This appeal calls upon us to decide whether a district court's sentence can be upheld as reasonable when it is based solely on the district court's policy disagreement with how the United States Sentencing Guidelines treat crack cocaine offenses as compared to powder cocaine offenses, notwithstanding Congress's repeated rejection of proposals to alter that treatment.

The federal statute governing drug offenses calls for mandatory minimum sentences to be imposed according to drug quantity as measured by weight, where the quantity needed to trigger each minimum varies by type of drug. Following and building on this structure, the Sentencing Guidelines provide sentencing ranges for offenses involving powder cocaine and crack cocaine according to a ratio of 100 to 1, such that a crime that involves a certain quantity of crack cocaine falls within the same sentencing range as a crime that involves 100 times that amount of powder cocaine. In this case, without making any adjustment for the particularities of the individual defendant or his specific offenses, the United States District Court for the Southern District of New York (Sweet, J.) found the 100:1 ratio untenable and instead simply applied the 20:1 ratio that the Sentencing Commission currently advocates but that Congress has repeatedly refused to adopt. The government filed the instant appeal, and we are now compelled to reverse. We hold that district courts do not have the authority to reject unilaterally the 100:1 ratio on policy grounds, and we remand for further proceedings.

According to a complaint filed in the Southern District of New York on May 29, 2003, Juan Castillo, also known as "Padilla," was part of a conspiracy operating out of an apartment in upper Manhattan to sell crack cocaine in the spring of 2003. The FBI agent who filed the complaint stated that, after speaking in April 2003 with a confidential informant who described Castillo's drug operation, the FBI agent and other law enforcement personnel executed a search warrant on April 24, 2003 and seized from the apartment in question a variety of drugs and drug paraphernalia, including quantities of crack cocaine and powder cocaine, scales for weighing narcotics, materials for making crack, and a notebook in which drug transaction records were kept. The complaint was filed after the FBI agent and other law enforcement personnel made further identifications of Castillo that linked him with the operation of the drug sales from the apartment.

On June 9, 2003, Castillo was arrested pursuant to this complaint, and in statements made to the arresting officers, Castillo admitted that he had been selling drugs for approximately a year and a half in New York. He stated that he was part of a group of four other individuals who sold drugs from the vicinity of that upper Manhattan apartment, explaining that his main job was to negotiate prices for purchasing kilogram quantities of cocaine and that he also at times acted as a look-out for his colleagues, warning them when he knew the police were in the area. Castillo told the officers the names of his colleagues.

On July 7, 2003, an indictment based on the above conduct was filed, charging Castillo with three counts: (1) participating in a conspiracy to distribute and possess with intent to distribute 5 kilograms and more of mixtures containing cocaine and 50 grams and more of mixtures containing crack cocaine; (2) distributing and possessing with intent to distribute more than a kilogram and a half of mixtures containing crack cocaine; and (3) distributing and possessing with intent to distribute more than 5 kilograms of mixtures containing cocaine.

Castillo, represented by counsel, met with the government on October 6, 2003 for a safety valve proffer, to attempt to qualify for relief from the mandatory minimum sentences called for by 18 U.S.C. § 3553(f) and Section 5C1.2 of the Sentencing Guidelines.1 During the meeting, Castillo told the government that he had dealt drugs from 1994 to 2003 and that over the course of his drug career he had progressed from selling marijuana to selling cocaine and crack. He also shared details of the drug operation he was a part of at the time of the arrest. Finally, he stated that he had distributed over 10 kilograms of crack cocaine in the New York area between 1994 and 2003.

On February 23, 2004, the government provided Castillo with a Pimentel letter setting forth the government's position about the application of the Sentencing Guidelines to Castillo's case. See United States v. Pimentel, 932 F.2d 1029, 1034 (2d Cir.1991). According to the government, the base offense level would be 38, and because the defendant seemed to qualify for safety valve relief, a two-level decrease in offense level would be warranted, reducing the total offense level to 36. The government further explained that because Castillo had no criminal history points, his criminal history category would be I. Based on these calculations, the Guidelines set forth a sentencing range of 188 to 235 months' imprisonment.

Castillo pleaded guilty to all three counts in the indictment, without a plea agreement, before a magistrate judge on March 24, 2004. The district court accepted Castillo's guilty plea by written order dated April 1, 2004.

A Pre-Sentence Report was prepared in advance of the May 17, 2005 sentencing. The PSR agreed with the calculations set forth in the Pimentel letter and additionally recommended that Castillo benefit from a three-level reduction in offense level for acceptance of responsibility, which would result in a Guidelines range of 135 to 168 months. The PSR recommended that Castillo be sentenced at the bottom of the range and noted that there were no mitigating circumstances in Castillo's case that might affect his sentence.

Both Castillo and the government submitted sentencing memoranda in advance of the sentencing. Castillo requested a low or non-Guidelines sentence on a number of grounds. First, he argued that the district court should sentence him for only the quantity of drugs to which he admitted in post-arrest statements and the guilty plea — which would place him at a level 27, resulting in a Guidelines range of 70-87 months — instead of the entire amount of drugs found during the search of the apartment. He argued that this lower level was especially warranted because he was not a leader of the conspiracy, did not own the apartment where the drugs were seized, and was not present during the seizure. Castillo next argued that the district court should take into account his lack of prior bad acts, his good family relationships, his efforts to improve himself through education during his confinement, and the fact that he would be additionally punished by deportation to the Dominican Republic at the end of his prison sentence. He urged the district court to account for the fact that he tried to provide substantial assistance to the government, even though as it turned out he did not meet the criteria for a formal letter from the government asking for a reduction in sentence on that ground.

Most important for the purposes of this appeal, however, was Castillo's argument that, in the wake of United States v. Booker, 543 U.S. 220, 125 S. Ct. 738, 160 L. Ed. 2d 621 (2005), the district court was "now free to ignore the harsher penalties imposed under the Guidelines for cases involving `crack' cocaine as opposed to cocaine, and find the appropriate guideline by treating the sentences with equal severity, as there never was any rational reason to treat the substances differently." Castillo contended that the disparity between sentences for offenses involving crack cocaine and powder cocaine violates 18 U.S.C. § 3553(a)(6), which calls on courts to avoid unwarranted sentencing disparities among defendants with similar records who committed similar offenses. He also argued that it is difficult to tell what substances constitute crack and what substances constitute powder cocaine and so the disparity is even more unfair.

The government rejected Castillo's argument that a low or non-Guidelines sentence was appropriate, urging the district court to sentence within the Guidelines range of 135 to 168 months. The district court should calculate Castillo's base offense level as 38, argued the government, because this level properly accounts for all of the post-arrest statements Castillo made with respect to the amount of cocaine he had sold, including statements during the proffer session. The government also discounted Castillo's argument that he should receive a reduction as a lower-ranked member of the conspiracy, emphasizing his serious responsibility and participation in the drug ring. Finally, the government argued that the disparity between crack and powder cocaine should not result in a lower sentence, where the disparity had repeatedly withstood court challenges, where the disparity was not a permissible ground for a departure, and where the factors set forth in 18 U.S.C. § 3553(a) supported a Guidelines sentence.

Before the sentencing proceeding, the district court issued a written opinion announcing Castillo's sentence. See United States v. Castillo, 2005 WL 1214280, 2005 U.S. Dist. LEXIS 9780, No. 03 Cr. 835(RWS) (S.D.N.Y. May 20, 2005). While the district court agreed with the government that the Guidelines range called for a term of imprisonment of 135 to 168 months, id. 2005 WL 1214280, at *4-5, 2005 U.S. Dist. LEXIS 9780, at *10-11, the district court instead imposed a sentence of 87 months' incarceration on each count, to be served concurrently, followed by a three-year term of supervised release. Id. 2005 WL 1214280, at *1, 2005 U.S. Dist. LEXIS 9780, at * 1.

