with whom Justice Stevens and Justice Ginsburg join, dissenting.
The Court has interpreted a federal sentencing statute in a manner that disadvantages almost 200,000 federal prison*494ers. See Pet. for Cert. 11, and n. 2. It adopts this reading despite the existence of an alternative interpretation that is more consistent with the statute’s text. Absent a clear congressional directive, the statute ought not to be read as the Court reads it. For the Court’s interpretation — an interpretation that in my submission is quite incorrect — imposes tens of thousands of years of additional prison time on federal prisoners according to a mathematical formula they will be unable to understand. And if the only way to call attention to the human implications of this case is to speak in terms of economics, then it should be noted that the Court’s interpretation comes at a cost to the taxpayers of untold millions of dollars. See id., at 11. The interpretation the Court adopts, moreover, will be devastating to the prisoners who have behaved the best and will undermine the purpose of the statute. These considerations, and those stated below, require this respectful dissent.
I
The federal sentencing statute at issue here provides:
“[A] prisoner who is serving a term of imprisonment of more than 1 year[,] other than a term of imprisonment for the duration of the prisoner’s life, may receive credit toward the service of the prisoner’s sentence, beyond the time served, of up to 54 days at the end of each year of the prisoner’s term of imprisonment, beginning at the end of the first year of the term, subject to determination by the Bureau of Prisons that, during that year, the prisoner has displayed exemplary compliance with institutional disciplinary regulations.... [Cjredit for the last year or portion of a year of the term of imprisonment shall be prorated and credited within the last six weeks of the sentence.” 18 U. S. C. § 3624(b)(1) (emphasis added).
According to the Court, the phrase “term of imprisonment” must mean “time actually served” the third time that it ap*495pears in this particular subsection. Ante, at 485. But the Court gives the phrase a different interpretation the first two times it is used in the very same sentence. This in itself indicates that something is quite wrong here.
Petitioners invite the Court to read “term of imprisonment” to mean “the sentence imposed.” Brief for Petitioners i. This, too, seems unworkable. And it can be acknowledged that the Court’s rejection of this interpretation is correct.
The choice, however, is not just between the Court’s reading and that offered by petitioners. There is a third possibility, one more consistent with the statute than either of these two alternatives.
A fair reading of the statute, and a necessary reading to accomplish its purpose best, is to interpret the phrase “term of imprisonment” to refer to the span of time that a prisoner must account for in order to obtain release. The length of the term is set at the outset by the criminal sentence imposed. The prisoner earns release when that term has been fully completed. Most of the term will be satisfied through time spent behind bars. Assuming the prisoner is well behaved, however, he may earn good time credits along the way; and those credits may substitute for actual prison time. Each year of the term comprises a full 365 days, which must be accounted for through a combination of prison time and credits. Thus conceived, a prisoner’s “term” is the administrative period along which progress toward eventual freedom is marked.
Consider the Court’s example of a prisoner subject to a 10-year sentence. See ante, at 477-479. The sentence is divided into ten 365-day segments. Each segment constitutes a year of the term. The prisoner will spend the first 365 days behind bars. In the statute’s words, he has reached “the end of the first year of the term.” Now is the time for credit to be awarded, and he may receive up to 54 days if sufficiently well behaved. Because he has already com*496pleted a full year of his term, those credits go toward completion of the next year. If, based on good behavior, he has earned the maximum of 54 days, he would need another 311 days behind bars before the second year of his term of imprisonment is at an end (because 54 + 311 = 365). If he has earned fewer than 54 days, a longer incarceration will be required to reach 365. Regardless, once the prisoner reaches the end of the second year of his term, he will again be eligible to receive good time credits.
This process repeats itself for the third year of the term, and so on. In the final year of his term (in this example, the 10th segment into which his term has been divided), the prisoner will receive credit in a prorated amount, to be awarded “within the last six weeks of the sentence.” This ensures that the prisoner does not reach the end of year 10, only to find that he has just earned 54 days of credit he no longer needs.
The controlling rule is that each year of the prisoner’s term — each of the 10 administrative segments — comprises 365 days that must be completed through a combination of service and credits. By combining actual prison time with the credits he has earned, a prisoner may complete a particular year of his term in less than 365 calendar days. As a result, credits may enable a well-behaved prisoner to complete his 10-year sentence before 10 calendar years have elapsed. For a 10-year (3,650-day) sentence, a prisoner will serve 3,117 days behind bars if he earns a maximum of approximately 533 credits. This is 63 more days of credit than under the Court’s reading — more than 6 additional credit days for every year of the sentence imposed.
