dissenting:
The majority gleans from the record in this case the assurance that the district court would have imposed the same term of supervised release regardless of whether the § 2K2.1(b)(5) enhancement applies in calculating Blackburn’s Guidelines sentence. The record is more equivocal than the majority claims, however, ánd the Supreme Court’s ruling in United States v. Booker, 543 U.S. 220, 125 S.Ct. 738, 160 L.Ed.2d 621 (2005), and our cases applying it have made clear that we should not speculate as to what a district court may do in resentencing a defendant except in very limited circumstances that are not present here. I find no clear indication in the record of the sentencing proceeding that the district court would adhere to the same sentence if we remanded this case. I would conclude, therefore, that any error in calculating Blackburn’s Guidelines sentence was not harmless and that Blackburn’s claim is not moot. Because I would reach the merits of this appeal, I respectfully dissent.
The majority claims, in dicta, that the Supreme Court’s ruling in United States v. Johnson, 529 U.S. 53, 120 S.Ct. 1114, 146 L.Ed.2d 39 (2000), “suggests that to reduce Blackburn’s term of supervised release as a means of offsetting excess prison time would disserve the objectives of supervised release.” The majority misconstrues Johnson, and I write on this issue because I do not want the majority’s dicta to create an improper suggestion to district courts. The Court’s statement in Johnson that “[t]he objectives of supervised release would be unfulfilled if excess prison time were to offset and reduce terms of supervised release” was made in the context of its conclusion that incarceration and supervised release are not interchangeable where the supervised-release term is mandatory.1 See 539 U.S. at 59-60, 123 S.Ct. 2142. Johnson stands only for the limited proposition that supervised release commences upon a defendant’s release from prison and that any reduction in a defendant’s mandatory term of supervised release for excess time served results not from the automatic operation of statute, but solely from the sentencing court’s exercise of discretion under 28 U.S.C. § 3583(e) to reduce supervised-release terms.2 Id. at 60, 123 S.Ct. 2142.
In United States v. Barresi, 361 F.3d 666 (2d Cir.2004), we explicitly rejected the reading of Johnson advanced by the majority here — that Johnson sweeps more broadly than its holding. We acknowledged that “if the three-year supervised-release term imposed on Barresi had been man*267dated by statute, the district court would have been correct that Barresi’s supervised-release term could not be reduced [at resentencing] to compensate for extra time spent in prison.” Id. at 674 (citing Johnson, 529 U.S. at 54-56, 120 S.Ct. 1114). We nevertheless remanded the case for resen-tencing although Barresi had already completed his prison term because he was serving a non-mandatory term of supervised release that the district court had discretion to reduce. Id. at 675. We instructed the district court to determine first what sentence of incarceration it would have imposed if it had not considered impermissible factors in upwardly departing from the Guidelines sentence. We also directed that if the district court would have imposed a prison term less than the sentence actually imposed, it “should then consider whether it wishes to exercise its discretion to reduce Barresi’s supervised-release term ... in order to compensate for the fact that Barresi completed his ... prison term before his re-sentencing took place.” Id. Barresi thus stands unequivocally for the proposition that, when resentencing a defendant who has completed his or her term of incarceration, a district court may compensate for any excess time the defendant was incarcerated by exercising its discretion to reduce the defendant’s non-mandatory term of supervised release.3 Barresi also makes clear that a district court may compensate a defendant who has served excess prison time and is serving a non-mandatory supervised-release term by either re-sentencing the defendant or exercising its authority under § 3583(e)(1).
This case is substantially similar to Bar-resi. The statute governing Blackburn’s sentence, 18 U.S.C. § 924(a)(2), does not provide a mandatory minimum term of supervised release. Because the maximum term of incarceration under § 924(a)(2) is ten years, Blackburn’s offense was a Class C felony, 18 U.S.C. § 3559(a)(3). The longest term of supervised release that the district court could have imposed was three years, 18 U.S.C. § 3583(b)(2), but the court could also have imposed a lesser term or no term at all under the post-Booker sentencing regime, and could do so now. If we were we to reach the merits and hold that the four-level enhancement under U.S.S.G. § 2K2.1(b)(5) does not apply in this case, under Barresi (and Johnson), the district court could use its discretion to reduce Blackburn’s term of supervised release to compensate him for any excess time he served in prison. Cf. Levine v. Apker, 455 F.3d 71, 76 (2d Cir.2006) (holding that a habeas petitioner’s challenge to a policy *268and rule of the Bureau of Prisons regarding the maximum time a prisoner could serve in a community corrections center was not moot where the petitioner had completed his prison term and was serving a term of supervised release because the district court “might,” on remand, “modify the length of [petitioner’s] supervised release”).