The district court did not point to anything in either the defendant's background or offense in support of this lower sentence. Indeed, in discussing the defendant's background, the district court simply presented without further comment a brief overview of certain biographical details, including the following facts: that Castillo was born in the Dominican Republic in 1977; that his parents separated when he was a child; that he lived with his father and step-mother and maintained minimal contact with his mother while he was growing up; that he emigrated to the United States with his family in 1989 and settled in the Bronx; that he had a happy childhood; that he completed 11 years of formal schooling; that he was not married but had fathered four children with three different women, one of whom he had been involved with for the past seven years; that he had no substance abuse issues; and that he had no significant past employment and had never filed a federal income tax return. Id. at 2005 WL 1214280, at *1-3, 2005 U.S. Dist. LEXIS 9780, at *4-6. Reviewing the offense conduct, the district court summarized the findings from the execution of the search warrant and Castillo's post-arrest statements and plea allocution, but said nothing to indicate that the court found this particular offense out of the ordinary. Id. 2005 WL 1214280, at *2-4, 2005 U.S. Dist. LEXIS 9780, at *6-7.

Instead, the reduced sentence was based on only one argument raised in Castillo's sentencing memorandum: the disparity between the Guidelines range for offenses involving crack cocaine and offenses involving powder cocaine. The district court concluded that this disparity could not be justified under § 3553(a)(6) and that a non-Guidelines sentence was thus warranted. "Since Booker," the district court explained, "a number of courts, concerned by the disparity between crack and cocaine powder sentences imposed under the Guidelines, have imposed non-Guidelines sentences in cases involving crack." Id. 2005 WL 1214280, at *5, 2005 U.S. Dist. LEXIS 9780, at * 12. The district court cited two such courts, and went on to quote the observation of one of them that "Courts, commentators and the Sentencing Commission have long criticized this disparity, which lacks persuasive penological or scientific justification, and creates a racially disparate impact in federal sentencing." Id. 2005 WL 1214280, at *5, 2005 U.S. Dist. LEXIS 9780, at *12-13 (quoting United States v. Smith, 359 F. Supp. 2d 771, 777 (E.D.Wis.2005)).

Noting that the Smith court had applied the 20:1 ratio between crack and powder cocaine recently recommended by the Sentencing Commission — a reduction from the 100:1 ratio currently in place — the district court followed suit, and recalculated Castillo's drug quantity according to this ratio. Id. 2005 WL 1214280, at *5-6, 2005 U.S. Dist. LEXIS 9780, at *13. Such a recalculation resulted in a base offense level of 34, which, minus the five levels for safety valve relief and acceptance of responsibility, resulted in a total offense level of 29. Id. 2005 WL 1214280, at *6, 2005 U.S. Dist. LEXIS 9780, at * 14. Taking into account Castillo's criminal history category of I, the Guidelines recommended a term of 87 to 108 months' imprisonment for this offense level. Id. In announcing a sentence of 87 months on each count to run concurrently, the district court chose the low point in the range without further explanation, and explained that the terms of the sentence were subject to modification at the sentencing hearing to take place later that day. Id. 2005 WL 1214280, at *6 2005 U.S. Dist. LEXIS 9780, at *15.

At the sentencing hearing, the defendant thanked the court for the sentence. The government, however, "willfully disagree[d]." The government referred the court to the government's sentencing submission, reiterating its position that the Second Circuit had held that the disparity between crack and powder cocaine was rational and that, far from avoiding sentencing disparities under § 3553(a)(6), the district court's actions actually would create a disparity with other defendants. The government maintained that the Guidelines sentence of 135 to 168 months was appropriate.

Notwithstanding the government's arguments, the district court reaffirmed the sentence announced in the written opinion.2 Castillo is currently serving his sentence.

On appeal, the government argues that the sentence — 48 months below the low end of what the district court found to be the applicable Guidelines range — was unreasonable. According to the government, the district court could not properly impose a non-Guidelines sentence simply because of its policy disagreement with the relevant Guidelines instead of its assessment of factors specific to Castillo and his crime. Such a sentence, goes the argument, cannot be reconciled with Booker because that case requires careful, defendant-specific consideration of the Guidelines along with the other § 3553(a) factors; it does not permit a categorical rejection of a particular section of the Guidelines. Moreover, the government continues, the sentence is contrary to § 3553(a)(6) because it creates unwarranted disparities between Castillo and other similarly situated defendants who continue to be sentenced according to the 100:1 ratio, and it is unsupported by the other § 3553(a) factors. Finally, the government argues that the district court's sentence cannot be squared with congressional intent to treat crack offenses much more severely than offenses based on powder cocaine. For all of these reasons, the government asks us to reverse and remand.

Castillo argues in response that the sentence was reasonable: The district court properly considered the factors in § 3553(a); committed no error in its calculation of the advisory Guidelines range; determined that a Guidelines sentence was not appropriate; and permissibly exercised its discretion to impose a non-Guidelines sentence. According to Castillo's reading, a district court must impose a non-Guidelines sentence when it believes that a Guidelines sentence would result in a sentence "greater than necessary," which, Castillo argues, is all that the district court did.

We agree that the sentence must be vacated and remanded for further proceedings. Before turning to the specifics of the case, however, we review the history of the Sentencing Guidelines concerning crack and powder cocaine.

The 100:1 ratio between crack and powder cocaine first came into being as a result of the Anti-Drug Abuse Act of 1986, Pub.L. No. 99-570, 100 Stat. 3207 (1986) ("1986 Act"). The ratio refers not to a comparison of the length of a sentence given for offenses involving the two drugs but to the quantity of each drug that is required to trigger certain sentences. The relevant part of the 1986 Act has been codified at 21 U.S.C. § 841(b) and calls for 5 kilograms of cocaine but only 50 grams of cocaine base (commonly referred to as "crack")3 to trigger a mandatory minimum of ten years, while 500 grams of cocaine but only 5 grams of cocaine base will trigger a mandatory minimum of five years. See 21 U.S.C. §§ 841(b)(1)(A)(ii)-(iii), 841(b)(1)(B)(ii)-(iii).

The speed with which the 1986 Act moved through Congress has often been noted. See, e.g., United States Sentencing Commission, Special Report to the Congress: Cocaine and Federal Sentencing Policy, February 1995 ("1995 Report"), at 116-17; Perry, 389 F.Supp.2d at 289 (collecting articles); Smith, 359 F.Supp.2d at 778 (collecting cases). Because of this speed, the legislative history is devoid of a formal committee report analyzing the Act's contents — indeed, no committee hearings took place — and the legislative history is based mostly on the statements of specific legislators. See 1995 Report at 117. While no formal discussion of the 100:1 ratio exists in the legislative history, it is noteworthy that several other ratios (including 50:1 and 20:1, the latter of which was apparently supported by the Reagan administration) were under consideration in different bills that were eventually rejected. Id. Although "the abbreviated, somewhat murky legislative history simply does not provide a single, consistently cited rationale for the crack-powder cocaine penalty structure," id. at 121, the Sentencing Commission has viewed the following five congressional conclusions as lying behind the sentencing disparity: (1) the "extraordinarily addictive" nature of crack cocaine, both in relative and absolute terms; (2) the higher correlation between crack cocaine use and the commission of other serious crimes than with other drugs; (3) the "especially perilous" physiological effects of crack cocaine; (4) the sense that "young people were particularly prone to using crack cocaine"; and (5) the increasingly widespread use of crack cocaine because of its high potency, low cost, and ease of manufacture, transportation, and administration. 1995 Report at 118.

In 1987, the Sentencing Commission issued the first edition of the Sentencing Guidelines. Section 2D1.1 of the Sentencing Guidelines incorporated the 100:1 ratio that Congress had imposed for mandatory minimums into a drug quantity table that tied quantities of drugs to base offense levels. U.S.S.G. § 2D1.1(c). For example, 150 kilograms of cocaine powder but only 1.5 kilograms of crack results in a base offense level of 38, 50 kilograms of cocaine but only 500 grams of crack triggers a base offense level of 36, and so on. The application notes to Section 2D.1 clearly indicate that the table is based on Congress's structure for mandatory minimums in § 841(b) and explain that because of the effort to link drug quantities to statutory equivalences, the ratios "do not necessarily reflect dosages based on pharmacological equivalents." See Application Note 10.

