as having some level of violence, including many of those offenses that courts have found to qualify as a ‘violent felony’ under section 924(e)(2)(B) [of the ACCA]. Prior violent convictions include those offenses classified as murder, unspecified manslaughter, non-negligent manslaughter, kidnapping, statutory rape, forcible sex offense, robbery, aggravated assault, simple assault, intimidating a witness, intimidation, hit and run with bodily injury, extortion, child abuse, burglary, arson, rioting, and any other unspecified violent offense.” Id. at 80 n.70. However, in this analysis, as in ACCA predicate determinations generally, burglary84×84. Burglary is an enumerated as a “violent felony” offense under the ACCA. 18 U.S.C. § 924(e)(2)(B)(ii). and robbery85×85. Robbery statutes satisfy the “physical force” requirement of the ACCA’s elements clause and are thus considered ACCA predicates if they “require the criminal to overcome the victim’s resistance.” Stokeling v. United States, 139 S. Ct. 544, 550 (2019). are considered “violent” offenses, committed by 42.6% and 42.0%, respectively, of people sentenced under the ACCA that year.86×86. Armed Career Criminals, supra note 7, at 32. Of the group, 59.3% were previously convicted of assault.87×87. Id. By contrast, only small percentages had committed prior homicides or rapes: 6.7% and 2.9%, respectively.88×88. Id. Additionally, 23.4% of people sentenced under the ACCA in fiscal year 2019 committed “other” “violent” offenses, including “kidnapping, intimidating a witness, extortion, child abuse, arson, rioting, and any other unspecified violent offense.” Id. at 32, 80 n.72.

C. Inconsistencies in Sentencing

Many scholars question whether such a lengthy period of incarceration is warranted for all people who qualify for ACCA sentencing.89×89. See, e.g., Barkow, supra note 10, at 227–28. As noted above, the ACCA imposes a minimum fifteen-year sentence. As a point of comparison, the median prison time served for murder in the United States is 17.5 years,90×90. Danielle Kaeble, Bureau of Just. Stat., Time Served in State Prison, 2018, at 2 tbl.1 (2018), https://www.bjs.gov/content/pub/pdf/tssp18.pdf [https://perma.cc/9ZCS-LTNF]). and the median amount of time served for any violent offense is 2.4 years.91×91. Id. Sixth Circuit Judge Stranch commented:

Perhaps one of the greatest harms is that indiscriminate criminalization erodes the faith of our citizens in the federal criminal justice system. That loss of faith in the system entrusted with societal justice reverberates through our communities, damaging our families, our schools, and our workplaces — components that are necessary to an effective social structure in our nation.92×92. United States v. Young, 766 F.3d 621, 634 (6th Cir. 2014) (Stranch, J., concurring).

Only drug offenses with maximum terms of imprisonment of ten years or more are considered “serious” and serve as ACCA predicates. In McNeill v. United States,93×93. 563 U.S. 816 (2011). the Supreme Court held that whether a prior conviction qualifies as a “serious drug offense” depends on “the maximum sentence applicable to a defendant’s previous drug offense at the time of his conviction for that offense.”94×94. Id. at 820 (emphasis added). This is problematic because recently many drug sentences have been reduced to less than ten years, but they continue to serve as ACCA predicates for those defen-dants who committed the offense before the reduction.

However, the McNeill Court noted that it was not addressing “a situation in which a State subsequently lowers the maximum penalty applicable to an offense and makes that reduction available to defendants previously convicted and sentenced for that offense.”95×95. Id. at 825 n.* (citing Brief for United States at 18 n.5, McNeill, 563 U.S. 816 (No. 10-5258); 18 U.S.C. § 3582(c)(2)). This creates strange anomalies. The Drug Law Reform Act of 200496×96. 2004 N.Y. Laws ch. 738, § 36 (codified at N.Y. Penal Law § 70.70 (McKinney 2021)). (“2004 Act”) decreased maximum penalties for first-time New York class B and C felony drug convictions to nine years and five-and-a-half years, respectively.97×97. Id. The Drug Law Reform Act of 200998×98. 2009 N.Y. Laws ch. 56, § 9 (codified at N.Y. Crim. Proc. Law § 440.46 (McKinney 2011)). (“2009 Act”) made these decreased penalties available to defendants convicted of a class B offense before the 2004 Act.99×99. Id. In 2014, the Southern District of New York held that New York class B drug felonies, which after the reduction carried a maximum penalty of only nine years, were not ACCA predicate offenses because the 2009 Act applied the 2004 Act’s reduction retroactively to offenses committed before the latter’s adoption.100×100. United States v. Calix, No. 13 CR 582, 2014 WL 2084098, at *15 (S.D.N.Y. May 13, 2014), aff’d, 787 F. App’x 4 (2d Cir. 2019).

Yet the Second Circuit held in Rivera v. United States101×101. 716 F.3d 685 (2d Cir. 2013). that though New York class C drug felonies now have a maximum five-and-a-half-year penalty, those committed prior to the 2004 Act remain ACCA predicates because the 2009 Act did not apply the 2004 Act’s reduction retroactively to defendants convicted of only class C offenses.102×102. Id. at 689–90. As a result, the less serious class C convictions (attempted drug sales) can be considered “serious” and thus count as ACCA predicates, while the more serious class B convictions (completed drug sales) are not considered “serious” and so fail to serve as ACCA predicates.

Imposition of mandatory minimums such as that provided by the ACCA is intended to decrease sentencing inconsistencies. However, in practice, this is not the case. Most cases are decided by plea agreements, especially when there is a harsh mandatory minimum sentence on the table.103×103. See Litman, supra note 9, at 56 (“[The] ACCA’s harsh mandatory minimum may lead many defendants to plead guilty to avoid more extensive prison time.”). Prosecutors still decide what offenses to charge. Moreover, prosecutors retain their discretion in deciding for which defendants they recommend a reduction for providing “substantial assistance,” which can lower a sentence below even a “mandatory” minimum.104×104. An offender who provides “substantial assistance” to the authorities may have an applicable mandatory minimum penalty waived and receive an additional offense level reduction under a guidelines departure provision. See 18 U.S.C. § 3553(e); U.S. Sent’g Guidelines Manual § 5K1.1 (U.S. Sent’g Comm’n 2016). The extent of a substantial assistance departure is determined by the sentencing court in light of factors enumerated in the guideline, and is not limited to a specific number of offense levels. See U.S. Sent’g Guidelines Manual § 5K1.1(a) (U.S. Sent’g Comm’n 2016) (listing five factors to be considered in determining the extent of a downward departure for substantial assistance). People are also sentenced under the ACCA unevenly among racial and geographic lines.