The majority acknowledges that the district court could reduce Blackburn’s supervised-release term at resentencing under Johnson and Barresi, and we therefore agree that challenges such as Blackburn’s are not always moot.4 It contends, however, that the record is clear that the district court would have imposed the maximum term of supervised release regardless of the applicability of the § 2K2.1(b)(5) enhancement. Believing that it is unlikely that the district court would reduce Blackburn’s supervised-release term on remand, the majority concludes that Blackburn’s claim is moot. I disagree with both of the majority’s conclusions: that the concept of mootness, rather than harmless error, should guide our analysis and that any sentencing error did not affect Blackburn’s sentence.
Given the majority’s acknowledgment that the district court could resentence Blackburn to a reduced supervised-release term, its ultimate conclusion should be not that the case is moot, but that any error in the Guidelines calculation was harmless. This is effectively the analysis the majority undertakes in combing the record of the sentencing proceeding, and it is the analysis we have found appropriate in considering Booker errors, see, e.g., United States v. Lake, 419 F.3d 111, 113-14 (2d Cir. 2005); United States v. Sharpley, 399 F.3d 123, 127 (2d Cir.2005). The majority, however, conflates its determination that any sentencing error in this case did not affect the proceedings — a harmless-error analysis — with the conclusion that relief is so remote and speculative as to make the case moot. In doing so, it ignores important distinctions between direct and collateral consequences of a conviction, on the one hand, and relief that is unlikely as opposed to remote, on the other.
In Spencer v. Kemna, 523 U.S. 1, 118 S.Ct. 978, 140 L.Ed.2d 43 (1998), the Supreme Court concluded that a habeas petitioner’s challenge to an order revoking his parole was moot where the petitioner completed the entire term of imprisonment underlying the parole revocation before the district court ruled on his habeas petition and could not establish a “collateral consequence” of his conviction. Id. at 7, 14-16. The Court rejected as too speculative several purported collateral consequences that involved two contingent events: (1) whether the subsequent collateral proceeding would ever occur, and (2) whether the revocation would be used against the defendant in that proceeding.5 *269Id. at 15-16. Spencer does not control the resolution of this case because Blackburn’s entire sentence has not expired and his supervised-release term is a direct consequence of his conviction. Moreover, the potential relief here is also direct, rather than contingent or collateral, because if we disagreed with the district court’s Guidelines ruling, we would remand for resen-tencing. The majority, however, appears to import implicitly the analysis of Spencer into this case by claiming that a ruling favorable to Blackburn on remand is “so remote and speculative” as to make the case moot. While the majority could take the reasonable, albeit erroneous, position that relief on remand is so unlikely as to make any error harmless, relief is hardly “remote” or “speculative” here, where a direct consequence of Blackburn’s conviction is at issue.
In short, the restrictions imposed by the terms of Blackburn’s supervised-release term constitute a concrete injury, caused by his conviction and sentence, and re-dressable (at least in part) by a resentenc-ing proceeding. See Spencer, 523 U.S. at 7, 118 S.Ct. 978. The majority therefore errs in determining that Blackburn’s claim is moot rather than concluding that any error was harmless. In any event, the record does not support the view that the district court’s Guidelines calculation, if erroneous, was harmless.
Although the district court noted the disagreement among the courts of appeals as to the applicability of § 2K2.1(b)(5) and stated that it gave Blackburn “credit” at sentencing for “some close legal issues,” the district court nonetheless applied the enhancement in calculating Blackburn’s offense level, which exposed Blackburn to a Guidelines range of 46 to 57 months’ imprisonment. The district court sentenced Blackburn to 37 months’ imprisonment— the top of the Guidelines range that would have applied without the four-level enhancement. But we should not conclude that the district court engaged in a disingenuous sleight of hand by applying the enhancement in its Guidelines calculation only to remove the enhancement in its post-Booker sentence. Although the sentence imposed may reflect the district court’s misgivings or uncertainty about the enhancement, the court nonetheless included the enhancement in its Guidelines calculation. The most reasonable inference from the sentencing transcript, in my view, is that the district court gave Blackburn some credit for what it perceived as a “close legal issue[ ],” but that the nine-month departure from Blackburn’s Guidelines range cannot be completely attributed to this issue. The district court clearly stated on the record that “[tjhere have been some close legal issues that I think needed to be addressed, that I give you credit for.” The *270district court thus indicated that several factors figured into its decision to impose a non-Guidelines sentence of 37 months. I take the district court at its word that the § 2K2.1(b)(5) enhancement was not the sole factor influencing its decision to impose a non-Guidelines sentence. It therefore appears possible — whether likely or not — that the district court may have sentenced Blackburn to a lesser term of incarceration had it not imposed the enhancement and that Blackburn’s sentence would have been “nontrivially” lower than 37 months. See United States v. Crosby, 397 F.3d 103, 118 (2d Cir.2005). The majority declines to challenge this reading of the sentencing transcript, although it acknowledges that Blackburn’s sentence reflects “the uncertain applicability of the enhancement.” At the very least, the sentencing transcript does not give us adequate assurance that the district court would have imposed the same prison term regardless of § 2K2.1(b)(5)’s applicability. It is this lack of assurance with respect to Blackburn’s prison term that makes this appeal not moot because, as we held in Barresi, the district court on remand could exercise its discretion to reduce Blackburn’s supervised-release term to compensate for any excess prison time.
