As a means of providing health care to the aged and disabled, Congress enacted the Medicare program in 1965. See Title XVIII of the Social Security Act, 79 Stat. 291, as amended, 42 U.S.C. § 1395 et seq. Under the program, providers of health care services can enter into agreements with the Secretary of Health and Human Services pursuant to which they are reimbursed for certain costs associated with the treatment of Medicare beneficiaries. To operate the program, the Secretary issued regulations imposing limits on the amount of repayment based on a range of factors designed to approximate the cost of providing general routine patient service. The question before us is whether the Secretary must afford the six petitioning hospitals an opportunity to establish that they are entitled to reimbursement for costs in excess of such limits.

* A.

A complex statutory and regulatory regime governs reimbursement, rough description of which is necessary background to this case. To begin, Congress has required the Secretary to repay the lesser of the "reasonable cost" or "customary charg[e]." See 42 U.S.C. § 1395f(b)(1). Rather than attempt to define "reasonable cost" with precision, Congress empowered the Secretary to issue appropriate regulations setting forth the methods to be used in computing such costs. See 42 U.S.C. § 1395x(v)(1)(A).1

Prior to 1972, the Secretary's regulations contemplated reimbursement of the entirety of a provider's services to Medicare patients unless its costs were found to be "substantially out of line" with those of similar institutions. See, e.g., 20 CFR § 405.451(c) (1967).2 In 1972, apparently fueled by concern that providers were passing on inefficient and excessive expenses, see H.R.Rep. No. 92-231, pp. 82-85 (1971); S.Rep. No. 92-1230, pp. 188-189 (1972), Congress amended the statute to specify that "reasonable costs" meant only those "actually incurred, excluding therefrom any part of incurred cost[s] found to be unnecessary in the efficient delivery of needed health services," 42 U.S.C. § 1395x(v)(1)(A), and to authorize the Secretary—as part of the "methods" of determining costs—to establish appropriate cost limits. See 42 U.S.C. § 1395x(v)(1)(A).

Accordingly, the Secretary promulgated regulations, updated yearly and establishing routine cost limits based on factors such as the type of health care provider (hospi tals, skilled nursing facility, etc.), type of services it rendered, its geographical location, size, and mix of patients treated. See 20 CFR § 405.460 (1975). Hospitals are divided in terms of bed size, and of whether they are urban—i.e., located in a Standard Metropolitan Statistical Area (SMSA)—or rural. As of 1979, the labor-related component of provider costs was to be determined by a wage index keyed to the hospital's location. See, e.g., 46 Fed.Reg. 33637 (1981).

The regulations generally provide that reimbursable costs must be within the cost limits. The regulations also allow for adjustments to the limits as applied to a provider's particular claim. A provider classified as a rural hospital can apply for reclassification as an urban one. 42 CFR § 413.30(d) (1992). An exemption from the applicable cost limits can be obtained under certain specified situations—e.g., when excess expenses are due to "extraordinary circumstances," when the provider is the sole hospital in a community, a new provider, or a rural hospital with fewer than fifty beds. § 413.30(e). In addition, exceptions are available for, inter alia, "atypical services," extraordinary circumstances beyond the provider's control, unusual labor costs, or essential community services. § 413.30(f).3

Two statutory provisions are of central importance to this litigation. First, apparently to protect providers' liquidity, the statute contemplates a system of interim, advance payments during the year. Specifically, the Secretary "shall periodically determine the amount which should be paid . . . and the provider of services shall be paid, at such time or times as the Secretary believes appropriate (but not less often than monthly) . . . the amounts so determined, with necessary adjustments on account of previously made overpayments or underpayments." 42 U.S.C. § 1395(g)(a). These interim payments by definition are only approximate ones, based on the provider's preaudit, estimated costs of anticipated services. See 42 CFR §§ 413.64(e), (f) (1992). Second, the regulations were required to "provide for the making of suitable retroactive corrective adjustments where, for a provider of services for any fiscal period, the aggregate reimbursement produced by the methods of determining costs proves to be either inadequate or excessive." 42 U.S.C. § 1395x(v)(1)(A)(ii) (clause (ii)).