The ACCA disproportionately affects people of color. Only 24.0% of all people released from federal custody between 2009 and 2011 were Black, but more than half (53.2%) of the people sentenced under the ACCA and released during the same time period were Black.105×105. Armed Career Criminals, supra note 7, at 42 tbl.5. Moreover, in fiscal year 2016, Black people convicted under the ACCA received longer average sentences than any other racial group — averaging 185 months, compared to 178 months for White people, 173 months for Hispanic people, and 147 months for people of “Other Race.”106×106. U.S. Sent’g Comm’n, Mandatory Minimum Penalties for Firearms Offenses in the Federal Criminal Justice System 43 (2018) [hereinafter Mandatory Minimums for Firearms]. Here, “Other Race” refers to Native American, Alaskan Native, Asian, or Pacific Islander. Id. at 68 n.58.

As Justice Sotomayor remarked in her dissent in a recent ACCA case, even without the ACCA, felons-in-possession are generally already subject to increased sentencing under the U.S. Sentencing Guidelines,107×107. However, under the Sentencing Guidelines a conviction may be time-barred, but no time limit exists for ACCA predicates. See U.S. Sent’g Guidelines Manual § 4B1.4, cmt. n.1 (U.S. Sent’g Comm’n 2016) (noting that the time periods in section 4A1.2 are inapplicable to the determination of whether a defendant is subject to an enhanced sentence under § 924(e) (the ACCA)). Under the ACCA, predicates must be committed “on occasions different from one another,” United States v. Hockenberry, 730 F.3d 645, 667 (6th Cir. 2013) (quoting 18 U.S.C. § 924(e)(1)), while the question under the Sentencing Guidelines is whether the prior offenses constitute a single sentence, see id. at 668 (“[W]e have held that the ACCA does not apply the same standards as § 4A1.2(a)(2) of the Guidelines.” (citing United States v. Birdsong, 330 F. App’x 573, 585–86 (6th Cir. 2009))); United States v. Ross, 569 F.3d 821, 823 n.3 (8th Cir. 2009) (noting that the analysis of separateness of prior convictions is different from the analysis of different occasions under the ACCA). because of their past convictions.108×108. Stokeling v. United States, 139 S. Ct. 544, 559 (2019) (Sotomayor, J., dissenting). Some ACCA offenders would be subject to the sentencing enhancement for career offenders at section 4B1.1. A person is subject to the career offender enhancement under the Guidelines when his instant offense of conviction is a felony crime of violence or a controlled-substance offense, he was at least eighteen years old at the time of the instant offense of conviction, and he has at least two prior felony convictions of either a crime of violence or a controlled-substance offense. U.S. Sent’g Guidelines Manual § 4B1.1 (U.S. Sent’g Comm’n 2016). In fiscal year 2019, 15.1% of people sentenced under the ACCA also were determined to be career offenders under the guidelines. Armed Career Criminals, supra note 7, at 25. Moreover, some people sentenced under the ACCA are also subject to an additional mandatory minimum penalty ranging from five years to life for being convicted under 18 U.S.C. § 924(c) for the use of a firearm during or in relation to a crime of violence or drug-trafficking crime. 18 U.S.C. § 924(c)(1)(A)–(C). Thus, as Professor Barkow points out, “[w]hether the ACCA applies is thus not a question of whether that prior conduct will result in greater punishment, but is instead a question of how much greater that punishment will be.”109×109. Barkow, supra note 10, at 235. The ACCA in a way double-counts prior drug offenses because the Supreme Court held that the maximum term of imprisonment, for the purpose of determining whether a prior offense qualified as a “serious drug offense” (prescribing a maximum sentence of ten years or more), must be determined with reference to applicable recidivist enhancements.110×110. United States v. Rodriquez, 553 U.S. 377, 393 (2008).

D. Costs to the Court System and Society

The ACCA imposes significant administrative costs on the judiciary. The Supreme Court mandated that courts use the “categorical approach” to determine if a prior offense serves as an ACCA predicate.111×111. Taylor v. United States, 495 U.S. 575, 602 (1990). Applying the categorical approach to recidivism statutes generally entails considering only “the statutory definitions of . . . prior offenses,” rather than “the particular facts underlying those convictions.”112×112. Id. at 600 (citing United States v. Chatman, 869 F.2d 525, 529 (9th Cir. 1989); United States v. Headspeth, 852 F.2d 753, 758–59 (4th Cir. 1988); United States v. Vidaure, 861 F.2d 1337, 1340 (5th Cir. 1988), cert. denied, 489 U.S. 1088 (1989); United States v. Sherbondy, 865 F.2d 996, 1006–10 (9th Cir. 1988)); see also Jennifer Lee Barrow, The Return of the Jury: Conduct-Based Sentencing for Recidivism, 2022 Wis. L. Rev. 785, 788 (discussing the Supreme Court’s articulated reasoning for requiring the categorical approach). It requires a judge to presume that a conviction rested upon the “minimum conduct”113×113. Moncrieffe v. Holder, 569 U.S. 184, 191 (2013). criminalized by the statute.114×114. See Barrow, supra note 113, at 805–06. “Violent felon[ies]” are ACCA predicates.115×115. 18 U.S.C. § 924(e)(1). However, if a “realistic probability”116×116. Gonzales v. Duenas-Alvarez, 549 U.S. 183, 193 (2007). exists that any person could violate the statute in a nonviolent way, the conviction under the statute fails to count as a “violent” offense, even if the conduct of the person who violated it was indisputably violent.117×117. See Barrow, supra note 113, at 806.

Determining whether federal and state offenses serve as ACCA predicates under the categorical approach is becoming increasingly complex, and such decisions are extensively litigated.118×118. See, e.g., United States v. Walton, 881 F.3d 768, 775 (9th Cir. 2018) (holding that a defendant’s conviction for armed robbery under Alabama law is not a violent felony); United States v. Jordan, 812 F.3d 1183, 1185 (8th Cir. 2016) (holding that aggravated assault under Arkansas law is not a violent felony). The Seventh Circuit explained that “[i]n recent years, federal courts have seen a floodtide of litigation over what qualifies as an ACCA predicate.”119×119. Dotson v. United States, 949 F.3d 317, 318 (7th Cir. 2020). Justice Alito lamented that “the ‘categorical approach’ to predicate offenses has created numerous splits among the lower federal courts, the resolution of which could occupy [the Supreme] Court for years.”120×120. Chambers v. United States, 155 U.S. 122, 133 (2009) (Alito, J., concurring in the judgment) (footnote omitted).