In the Anti-Drug Abuse Act of 1988, Congress further differentiated the punishment of crack and powder cocaine offenses by creating a mandatory minimum penalty for simple possession of crack cocaine, under which possession of over five grams of crack is punishable by a minimum of five years in prison. See 21 U.S.C. § 844; United States Sentencing Commission, Special Report to the Congress: Cocaine and Federal Sentencing Policy, April 1997 ("1997 Report"), at 3. The only mandatory minimum for simple possession of a controlled substance, it is distinguishable from the penalties for simple possession of powder cocaine, for which possession in any quantity is a misdemeanor subject to a maximum penalty of one year in prison. Id.

The imposition of the 100:1 ratio in the mandatory minimums and throughout the entire sentencing structure was a controversial decision. Apparently concerned about the over-inclusivity of the mandatory minimums, Congress created the so-called "safety-valve provision" in the Mandatory Minimum Sentencing Reform Act of 1994, under which defendants could obtain relief from the mandatory minimums if they met certain requirements. See Pub.L. No. 103-322, tit. VIII, § 80001(a), 108 Stat. 1796, 1985-86, codified at 18 U.S.C. § 3553(f); see also United States v. Reynoso, 239 F.3d 143, 148 (2d Cir.2000) (explaining purposes of the Act). Later that year, Congress directed the Sentencing Commission to study federal sentencing policy on all forms of cocaine and to report back to Congress with recommendations on whether the current structure with the disparate penalties for crack and powder cocaine should be retained or modified. See Violent Crime Control and Law Enforcement Act of 1994, Pub.L. No. 103-322, § 280006, 108 Stat. 1796 (1994).

The 1995 Report resulted from this request. In that report, the Sentencing Commission identified a number of concerns with the 100:1 ratio. In particular, the Sentencing Commission noted that, even though it found no evidence that the differential penalties had been racially motivated, African-Americans were nonetheless bearing the brunt of the higher sentences for crack cocaine. 1995 Report at xii. The Sentencing Commission was also troubled that the 100:1 ratio meant that low-level street dealers of crack were being sentenced far more severely than the high-level powder-cocaine suppliers who had sold the street dealers the raw material to make the crack in the first place. Id. at xii-xiii. The Commission concluded that, while some differential was warranted between crack and powder cocaine, the 100:1 ratio was disproportionate given the relative harms of the two forms of the drug, and noted that, to the extent that some of the specific ills associated with crack use were already taken into account through other enhancements such as specific offense characteristics, the use of the 100:1 ratio might result in what is in effect double punishment. Id. at xiii-xiv. While the Commission did not propose any particular revision to the Guidelines or the statute in terms of the ratio, the Commission stated its intention to study the matter further and to present revised Guidelines relevant to cocaine offenses within the next year. Id. at xv.

Several months after the 1995 Report was issued — apparently concluding that the goals of the crack/powder differential could best be addressed by adjusting specific enhancements instead of maintaining the differential — the Commission proposed a revision to the Guidelines that would eliminate the differential entirely. Under this proposal, the term "cocaine" would be re-defined to encompass all forms of that drug, including crack as well as powder, thus reducing the proposed penalty range associated with a particular quantity of crack. See 60 Fed. Reg. 25074, 25075-76 (May 10, 1995). Additionally, seeking to target ills that were more associated with crack use than with powder use, the proposal included increased enhancements for use of a weapon in connection with controlled substance offenses and an application note that would specify that bodily injury to a victim would be grounds for an upward departure. Id. at 25076-77. This proposal did not have the unanimous support of the Sentencing Commission, however. Three of the seven members on the Commission dissented from this recommendation, arguing that some differential was required because sentencing enhancements could not entirely account for the different harms between the two forms of the drug. See 1997 Report at 1. Notwithstanding this lack of unanimity, the proposed changes were to go into effect pursuant to 28 U.S.C. § 994(p) in early November 1995, unless Congress chose to modify or disapprove the recommended changes before that time.

Congress chose the latter route. After the House Committee on the Judiciary's Subcommittee on Crime held a day of hearings on the Commission's recommended changes, Congress rejected the Commission's proposal to do away with the sentencing disparity on October 30, 1995. Pub.L. No. 104-38, § 1, 109 Stat. 334 (Oct. 30, 1995); see also H.R.Rep. No. 104-272 at 3-5 (1995), reprinted in 1995 U.S.C.C.A.N. 335, 337-38 (describing hearings). Recognizing the Commission's strong feeling that the 100:1 ratio was not justified, Congress directed the Commission to propose new revisions of the crack/powder Guidelines, but cabined the authority of the Commission with instructions that "the sentence imposed for trafficking in a quantity of crack cocaine should generally exceed the sentence imposed for trafficking in a like quantity of powder cocaine" and that the Commission's recommendations should "propose revision of the drug quantity ratio of crack cocaine to powder cocaine under the relevant statutes and guidelines in a manner consistent with the ratios set for other drugs and consistent with the objectives set forth in section 3553(a) of title 28 United States Code." Pub.L. No. 104-38, at §§ 2(a)(1)(A), 2(a)(2). Upon signing this bill into law, President Clinton took the opportunity to state specifically that it was inappropriate to "dramatically reduc[e] the penalties for crack," given the "devastating impact [of crack] on communities across America, especially inner-city communities," while acknowledging that "[s]ome adjustment" of the "substantial disparity between sentences for crack as compared to equal amounts of powder cocaine" was warranted. See Presidential Statement on Signing S. 1254, 1995 WL 634347 (Oct. 30, 1995).

The 1997 Report was the Commission's response to Congress's 1995 directive. In that report, comparing current federal sentencing policy to the accepted goals of the policy, the Commission explained that both crack and powder cocaine are dangerous but that federal sentencing policy should reflect the greater dangers associated with crack; that the current mandatory minimum structure did not accurately target the "mid-level" or "serious" traffickers whom Congress sought to punish heavily; that the current sentencing structure for crack offenses was not an efficient use of limited federal resources; and that the current sentencing policy resulted in a public perception of unfairness and inconsistency because most offenders convicted of crack distribution were African-American while most crack users were white, and because powder cocaine can be easily converted into crack cocaine so sentences received were often linked to the government's decision of when to seize and arrest. 1997 Report at 3-8.

To address these discrepancies between goals and results, the Commission recommended that Congress revise the statutory mandatory minimum scheme by increasing the quantity of crack that triggers a five-year mandatory minimum from 5 grams to somewhere between 25 and 75 grams and by decreasing the quantity of powder cocaine that triggers a five-year mandatory minimum from 500 grams to somewhere between 125 and 375 grams — effectively recommending that the ratio be reduced from 100:1 to 5:1, if the top and bottom range for each form of the drug were used similarly. Id. at 9. The Commission called on Congress to adopt the new ratio "as soon as possible," and offered its expertise to aid both Congress and the Executive Branch "at any time." Id. at 9-10. Finally, the Commission expressed its belief that the mandatory minimum for simple possession of crack cocaine could not be supported for the same reasons that the 100:1 ratio was inappropriate, and recommended that the penalty for simple possession of crack cocaine be revised to be the same as the penalty for simple possession of powder cocaine. Id. at 10.

Unlike in 1995, the Commission did not propose revisions to the Guidelines that Congress would have to decline if it did not want them to go into effect. Notwithstanding support from the Clinton administration (which proposed revising the ratio to 10:1 by raising the crack trigger to 25 grams and lowering the cocaine trigger to 250 grams) and from nearly thirty federal judges who had been former United States Attorneys (who submitted a letter to Congress recommending that the disparity be eliminated or drastically reduced), the 1997 Report did not bring about the change that the Sentencing Commission had recommended. See Perry, 389 F.Supp.2d at 302 n. 24 (summarizing reaction to 1997 Report); 2002 Report at 2 & n. 7 (same). While the report did prompt the introduction of a number of bills later that year that would have equalized the 100:1 ratio, nothing passed. See 2002 Report at 3 and nn. 11-12.