Petitioners are six Nebraska hospitals certified as "providers" of health care services and classified as "rural" for Medicare purposes. Between 1980 and 1984, their costs exceeded the corresponding cost limits. Pursuant to 42 U.S.C. § 1395oo, they filed an appeal to the Provider Reimbursement Review Board (PRRB) in which they challenged the validity of the applicable cost limits on two grounds. First, they claimed that the wage index that was used to calculate reasonable cost of labor did not account for the use of part-time employees. Because petitioners used a greater proportion of part-time employees than the national average, this had the effect of artificially lowering their index values. In support of their claim, they pointed to Congress' decision in 1983 ordering the Secretary to conduct a wage index study to consider the distortion due to part-time employment, Medicare and Medicaid Budget Reconciliation Amendments of 1984, Pub.L. 98-369, § 2316(a), 98 Stat. 1081, followed by the Secretary's own revision of the wage index in 1986 which accounted for part-time employees, 51 Fed.Reg. 16772 (1986), and to Congress' directive that the revised index be applied to discharges occurring after May 1, 1986. Medicare and Medicaid Budget Reconciliation Amendments of 1985, Pub.L. 99-272, § 9103(a), 100 Stat. 156. Second, they asserted that under the cost limits a rural hospital could not show that it incurred the same wage costs as its urban counterparts when in fact its location next to urban hospitals forced it to compete for employees by offering equivalent compensation. Petitioners also complained that the cost limits were applied conclusively rather than presumptively. Invoking clause (ii), which provides for "suitable retroactive corrective adjustments," they argued that they were entitled to reimbursement of all costs they could show to be reasonable, even if they were in excess of the applicable cost limit.4

Because the PRRB believed that it lacked the authority to award the desired relief, it granted petitioners' request for expedited judicial review. See 42 U.S.C. § 1395oo (f)(1). Adhering to the Eighth Circuit's decision in St. Paul-Ramsey Medical Center v. Bowen, 816 F.2d 417 (1987), the District Court ruled for petitioners, holding that clause (ii) compelled the Secretary to reimburse all costs shown to be reasonable, regardless of whether they surpassed the amount calculated under the cost limit schedule.5

The United States Court of Appeals for the Eighth Circuit reversed. Good Samaritan Hospital v. Sullivan, 952 F.2d 1017 (1991). The court relied on our decision in Bowen v. Georgetown University Hospital, 488 U.S. 204, 109 S.Ct. 468, 102 L.Ed.2d 493 (1988), in which we held that clause (ii) does not permit retroactive rulemaking. 952 F.2d, at 1023. It reasoned that petitioners' request for adjustments to correct "inequalities in the system . . . would amount to a retroactive change in the methods used to compute costs that, after Georgetown, is invalid." Id., at 1024. Instead, the Court of Appeals adopted the Secretary's more modest view of clause (ii) as permitting only a "year-end book balancing of the monthly installments" with the amount determined to be "reasonable" under the applicable regulations. Ibid. Under this approach, clause (ii) establishes the mechanism through which the total of the interim payments extended pursuant to § 1395g (which merely purport to be estimates of actual costs) are reconciled with the postaudit amounts determined at year's end to be owed under the methods determining allowable costs.6 We granted certiorari to resolve a conflict among the Courts of Appeals.7 506 U.S. ----, 113 S.Ct. 319, 121 L.Ed.2d 240 (1992).

The starting point in interpreting a statute is its language, for "[i]f the intent of Congress is clear, that is the end of the matter." Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 842, 104 S.Ct. 2778, 2781, 81 L.Ed.2d 694 (1984). See also NLRB v. Food & Commercial Workers, 484 U.S. 112, 123, 108 S.Ct. 413, 416, 98 L.Ed.2d 429 (1987). Clause (ii) instructs the Secretary to "provide for the making of suitable retroactive corrective adjustments where, for a provider of services for any fiscal period, the aggregate reimbursement produced by the methods of determining costs proves to be either inadequate or excessive." Petitioners argue that the mandate is clear: The methods for determining reasonable costs having been determined pursuant to § 1395x(v)(1)(A), clause (ii) must be read to mean that such methods nonetheless might yield "inadequate or excessive" amounts in any particular instance. Where such is the case, it is submitted, the clause mandates a correction that will provide full reimbursement for reasonable costs.