Defendants litigate ACCA applicability under the categorical approach, not only for “violent felon[ies]” but also for “serious drug offense[s].”121×121. See, e.g., Shular v. United States, 140 S. Ct. 779, 783 (2020) (quoting 18 U.S.C. § 924(e)(1)). For example, in United States v. Cantu,122×122. 964 F.3d 924 (10th Cir. 2020). a person’s ACCA conviction was overturned because one of his predicate drug convictions was for violating an Oklahoma statute that, in addition to prohibiting various substances controlled under federal law, also criminalized three substances not controlled under federal law.123×123. Id. at 928, 934. Thus, the Oklahoma conviction failed to count as a “serious drug offense,”124×124. 18 U.S.C. § 924(e)(1). even though the convicted person’s conduct involved methamphetamines, a controlled substance under federal law.125×125. Cantu, 964 F.3d at 928. The categorical approach creates a lack of uniformity because another person who engaged in conduct involving methamphetamines identical to the conduct of the convicted person in Cantu, in a state where the statute criminalized only controlled substances under federal law, would be subject to the ACCA’s fifteen-year minimum sentence.126×126. Cf. Barrow, supra note 113, at 790, 806–07 (explaining the uniformity problem caused by the categorical approach’s application to the ACCA-enumerated offense of burglary).

Relatedly, the ACCA also imposes notable costs on the criminal justice system overall. Judge Stranch opined: “Society pays a great price when Congress over-criminalizes conduct. . . .  Prisons operate substantially over capacity, and the Department of Justice is forced to cut funding from other priorities to pay for prison growth.”127×127. United States v. Young, 766 F.3d 621, 633–34 (6th Cir. 2014) (Stranch, J., concurring) (citing Rhonda McMillion, Bipartisan Push Is On for Sentencing Reform, ABA J. (Feb. 1, 2014, 7:40 AM), https://www.abajournal.com/magazine/article/bipartisan_push_is_on_for_sentencing_reform [https://perma.cc/D73R-A6NF]). Recently, in an executive order, President Biden noted that “[m]ore than two million people are currently incarcerated in the United States, including a disproportionate number of people of color” and that “[t]here is broad consensus that our current system of mass incarceration imposes significant costs and hardships on our society and communities and does not make us safer.”128×128. Exec. Order No. 14,006, 86 Fed. Reg. 7483, 7483 (Jan. 26, 2021).

While ACCA cases make up only a small portion of the annual docket,129×129. Armed Career Criminals, supra note 7, at 29. people sentenced under the ACCA tend to be disproportionately represented in the federal prison population given the lengthy sentences they receive.130×130. Id. Granted, the number of people sentenced under the ACCA decreased by half between fiscal years 2010 and 2019.131×131. Id. at 59. While 590 people were sentenced under the ACCA in fiscal year 2010, only 312 people were sentenced under the ACCA in fiscal year 2019.132×132. Id.

However, as a result of the fifteen-year minimum sentence, as of June 27, 2020, there were 3572 people sentenced under the ACCA in federal Bureau of Prisons (BOP) custody, accounting for 2.5% of the federal prison population serving a sentence for a federal conviction.133×133. Id. at 29. That is, 2.5% of the offenders to which the U.S. Sentencing Commission could match Commission records. Id. at 80 n.66. There were 160,823 prisoners in federal custody, as of June 27, 2020. Id. Of those, 144,122 were sentenced for federal offenses, while 16,702 were sentenced in the courts of the District of Columbia, were military offenders, or were pretrial offenders. Id. The Sentencing Commission was able to match records to 142,393 of the 144,122 offenders serving sentences for federal convictions. Id.

III. This Essay’s Study

Part III begins with a description of this Essay’s methodology, followed by the demographics of this Essay’s study sample and then the demographics of people more recently sentenced under the ACCA. The results of this Essay’s study follow, including a discussion of recidivism overall and recidivism of people who committed violent and drug offenses.

A. Methodology

This Essay contributes to ACCA recidivism analysis by studying all people sentenced under the ACCA who were released in 2009, 2010, and 2011 during an eight-year follow-up period.134×134. That is, all ACCA offenders whose ACCA sentences were not vacated; who were U.S. citizens; who possessed valid FBI numbers; who were “[n]ot reported dead, escaped, or detained”; and whose presentence investigation reports were submitted to the U.S. Sentencing Commission. Armed Career Criminals, supra note 7, at 41. The purpose of this Essay’s empirical analysis is to determine if people sentenced under the ACCA with fewer than three “violent felony” predicates recidivated less than people with three or more prior “violent felon[ies],”135×135. 18 U.S.C. § 924(e)(1). to inform whether Congress should modify the ACCA accordingly.

Recidivism “refers to a person’s relapse into criminal behavior, often after the person receives sanctions or undergoes intervention for a previous crime.”136×136. Recidivism, supra note 2. Recidivism is typically “measured by criminal acts that resulted in [the] rearrest, reconviction,” or reincarceration of a person during a specified period.137×137. Id. This Essay uses rearrest as the measure, defined as “any rearrest for a new crime or for an alleged violation of the conditions of [the offender’s] supervision over a specified follow-up period”138×138. Armed Career Criminals, supra note 7, at 41. — eight years for this study. This study counted all arrests, including felonies, misdemeanors, and supervision violations but excluded minor traffic offenses.139×139. Id. at 41.

Any method of calculating recidivism has its disadvantages. All methods may be underinclusive because they fail to account for undetected crime. Rearrest may be overinclusive because not every person arrested has actually committed a crime.140×140. See Jeffrey Fagan & Martin Guggenheim, Preventive Detention and the Judicial Prediction of Dangerousness for Juveniles: A Natural Experiment, 86 J. Crim. L. & Criminology 415, 427 (1996). The measure of rearrest may be particularly overinclusive in communities of color that experience more police presence.141×141. See Prescott et al., supra note 33, at 1666; see also Robin Smyton, How Racial Segregation and Policing Intersect in America, Tufts Now (June 17, 2020), https://now.tufts.edu/articles/how-racial-segregation-and-policing-intersect-america [https://perma.cc/86XA-GY6M]. Additionally, reconviction and reincarceration data may be underinclusive because of shortcomings in reporting.142×142. See Armed Career Criminals, supra note 7, at 81 n.87.

Most recidivism studies depend on the accuracy of publicly available information,143×143. See, e.g., Prescott et al., supra note 33, at 1683. but the author entered into an agreement with the FBI and the U.S. Sentencing Commission to obtain the nonpublic criminal-history information and recidivism data used in this study.144×144. See Armed Career Criminals, supra note 7, at 81 n.83.

Additionally, many recidivism studies involve follow-up periods of no more than a few years,145×145. See Kurlychek et al., supra note 33, at 79. but this study tracks people for eight years after release.146×146. Armed Career Criminals, supra note 7, at 41. Granted, half of the people in this study who were rearrested faced arrest within the first year and a half after release.147×147. Id. at 43. But some recidivism is missed by short follow-up periods.148×148. A study of more than 88,000 people who were first arrested in New York in 1980 and tracked for the next twenty-five years showed that the time it takes for a person’s chance of being rearrested to approximate that of a person with no criminal record is about ten to thirteen years. Alfred Blumstein & Kiminori Nakamura, Opinion, Paying a Price, Long After the Crime, N.Y. Times (Jan. 9, 2012), https://www.nytimes.com/2012/01/10/opinion/paying-a-price-long-after-the-crime.html [https://perma.cc/E355-7CDU]. Approximately one-third of the recidivism discovered by this Essay’s empirical analysis would have been missed if the follow-up period had been only three years after release, as opposed to eight years.149×149. Approximately 40% of the study group had been rearrested at three years after release, while 59% of the study group had been rearrested at eight years after release. Armed Career Criminals, supra note 7, at 43 fig.12.

