concurring:
Because the district court committed no procedural error and sentenced Apodaca to the lifetime term of supervised release recommended by the Guidelines, our precedents require us to affirm Apodaca’s sentence. See, e.g., United States v. Carty, 520 F.3d 984, 993-94 (9th Cir.2008) (en banc). I therefore concur in the judgment and almost all of the opinion.
I write separately to state my view that the applicable statute, 18 U.S.C. § 3583(k), and Guidelines policy statement, U.S.S.G. § 5D1.2(b)(2), grossly overestimate the risk that defendants like Apodaca, who are convicted only of possessing child pornography downloaded from the Internet, and who have no prior contact child sex abuse convictions, will commit contact sex offenses against children. Compare, e.g., United States v. Williams, 636 F.3d 1229, 1231-32 (9th Cir.2011) (prior child sexual assault convictions). The routine imposition of lifetime terms of supervised release on Internet-only child pornography offenders departs from Congress’s purpose in enacting § 3583(k) and ignores the best available empirical evidence. This practice results in onerous sentences imposed without individualized attention to the “history and characteristics of the defendant,” 18 U.S.C. § 3553(a)(1), or to the “need for the sentence imposed to protect the public from further crimes of the defendant.” Id. § 3553(a)(2)(C).
A lifetime term of supervised release is extremely rare. Less than one percent of the federal defendants sentenced between fiscal years 2005-09 received such a term. U.S. Sent’g Comm’n, Federal Offenders Sentenced to Supervised Release 58-60 & nn. 252-55 (2010), available at http://www. ussc.gov/Research/Research_Publications/ SupervisecLRelease/20100722_SupervisecL Release.pdf. Congress has provided for lifetime supervision for defendants convicted of drug trafficking offenses, see 21 U.S.C. §§ 841(b), 960(b); terrorism offenses, see 18 U.S.C. §§ 2332b(g)(5)(B), 3583(j); and various sex offenses, see id. § 3583(k). For all other crimes, the maximum term of supervision is five years. Id. § 3583(b). The “overwhelming majority” — more than 95 percent — of those who received lifetime supervision were convicted of sex offenses. U.S. Sent’g Comm’n, supra, at 58-59.
When Congress enacted § 3583(k), it had in mind sex offenders who had committed contact offenses against children and were perceived to pose high risks of recidivism. See PROTECT Act of 2003, Pub.L. No. 108-21, § 101(3), 117 Stat. 650, 651-52; H.R. CONF. REP. NO. 108-66, at 49-50 (2003) (defending lifetime supervised release as necessary to monitor “the perpetrators of child sexual abuse crimes, whose criminal conduct may reflect deep-seated aberrant sexual disorders that are not likely to disappear within a few years of release from prison”), reprinted in 2003 U.S.C.C.A.N. 683, 684. See also, e.g., 149 Cong. Rec. S5113-01, 5132 (Apr. 10, 2003) (statement of Sen. Sessions) (“[Sjexual offenders and predators are repeat offenders. A 1998 study of sexual recidivism factors for child molesters showed that 43 percent of offenders sexually reoffended within a 4-year follow-up period. Almost half of the people arrested as child molesters reoffended in a sexual abuse case within 4 years.... We put in a provision that *1086would allow lifetime supervision---- The theory behind it is simply this: science and history tell us that child molesters are repeat offenders.”).
The legislative history of a bill passed by the House in the previous Congress, which was the genesis of § 3583(k), confirms this understanding. See Lifetime Consequences for Sex Offenders Act of 2002, H.R. 4679, 107th Cong. (2002); H.R. CONF. REP. 108-66, at 50. The House Report on H.R. 4679 focused on the risk of recidivism posed by violent sexual offenders and persons convicted of sexual assault. See H.R. REP. NO. 107-527, at 2 (2002). Likewise, debate on the 2002 bill before the House Judiciary Committee indicates that Representatives were primarily concerned with the threat of recidivism by contact sex offenders against children. See, e.g., Lifetime Consequences for Sex Offenders Act of 2002: Markup Transcript on H.R. 4679 Before the H. Comm, on the Judiciary, 107th Cong. 12 (2002) (statement of Rep. Green)(“This is giving judges the tools to prevent sex crimes against kids, because we know that sex crimes against kids — those monsters who perpetrate sex crimes against kids have among the highest recidivism rates of any type of offender.”), available at http://judiciary. house.gov/legacy/107-527.pdf.; id. at 23 (statement of Rep. Lofgren) (“[Ejspecially in the case of child molesting, where the recidivism rate is almost 100 percent, I think just releasing a child molester is a very dangerous thing to do.”).