In contrast, the Secretary asserts that the "aggregate reimbursement" refers to the sum total of the interim payments made pursuant to § 1395g. These payments are, of course, based on the methods chosen by the Secretary to determine reasonable costs, but they are only anticipatory estimates of what the providers' reimbursable costs will be, made before all relevant data is available. At year's end, when the provider's reimbursable costs for services actually provided during that year are on hand, the pre-audit "aggregate" of the interim payments can be compared to the postaudit amounts due under the methods. Because the interim payments might have been erroneously calculated, their total might not match amounts owed, and adjustments must be performed to reconcile the two. See 42 CFR § 413.64(e), (f) (1992).

In our view, the language of clause (ii) does not itself clearly settle the issue before us. The clause is ambiguous in two respects. First, the "aggregate reimbursement produced by the methods of determining costs" could mean either (in petitioners' view) the amount due given proper application of the Secretary's regulations, or (in the Secretary's) the total of the interim payments, themselves derived from application of the methods to rough, incomplete data. Second, the clause refers to "inadequate" and "excessive" reimbursements, but without at any point stating the standard against which inadequacy or excessiveness is to be measured. Petitioners contend that the implicit referent must be the reasonable costs as established by the providers, without regard to the methods; the Secretary concludes that it must be the reasonable costs as determined by the agency applying the methods.

Each of the conflicting constructions is plausible but each has its difficulty. Petitioners contend that although the interim reimbursements might lead to inaccurate repayments, they are not part of the methods of determining costs to which § 1395x(v)(1)(A) refers, but rather payment methods governed by § 1395g. Moreover, the book-balancing role the Secretary would have us assign to clause (ii) arguably is already performed by § 1395g, which mandates periodic reimbursement "prior to audit or settlement by the General Accounting Office . . . with necessary adjustments on account of previously made overpayments or underpayments." The Secretary counters that, while clause (ii) is directed at year-end adjustments and designed to ensure that providers are reimbursed their reasonable costs, § 1395g addresses periodic adjustments to be made during the course of the fiscal year; § 1395g thus has its own role to play and is not surplusage.8

The Secretary also argues that words such as "corrective" and "adjustments" more readily evoke the simple mathematical rectifications that she contemplates than the complex process of revisiting applicable methods and comparing the amounts paid with an ill-defined standard of "reasonable" costs that is called for by petitioners' approach.9 It is true that § 1395x(v)(1)(A) defines reasonable cost as "the cost actually incurred, excluding therefrom any part of incurred cost found to be unnecessary in the efficient delivery of needed health services," and petitioners contend that this is the yardstick against which reimbursements must be measured. But the statute proceeds to explain that reasonable cost "shall be determined in accordance with regulations establishing the method or methods to be used." In similar fashion, the 1972 amendments allow for the provision of "limits on the direct or indirect overall incurred costs or incurred costs of specific items or services or groups of items or services to be recognized as reasonable. " Ibid. (emphasis added). In short, aside from the implementing agency's determination pursuant to its regulations, as to which Congress granted broad discretion, there is no available standard of reasonableness that could form a ready basis for "correct[ion]" or "adjustmen[t]."10

Because both the parties and the Court of Appeals are of the view that Georgetown is controlling, we turn our attention for a moment to our decision in that case. In 1983, a District Court struck down the Secretary's 1981 new cost rule for failure to comply with notice and comment requirements. After following proper procedures, the Secretary promulgated the same rule in 1984 and sought to apply the method retroactively for the time it had been held invalid. 488 U.S., at 206-207, 109 S.Ct., at 470-471. Drawing on the authority of clause (ii), the Secretary thus began to recoup "overpayments" claimed to have been made to hospitals as a result of the District Court's decision. The precise question we faced was whether clause (ii) permitted such retroactive rulemaking. We held that it did not. As we explained, although clause (ii) "permits some form of retroactive action [it does not] provid[e] authority for retroactive promulgation of cost-limit rules." Id., at 209, 109 S.Ct., at 472. Rather,