This Essay compares the recidivism rates of three categories of people sentenced under the ACCA: (1) 616 people (83.4% of the total) with at least three prior “violent felonies” (the “violent group”); (2) 33 people (4.5% of the total) with no prior “violent felonies” (only prior “serious drug offenses”) (the “drug-only group”); and (3) 90 people (12.2% of the total) with one or two prior “violent felonies” (and one or more “serious drug offenses”) (the “mixed group”).150×150. 18 U.S.C. § 924(e)(1); Armed Career Criminals, supra note 7, at 47, 53 fig.17.

The author chose these categories to demonstrate the effects of various proposals for modifications to the ACCA. If “serious drug offense[s]”151×151. 18 U.S.C. § 924(e)(1). were eliminated as ACCA predicates, then only offenders in the “violent group” would remain subject to the ACCA. If the ACCA required at least one “violent felony,”152×152. Id. then only people in the “violent group” or the “mixed group” would remain subject to the ACCA.

B. Demographics of This Essay’s Study Sample

During 2009, 2010, and 2011, 884 people initially sentenced under the ACCA were released from federal custody.153×153. Armed Career Criminals, supra note 7, at 41. Of that total, 849 were U.S. citizens,154×154. The seventeen noncitizens were removed from the data set because reliable recidivism information likely could not be obtained for them. It is unlikely the removal of the noncitizens affected the study’s results. possessed valid FBI numbers,155×155. For this study, recidivism data could be obtained only for offenders with valid FBI numbers. had their presentence investigation report submitted to the U.S. Sentencing Commission,156×156. The presentence investigation report was used to determine the criminal history of the offender, in order to determine which group the offender was a part of (“violent,” “drug-only,” “mixed”). and were not reported dead, escaped, or detained.157×157. See Armed Career Criminals, supra note 7, at 41, 81 n.84. Of those 849, 110 had their ACCA sentences vacated.158×158. In the study, 136 offenders had been resentenced after their initial sentencing, id. at 81 n.85, though this was because of a variety of reasons, some having no impact on the offender’s status as an armed career criminal, see id. To determine whether resentencing affected an offender’s armed-career-criminal status, the Commission matched the records of the ACCA offenders to records in its resentencing datafiles, focusing on whether intervening ACCA-related case law likely impacted the offender’s armed career criminal status at resentencing. Id. Where the examination indicated that the proceedings resulted in a legal redetermination that an offender was no longer subject to the ACCA, the offender was removed from the study in order to focus it on a more precisely and accurately designated group of ACCA offenders as defined by current law. Id. Those offenders who were removed from the study constitute the 110 offenders whose ACCA sentences were vacated. Id. This left 739 people for this study group.159×159. Id. at 41.

Almost all of the people in this study group were male (99.5%).160×160. Id. at 42 tbl.5. The largest racial/ethnic group represented was Black (53.2%), followed by White (40.9%).161×161. Id. In addition, 4.7% of the study group was Hispanic, and 1.2% were another race.162×162. Id. At sentencing, the group members had a median age of 36 years.163×163. Id. at 42. Their median age at release was 47.164×164. Id.

The “drug-only group” contained the greatest proportion of people of color (88.3%). Black people comprised 76.5% of the “drug-only group,” and 11.8% of the “drug-only group” was Hispanic. The “mixed group” was 56.9% Black, 11.8% Hispanic, and 2.0% another race, for a total of 70.7% of people of color. The “violent group” contained the smallest proportion of people of color (55.6%, consisting of 51.2% Black, 3.2% Hispanic, and 1.2% another race).

C. Demographics of More Recent ACCA Offenders

The demographics of this Essay’s study sample are fairly consistent with more recent demographics of people sentenced under the ACCA, except that the percentage of people of color sentenced under the ACCA appears to be increasing. In fiscal year 2010, 70.6% of people sentenced under the ACCA were people of color165×165. See id. at 79 n.52. : 63.8% Black, 5.3% Hispanic, and 1.5% another race.166×166. See id.

By fiscal year 2016, more than three-quarters (76.3%) of people sentenced under the ACCA were people of color167×167. See Mandatory Minimums for Firearms, supra note 107, at 38. : 70.4% were Black,168×168. Id. 4.3% were Hispanic,169×169. Id. and 1.6% were another race.170×170. Id. That year, the proportion of Black people convicted under the ACCA (70.4%) was much larger than the percentage of Black people convicted of an offense carrying any mandatory minimum penalty (29.7%).171×171. Id.

In fiscal year 2019, the percentage of people of color sentenced under the ACCA climbed to 84.3%.172×172. See Armed Career Criminals, supra note 7, at 22, 79 n.52. Nearly three-quarters (73.7%) of people sentenced under the ACCA were Black.173×173. Id. The percentage of Hispanic people sentenced under the ACCA approximately doubled (from fiscal years 2010 and 2016) to 9.6%.174×174. Id. And 1.0% were another race.175×175. Id.

Again, that year, the percentage of Black people sentenced under the ACCA (73.7%) was much greater than the proportion of Black people convicted of mandatory minimum–carrying offenses overall (27.6%).176×176. See U.S. Sent’g Comm’n, Quick Facts: Mandatory Minimum Penalties (FY 2019), https://www.ussc.gov/sites/default/files/pdf/research-and-publications/quick-facts/Quick_Facts_Mand_Mins_FY19.pdf [https://perma.cc/55NT-XKXE]. Surprisingly, the percentage of Black people sentenced under ACCA has increased over the past ten years, while the racial composition of the population convicted of federal crimes overall has remained consistent, with Black people comprising approximately one-fifth of that population.177×177. See Armed Career Criminals, supra note 7, at 79 n.52 (“[I]n [fiscal year] 2019, 20.2% of federal offenders were Black, 19.9% were White, 56.3% were Hispanic, and 3.6% were Other races.”). Here, “Other race” refers to Native American, Alaskan Native, Asian, or Pacific Islander. Id. at 79 n.53.

In fiscal year 2016, people sentenced under the ACCA were evenly distributed throughout the age brackets between 26 and 50.178×178. Mandatory Minimums for Firearms, supra note 107, at 39. Fewer people were older than 50 years of age or younger than 26 years of age.179×179. Id. The largest group belonged to age bracket 41 to 50.180×180. Id. In fiscal year 2019, people sentenced under the ACCA ranged from 21 to 74 years of age, with a median age of 38 years.181×181. Armed Career Criminals, supra note 7, at 22 tbl.2.