Current empirical literature casts serious doubt on the existence of a substantial relationship between the consumption of child pornography and the likelihood of a contact sexual offense against a child. For example, one recent study that followed 231 child pornography offenders for six years after their initial offenses found that only nine persons, or 3.9 percent of the sample, committed even a non-contact sexual offense; only two persons, or 0.8 percent of the sample, committed a contact offense. Jérome Endrass et al., The Consumption of Internet Child Pornography and Violent Sex Offending, 9 BMC Psychiatry 43 (2009), available at http://www. ncbi.nlm.nih.gov/pmc/articles/PMC2716325/ pdfil471-244X-9-43.pdf. Endrass et al. concluded that “the consumption of child pornography alone does not seem to represent a risk factor for committing hands-on sex offenses ... at least not in those subjects without prior convictions for hands-on sex offenses.” Id. See also Michael C. Seto & Angela W. Eke, The Criminal Histories and Later Offending of Child Pornography Offenders, 17 Sexual Abuse 201, 208 (2005) (“[0]ur finding does contradict the assumption that all child pornography offenders are at very high risk to commit contact sexual offenses involving children.”).
Other studies reinforce the conclusion that individuals who have only possessed and/or viewed child pornography present substantially lower risks of harm than do individuals who have committed contact sex offenses. See, e.g., L. Webb, J. Craissati & S. Keen, Characteristics of Internet Child Pornography Offenders: A Comparison with Child Molesters, 19 Sexual Abuse 449, 463 (2007) (finding Internet-only offenders “significantly less likely to fail in the community than child molesters,” and concluding that “by far the largest subgroup of internet offenders would appear to pose a very low risk of sexual recidivism”), available at http://sax. sagepub.com/contenV19/4/449.full.pdf+ html; Seto & Eke, supra, at 207 & tbl.3 (finding that only 1.3 percent of Internet-only offenders in the sample recidivated with contact sex offenses, in contrast to 9.2 percent of persons with prior Internet and contact sex offenses).
*1087The government points to a single study that it argues demonstrates a causal link' between viewing child pornography and sexually assaulting children. Michael L. Bourke & Andres E. Hernandez, The ‘Butner Study’ Redux: A Report on the Incidence of Hands-On Child Victimization by Child Pornography Offenders, 24 J. Family Violence 183 (2009). But one of the study’s authors has disavowed the government’s citation of his work to support this claim, and has cautioned that “the argument that the majority of [child pornography] offenders are indeed contact sexual offenders and, therefore, dangerous predators ... simply is not supported by the scientific evidence.” Andres E. Hernandez, Psychological and Behavioral Characteristics of Child Pornography Offenders in Treatment, at 4 (Apr.2009) (unpublished manuscript) (emphasis deleted), available at http://www.iprc.unc.edu/G8/ Hernandez_position_paper_GlobaL Symposium.pdf. The author has explained that, because the study subjects were volunteer participants in a prison sex offender treatment program, the observed correlation between Internet child pornography consumption and contact sex offenses cannot be generalized. Id. at 9. Moreover, the underlying Butner study reported that “[t]he vast majority of our subjects indicated they committed acts of hands-on abuse prior to seeking child pornography via the Internet.” Id. (emphasis in original). See also Melissa Hamilton, The Efficacy of Severe Child Pornography Sentencing: Empirical Validity or Political Rhetoric?, 22 Stan. L. & Pol’y Rev. (manuscript at 42-48) (forthcoming 2011) (pointing out further problems with the Butner study, including sample bias and the possibility that study subjects over-reported their history of sexual offenses in order to curry favor with clinicians and avoid expulsion from the treatment program), available at http://papers.ssrn.com/sol3/papers.cfm? abstracted=1689507.
A growing body of literature suggests that the Internet — by rendering child pornography more accessible, affordable, and anonymously obtainable than in the past— has “facilitate^., a new kind of crime.” Andreas Frei et al., Paedophilia on the Internet — A Study of 33 Convicted Offenders in the Canton of Lucerne, 135 Swiss Med. Weekly 488, 492 (2005), available at http://www.smw.ch/docs/pdf200x/2005/33/ smw-11095.pdf. See also Endrass et al., supra, at 44; Webb, Craissati, & Keen, supra, at 450. There is no evidence that Congress or the Commission considered empirical evidence relevant to Internet-only offenders when they aggregated such offenders with contact sexual offenders under § 3583(k) and U.S.S.G. § 5D1.2(b)(2). Rather, Congress was contemplating quite different criminals: “people who are convicted of pedophilia” for whom “the recidivism rate ... is close to 100 percent.” Markup Transcript on H.R. 1679, supra, at 18 (statement of Rep. Sensenbrenner). That estimate of the risk of subsequent contact sex offenses may be high even with respect to contact sex offenders; with respect to Internet-only offenders like Apodaca, it is simply wrong.