Reading “term of imprisonment” this way is consistent with all parts of the statute. The prisoner receives his credit “at the end of each year of [his] term of imprisonment,” a process that “begin[s] at the end of the first year of the term.” Credit is only awarded if the prisoner has proven well behaved “during that year.” This interpreta*497tion fulfills the “objective of §3624” — rewarding a prisoner for exemplary conduct during the preceding year. See ante, at 482.
This approach also has a textual integrity that the Court’s reading does not: It gives “term of imprisonment” the same meaning each time it is used by the statute. Every time it appears in § 3624(b)(1), “term of imprisonment” refers to the administrative period that a prisoner must complete in order to earn his freedom. The Court, by contrast, would read this phrase to mean “time actually served” the third time it is used, but “the sentence imposed” the first two times it is used (“ ‘a prisoner who is serving a term of imprisonment of more than l ’year[,] other than a term of imprisonment for the duration of the prisoner’s life’ ”). See ante, at 483-485. The Court’s interpretation thus runs afoul of the “ ‘presumption that a given term is used to mean the same thing throughout a statute.’ ” Ante, at 483-484 (quoting Brown v. Gardner, 513 U. S. 115, 118 (1994)). The inconsistency here is particularly egregious because all three uses appear in the same sentence. See id., at 118 (“[The] presumption [is] surely at its most vigorous when a term is repeated within a given sentence”).
The Court responds by noting another part of the statute, a provision stating that prisoners shall receive clothing, money, and transportation “[u]pon the release of [the] prisoner on the expiration of the prisoner’s term of imprisonment.” § 3624(d). A prisoner is released at the end of his actual time behind bars, says the Court, and so “term of imprisonment” must here refer to time actually served. Yet release also comes at the end of a prisoner’s “term” in the sense described above — that is, when the balance of the sentence has been reduced to zero through a combination of prison time and good time credits. Indeed, this administrative use of the phrase fits well with the word “expiration,” which in its most natural sense in this context refers to the close of a formal accounting period. See Black’s Law Die-*498tionary 619 (8th ed. 2004) (“[a] coming to an end; esp., a formal termination on a closing date”). By contrast, it is awkward at best to say, as the Court would have it, that a prisoner’s actual time behind bars is something that “expires.”
The Court’s approach produces yet another oddity. The statute requires that prorated credit be awarded for “the last year or portion of a year of the term of imprisonment.” One might naturally assume that the last year of a 10-year term would be year 10. That is how things work under the approach described above, in which a 10-year sentence is subdivided into 10 administrative segments.
But under the Court’s reading, a prisoner serving a 10-year sentence will never reach year 10 of his term; year 10 simply does not exist. According to the Court, year nine is the final year, and even year nine is not a full year: It lasts “no more than 298 days.” Ante, at 478. If this sounds confusing, it will be all the more so to the prisoner who has just received his sentence and turns to the statute books to figure out when to expect his freedom.
The Court does not even attempt to defend these flaws. Instead, it points to four supposed defects in the approach described above. None withstands examination.
First, the Court notes that the statute requires the release of a prisoner “upon The expiration of the prisoner’s term of imprisonment, less any time credited’ for good behavior.” Ante, at 489 (quoting § 3624(a)). But if “term of imprisonment” truly refers to the entire span that a prisoner must complete to earn his freedom — a period that accounts both for actual time and for good time credits — then why would the “less any time credited” language be appropriate? The answer is that this provision — which appears at the very beginning of the section entitled “Release of a prisoner”— announces to a prisoner when release may be expected: when the prisoner’s term expires, taking into account credit days “as provided in subsection (b).” § 3624(a) (boldface deleted). This use of language is common. A debtor who *499says “I will write a check for what I owe you, less what you owe me” is simply saying “I will pay what I owe, taking into account your debts to me.” Perhaps the same meaning could have been conveyed using different words, but this is hardly probative.
Second, the Court alleges that the above approach conflicts with the statute’s requirement that credit be awarded “at the end of each year” based upon behavior “during that year.” After all, if a year of the term can be satisfied in part through credit, then it may last less than a full calendar year. Yet the statute does not require that credit be awarded at the end of a calendar year for good behavior during a calendar year. What it requires is that credit be awarded “at the end of each year of the prisoner’s term of imprisonment” for good behavior “during that year.” And this is precisely what the above approach does.