"clause (ii) directs the Secretary to establish a procedure for making case-by-case adjustment to reimbursement payments where the regulations prescribing computation methods do not reach the correct result in individual cases. The structure and the language of the statute require the conclusion that the retroactivity provision applies only to case-by-case adjudication, not to rulemaking." Ibid. (footnote omitted). As we further stated, "nothing in clause (ii) suggests that it permits changes in the methods used to compute costs; rather, it expressly contemplates corrective adjustments to the aggregate amounts or reimbursement produced pursuant to those methods." Id., at 211, 109 S.Ct., at 473 (emphasis in original).

But while Georgetown eliminated across-the-board, retroactive rulemaking from the scope of clause (ii), it did not foreclose either of the two interpretations urged in this case: case-by-case adjustments based on a comparison of interim payments with "reasonable" costs as determined by the Secretary; and case-by-case adjustments based on a comparison of amounts due under the regulations with "reasonable" costs as demonstrated by the provider. Cf. id., at 209, n. 1, 109 S.Ct., at 472, n. 1.

Confronted with an ambiguous statutory provision, we generally will defer to a permissible interpretation espoused by the agency entrusted with its implementation. See National Railroad Passenger Corp. v. Boston & Maine Corp., 503 U.S. ----, ----, 112 S.Ct. ----, ----, --- L.Ed. ---- (1992); Department of Treasury, IRS v. FLRA, 494 U.S. 922, 933, 110 S.Ct. 1623, 1629, 108 L.Ed.2d 914 (1990); K mart Corp. v. Cartier, Inc., 486 U.S. 281, 291-292, 108 S.Ct. 1811, 1817-1818, 100 L.Ed.2d 313 (1988). Of particular relevance is the agency's contemporaneous construction which "we have allowed . . . to carry the day against doubts that might exist from a reading of the bare words of a statute." FHA v. The Darlington, Inc., 358 U.S. 84, 90, 79 S.Ct. 141, 145, 3 L.Ed.2d 132 (1958). See also Aluminum Co. of America v. Central Lincoln Peoples' Utility Dist., 467 U.S. 380, 390, 104 S.Ct. 2472, 2479, 81 L.Ed.2d 301 (1984).

In this case, the regulatory framework put in place by the agency in furtherance of the Medicare program supports the book-balancing approach to clause (ii). Nowhere in the regulations was there mention of a mechanism for implementing the kind of substantive recalculation and deviation from approved methods suggested by petitioners. On the other hand, the regulations provided on more than one occasion for the year-end book balancing adjustment that, in the Secretary's opinion, is mandated by clause (ii). For instance, 20 CFR § 405.451(b)(1) (1967) stated:

"These regulations also provide for the making of suitable retroactive adjustments after the provider has submitted fiscal and statistical reports. The retroactive adjustment will represent the difference between the amount received by the provider during the year . . . and the amount determined in accordance with an accepted method of cost apportionment to be the actual cost of services rendered to beneficiaries during the year."11

Use of the words "suitable retroactive adjustment," borrowed from clause (ii), demonstrates the agency's understanding. As we wrote in Georgetown, "[i]t is clear from the language of these provisions that they are intended to implement the Secretary's authority under clause (ii)." 488 U.S., at 211, n. 2, 109 S.Ct., at 473, n. 2 (emphasis added). What is more, "[t]hese are the only regulations that expressly contemplate the making of retroactive corrective adjustments." Id., at 212, 109 S.Ct., at 473 (emphasis added). From the outset, then, the agency viewed clause (ii) as a directive for retroactive adjustment of payments for allowable costs, as determined by the methods.