Congress has directed that, in setting a term of supervised release, a sentencing court must select a term “sufficient, but not greater than necessary, to,” inter alia, “protect the public from further crimes of the defendant.” 18 U.S.C. § 3553(a)(2)(C); id. § 3583(c). It is inconsistent with this directive to treat all defendants convicted of one of the disparate group of sex' offenses enumerated in § 3583(k) as equally dangerous and therefore equally deserving of lifetime supervision. It is also inconsistent with the need to avoid unwarranted similarities among sentences for defendants convicted of dissimilar conduct. See Gall v. United States, 552 U.S. 38, 55, 128 S.Ct. 586, 169 L.Ed.2d 445 (2007). Sentencing Guideline § 5D1.2(b)(2) recom*1088mends lifetime supervised release for any defendant convicted of a “sex offense”— whether the offense is the violent rape of a child by a defendant with several sex abuse priors or the downloading of child pornography over the Internet by a first-time offender. It is difficult to reconcile this Guideline with the requirement of § 3553(a)(1) that a sentencing court focus on the particular characteristics of the offense and the offender. See United States v. Dorvee, 616 F.3d 174, 187 (2d Cir.2010); Carty, 520 F.3d at 991.
The policy statement at U.S.S.G. § 5D1.2(b)(2) does not “exemplify the [Sentencing] Commission’s exercise of its characteristic institutional role.” Kimbrough v. United States, 552 U.S. 85, 109, 128 S.Ct. 558, 169 L.Ed.2d 481 (2007). The Sentencing Commission added the policy statement when the statutory maximum term of supervised release was five years, not life. Compare U.S.S.G. app. C, amend. 615 (2001) with 18 U.S.C. § 3583(b) (2000). The Commission defined the term “sex offense” not by attempting to classify various types of sex offenders with respect to their relative risks of recidivism or their needs for ongoing supervision and treatment, but by adopting Congress’s definition of that term. Compare U.S.S.G. app. C., amend. 701 (2007) (codified at U.S.S.G. § 5D1.2 cmt. n. 1) with PROTECT Act of 2003, Pub.L. No. 108-21, § 101(3), 117 Stat. 651, 652 (codified at 18 U.S.C. § 3583(k)) and Adam Walsh Child Protection and Safety Act of 2006, Pub.L. No. 109-248, § 141(e)(2), 120 Stat. 587, 603 (codified at 18 U.S.C. § 3583(k)). As was true of the crack Guidelines at issue in Kimbrough, Congress, in enacting § 3583(k), relied on “assumptions about the ... prevalence of certain harmful conduct associated with” the consumption of child pornography “that more recent research and data no longer support.” 552 U.S. at 97, 128 S.Ct. 558 (internal quotation marks omitted).
The Commission has declared that a reassessment of the risk of recidivism posed by child pornography offenders is a priority in the coming year. See Notice of Final Priorities, 75 Fed.Reg. 54,699, 54,699-700 (Sept. 8, 2010). Cf. Kimbrough, 552 U.S. at 99-100, 128 S.Ct. 558 (discussing Commission’s proposed amendments to the 100:1 crack/powder sentencing ratio). As the law-now stands, Congress has permitted and the Sentencing Commission has recommended a lifetime term of supervised release for every child pornography offender. The evidence demonstrates that in many cases, such a sentence is not justifiable on the grounds that the defendant poses an elevated risk of committing contact sex offenses. I hope that Congress or the Commission will address the undifferentiated treatment of the dissimilar groups of sex offenders covered by § 3583(k) and U.S.S.G. § 5D1.2(b)(2) to ensure that terms of supervised release are “no greater than necessary” to achieve the purposes of sentencing, see 18 U.S.C. § 3553(a), while keeping in mind that supervised release “fulfills rehabilitative ends, distinct from those served by incarceration.” United States v. Johnson, 529 U.S. 53, 59, 120 S.Ct. 1114, 146 L.Ed.2d 39 (2000).