In fiscal year 2019, 99.4% of people sentenced under the ACCA were United States citizens.182×182. Id. In fiscal year 2016, all but one of the 304 people convicted under the ACCA were male.183×183. Mandatory Minimums for Firearms, supra note 107, at 36, 38. Likewise, in fiscal year 2019, 98.7% of people convicted under the ACCA were male.184×184. Armed Career Criminals, supra note 7, at 22 tbl.2.

D. ACCA Recidivism Overall in This Study

Through the eight years following release, 436 of the 739 people in the sample set were rearrested, for a recidivism rate of 59.0%.185×185. Id. at 43. This recidivism rate is lower than the 67.5% recidivism rate the U.S. Sentencing Commission found over an eight-year follow-up period for eighty U.S. citizen ACCA offenders released in 2005, Firearms Recidivism Report for 2005 Releasees, supra note 32, at 2, 34, and the 71.6% recidivism rate the Commission found over the same follow-up period for 311 citizen ACCA offenders released in 2010, U.S. Sent’g Comm’n, Recidivism of Federal Firearms Offenders Released in 2010, at 2, 45 (2021) [hereinafter Firearms Recidivism Report for 2010 Releasees], https://www.ussc.gov/sites/default/files/pdf/research-and-publications/research-publications/2021/20220209_Recidivism-Firearms.pdf [https://perma.cc/9Q48-6QT3]. This is likely because neither of the previous studies removed offenders whose ACCA sentences were vacated. Also, they were both studies of small numbers of ACCA offenders. For those who reoffended, the median number of rearrests during the eight-year follow-up period was three.186×186. Armed Career Criminals, supra note 7, at 43. This recidivism rate is lower than the 68.1% recidivism rate that a previous U.S. Sentencing Commission study found for the 3,446 individuals sentenced for federal firearms offenses released in 2005187×187. See id.; Firearms Recidivism Report for 2005 Releasees, supra note 32, at 4, 34. and the 69.0% recidivism rate the Commission found for the 5659 federal firearms offenders released in 2010,188×188. Firearms Recidivism Report for 2010 Releasees, supra note 186, at 5–6. during the same eight-year follow-up period. It is higher than the 46.3% recidivism rate for the 21,554 people sentenced for federal crimes other than firearms offenses and released in 2005189×189. Firearms Recidivism Report for 2005 Releasees, supra note 32, at 3, 14. Both the firearms offenders and non-firearms offenders were originally sentenced between fiscal year 1990 and the first quarter of fiscal year 2006. Id. at 3. and the 45.1% recidivism rate for the 26,476 people sentenced for crimes other than firearms offenses and released in 2010,190×190. Firearms Recidivism Report for 2010 Releasees, supra note 186, at 5–6. through the same follow-up period. However, it is lower than the 76.6% recidivism rate of people released from state prison in 2005, during just a five-year follow-up period.191×191. Matthew R. Durose et al., Bureau of Just. Stat., Recidivism of Prisoners Released in 30 States in 2005: Patterns from 2005 to 2010, at 7 (2014), https://www.bjs.gov/content/pub/pdf/rprts05p0510.pdf [https://perma.cc/RJ3Q-BPZW]. The U.S. Sentencing Commission found a 44.9% recidivism rate for federal offenders released in 2005, using a comparable five-year follow-up period and including only federal offenders released from prison (as opposed to those sentenced to only probation or a fine). U.S. Sent’g Comm’n, Recidivism Among Federal Offenders: A Comprehensive Overview 15 (2016) [hereinafter Recidivism Comprehensive Overview]. Some key differences between state and federal offenders may explain the different recidivism rates. Only 6.8% of the inmates released from federal prison had a violent commitment offense, while a quarter (25.7%) of the state inmates did. Id. at 32 n.40. The federal offenders were 60.1% under 40 and 85.9% male, but the state offenders were 68.5% under 40 and 89.3% male. Id. The state offender study also included noncitizens, while the federal offenders studied were all U.S. citizens. Id. It is important to remember that ACCA predicates are usually state law offenses.

Interestingly, this 59.0% ACCA recidivism rate is lower than the 74.2% recidivism rate the Commission found for 2231 people convicted of “prohibited class”192×192. That is, “prohibited class” of people, such as felons, as opposed to “prohibited [type of] weapon.” See Firearms Recidivism Report for 2005 Releasees, supra note 32, at 24. “Prohibited class” offenders were sentenced under U.S. Sentencing Guidelines section 2K2.1. Id. These offenders were those prohibited from possessing a firearm, “straw purchasers,” or offenders who transferred a firearm or ammunition to a prohibited person. Id. Under section 2K2.1, “[t]he Base Offense Level for Prohibited Class offenders is set at 14, 20, or 24 depending on the number of felony convictions, if any, in the offender’s past for a ‘crime of violence’ or ‘controlled substance offense.’” Id. at 9 (footnotes omitted); U.S. Sent’g Guidelines Manual § 2K2.1 (U.S. Sent’g Comm’n 2016). firearms offenses released in 2005193×193. See Firearms Recidivism Report for 2005 Releasees, supra note 32, at 24, 27. and the 74.5% recidivism rate the Commission found for the 3655 “prohibited class” firearms offenders released in 2010,194×194. Firearms Recidivism Report for 2010 Releasees, supra note 186, at 5–6. during an eight-year follow-up period. Like the instant offense under the ACCA, most of those “prohibited class” firearms offenses are violations of 18 U.S.C. § 922(g).195×195. See id. at 8. In the absence of the ACCA penalty, 18 U.S.C. § 922(g) imposes no minimum sentence and prescribes a maximum sentence of ten years.196×196. 18 U.S.C. § 924(a)(2). It makes little sense to punish people convicted under the ACCA, a subgroup of the 18 U.S.C. § 922(g) offenders, more harshly when their recidivism rate is less than the rate for others who violate 18 U.S.C. § 922(g).

For the people in this study who recidivated, the median time to rearrest was 16 months, or in other words, half of the people who were rearrested faced arrest less than one and a half years after their initial release from prison or placement on probation, after serving their ACCA sentence.197×197. Armed Career Criminals, supra note 7, at 43, 44. During their first year after release, 24.2% of people recidivated for the first time.198×198. Id. at 44. Each year following release, a smaller proportion of people were rearrested. For instance, 13.1% of people reoffended during their second year after release, and 8.0% recidivated during their third year after release.199×199. Id. Only 1.1% of people were re-arrested during their eighth year after release.200×200. Id.

These findings are similar to the 12-month median time to rearrest the U.S. Sentencing Commission found for eighty U.S. citizens sentenced under the ACCA who were released in 2005, during an eight-year follow-up period, and are also similar to the 17-month median time to rearrest for all people convicted for firearms offenses released in 2005, through the same follow-up period.201×201. Firearms Recidivism Report for 2005 Releasees, supra note 32, at 34. In contrast, the Sentencing Commission found a 21-month median time to rearrest for all people convicted of federal crimes released in 2005 during the same period.202×202. Recidivism Comprehensive Overview, supra note 192, at 15.