In the aftermath of the 1972 amendments adding the cost limit provision, the agency appears to have ascribed the same role to clause (ii), namely to retroactively correct the difference between interim payments and reasonable costs—only, as a result of the amendments, the adjustment would now be based on the new definition of reasonable costs, which includes the cost limits that as a general rule were not to be exceeded. As previously described, however, the regulations promulgated by the Secretary permitted various exceptions, exemptions, and adjustments to the limits. See 20 CFR § 405.460(f) (1975); supra, at ----. A provider could obtain a reclassification "on the basis of evidence that [its] classification is at variance with the criteria specified in promulgating limits." 20 CFR § 405.460(f)(1) (1975). Exemptions for sole community hospitals have expanded to include new providers, rural hospitals with less than 50 beds; exceptions now extend to atypical services, circumstances such as strikes or floods, educational services, essential community services, unusual labor costs. See 42 CFR § 413.30 (1992). The agency's development—and continued augmentation —of a list of situations in which the cost limits would be waived is difficult to harmonize with an interpretation of clause (ii) that would give a provider the right to contest the application of any particular and statutorily authorized method to its own circumstances. Rather, it is consistent with a view that the cost limits by definition entailed generalizations that would benefit some providers while harming others, and with a desire to refine these approximations through the Secretary's creation of exceptions and exemptions.12

Petitioners argue that any deference to the agency's current position is unwarranted in light of its shifting views on the matter. It is true that over the years the agency has embraced a variety of approaches. Compare, e.g., Regents of Univ. of California v. Heckler, 771 F.2d 1182 (CA9 1985) (agency contends that clause (ii) permits only book-balancing); Whitecliff v. United States, 210 Ct.Cl. 53, 536 F.2d 347 (1976) (same), with Georgetown, supra (agency argues that clause (ii) allows retroactive rulemaking). In response, the Secretary attributes such inconsistency to the lower courts' erroneous interpretations of clause (ii). If providers could obtain substantive retroactive adjustments in the event of alleged underpayment, the argument goes, then so, in the face of alleged underpayment, would the agency. However, in the aftermath of Georgetown, she notes that the agency returned to its earlier position.

The Secretary is not estopped from changing a view she believes to have been grounded upon a mistaken legal interpretation. See Automobile Club of Michigan v. Commissioner, 353 U.S. 180, 180-183, 77 S.Ct. 707, 707-709, 1 L.Ed.2d 746 (1957). Indeed, "[a]n administrative agency is not disqualified from changing its mind; and when it does, the courts still sit in review of the administrative decision and should not approach the statutory construction issue de novo and without regard to the administrative understanding of the statutes." NLRB v. Iron Workers, 434 U.S. 335, 351, 98 S.Ct. 651, 660, 54 L.Ed.2d 586 (1978). See also NLRB v. Curtin Matheson Scientific, Inc., 494 U.S. 775, 787, 110 S.Ct. 1542, 1549, 108 L.Ed.2d 801 (1990); NLRB v. J. Weingarten, Inc., 420 U.S. 251, 265-266, 95 S.Ct. 959, 967-968, 43 L.Ed.2d 171 (1975). On the other hand, the consistency of an agency's position is a factor in assessing the weight that position is due. As we have stated, "[a]n agency interpretation of a relevant provision which conflicts with the agency's earlier interpretation is 'entitled to considerably less deference' than a consistently held agency view." INS v. Cardoza-Fonseca, 480 U.S. 421, 446, n. 30, 107 S.Ct. 1207, 1221, n. 30, 94 L.Ed.2d 434 (1987) (quoting Watt v. Alaska, 451 U.S. 259, 273, 101 S.Ct. 1673, 1681, 68 L.Ed.2d 80 (1981)). How much weight should be given to the agency's views in such a situation, and in particular where its shifts might have resulted from intervening and possibly erroneous judicial decisions and its current position from one of our own rulings will depend on the facts of individual cases. Cf. FEC v. Democratic Senatorial Campaign Committee, 454 U.S. 27, 37, 102 S.Ct. 38, 44 (1981).