The disparity did, however, remain on the radar screen. There was generally bipartisan support for the idea that the 100:1 ratio was too great, although not consensus as to the best way to reduce it. Bills to reduce the ratio — some by raising the quantities for crack that would trigger the minimum, others by lowering the quantities of powder that would trigger the minimum, still others by doing both — continued to be introduced over the next few years. See 2002 Report at 3-4 & nn. 11-12. For example, in 2000, Senator Spencer Abraham (R-Michigan) introduced an amendment to bankruptcy legislation that would have lowered the ratio to 10:1, by leaving the quantity of crack that triggered the five-year minimum intact but by lowering the quantity of powder cocaine from 500 grams to 50 grams — in effect increasing the sentences for powder cocaine rather than reducing them for crack. See 2002 Report at 3-4, referring to Bankruptcy Reform Act of 2000, H.R. 833, 106th Cong. (2000). This legislation passed the Senate in 2000 by one vote but went no further. See id.

In 2001, Senators Jeff Sessions (R-Alabama) and Orrin Hatch (R-Utah) introduced the Drug Sentencing Reform Act of 2001, which proposed to lower the ratio to 20:1, by decreasing the amount of powder cocaine and increasing the amount of crack cocaine necessary to trigger each mandatory minimum, thereby making crack cocaine sentencing somewhat more lenient while strengthening the penalties for powder. See S. 1874, 107th Cong. (2001); see also 147 Cong. Rec. S13961, S13965 (Dec. 20, 2001) (Statement of Sen. Sessions) (noting that "the 100-to-1 disparity in sentencing between crack cocaine and powder cocaine, which falls the hardest on African-Americans, is not justifiable" and asking colleagues "to cast aside the politics of the Left and the Right and to support this bill on the merits as a matter of plain, simple justice"). This bill was referred to the Senate Committee on the Judiciary but was never voted upon.

In a further attempt to move the issue forward, Senators Patrick Leahy (D-Vermont) and Hatch, the then-Chairman and Ranking Member of that committee, wrote the Commission to ask for another report on the penalty structures for cocaine offenses. See 2002 Report at 4. The Commission submitted the 2002 Report in response to this request. See id.

In the 2002 Report, the Commission "firmly and unanimously" expressed its belief "that the current federal cocaine sentencing policy is unjustified and fails to meet the sentencing objectives set forth by Congress in both the Sentencing Reform Act and the 1986 Act." 2002 Report at 91. The beliefs about the relative harmfulness of the two forms of the drug and the existence of problems associated with crack cocaine use could no longer be supported by the evidence, the Commission concluded. According to the Commission, the ratio "greatly overstates" the comparative problems with crack with respect to addiction levels, prenatal exposure, and propensity of youth to use crack cocaine; the penalties for crack cocaine offenses have the greatest effect on low-level street dealers of relatively small amounts, instead of hitting serious dealers the hardest; the penalties overstate the seriousness of most crack cocaine offenses; and the severity of the penalties lead to the troubling perception of "improper racial disparity." Id. at 93-103. The Commission added, however, that the data are not available to conclude whether the disparate impact of the crack sentences on black traffickers is actually disproportionate to the number of black traffickers. Id. at 103. The Commission recommended "disentangling some of the harms accounted for in the 100-to-1 drug quantity ratio" by taking the following steps:

(1) using specific sentencing enhancements to target the minority of offenders who engage in the most harmful conduct that concerned Congress in 1986; (2) decreasing the residual quantity-based penalties that apply to all crack cocaine offenders accordingly (to at least 25 grams for the five-year mandatory minimum penalty, and at least 250 grams for the ten-year mandatory minimum penalty); and (3) maintaining at current levels the quantity-based penalties for powder cocaine offenses.

Id. at 91-92; see also id. at 104-111 (expanding on recommendations). This recommendation would reduce the ratio to 20:1. Echoing its efforts in 1997, the Commission also urged Congress to repeal the five-year mandatory minimum penalty for simple possession of five grams or more of crack cocaine. Id. at 109.

In making these recommendations, the Commission acknowledged the view of the Department of Justice ("DOJ") that the current sentencing scheme for crack cocaine offenses should remain in place. Id. at 92. The Commission reasoned, however, that DOJ's concern that crack was associated with more ills than was powder cocaine did not mean that the 100:1 ratio must stay in place; instead, the Commission argued, that ratio was overinclusive, and DOJ's goals would be better met by reducing the ratio and clarifying the enhancements in the Guidelines. Id. In that respect, the Commission noted that the first draft of the Sentencing Guidelines had not been promulgated when Congress first put in place the mandatory minimums, such that the only tool available at that time was the blunt mandatory minimum scheme. Id. at 91. Because the "finely calibrated" Guidelines were now in place, however, the Commission believed that mandatory minimum scheme itself needed rethinking. Id. While the Commission presented a model Guidelines revision to Congress with the Report, the revision assumed that certain corresponding statutory changes would be made, and so was not a formal proposal that would take effect unless congressionally rejected. See Appendix A to the 2002 Report.

A number of bills were proposed in the wake of the 2002 Report. See, e.g., H.R. 345, 108th Cong. (2003) (proposing that the penalties for crack and powder cocaine be equalized by increasing the penalties for powder offenses); H.R. 1435, 108th Cong. (2003) (proposing that the penalties for crack and powder cocaine be equalized by reducing the penalties for crack offenses). Indeed, just last month Senator Sessions — joined by Senators Mark Pryor (D-Arkansas), John Cornyn (R-Texas), and Ken Salazar (D-Colorado), all former state Attorneys General — introduced the Drug Sentencing Reform Act of 2006, which proposed to reduce the disparity to 20:1 by reducing the penalty for crack cocaine while raising the penalty for powder cocaine. See S. 3725, 109th Cong. (2006); Press Release of Senator Sessions, Sens. Sessions, Pryor, Cornyn and Salazar Introduce Drug Sentencing Reform Act. However, while bills continue to be introduced that would reduce or equalize the ratio in some fashion, nothing to this end has been adopted. Thus, the mandatory minimums and 100:1 ratio in the Guidelines remain in place.

Finding little success in the legislative arena, opponents of the 100:1 ratio turned to the judiciary over the course of the 1990s. Yet, as a 2002 analysis of federal sentencing law written by the Department of Justice explains,

Every appellate court that has heard a challenge to the crack and powder cocaine sentencing structure has upheld it as constitutional. Defendants challenged the federal sentencing scheme under the Equal Protection and Due Process clauses, and the Eighth Amendment. Defendants have also asserted that the federal sentencing statutes are unconstitutionally vague. These constitutional challenges to the federal sentencing scheme have failed.

Department of Justice, Federal Cocaine Offenses: An Analysis of Crack and Powder Penalties (Mar. 19, 2002) ("2002 DOJ Report"), at 14.

This Circuit's precedents illustrate this phenomenon. In particular, as to challenges brought under the Equal Protection Clause — challenges whose underlying concern the district court here echoed in commenting that the crack/powder disparity has a racially disparate effect — this Court has upheld the crack/powder disparity under rational basis review, see United States v. Stevens, 19 F.3d 93, 97 (2d Cir. 1994) ("[W]e believe that treatment of one gram of crack cocaine as the equivalent of 100 grams of powder cocaine is rationally related to the legitimate governmental purpose of protecting the public against the greater dangers of crack cocaine. . . ."); concluded that "Congress and the Sentencing Commission did not enact the 100 to 1 ratio with a discriminatory intent," see United States v. Moore, 54 F.3d 92, 99 (2d Cir.1995) (rejecting defendant's effort to argue that heightened scrutiny applies); and declined to treat crack and cocaine traffickers as a suspect class, see United States v. Coleman, 166 F.3d 428 (2d Cir. 1999) (per curiam), cert. denied, 526 U.S. 1138, 119 S. Ct. 1794, 143 L. Ed. 2d 1021 (1999) (rejecting defendant's argument that the policy positions of the Sentencing Commission and then-Attorney General Janet Reno in support of eliminating or reducing the ratio should trigger intermediate scrutiny).