Third, the Court frets that, under the approach above, prisoners will earn credit at different rates during a single sentence. It admonishes that “[t]he use of different rates finds no support in the statute.” Ante, at 490. This response is telling. The statute, in fact, prescribes no particular rate — and certainly no formula based on a rate — except as embodied in one clear directive: Prisoners are eligible to earn “up to 54 days at the end of each year of the prisoner’s term of imprisonment.” As to that command, the above approach is perfectly faithful.
Fourth, the Court suggests that the above approach causes credit to vest immediately, contrary to the statute. Again, this is not true. As per the statute, credit only vests “on the date the prisoner is released from custody,” § 3624(b)(2), meaning that it can be revoked at any time before that date. This gives prisoners approaching their release date an extra incentive to behave.
As a fallback, the Court wonders what would happen if a prisoner misbehaved on the final day of his 10-year sentence. Would the Bureau of Prisons (BOP) be forced to “retroac*500tively adjust the duration of all of his [term years] to 365 days”? Ante, at 491. The answer is what one might suppose: A prisoner whose credits are revoked will find himself precisely where he would have been if those credits had never been earned. All years of the term remain 365 days, as they always have. But a misbehaving prisoner who had formerly earned, say, 500 credits will find himself without the benefit of those 500 days. That will leave him with more of his term to complete — 500 days more, to be precise. If he behaves well again, he can resume earning credit for the remainder of his term, but he has lost the opportunity to earn credits for any prior years. See § 3624(b)(1). This is not at all confusing for a prisoner; and certainly it is as straightforward, if not more so, than the Court’s approach. The Court’s view causes a prisoner’s “term of imprisonment” to shrink over time according to an algebraic formula, only to expand again if he misbehaves.
Finally, the Court speculates that BOP might find the above approach difficult to administer. The Court identifies no basis for this claim, nor does one exist. The information used to calculate a prisoner’s term under the above approach is the same as it is under the Court’s approach. True, a prisoner may become eligible to be awarded credit on different calendar days during the course of his term. But under the Court’s approach, this also happens when awarding credit in the final year. And, it goes without saying, federal prisoners begin their incarceration on different calendar days anyway, so that under any approach, BOP will be forced to evaluate prisoners throughout the calendar year.
II
The Court’s reading of § 3624(b)(1), therefore, is less consistent with the text than the reading explained above. But even if these interpretations were in equipoise, under any fair application the rule of lenity should tip the balance in petitioners’ favor. When a penal statute is susceptible of *501two interpretations, the one more favorable to the defendant must be chosen unless “text, structure, and history... establish that the [harsher] position is unambiguously correct.” United States v. Granderson, 511 U. S. 39, 54 (1994). Resolving ambiguity in favor of lenity ensures that statutes provide “fair warning[,] ... in language that the common world will understand, of what the law intends to do if a certain line is passed.” United States v. Bass, 404 U. S. 336, 348 (1971) (internal quotation marks omitted). The rule thus applies “not only to interpretations of the substantive ambit of criminal prohibitions, but also to the penalties they impose.” Bifulco v. United States, 447 U. S. 381, 387 (1980).
The Court assumes without deciding that § 3624(b) is penal in nature. See ante, at 488. No assumption is necessary: The statutory provision awarding good time credits “in fact is one determinant of [a] prison term,” so that a prisoner’s “effective sentence is altered once this determinant is changed.” Weaver v. Graham, 450 U. S. 24, 32 (1981). In Weaver, the Court considered whether an amendment to Florida’s statutory formula for calculating good time credits implicated the Ex Post Facto Clause. The Court concluded that it did, as the new statute “substantially alter[ed] the consequences attached to a crime already completed, and therefore change[d] ‘the quantum of punishment.’ ” Id., at 33 (quoting Dobbert v. Florida, 432 U. S. 282, 294 (1977)). For the same reason, the penal effect of § 3624(b)(1) is substantial enough to implicate the rule of lenity. We should not disadvantage almost 200,000 federal prisoners unless Congress clearly warned them they would face that harsh result.