Nor does the record reflect the district court’s unambiguous intent to impose a three-year term of supervised release notwithstanding a potential error in calculating Blackburn’s Guidelines range. The majority divines from the district court’s concern that Blackburn change his lifestyle a sufficiently clear expression of intent that it would not reduce Blackburn’s term of supervised release to compensate for the fact that Blackburn completed his prison sentence before resentencing could occur. This conclusion is wholly unwarranted. Our recent Sixth Amendment jurisprudence teaches us that it is not the role of appellate courts to predict what a district court may do when resentencing a defendant. As we noted in Crosby, “[p]er-haps in some cases an appellate court could make an educated guess as to the likely outcome of a remand [for resentenc-ing], but that guess might be wrong, absent a clear indication at the original sentencing supporting the inference that the same sentence would have been imposed ...,” 397 F.3d at 117-18.
I acknowledge that a sentencing error would be harmless, and Blackburn’s claim might be moot, if the district court had sentenced Blackburn to a mandatory minimum sentence. See Sharpley, 399 F.3d at 126-27 (finding district court’s use of the Guidelines as a mandatory regime harmless error because the court imposed the mandatory minimum sentence). It is less clear that a sentencing error would be harmless, or the claim moot, if the district court, after considering the factors set forth at 18 U.S.C. § 3553(a), made clear that it would impose the same prison sentence or term of supervised release regardless of the propriety of its Guidelines calculation. See United States v. Bah, 439 F.3d 423, 431 (8th Cir.2006) (declining to find a post-Booker Guidelines calculation error harmless in the light of the district court’s responsibility to determine the appropriate Guidelines range even though the district court announced an identical alternative sentence that was not based on an alternative Guidelines calculation but was intended to cover any potential calculation errors); cf. Crosby, 397 F.3d at 111-12 (noting that “to fulfill th[e] statutory duty to ‘consider’ the Guidelines, a sentencing judge will normally have to determine the applicable Guidelines range” except in some circumstances involving difficult factual determinations, such as complex loss calculations, or unclear policy statements regarding the availability of a departure); United States v. Rattoballi, 452 F.3d 127, 132-33 *271(2d Cir.2006) (“[T]he guidelines cannot be called just ‘another factor’ in the statutory list, 18 U.S.C. § 3553(a), because they are the only integration of the multiple factors and, with important exceptions, their calculations were based upon the actual sentences of many judges.” (internal quotation marks omitted)). Here, the majority acknowledges that the district court never stated that it would impose the same sentence regardless of the applicability of the § 2K2.1(b)(5) enhancement. It nevertheless goes on to make the very “educated guess” we found objectionable in Crosby in concluding not that any error was harmless, but that Blackburn’s claim is moot. I do not think that the district court expressed its intentions so clearly “as to obviate the need for a remand for any clarification.”
As we have noted in the Sixth Amendment context, “it will usually not be easy to divine with certainty that the sentencing judge would have imposed the same sentence” notwithstanding error. Lake, 419 F.3d at 113. I do not think this is one of the “ ‘rare’ case[s] ... where we can confidently say that a [potential] sentencing error was harmless.”6 Id. at 114. “The applicable guideline range provides the frame of reference against which the judge chooses an appropriate sentence,” id., and I cannot conclude that the range identified by the district court, if erroneous, did not affect Blackburn’s sentence. Instead, because this case is not moot in the light of Barresi and the sentencing transcript gives us no clear indication that the district court would have imposed the same prison term and the same period of supervised release regardless of any error in calculating Blackburn’s Guidelines range, I would conclude that any sentencing error was not harmless.