Separately, a panel of the Second Circuit also held that "the harsher penalties for crack crimes present no basis for downward departure." United States v. Haynes, 985 F.2d 65, 70 (2d Cir.1993) (rejecting defendants' argument that "because most crack users are African-Americans — while most cocaine users are white — the enhanced crack penalties unfairly punish African-Americans and should be a ground for downward departure" on the ground that "the enhanced penalties for crack reflect a rational and specific congressional aim of deterring drug transactions involving crack"). Other circuits reached the same conclusion, even after the Sentencing Commission issued the 1995 Report that expressed its dissatisfaction with the ratio. See United States v. Gaines, 122 F.3d 324, 329-30 (6th Cir. 1997) (finding no authority for downward departure based on 1995 Report because "Congress made a deliberate and informed decision to keep the 100:1 ratio and not to adopt the 1:1 ratio" and "[w]hen Congress and the Sentencing Commission disagree on matters of sentencing policy, Congress trumps"); United States v. Berger, 103 F.3d 67, 71 (9th Cir.1996) (agreeing with other courts' rejection of "the notion that a district court may override the express intention of Congress regarding penalties for crack cocaine and powder cocaine" and approvingly citing another court's conclusion that "[i]t is not the province of this Court to second guess Congress's chosen penalty") (internal citations and quotation marks omitted); United States v. Sanchez, 81 F.3d 9, 11 (1st Cir.1996) ("[W]e cannot blind our eyes to the fact that the Congress shot down the Commission's recommendation [to eliminate the 100:1 ratio]"); United States v. Lewis, 90 F.3d 302, 306 (8th Cir.1996) ("It is not for [the courts] to decide whether the 100:1 ratio is wise or equitable; that is a question for the popularly chosen branches of government."); United States v. Alton, 60 F.3d 1065, 1071 (3d Cir.1995) ("We defer to Congress and the Sentencing Commission to address the related policy issues [involving the disparate impact of crack cocaine penalties on African-Americans] and to consider the wisdom of retaining the present sentencing scheme").

With constitutional challenges foreclosed and departure authority rejected, by the end of the 1990s there seemed to be little role that the courts could play with respect to the crack/powder ratio, other than implementing it as written. All of this was put into question, however, after the Supreme Court issued its decision in Booker in January of 2005.

In Booker, the Supreme Court considered two different cases involving sentences for crack and powder cocaine offenses. The first defendant, Freddie Booker, had been convicted by a jury of possession with intent to distribute at least 50 grams of crack, which, under the Guidelines, would have given rise to a maximum sentence of 21 years and 10 months. 543 U.S. at 227, 125 S. Ct. 738. Instead, the district court sentenced him to 30 years, after finding by a preponderance of the evidence at a post-trial proceeding that he had possessed an additional 566 grams of crack and was guilty of obstructing justice, findings that triggered a Guidelines range of 30 years to life. Id. The second defendant, Duncan Fanfan, had been convicted by a jury of conspiracy to distribute and to possess with intent to distribute at least 500 grams of cocaine. Id. at 228, 125 S. Ct. 738. Under the Guidelines, conviction on the basis of these facts would have given rise to a sentence of imprisonment ranging from 5 to 6 years. Id. In a post-trial hearing, however, the district court found by a preponderance of the evidence that Fanfan was responsible for 2.5 kilograms of cocaine and 261.6 grams of crack, and also found that he was a leader of the conspiracy. Id. On the basis of these additional facts, the Guidelines called for a sentence of 15 or 16 years (although the district court ultimately did not impose this higher sentence). Id. at 228-29, 125 S. Ct. 738.

While the effects of the ratio are clearly apparent from these two sentences — additional factfinding aside, Booker's conviction on the basis of a much smaller quantity of crack cocaine resulted in a much longer sentence — the question before the Supreme Court did not involve the legality of the sentencing disparity between crack and powder cocaine. Instead, the issues were the constitutionality of judicial fact-finding at sentencing in light of the Sixth Amendment's guarantee of a jury trial and the continued vitality of the Guidelines if judicial fact-finding were deemed to be a constitutional problem. Id. at 229, 125 S. Ct. 738.

The Supreme Court answered these questions in two separate opinions. In the substantive opinion, the Court concluded that, to comply with the Sixth Amendment, "[a]ny fact (other than a prior conviction) which is necessary to support a sentence exceeding the maximum authorized by the facts established by a plea of guilty or a jury verdict must be admitted by the defendant or proved to a jury beyond a reasonable doubt." Id. at 244, 125 S. Ct. 738 (Stevens, J., opinion for the Court). Separately, in the remedial opinion, the Court held that the proper way to implement the constitutional holding was to excise the provision of the Sentencing Guidelines that made its application mandatory, as well as the provision that set forth standards of review on appeal. Id. at 245, 125 S. Ct. 738 (Breyer, J., opinion for the Court). As we explained in United States v. Crosby, 397 F.3d 103 (2d Cir.2005), "Booker/Fanfan holds that judicial fact-finding that increases [the] maximum lawful penalty, if required by the Guidelines, is prohibited. As a result of the Remedy Opinion in Booker/Fanfan, however, the maximum lawful sentence is the statutory maximum sentence, and because judicial fact-finding under advisory guidelines cannot increase that lawful maximum, judicial fact-finding now encounters no Sixth Amendment difficulties." Id. at 109 n. 6.

In the post-Booker world, then, district courts have the authority to give two different types of sentences: Guidelines sentences (with or without a permissible departure) and nonGuidelines sentences. See Crosby, 397 F.3d at 111-12 and n. 9. All sentences are to be imposed in accordance with 18 U.S.C. § 3553(a), which calls on district judges to "consider" a variety of factors, including the applicable Guidelines range and certain factors specific to the defendant. Id. at 111-114.

After Booker thus changed the landscape of federal sentencing, a number of district judges who had been disturbed over the years by the effects of the 100:1 ratio began to sentence defendants convicted of crack offenses to non-Guidelines sentences. These district judges believed that they were bound not to apply the 100:1 ratio if, after calculating the Guidelines range and considering the § 3553(a) factors, they concluded that the ratio resulted in a sentence "greater than necessary" to reflect the sentencing interests of § 3553(a)(2) and that the Guidelines sentence resulted in "unwarranted sentence disparities" in violation of § 3553(a)(6). They relied greatly on the Commission's repeated efforts to lower the ratio and on the idea that, while the mandatory minimums may be statutory and therefore binding, the ratio embedded in the drug quantity table in the Guidelines is, after Booker, only advisory. Most of the district courts chose to apply the 20:1 ratio recommended by the Commission in 2002, but at least one applied a 10:1 ratio. See, e.g., United States v. Fisher, ___ F.Supp.2d ___, 2005 WL 2542916 at *5-7, 2005 U.S. Dist. LEXIS 23184, at *24-25 (S.D.N.Y. Oct. 11, 2005) (applying 10:1 ratio); United States v. Perry, 389 F. Supp. 2d 278, 307-08 (D.R.I.2005) (applying 20:1 ratio); United States v. Beamon, 373 F. Supp. 2d 878, 887 (E.D.Wis.2005) (same).4

At the same time, other district judges considered defendants' requests to sentence without relying on the 100:1 ratio but declined to do so. Even while expressing discomfort or dismay about the heightened sentences for crack offenses, these judges explained that it was up to Congress to set sentencing policy and that they could not deem this disparity "unwarranted" within the meaning of the federal sentencing statute, given Congress's repeated refusal to approve a lower ratio. They therefore concluded that consideration of the § 3553(a) factors with respect to a particular defendant may result in an individualized sentence below the Guidelines range, but that the 100:1 ratio is the ratio that must be used to calculate the range itself. See, e.g., United States v. Doe, 412 F. Supp. 2d 87 (D.D.C.2006); United States v. Tabor, 365 F. Supp. 2d 1052 (D.Neb.2005).