In the circumstances of this case, where the agency's interpretation of a statute is at least as plausible as competing ones, there is little, if any, reason not to defer to its construction. We should be especially reluctant to reject the agency's current view which, as we see it, so closely fits "the design of the statute as a whole and . . . its object and policy." Crandon v. United States, 494 U.S. 152, 158, 110 S.Ct. 997, 1001, 108 L.Ed.2d 132 (1990).

Section 1395 explicitly delegates to the Secretary the authority to develop regulatory methods for the estimation of reasonable costs. See 42 U.S.C. § 1395x(v)(1)(A).13 To be sure, by virtue of their being generalizations, they necessarily will fail to yield exact numbers—to the detriment of health care providers at times, to their benefit at others.14 Presumably, the methods could use a more exact mode of calculating depreciation, cf. Daughters of Miriam Center for the Aged v. Mathews, 590 F.2d 1250 (CA3 1978), account for proximity to a college or university because it can distort the wage index, cf. Austin, Texas, Brackenridge Hospital v. Heckler, 753 F.2d 1307, 1316 (CA5 1985), or to a high-crime zone in which heightened, and expensive, security is called for. All of these variables, and many others, affect actual costs; factoring them in the methods undoubtedly would improve their accuracy. But "[w]here, as here, the statute expressly entrusts the Secretary with the responsibility for implementing a provision by regulation, our review is limited to determining whether the regulations promulgated exceeded the Secretary's statutory authority and whether they are arbitrary and capricious." Heckler v. Campbell, 461 U.S. 458, 466, 103 S.Ct. 1952, 1956, 76 L.Ed.2d 66 (1983) (footnote and citations omitted).

Besides being textually defensible, the Secretary's restrictive reading of clause (ii) comports with this broad delegation of authority. Congress saw fit to empower the agency to devise methods to estimate actual costs, and the agency has opted for the use of certain generalizations, with additional fine-tuning by way of exceptions, exemptions, reclassifications, and by making allowances for possible variations in costs consistent with efficiency. See n. ----, supra.15 What the agency forbids is the kind of wide range, ad hoc reassessments of the accuracy of the chosen methods implicit in petitioners' interpretation. Indeed, and for all practical purposes, petitioners' contention is that the methods chosen by the agency did not take into account sufficient variables, namely the proportion of part-time workers and proximity to urban centers. It is, in all but name, a challenge to the validity of the methods albeit in an individual case—including the cost limits, the exceptions and the exemptions, and to their adequacy as gauges of reasonable costs. The Secretary has construed the statute to allow such attacks, not via clause (ii), but rather, in keeping with the broad authority with which she is possessed, by way of the arbitrary and capricious provision of the Administrative Procedure Act, 5 U.S.C. § 706.16

The issue is not without its difficulties whichever way we turn. Though not the sole permissible one, the agency's interpretation of clause (ii), manifested in regulations promulgated soon after enactment and expressed today, "give[s] reasonable content to the statute's textual ambiguities." Department of Treasury, IRS v. FLRA, 494 U.S., at 933, 110 S.Ct., at 1629. The judgment of the Court of Appeals is

Affirmed.

Justice SOUTER, with whom Justice STEVENS and Justice SCALIA join, dissenting.

In the Court's view, the contrasting interpretations of clause (ii) proffered by the petitioners and the Secretary are in such equipoise that even slight deference to the Secretary is enough to tip the balance her way. As I read it, however, the language of clause (ii) plainly favors the petitioners.

The Court focuses on two portions of clause (ii). First, it says, the phrase "aggregate reimbursement produced by the methods of determining costs" may be understood, not only as the petitioners would read it, but as the Secretary does: "the total of the interim payments . . . derived from application of the methods [of determining costs] to rough, incomplete data." Ante, at ____. Second, the Court finds that "inadequate or excessive" may well mean, as the Secretary suggests, inadequate or excessive as measured against "the reasonable costs as determined by the [Secretary] applying the methods [of determining costs]." Ibid. I think the language of clause (ii) precludes these readings.