The most common recidivism offense was assault, committed by 28.2% of the people in this study.203×203. Armed Career Criminals, supra note 7, at 45 fig.13. The second-most-common new offense category for people sentenced under the ACCA was public order offenses,204×204. For purposes of this study, “public order” offenses include offenses that “involve the public or occur in public areas, vice crimes, tampering with evidence, and obstruction of justice.” Id. at 80 n.68. This category includes the specific offenses of “disorderly conduct, public drunkenness, gambling, gang participation, harassment, possessing contraband in prison, resisting arrest without violence, and perjury.” Id. followed by drug trafficking, committed by 13.8% and 8.5% of people, respectively.205×205. Id. at 45 fig.13. Only 3.7% and 4.6% of the people in this study committed homicide or rape, respectively.206×206. Id. This is relevant, considering the purpose of the ACCA. Justice Sotomayor has explained: “The ACCA . . . does not look to past crimes simply to get a sense of whether a particular defendant is generally a recidivist; rather, it looks to past crimes to determine specifically ‘the kind or degree of danger the offender would pose were he to possess a gun.’”207×207. Stokeling v. United States, 139 S. Ct. 544, 559 (2019) (Sotomayor, J., dissenting) (quoting Begay v. United States, 553 U.S. 137, 146 (2008)).

E. Recidivism Differences Between the Violent and Drug Groups

Rates of recidivism varied depending on the nature of the person’s predicate offenses. People with no “violent felon[ies]” as predicates (33 “drug-only offenders”) recidivated at a rate of 36.4%.208×208. Armed Career Criminals, supra note 7, at 57. All of the recidivism data came from the U.S. Sentencing Commission’s recidivism datafiles. The datafiles were used to examine the types and extent of rearrests of ACCA offenders during the eight-year follow-up period. People with one or two “violent felon[ies]” (and one or more “serious drug offense[s]”) (“mixed group” with 90 people) recidivated at a rate of 48.9%.209×209. See id. at 42, 53–54, 57. People with three or more “violent felony” predicates (616 “violent group” members) recidivated at a rate of 61.7%.210×210. Id. The author chose the primary categories (using number of “violent felonies”) to demonstrate the effects of various proposals for modifications to the ACCA. A previous Sentencing Commission study demonstrated that violent offenders tended to recidivate more than drug offenders. U.S. Sent’g Comm’n, Recidivism Among Federal Violent Offenders 3 (2019) [hereinafter Recidivism Among Violent Offenders], https://www.ussc.gov/sites/default/files/pdf/research-and-publications/research-publications/2019/20190124_Recidivism_Violence.pdf [https://perma.cc/83J8-R94A].

The author chose those categories to show what would happen if drug offenses were eliminated as predicates; only the “violent group” would remain subject to the ACCA. If at least one “violent felony” was required for ACCA applicability, then only the “violent group” and the “mixed group” would be subject to the ACCA.

No difference in demographics, such as age or race, appears to explain the lower recidivism rate for those with fewer than three prior violent offenses. A Sentencing Commission study of all people convicted of federal crimes released in 2005 found that of all characteristics, age is the most significant predictor of recidivism.211×211. See U.S. Sent’g Comm’n, The Effects of Aging on Recidivism Among Violent Offenders 23–24 (2017); U.S. Sent’g Comm’n, Report to the Congress: Career Offender Sentencing Enhancements 41 (2016) [hereinafter Career Offender Report to the Congress], https://www.ussc.gov/sites/default/files/pdf/news/congressional-testimony-and-reports/criminal-history/201607_RtC-Career-Offenders.pdf [https://perma.cc/ZP2D-UPXW]. That study and many others212×212. See, e.g., Prescott et al., supra note 33, at 1675, 1692. have found that younger people recidivate at much higher rates than older people.

The mean age at release for the “drug-only group” in this study is 40 years, while the mean age at release for the “mixed group” is 45.7 years, and the mean age at release for the “violent group” is 48.2 years. It would have been expected that the relative youth of the “drug-only group” and the “mixed group” would correspond with a higher recidivism rate, but in fact, the recidivism rates of both groups were lower than that of the older “violent group.” There may not be a large difference in the mean ages of the three groups, but the lower recidivism rate of the slightly younger groups at least suggests that age fails to explain the lower recidivism rate of the people with fewer than three violent offenses.

The same previous Sentencing Commission study of all people convicted of federal offenses released in 2005, examined the recidivism for different racial groups. Black people were rearrested at the highest rate (59.1%), followed by people of “Other Race” (49.4%), Hispanic people (49.1%), and White people (41.7%).213×213. Recidivism Comprehensive Overview, supra note 192, at 24. However, it appears that much of the difference in recidivism rates was attributable to the higher average “Criminal History Category” of the Black people in the study, because the same study found criminal history to be closely correlated with recidivism rates.214×214. Id. An individual’s “Criminal History Category,” which can range from one to six, is determined by “the recency and severity of an offender’s prior sentences and supervision status.” Id. at app. B n.2 (citing U.S. Sent’g Guidelines Manual § 4A1.1 (U.S. Sent’g Comm’n 2016)). Additionally, rearrest rates may be higher among Black people if they live in communities that generally experience more police presence.215×215. See Prescott et al., supra note 33, at 1666; see also Smyton, supra note 142.

Even considering these previous findings, racial differences fail to explain the lower recidivism rates of the “drug-only group” and the “mixed group.” The “drug-only group,” which contained the highest percentage of people of color, had the lowest recidivism rate.216×216. Black offenders comprised 76.5% of the “drug-only offenders” and 11.8% of offenders in the group were Hispanic, for a 88.3% total percentage of people of color. The “mixed group” was comprised of 56.9% Black offenders, 11.8% Hispanic offenders, and 2% offenders of another race for a 70.7% total percentage of people of color. The “violent offenders” were 55.6% people of color (51.2% Black, 3.2% Hispanic, and 1.2% of another race). The “violent group,” which contained the lowest percentage of people of color, had the highest recidivism rate. The failure of demographics to explain the lower recidivism rates of the “drug-only group” and the “mixed group” suggests the absence of violence in their criminal history is what explains the lower recidivism.

A recent Sentencing Commission study evaluated the relationship between incarceration length and recidivism, using three models.217×217. U.S. Sent’g Comm’n, Length of Incarceration and Recidivism (2020), https://www.ussc.gov/sites/default/files/pdf/research-and-publications/research-publications/2020/20200429_Recidivism-SentLength.pdf [https://perma.cc/66HZ-6PYH]. All three of the models found statistically significant lower recidivism rates for people convicted of federal crimes and incarcerated for more than 120 months, as compared with people with similar characteristics who served shorter sentences.218×218. Id. at 4. Two of the three models found that people incarcerated for more than 120 months recidivated 30% less, and another model found those people recidivated 45% less.219×219. Id. at 30. The third model had a smaller sample size. Id. Two models found that people incarcerated between 60 and 120 months recidivated 17% less than similarly situated people who served less than 60 months.220×220. Id.