When the government appealed the case at bar, no court of appeals had yet ruled on whether district courts were free to calculate the sentence using a ratio other than the 100:1 ratio provided in the Guidelines. Shortly thereafter, the Seventh and Eighth Circuits ruled that district courts did not err by sentencing within the Guidelines range for crack offenses, notwithstanding defendants' arguments that the ratio gave rise to unreasonable sentences. See United States v. Gipson, 425 F.3d 335, 337 rehearing denied, 431 F.3d 993 (7th Cir.2005); United States v. Cawthorn, 429 F.3d 793, 802-03 (8th Cir.2005). In the last few months, the First, Fourth, and Eleventh Circuits have reached the specific question presented here. See United States v. Pho, 433 F.3d 53 (1st Cir.2006); United States v. Eura, 440 F.3d 625 (4th Cir.2006); United States v. Williams, 456 F.3d 1353 (11th Cir.2006). All have ruled that a district court errs when it substitutes its own ratio for the 100:1 ratio set forth in the Guidelines and approved by Congress, a conclusion we now join.

Preliminarily, we must decide whether the district court erred in imposing a non-Guidelines sentence after granting safety valve relief, notwithstanding the requirement in the safety valve statute that "the court shall impose" a Guidelines sentence if the criteria for safety valve relief are met. 18 U.S.C. § 3553(f) (emphasis added). In Crosby, we identified this provision as one that Booker did not explicitly excise (because it was not at issue in that case) but nonetheless left in jeopardy, given the reasons that led the Booker remedial majority to excise the mandatory application of the Guidelines in § 3553(b)(1). See Crosby, 397 F.3d at 110 n. 8 (identifying § 3553(b)(2), § 3553(e), and § 3553(f) as subsections that require imposition of Guidelines sentences in certain situations but that Booker did not explicitly address). Subsequently, in United States v. Selioutsky, 409 F.3d 114 (2d Cir.2005), we held that the logic of Booker applied to § 3553(b)(2). We therefore excised the language of mandatory imposition of a Guidelines sentence in that subsection (which concerns sentences for child crimes and sexual offenses), making the Guidelines range applicable to that subsection advisory. Id. at 117. More recently, in United States v. Holguin, 436 F.3d 111 (2d Cir.2006), we considered but ultimately did not need to resolve the question of whether this logic also applied to § 3553(f). Id. at 116.

While neither party raises the issue here, we find it squarely presented by the facts of the case, and we hereby resolve it in the only sensible way: Following the logic of Booker and Selioutsky, we hold that § 3553(f) does not require the imposition of a Guidelines sentence if the district court finds the defendant eligible for safety valve relief. Defendants eligible for safety valve relief may accordingly avoid being sentenced under statutory mandatory minimums and may instead receive the benefit of the advisory Guidelines regime.

In the post-Booker regime, we review a district court's sentence for "reasonableness." See Crosby, 397 F.3d at 114. Reasonableness is defined not only by the length of the sentence but also by the process the district court used to determine the sentence; in other words, a sentence must be substantively reasonable as well as procedurally reasonable. Id. at 114-15. As to the latter, "[i]f a sentencing judge committed a procedural error by selecting a sentence in violation of applicable law, and that error is not harmless and is properly preserved or available for review under plain error analysis . . ., the sentence will not be found reasonable." Id. at 114. The question for us is whether the district court's imposition of a sentence that rejected the 100:1 ratio purely on generalized policy grounds, rather than on the basis of factors specific to Castillo, satisfies this standard of procedural reasonableness.

District court sentencing after Booker centers around 18 U.S.C. § 3553(a), which calls on the district court to "impose a sentence sufficient, but not greater than necessary, to comply with the purposes set forth in paragraph (2) of this subsection" and to "consider" the following factors "in determining the particular sentence to be imposed":

(1) the nature and circumstances of the offense and the history and characteristics of the defendant;

(2) the need for the sentence imposed —

(A) to reflect the seriousness of the offense, to promote respect for the law, and to provide just punishment for the offense;

(B) to afford adequate deterrence to criminal conduct;

(C) to protect the public from further crimes of the defendant; and

(D) to provide the defendant with needed educational or vocational training, medical care, or other correctional treatment in the most effective manner;

(3) the kinds of sentences available;

(4) the kinds of sentence and the sentencing range established for —

(A) the applicable category of offense committed by the applicable category of defendant as set forth in the guidelines . . . issued by the Sentencing Commission . . . and . . . that . . . are in effect on the date the defendant is sentenced . . .

(5) any pertinent policy statement —

(A) issued by the Sentencing Commission . . . and . . . that . . . is in effect on the date the defendant is sentenced;

(6) the need to avoid unwarranted sentence disparities among defendants with similar records who have been found guilty of similar conduct; and

(7) the need to provide restitution to any victims of the offense.

U.S.C. § 3553(a); see Crosby, 397 F.3d at 111-15 (describing process of post-Booker sentencing).

Castillo relies on several arguments under this framework to justify the reasonableness of the district court's sentence. All are unavailing.

According to Castillo, because Booker made the Guidelines only advisory, virtually everything about the Guidelines is open to fresh interpretation. All Booker requires, says Castillo, is that the Guidelines be "considered" under § 3553(a)(4), and if the district court considers the Guidelines range and then ultimately imposes a different sentence in light of the other § 3553(a) factors, Booker is satisfied, regardless of the steps the district court took to create that non-Guidelines sentence. In Castillo's view, then, it does not matter whether the court reached its conclusion on the basis of a generalized policy disagreement with the Guidelines or on factors specific to the particular defendant and his offense conduct.

We think this argument misconstrues Booker's treatment of the Guidelines. Booker's central concern was with the mandatory application of the Guidelines. Nothing in Booker specifically authorizes district judges to rewrite different Guidelines with which they generally disagree, which is effectively what district judges do when they calculate a sentence with a 20:1 or 10:1 ratio instead of the 100:1 ratio in the drug sentencing table. Instead, the focus of the Booker remedy opinion is on allowing district judges the flexibility to tailor sentences for each individual defendant against the backdrop of the Guidelines scheme as approved by Congress. See Booker, 543 U.S. at 264-65, 125 S. Ct. 738 (Breyer, J.) (explaining that the "features of the remaining system, while not the system Congress enacted, nonetheless continue to move sentencing in Congress' preferred direction, helping to avoid excessive sentencing disparities while maintaining flexibility sufficient to individualize sentences where necessary") (emphasis added); Crosby, 397 F.3d at 113-14 (noting that Booker does not permit district judges to "return to the sentencing regime that existed before 1987 and exercise unfettered discretion to select any sentence within the applicable statutory maximum and minimum"; instead, "the Supreme Court expects sentencing judges faithfully to discharge their statutory obligation to `consider' the Guidelines and all of the other factors listed in section 3553(a). We have every confidence that the judges of this Circuit will do so, and that the resulting sentences will continue to substantially reduce unwarranted disparities while now achieving somewhat more individualizedjustice.") (emphasis added).