Clause (ii) identifies its subject, "aggregate reimbursement," as the figure "produced by the methods of determining costs." Thus, once we know what "the methods of determining costs" are, we should be able to discover the nature of the "aggregate reimbursement" that is "produced by" those methods. Section 1395x(v)(1)(A) makes it clear that "methods" refers to the regulations implementing the statutory mandate to pay providers of services "the cost actually incurred, excluding therefrom any part of incurred cost found to be unnecessary in the efficient delivery of needed health services." The first sentence of § 1395x(v)(1)(A), which together with § 1395hh authorizes the Secretary to issue such regulations, identifies them as "regulations establishing the . . . methods to be used . . . in determining . . . costs." And clause (i) of § 1395x(v)(1)(A) uses the exact same phrase as clause (ii): the regulations shall take into account both direct and indirect costs, it says, so that "under the methods of determining costs," patients who are not Medicare beneficiaries will not subsidize beneficiaries, nor will beneficiaries subsidize nonbeneficiaries. Thus, "the methods of determining costs" are not procedures for estimating costs to make interim payments; rather, they are the means for figuring the actual "reasonable cost of . . . services."

The Secretary appears not to dispute this, but contends, in the Court's words, that the phrase "produced by the methods of determining costs" actually means "derived from application of the methods to rough, incomplete data." Ante, at ____. In other words, as the Government asserted at oral argument, "what you're really doing is taking estimated data but running them through the same methods that you're eventually going to run the final data through in order to get a result." Tr. of Oral Arg. 31-32. There is, however, an obvious difficulty with this proposed interpretation: the complete lack of any reference to "incomplete" or "estimated" data in clause (ii). Two less obvious difficulties are even more telling.

First, nothing in Title XVIII of the Social Security Act specifies that interim payments should be calculated by applying to estimated data the complete, detailed methodology for reaching a final reasonable cost figure; the Secretary's own regulations, indeed, suggest just the opposite. "The interim payment," states the relevant regulation, "may be related to the last year's average per diem, or to charges, or to any other ready basis of approximating costs." 42 CFR § 413.60(a) (1992). And for purposes of devising preliminary estimates, this makes perfect sense; working through a permissible method for determining costs in all its detail may not improve the quality of an estimate if the raw figures used are mostly guesswork. But this divergence of methods for calculating interim payments and methods for determining reasonable cost casts doubt on the Secretary's proffered interpretation of "produced by the methods of determining costs." If interim, estimated payments may in fact be calculated without strict adherence to the methods of determining costs, it is hard to see why Congress would choose to identify a series of interim payments as "the aggregate reimbursement produced by the methods of determining costs."

Second, the Secretary's interpretation assumes that "the methods of determining costs" are no more than a series of equations, which can be applied as readily to final, audited cost figures as to mere projections. But the statute suggests that the term "methods" is not to be understood so narrowly. In the words of the statute, for example, the regulations establishing the methods may not only "provide for determination of the costs of services on a per diem, per unit, per capita, or other basis"; they may also "provide for the use of estimates of costs of particular items or services." § 1395x(v)(1)(A). Thus, as the statute conceives of them, the methods encompass not only a set of equations, but a set of determinations about whether to use actual costs or cost estimates for particular items or services. This set of determinations is relevant, of course, not to reckoning interim payments, but to calculating the final reimbursement due the provider of health services. Accordingly, a figure that is "produced by the methods of determining costs" should, absent some contrary indication, be the final figure.