People sentenced under the ACCA are subject to a fifteen-year mandatory minimum, but may be sentenced to less than fifteen years (180 months) if they provide “substantial assistance” to authorities. It is difficult to assess how that recent Sentencing Commission study affects the recommendations of this Essay because the study declined to differentiate between people who served more than 120 months — the maximum sentence for the ACCA’s instant offense, § 922(g) — when the person is not subject to the ACCA. Even if a general principle can be inferred from the Commission’s study — that longer incarceration periods generally lead to lower recidivism rates — this principle fails to explain the difference between the recidivism rates of the different groups. Courts imposed the longest sentences (180.5 months on average) on the “violent group,” which recidivated the most. Courts imposed an average of 157.3 months on the “mixed group,” which recidivated at a rate in between the other two groups. Courts imposed only an average of 152.6 months on the “drug-only group,” which recidivated the least.

The “drug-only group” and the “mixed group” likely received lower sentences than the “violent group” because higher percentages of them received reductions for providing “substantial assistance.” While 36.4% of the “drug-only group” and 36.0% of the “mixed group” received “substantial assistance” relief, only 22.0% of the “violent group” obtained such relief.

The possibility exists that the “drug-only group” and the “mixed group” only recidivated less because they served lengthy incarceration terms. Though the “violent group” served the longest average sentences, perhaps people who commit drug offenses are benefitted more by incarceration than people who commit violent offenses. However, some other studies confirm that people who commit drug offenses recidivate less than people who commit violent offenses.221×221. See, e.g., Career Offender Report to the Congress, supra note 212, at 40–41.

Using data from its 2016 recidivism study of more than 25,000 people released from federal custody in 2005, in its 2016 Report to the Congress: Career Offender Sentencing Enhancements, the Sentencing Commission noted that people sentenced under the career offender guideline who committed only drug-trafficking offenses recidivated at lower rates than people sentenced under the guideline who committed at least one violent offense.222×222. Id. For purposes of that analysis, the U.S. Sentencing Commission identified offenses generally accepted as involving some level of violence, including many offenses that courts have found to qualify as “crimes of violence” under the career offender guideline. See id. at 52–55. The career offender guideline applies to an offender who commits a felony crime of violence or controlled-substance offense when he has at least two prior felony convictions for either a crime of violence or a controlled-substance offense. Id. at 14–15. The report also found that career offenders who had committed at least one violent offense have more serious and extensive criminal histories and were more likely to commit another violent offense in the future, compared to career offenders who had only drug-trafficking prior offenses. Id. at 30, 40–41.

However, differences exist between the ACCA and the career offender provision of the Sentencing Guidelines. Predicate career offender “controlled substance” offenses need only be punishable by a prison term of more than one year, while ACCA predicate “serious drug offense[s]” must be punishable by imprisonment of ten years or more. Additionally, under the ACCA, the instant offense involves firearms, while a person sentenced under the career offender provision may possess no prior weapons offenses.

Previous studies vary in concluding whether people who have committed violent offenses generally reoffend more than people who have committed nonviolent offenses. A Sentencing Commission study of people released in 2005 found that those who committed violent federal offenses recidivated much more (63.8%) than those who committed nonviolent federal offenses (39.8%).223×223. Recidivism Among Violent Offenders, supra note 211, at 3. However, three Bureau of Justice Statistics studies224×224. One of the studies concerned offenders released during the same year (2005) as the 2016 Sentencing Commission study. See Prescott et al., supra note 33, at 1669 tbl.4 & n.108 (citing Durose et al., supra note 192). found that people who committed state violent offenses recidivated less than people who committed state nonviolent offenses.225×225. Id. at 1669 tbl.4, 1671–72 (summarizing Bureau of Justice Statistics findings: in a study of state offenders released in 1983, recidivism rates of 59.6% for violent offenders and 62.5% for all offenders; in a study of state offenders released in 1994, recidivism rates of 61.7% for violent offenders and 67.5% for all offenders; and in a study of state offenders released in 2005, 71.3% for violent offenders and 76.6% for all offenders).

It was unclear whether people sentenced under the ACCA behave more like people who committed federal crimes or people who committed state crimes because the ACCA is a strange hybrid. The instant offense, 18 U.S.C. § 922(g), is a federal offense, but ACCA predicates are usually state law offenses.

F. Discussion of Results

No perfect method of predicting future recidivism exists. However, studying past recidivism may be the most effective predictor of future recidivism.226×226. See Ryan King & Brian Elderboom, Improving Recidivism as a Performance Measure 2, Urb. Inst. (2014), https://www.urban.org/sites/default/files/publication/23031/413247-Improving-Recidivism-as-a-Performance-Measure.PDF [https://perma.cc/N5U4-9KVG] (“[R]ecidivism is an important performance measure for justice agencies and should be at the heart of any effort to evaluate [Justice Reinvestment Initiative] outcomes.”); Prescott et al., supra note 33, at 1662 (“[W]e believe that recidivism data provide useful information that can allow policymakers to at least roughly understand the potential stakes of their decisions. Moreover, there is no necessarily better way to understand those stakes . . . .”). Moreover, this Essay’s study possesses certain inherent advantages that make its recidivism results more reliable.

One reliability problem facing recidivism studies that include released state prisoners is that many states use a parole system. Prisoners may be released early for good behavior while incarcerated and in consideration of other factors. Parole boards deliberately attempt to release people earlier who they estimate will be less likely to reoffend. Any recidivism study of people released from state prison is tainted by this fact. This situation makes it less likely that the recidivism rates of the released persons studied will be predictive of the potential recidivism rates of those who were not released. Therefore, one would need to exercise caution in using those recidivism rates to justify releasing others more quickly.227×227. See Prescott et al., supra note 33, at 1661 (“The core limitation is that we can only evaluate the postrelease behavior of people who have, in fact, been released from incarceration . . . .  It is worth remembering that with almost any study of postrelease recidivism, the people studied are not a random sample of the prison population but a sample of those who were released: people who did not receive life sentences, who may have behaved well in prison . . . .”).

However, the federal system no longer uses parole. People sentenced in federal court serve almost the entirety of their sentences. Thus, the released people in this study are more of a random sample, instead of a sample influenced by which people parole boards choose to release. This Essay’s study group was limited to people sentenced in federal court and almost all of the comparison recidivism rates used in this Essay are also of people sentenced in federal court.

Studying recidivism using people sentenced under a mandatory minimum, like the ACCA, also possesses advantages. One difficulty frequently encountered in comparing recidivism rates is sentence variation. As discussed previously, length of incarceration could positively or negatively affect recidivism.228×228. Id. at 1661–62 (“More time in prison might discourage future crime through rehabilitation or specific deterrence, but it could also encourage future crime by those imprisoned (for example, by making reentry into society more difficult).”).