This distinction between policy decisions as embedded in the Guidelines and judicial decisions as imposed on a case-by-case basis is at work in the language of § 3553(a), which Booker left in place. In § 3553(a)(2), the district court is instructed to consider "the need for the sentence imposed to reflect the seriousness of the offense," while in § 3553(a)(4), the district court is instructed to consider the sentencing range for "the applicable category of offense" (emphasis added) as set forth in the Guidelines. The differing language between § 3553(a)(2) and § 3553(a)(4) reflects the difference between one particular defendant's crime and the larger genre of offenses into which it falls. Indeed, contrasting the language of § 3553(a)(2) with the language of § 3553(a)(4) clearly indicates that the Sentencing Commission and district courts have two different roles with respect to the Guidelines. Section 3553(a)(4) is the domain of the Sentencing Commission, whose task is to determine a sentencing range for the category of crime — here, crack as opposed to powder cocaine offenses.5 In contrast, § 3553(a)(2) is the domain of the district court, whose job is to consider, given the seriousness of the category of the crime as reflected in the Guidelines sentencing range, where the defendant's particular offense fits with respect to that range — that is, whether it falls within the range (and, if so, where), or whether a non-Guidelines sentence is appropriate. Cf. United States v. Anati, 457 F.3d 233, 237-38 (2d Cir. 2006) (contrasting § 3553(a)(2) with § 3553(a)(4) and suggesting that "it would be doubtful if a judge could enhance [a sentence] because of a personal view as to how much more serious the category of heroin offenses is than the category of cocaine offenses" because this would "ignore the Commission's precisely calibrated assessment" as to the relative harms of those offenses, but that "perhaps a judge may consider that the particular circumstances under which the defendant commits a heroin offense deserved an enhanced sentence. . . .").

This interpretation of Booker and § 3553(a) comports with a traditional understanding of the separation of powers. "In our system, so far at least as concerns the federal powers, defining crimes and fixing penalties are legislative, not judicial, functions." United States v. Evans, 333 U.S. 483, 486, 68 S. Ct. 634, 92 L. Ed. 823 (1948). In contrast, judges, who deal with individual cases and controversies, impose particularized sentences on a specific defendant. While Congress has delegated its authority to fix penalties to the Sentencing Commission (whose Guidelines can then be accepted, modified, or rejected by Congress), and while judges may not mandatorily apply the Guidelines, nothing in Booker empowers judges to define penalties for categories of crimes. See also Pho, 433 F.3d at 61-62 (holding that judges' "newfound discretion" after Booker did not change the traditional "distribution of authority over sentencing policy" and that district courts may "exercise discretion in fashioning sentences . . . only within the ambit of the individualized factors spelled out in section 3553(a)"); Eura, 440 F.3d at 633 (envisioning the range of ratios and the multiplicity of different mechanisms to reach those ratios that district courts might adopt, and explaining that "[t]hese scenarios tell us that sentencing courts should not be in the business of making legislative judgments concerning crack cocaine and powder cocaine").

To be sure, when sentencing a defendant for a crack-related crime — as with any other crime — a district court judge may consider "the judge's own sense of what is a fair and just sentence under all the circumstances." United States v. Jones, 460 F.3d 191, No. 05-2289-cr, 2006 WL 2167171, at *3, 2006 U.S.App. LEXIS 19789, at *9 (2d Cir. Aug. 2, 2006); see also id. (describing such individualized judgment as part of "the historic role of sentencing judges"); cf. United States v. Rattoballi, 452 F.3d 127, 133 (2d Cir.2006) (emphasizing importance at sentencing of considering factors "personal to a particular defendant"). But nothing in Booker suggests that it is the task of district court judges to pronounce broad policy choices rather than specific sentences based on the specific facts of a case.

Castillo also attempts to rely on § 3553(a)(6), which instructs district courts to consider "the need to avoid unwarranted sentence disparities among defendants with similar records who have been found guilty of similar conduct" in determining a defendant's sentence. Castillo argues that this provision requires the district court to impose a non-Guidelines sentence when it believes that a Guidelines sentence would produce an unwarranted disparity. Setting aside the issue of whether the obligation to impose a non-Guidelines sentence also includes the authority to revise the Guidelines by supplying a new ratio — which, as we have just explained, cannot be the case — we note that courts do not operate in a vacuum in determining what constitutes an "unwarranted sentence disparity" under this provision of the statute.

Just as with any statute, the role of the judiciary is to determine what Congress meant by this statutory phrase. We have no authority to substitute our policy preferences for that of the legislative branch. Rather, our interpretation must be faithful to Congress's meaning as embedded in the words of its statute. While the 100:1 ratio clearly produces a disparity, it is one that Congress has mandated, one that Congress has continually refused to alter, despite the Sentencing Commission's various proposals for eliminating or reducing the ratio. "[A] sentencing disparity intended by Congress is not unwarranted." United States v. Duhon, 440 F.3d 711, 720 (5th Cir.2006) (holding that sentencing disparity between co-defendants created by departures for substantial assistance was intended by Congress and may therefore not be rejected under § 3553(a)(6)). Moreover, § 3553(a)(6) cautions against unwarranted disparities only to the extent that they stem from different sentences given to "defendants with similar records who have been found guilty of similar conduct," and Congress has time and again clarified that in its view crack and powder cocaine offenses are not "similar conduct." No amount of citation to the Sentencing Commission's reports that are critical of the differential treatment of crack and powder cocaine can hide that fact, and in the end, we are constrained by the public policy choice Congress has made. Given this history, the district court erred by relying on its own sense that the types of offenses are similar. See Pho, 433 F.3d at 64 ("Although the district court's approach does ameliorate the disparity in sentences for crack and powdered cocaine offenses . . . Congress intended that particular disparity to exist, and federal courts are not free to second-guess that type of decision.").

Castillo further argues that the 100:1 ratio is not binding because it is not statutory, and therefore Congress's intentions are not the key consideration. See, e.g., Fisher, 2005 WL 2542916, at *9, 2005 U.S. Dist. LEXIS 23184, at *27 (distinguishing between 100:1 ratio in mandatory minimum, where it is binding, and in Guidelines, where it is not). Again, this argument confuses the newly advisory nature of the Guidelines with the malleability of its provisions. There is certainly no dispute that, after Booker, district courts need not sentence within the range produced by the 100:1 ratio in the Guidelines, as long as the specific facts of the case when considered in light of § 3553(a) justify it. This freedom does not, however, mean that a court may impose a different ratio as a policy matter.

It is also worth emphasizing that the ratio in the Guidelines stems from and is closely pegged to the statutory ratio, making the statutory mandatory minimum for each drug's trigger amount the bottom of the Guidelines range for that quantity. In other words, Guidelines level 26 is the base offense level for the five-year mandatory minimum trigger, and produces a range that starts close to 60 months, while Guidelines level 32 is the base offense level for the ten-year mandatory minimum trigger, and produces a range that starts close to 120 months. See U.S.S.G. § 2D1.1, comment (Background). Each incremental increase in range and quantity is "proportional to the drug levels established by statute." Id. The imposition of the ratio in the Guidelines was not plucked out of thin air, then, without regard to any statutory basis. In fact, Congress expressed its reluctance, through the House Committee on the Judiciary's Report to the 1995 bill that rejected the Commission's initial attempt to eliminate the disparity, to lowering the ratio without making corresponding changes to the mandatory minimums:

It is important to note that if the Commission's guideline amendments went into effect without Congress lowering the current statutory minimum penalties, it would create gross sentencing disparities. Sentences just below the statutory minimum would be drastically reduced, but mandatory minimums would remain much higher. For example, an offender convicted of distributing 5 grams of crack would, under the statutory mandatory minimum penalty, face a mandatory prison term of 5 years; however, an offender convicted of distributing 4.9 grams of crack could, under the Commission's amendment to the guidelines, receive a sentence within a range of 0-6 months of imprisonment. The Commission's crack-related guideline amendments would establish penalties for crimes that stand in sharp contrast with statutory mandatory minimum penalties.

H.R.Rep. No. 104-272 at 3-5 (1995), reprinted in 1995 U.S.C.C.A.N. 335, 337.6 To be sure, some disparities will always exist in a post-Booker world where non-Guidelines sentences are available. But this is a clear statement of Congressional belief that changing the Guidelines ratio without changing the mandatory minimums would result in an unwarranted disparity, while retaining the ratio at 100:1 would not.