Ill
The Government — although not the Court — argues that we should embrace its interpretation out of deference to BOP. BOP has been charged by the Attorney General with responsibility for “ [approving inmate disciplinary and good time regulations.” 28 CFR §0.96(s) (2009). BOP has long *502followed the same credit-calculation method now advocated by the Court. The Government argues that we should defer to BOP’s choice as a permissible exercise of its delegated responsibility.
This argument fails on multiple levels. There is no indication that BOP has exercised the sort of interpretive authority that would merit deference under Chevron U. S. A. Inc. v. Natural Resources Defense Council, Inc., 467 U. S. 837 (1984). The statute does not create a legislative gap for BOP to fill. To the contrary, the procedures that govern the timing of credit awards are spelled out in great detail. Cf. Lopez v. Davis, 531 U. S. 230, 241-242 (2001) (where statute says that BOP “may” grant early release to certain prisoners, without specifying further criteria, Congress deliberately created a “statutory gap”). The statute even goes so far as to explain what to do “[i]f the date for a prisoner’s release falls on a Saturday, a Sunday, or a legal holiday.” § 3624(a). This legislative specificity as to timing contrasts with other provisions that do delegate authority to BOP. E. g., § 3624(b)(1) (awarding of credit is “subject to determination” by BOP that the prisoner “has displayed exemplary compliance with institutional disciplinary regulations”).
BOP has not claimed that its view is the product of any “formal administrative procedure tending to foster the fairness and deliberation that should underlie a pronouncement” with the force of law. United States v. Mead Corp., 533 U. S. 218, 230 (2001). In 2005, BOP made final an administrative rule adopting its preferred methodology. 70 Fed. Reg. 66752 (adopting 28 CFR § 523.20). But when pressed during an earlier stage of this litigation, BOP conceded that it had “failed to articulate in the administrative record the rationale upon which it relied when it promulgated” the rule. Tablada v. Thomas, 533 F. 3d 800, 805 (CA9 2008). The Court of Appeals accepted BOP’s concession, ibid., and that aspect of its ruling has not been appealed.
*503As a fallback position, the Government argues that BOP’s interpretation should receive at least some deference under Skidmore v. Swift & Co., 323 U. S. 134 (1944). But under Skidmore, an agency decision only merits “respect proportional to its ‘power to persuade.’” Mead, supra, at 235 (quoting Skidmore, supra, at 140). BOP's position is of long standing, but the administrative record is noteworthy for what it does hot contain — namely, any reasoned justification for preferring BOP’s methodology over statutorily permissible alternatives. BOP has consistently adhered to its mistaken belief that its approach is the only one that can be squared with the text. See 62 Fed. Reg. 50786 (1997) (explanation to interim rule asserting that the correct methodology “had been clearly stated by statute since the implementation of the Sentencing Reform Act of 1984”). For example, at no point did BOP consider, much less consciously reject, the interpretation outlined here. Cf. Reno v. Koray, 515 U. S. 50, 60-61 (1995) (deferring to BOP’s reasoned decision to reject one interpretation in favor of another). An agency need not consider all possible alternatives. But deference is not owed to an agency view, however consistently held, that from the start has been premised on legal error. See Mead, supra, at 228; Skidmore, supra, at 140.
* * *
The straightforward interpretation urged here accords with the purpose of the statute, which is to give prisoners incentive for good behavior and dignity from its promised reward. Prisoners can add 54 days to each year. And when they do so, they have something tangible. In place of that simple calculation, of clear meaning, of a calendar that can be marked, the Court insists on something different. It advocates an interpretation that uses different definitions for the same phrase in the same sentence; denies prisoners the benefit of the rule of lenity; and caps off its decision with an *504appendix that contains an algebraic formula to hang on a cell wall.
To a prisoner, time behind bars is not some theoretical or mathematical concept. It is something real, even terrifying. Survival itself may be at stake. See Dept, of Justice, Bureau of Justice Statistics, C. Mumola, Suicide and Homicide in State Prisons and Local Jails (NCJ 210036, Aug. 2005), online at http://bjs.ojp.usdoj.gov/content/pub/pdf/shsplj.pdf (as visited June 2,2010, and available in Clerk of Court’s case file) (prison homicide rates); National Prison Rape Elimination Commission Report, p. 4 (June 2009) (citing a national survey estimating that 60,500 state and federal prisoners had been sexually abused during the preceding year). To this time, the Court adds days — compounded to years. We should not embrace this harsh result where Congress itself has not done so in clear terms. I would reverse the judgment of the Court of Appeals.