Because I think we should reach the merits of Blackburn’s appeal, I respectfully dissent from the dismissal of this appeal.
. The gist of the Court's analysis in Johnson was that where Congress has mandated a particular term of supervised release, which serves ends distinct from those served by incarceration, Congress's objective would be unfulfilled if a court used excess prison time to offset a supervised-release term. Nothing in Johnson speaks to a district court's determination of what discretionary term of supervised release is appropriate in a particular case.
. The Court noted that even where the term of supervised release is mandated by statute, a district court “may terminate an individual's supervised release obligations ‘at any time after the expiration of one year of supervised release.”' 529 U.S. at 60, 120 S.Ct. 1114 (quoting 18 U.S.C. § 3583(e)(1)).
. Nothing in Barresi is inconsistent with the holding of United States v. Mercurris, 192 F.3d 290 (2d Cir.1999). In Mercurris, although the defendant had been sentenced to a term of supervised release following his incarceration, he was deported upon completion of his prison term. Id. at 293. We found the defendant's challenge to the length of his sentence moot in the light of his release from prison because, inter alia, he had "only a quixotic chance of legally returning to the United States.” Id. at 294. Mercurris thus stands for the limited proposition that a defendant who is deported following his or her release from prison and who faces a remote chance of ever being subject to the supervised-release portion of his or her sentence cannot challenge the length of his or her sentence. See United States v. DeLeon, 444 F.3d 41, 55-56 (1st Cir.2006) (citing Mercurris for the proposition that where a defendant is sentenced to a term of supervised release but deported upon completion of a term of incarceration, a claim of sentencing error is moot in the light of the remote chance that he or she will return legally to the United States and be subject to supervised release); cf. United States v. Lares-Meraz, 452 F.3d 352, 354-56 (5th Cir.2006) (finding that the Booker challenge of a defendant who was sentenced to a supervised-release term but deported upon completion of his incarceration was not moot, but that the error was harmless).
. Every federal court of appeals that has considered the issue has held that a challenge to the length of a defendant's sentence is not moot when the defendant has completed a term of incarceration but is serving a non-mandatory term of supervised release. See Johnson v. Pettiford, 442 F.3d 917, 918 (5th Cir.2006); United States v. Allen, 434 F.3d 1166, 1170 (9th Cir.2006); United States v. Larson, 417 F.3d 741, 747 (7th Cir.2005); United States v. Castro-Rocha, 323 F.3d 846, 847 n. 1 (10th Cir.2003); United States v. McCoy, 313 F.3d 561, 564 (D.C.Cir.2002) (en banc); United States v. Molak, 276 F.3d 45, 48-49 (1st Cir.2002); Dawson v. Scott, 50 F.3d 884, 886 n. 2 (11th Cir.1995); see also Jago v. Van Curen, 454 U.S. 14, 21 n. 3, 102 S.Ct. 31, 70 L.Ed.2d 13 (1981) (per curiam) (finding case not moot where petitioner was on parole).
. One purported collateral consequence considered in Spencer — whether the petitioner's revocation could be used against him in a future parole proceeding — was “no longer contingent” because petitioner had been convicted of another crime and was serving a prison term at the time the Supreme Court *269heard his appeal. 523 U.S. at 14, 118 S.Ct. 978. The Court rejected this argument against mootness, noting that the state parole statute gives the parole board almost “unlimited discretion” in determining whether to grant parole release and that a prior parole revocation is therefore “simply one factor, among many” to be considered by the parole board. Id. The potential use of the petitioner’s revocation in a future parole proceeding at issue in Spencer was a collateral consequence because the petitioner's sentence had expired. By contrast, the length of Blackburn's sentence is a direct consequence of his conviction. Moreover, putting this distinction aside, while the Guidelines are advisory post-Booker, the district court's discretion in imposing sentence is cabined by the factors set forth at § 3553(a) and is therefore far from "unlimited.” See United States v. Hamdi, 432 F.3d 115, 118-21 (2d Cir.2005) (finding challenge to length of sentence by defendant who had completed his sentence not moot where collateral consequence was availability of a discretionary waiver of inadmissibility and a lower sentence "would bear on two of the three ... criteria” for the exercise of discretion).
. In Lake, we noted the difficulty in labeling a Booker error harmless given that the Guidelines are now advisory and that district courts must consider all of the factors set forth at 28 U.S.C. § 3553(a) in imposing sentence. 419 F.3d at 114. Here, it is difficult to determine that the district court considered the length of Blackburn's supervised-release term to be of such independent significance that the court would have imposed the same term even if its Guidelines calculation was erroneous and the prison term it imposed was therefore excessive.