We note additionally that the district court's approach, writ large, would tend to produce greater disparities judge-by-judge around the country, which would surely not serve the purpose of § 3553(a)(6). See United States v. Joyner, 924 F.2d 454, 460 (2d Cir.1991) (explaining that the purpose of § 3553(a)(6) was to "eliminate unwarranted disparities nationwide"). Some judges would likely continue to apply the 100:1 ratio; others would apply the 20:1 ratio most recently proposed by the Commission; still others might use the 10:1 ratio offered by the Clinton administration, or the 5:1 ratio earlier favored by the Commission, or eliminate the ratio entirely on the theory that any disparity is unwarranted. There are also several different paths to adjusting the ratio: the triggering levels for both crack and powder cocaine could be modified, the former raised and the latter lowered; or the triggering level for crack could be raised to the higher amount for powder cocaine; or the amount for powder cocaine could be lowered to the small amounts for crack. See, e.g., Eura, 440 F.3d at 633; Pho, 433 F.3d at 63; Tabor, 365 F.Supp.2d at 1060-61. Depending on different judges' policy preferences about the relative harms of each form of the drug and of drug crimes in general, outcomes could differ widely.7

Castillo attempts to answer this concern with reference to the statement in Booker that "[w]e cannot and do not claim that use of a `reasonableness' standard will provide the uniformity that Congress originally sought to secure." Booker, 543 U.S. at 263, 125 S. Ct. 738 (Breyer, J.). Castillo reads into this statement the idea that because less uniformity in a post-Booker world is to be expected, the fact that different defendants will be sentenced under different ratios based on individual judges' beliefs about what constitute unwarranted sentencing disparities is an acceptable result. But even if less uniformity is to be expected, that does not mean that it should result from individual judges' revisions to the Guidelines en route to imposing non-Guidelines sentences rather than from individual judges' particularized assessments of specific defendants. Further, it ignores the fact that Booker crafted the remedy it did as the best way of protecting Congress's sentencing goals in light of the Court's constitutional holding. See Booker, 543 U.S. at 253-54, 263-65, 125 S. Ct. 738 (Breyer, J.). That some additional lack of uniformity is to be expected does not mean that efforts towards uniformity should be abandoned.

Indeed, if we uphold the practice of district courts' applying different ratios here, would that not provide a rationale, contrary to Congress's intent, for courts to revise other parts of the Guidelines with which they disagree? The entire drug quantity table could be effectively discarded as courts express differing opinions as to the relative harmfulness of different drugs. See Tabor, 365 F.Supp.2d at 1061 and n. 15 (expressing view that Guidelines range for methamphetamine trafficking could with good reason be increased). Similarly, a court might take issue with the loss amount table for economic crimes and adjust the relevant ranges. See U.S.S.G. § 2B1.1. Or, as one judge has explained, because of the statutory framework Congress has created for the two distinct crimes of (1) carrying a concealed dangerous weapon on an aircraft, 49 U.S.C. § 46505, and (2) possessing a firearm or other dangerous weapon in a federal facility other than a federal court, 18 U.S.C. § 930, and because of the Guidelines ranges that track those statutory provisions,

[a] person carrying a concealed dangerous weapon faces a significantly greater advisory Guidelines sentence (relatively speaking) if he is on an airplane rather than, say, in the U.S. Capitol. Would it be proper for a sentencing judge to conclude that section 3553(a)(6) authorizes judicial correction of this "disparity" based solely on a judicial conclusion that the two forms of conduct are "similar," and notwithstanding the clear expression by Congress in setting the statutory maximums that the two offenses are dissimilar (i.e., that an airplane offense is more serious than a federal-facility offense)? Certainly not. It is equally improper for federal judges categorically to correct what they, perhaps correctly, perceive to be a disparity between crack and powder cocaine sentences when Congress has to date concluded otherwise.

Doe, 412 F.Supp.2d at 95-96.

It is true that, in contrast to the ranges for methamphetamine, loss amount, and concealed dangerous weapons offenses, the Sentencing Commission has specifically and repeatedly spoken about what it perceives to be the unwarranted disparity between crack and powder cocaine offenses. But the Commission gets its authority only from Congress. Congress is the ultimate arbiter of federal sentencing policy, and Congress has specifically and repeatedly failed to act on the Commission's recommendations. Against this background, we cannot say that the 100:1 ratio produces an unwarranted disparity within the meaning of § 3553(a)(6).

Both Castillo and the government call our attention to other subsections of § 3553(a). But nothing in these provisions allows judges to reach a sentence by revising the Guidelines structure itself instead of specifically considering the facts of a defendant's case against the background of the Guidelines as written. We have already explained that provisions such as § 3553(a)(2)(A) refer to the seriousness of a particular set of factual circumstances, not to the seriousness of the broad category of offenses as a general policy matter, which is addressed by the Sentencing Guidelines under § 3553(a)(4)(A).

With respect to the particular sentence imposed in this case, the district court made no attempt to apply any of the case-specific factors from §§ 3553(a)(1) or (a)(2). The district court did not say anything about why the sentence imposed would "provide just punishment for the offense" or why it was justified by "the nature and circumstances of the offense" under § 3553(a)(2)(A). The district court did not suggest that "the history and characteristics of the defendant" — whom the district court acknowledged had a fairly happy and stable childhood — were a mitigating factor that should result in a lower sentence under § 3553(a)(1), nor why this sentence would "afford adequate deterrence to criminal conduct" under § 3553(a)(2)(B). Similarly, the district court was silent as to how this punishment was calculated "to protect the public from further crimes of the defendant" under § 3553(a)(2)(C). There was simply nothing in the district court's opinion to justify the sentence other than the district court's policy disagreement with the Guidelines and its erroneous reliance on reducing "disparity" under § 3553(a)(6). This was not enough to sustain the sentence.

Castillo expresses concern that the approach we adopt gives undue weight to the Guidelines. But requiring the judge to reach a sentence by first considering the Guidelines range under the Guidelines as actually written does not require the judge to give such undue weight. Of course we do not mean to suggest that a non-Guidelines sentence is never permissible in cases involving crack cocaine. "`[C]onsideration' does not mean mandatory adherance." Jones, 2006 WL 2167171, at *2, 2006 U.S.App. LEXIS 19789, at *7. After Booker, judges may ultimately reject a sentence within the Guideline range if that rejection is based on all of the § 3553(a) factors, specifically considered in light of the facts of the particular defendant's case. As the Eleventh Circuit recently explained, "a sentence below the Guidelines range may be reasonable, so long as it reflects the individualized, case-specific factors in § 3553(a)." Williams, 456 F.3d at 1369. But without any justification for why the § 3553(a) factors lead to a below-Guidelines sentence, and with the non-Guidelines sentence based only on the district court's generalized policy disagreement with the Guidelines, the sentence cannot be affirmed as "reasonable."

With respect to the central issue in this case — the relative merits of the ratio for crack and powder cocaine — we are, as we have stated, without license to usurp the policy role of the legislative and executive branches. That said, we would be blind to the thoughtful policy discussions of the last dozen years if we did not acknowledge what our survey of Sentencing Commission reports and recommendations, as well as various legislative proposals across the political spectrum, reveals: that the district court is surely not alone in its concern that the current ratio is too great.

Yet what that ratio should be — and indeed, any change, if it is to come — can result only from legislative direction. For the foregoing reasons, we are compelled to conclude that we see nothing in § 3553(a) or in Booker more generally that authorizes district courts to sentence defendants for offenses involving crack cocaine under a ratio different from that provided in the Sentencing Guidelines. That is not to say that district courts must always sentence within the ratio provided by the Guidelines; that would indeed be error under Booker. But we join the First, Fourth, and Eleventh Circuits in holding that district courts may give non-Guidelines sentences only because of case-specific applications of the § 3553(a) factors, not based on policy disagreements with the disparity that the Guidelines for crack and powder cocaine create. See Pho, 433 F.3d at 64-65; Eura, 440 F.3d at 633-34; Williams, 456 F.3d at 1366-67. We therefore reverse and remand for further proceedings.