Fortunately, because of the ACCA’s mandatory minimum sentence, the average time incarcerated for the “violent group” and the “drug-only group” only differed by a little more than two years (27.9 months). This makes it less likely that the difference in recidivism could be explained by differences in time incarcerated.

Demographics also fail to explain the difference in recidivism between the violent offenders, the drug-only offenders, and the mixed group. While studies show that younger people recidivate more, the younger ACCA “drug-only group” and “mixed group” recidivated less than the older “violent group.” While Black people who have been convicted of federal crimes are rearrested more than such persons of other racial groups (perhaps due to greater policing in neighborhoods with predominantly Black populations), in this study the “drug-only group” contained the most people of color and also had the lowest recidivism rate. All of these factors lend credibility to the Essay’s claim that the differences in recidivism rates are because of the predicate offenses the person committed (violent offenses versus drug offenses).

The conclusion that people with ACCA drug predicates reoffend less than people with ACCA violent predicates is further supported by findings about the criminal history of people who commit violent offenses, generally. U.S. Sentencing Commission studies show that people who commit drug offenses tend to have less extensive criminal histories than people who commit violent offenses. The Commission’s study of all people sentenced in federal court and then released or placed on probation in 2005, previously described in section III.A, showed that “career offenders,” under the Sentencing Guidelines, who committed only prior drug offenses had less extensive and less serious criminal histories than “career offenders” who committed at least one violent offense.229×229. 2011 Report to Congress, supra note 39, at 30, 43. Additionally, people sentenced under the ACCA in fiscal year 2019, who committed only drug-trafficking offenses had less extensive criminal histories than people sentenced under the ACCA that year who committed one or more violent offenses.230×230. Armed Career Criminals, supra note 7, at 36. The same study of those released or placed on probation in 2005 also showed that a person’s criminal history is closely correlated with recidivism.231×231. Recidivism Comprehensive Overview, supra note 192, at 5.

IV. Recommendations

Congress has expressed a willingness to revisit mandatory minimums in general, and even drug penalties specifically. For example, the First Step Act was a bipartisan look at mandatory minimums, and early drafts included changes to the ACCA. Also, some in Congress have suggested restructuring the ACCA. In 2018, two Senators proposed the Restoring the Armed Career Criminal Act, which would replace the “violent felony” and “serious drug offense” ACCA predicates with a single category — “serious felonies,” defined as any crime punishable by imprisonment of ten years or longer.232×232. Press Release, Sen. Tom Cotton, Cotton, Hatch Introduce the Restoring the Armed Career Criminal Act (Aug. 1, 2018), https://www.cotton.senate.gov/news/press-releases/cotton-hatch-introduce-the-restoring-the-armed-career-criminal-act [https://perma.cc/4326-DVQ2].

In the 2016 report Career Offender Sentencing Enhancements, the Sentencing Commission recommended greater differentiation by the Guidelines between “career offenders” with different types of criminal records. While the career offender provisions in the Guidelines and the ACCA vary in important respects, the Restoring the Armed Career Criminal Act, rather than further distinguishing between different types of criminal records, eliminates any differentiation between crimes with ten-year or longer sentences. Some examples of Arkansas crimes with maximum sentences of ten years233×233. Ark. Code Ann. § 5-4-401(a)(4) (2021) (“For a Class C felony, the sentence shall be not less than three (3) years nor more than ten (10) years . . . .”). that could perhaps be undesirable predicates for an “Armed Career Criminal Act” are: nonsupport (of a spouse or child more than $10,000),234×234. Id. § 5-26-401(b)(2)(B) (classifying offense as a class C felony if more than $10,000 but less than $25,000 is owed in child support). interference with custody (of a minor),235×235. Id. § 5-26-503(b) (classifying offense as a class C felony). and manufacturing or altering personal ID of a person (less than twenty-one years old).236×236. Id. § 5-27-502(b)(1)(A) (classifying offense as a class C felony).

Based upon this Essay’s recidivism analysis of nonpublicly available criminal history data, Congress should eliminate “serious drug offenses” as ACCA predicates. While people with three or more “violent felony” predicates (“violent group”) recidivated at a rate of 61.7%, people with one or two “violent felony” predicates (and one or more “serious drug offenses) (“mixed group”) recidivated at a rate of 48.9%, and people with no “violent felony” predicates (“drug-only group”) recidivated at a rate of 36.4%.237×237. Armed Career Criminals, supra note 7, at 54 tbl.8, 57. The recidivism rates for both the “mixed group” and the “drug-only group” are less than the 49.3% identical, overall recidivism rates for all people convicted of federal crimes released in 2005238×238. Recidivism Comprehensive Overview, supra note 192, at 5. and 2010.239×239. Firearms Recidivism Report for 2010 Releasees, supra note 186, at 4.

The “mixed group” and the “drug-only group” recidivated less even though their average ages at release were younger (45.7 and 43.0 years, respectively) than the average age at release for the “violent group” (48.2 years). A previous Sentencing Commission study showed an inverse relationship between age and recidivism, so it may have been expected that the more youthful “drug-only group” and “mixed group” would have had higher recidivism rates.

Eliminating the “serious drug offense” ACCA predicates would particularly reduce the percentage of incarcerated people of color. The ACCA has a disproportionate impact on people of color; a higher percentage of people sentenced under the ACCA are Black, as compared to the racial makeup of people convicted of federal crimes overall. Only 24.0% of all people released after federal sentences between 2009 and 2011 were Black, but more than half of the people released after ACCA sentences during the same time period were Black. Moreover, the “drug-only group” and the “mixed group,” which would no longer be subject to the ACCA under this Essay’s recommendations, contained higher percentages of people of color (88.3% and 70.7%, respectively) than the “violent group” (55.6%). The Essay makes no recommendation as to modifications of the “violent offense” predicates.

Conclusion

Many scholars and jurists criticize the ACCA, but now for the first time, this Essay provides an empirical basis for making a change. Using nonpublicly available criminal history data, this Essay compares the recidivism rate of people with fewer than three ACCA “violent felony” predicates (and one or more “serious drug offense” predicates) to those with at least three “violent felony” predicates. Based on this comparison, this Essay recommends that Congress eliminate drug offenses as ACCA predicates because people with fewer than three violent ACCA predicates recidivate at a rate no greater than the federal population as a whole. Implementing this change would reduce the ACCA’s administrative cost to the judiciary and the cost to the criminal justice system overall and particularly would reduce the incarceration of people of color.

 


The Harvard Law Review has not independently reviewed the data and analyses described herein.
* Climenko Fellow and Lecturer on Law, Harvard Law School. Former Supreme Court Fellow, placed at the U.S. Sentencing Commission. Many thanks to Professors Elizabeth Kamali and Andrea Roth; the Supreme Court Fellows program; and the staff of the U.S. Sentencing Commission for helpful discussion and comments. All views represented are the author’s alone.