The Court asserts that a contrary indication may be found in the use of the adjective "aggregate" to modify "reimbursement." " 'Aggregate,' " says the Court, "signifies 'sum total' and its use therefore might suggest that Congress had in mind the outcome of adding up the interim payments." Ante, at ____, n. 9 (citation omitted). I find no such suggestion in the statute's use of that term, for "aggregate," unlike, say, "cumulative," carries no necessary connotation of addition over time. More importantly, there is a far better explanation for the use of the term "aggregate." A health care provider will, over the course of a fiscal year, provide many different kinds of services to Medicare beneficiaries. Part A Medicare benefits, for example, cover, among other things, "inpatient hospital services," see 42 CFR § 409.5 (1992), a term that encompasses everything from bed and board, nursing services, and use of hospital facilities to medical social services, drugs, biologicals, supplies, appliances and equipment, certain other diagnostic and therapeutic services, and medical or surgical services provided by certain interns or residents-in-training. § 409.10(a). The statute plainly contemplates the use of different methods to determine the costs of these various services, see 42 U.S.C. § 1395x(v)(1)(A) (stating that the regulations "may provide for using different methods in different circumstances"), and the Secretary has indeed provided for a number of different methods. For instance, under the Secretary's "[d]epartmental method" for apportioning costs, the provider's cost of "routine services" is apportioned between Medicare and non-Medicare patients on an average cost per diem basis, whereas the cost of "ancillary" services is apportioned on the basis of the ratio of Medicare beneficiary charges to total patient charges in each department. See 42 CFR § 413.53(a)(1) (1992). The combined reimbursement for all of the different services performed by a health care provider, as calculated under all of the different methods allowed by the statute and specified in the regulations and other materials published by the Secretary, may aptly be labeled the "aggregate reimbursement."

As I thus read the statute, the term "aggregate" is important in making it clear not only that the "reimbursement" considered in clause (ii) is the total amount received by a provider for all of the services it has rendered to Medicare beneficiaries, but that the amount received should be considered only as a whole. This focus on the total amount received means that a provider who shows that a method results in an understating of the reasonable cost of a particular service will not necessarily be entitled to a "retroactive corrective adjustmen[t]" to recover that particular cost, for the Government may be able to show that the same method, or another method used by the provider, has overstated other costs. (By the same token, of course, the Government will not always deserve an adjustment when it shows that a method has overstated a particular cost.) The text's direction to look only at the total reimbursement also means that the provider will not be entitled to the prospective application of a more accurate method of its own devising, an insight into the statute that is hardly new; as the Court acknowledges, see ante, at ____, we recognized in Bowen v. Georgetown University Hospital, 488 U.S. 204, 211, 109 S.Ct. 468, 473 (1988) (emphasis in original), that "nothing in clause (ii) suggests that it permits changes in the methods used to compute costs; rather, it expressly contemplates corrective adjustments to the aggregate amounts of reimbursement produced pursuant to those methods."

This emphasis on the total, aggregate reimbursement received by the health-care provider makes sense in light of the broader goals of the Medicare program, addressing as it does Congress's concern that Medicare neither subsidize, nor be subsidized by, non-Medicare patients. See § 1395x(v)(1)(A)(i). As long as the aggregate Medicare reimbursement to a health-care provider equals its total reasonable costs of providing services to Medicare beneficiaries, that goal has been attained; the details of the methods used do not matter. Thus, I can find no ambiguity in the phrase "aggregate reimbursement produced by the methods of determining costs"; it refers univocally to the total, final amount due to a provider for services rendered to Medicare beneficiaries under the regulations promulgated by the Secretary.

The Court also finds ambiguity in the direction stated in clause (ii) to provide for an adjustment if the reimbursement proves to be "inadequate or excessive." While I agree with the Court that clause (ii) does not itself "at any point stat[e] the standard against which inadequacy or excessiveness is to be measured," ante, at ____, the absence of an explicit reference to a standard in clause (ii) does not keep us from looking for other textual clues about that standard. In this case, the strongest textual clue is found in the immediate neighbor of clause (ii), clause (i). Together, clauses (i) and (ii) form the fourth and last sentence of § 1395x(v)(1)(A). Whereas the third sentence of § 1395x(v)(1)(A) is permissive, the fourth sentence is mandatory; it concerns those things that the Secretary's regulations "shall" take into account or for which they "shall" provide. Clause (i) requires the regulations to take into account "both direct and indirect costs of providers of services" so that "the necessary costs of efficiently delivering covered services to individuals covered by the insurance programs established by this subchapter will not be borne by individuals not so covered, and the costs with respect to individuals not so covered will not be borne by such insurance programs." § 1395x(v)(1)(A)(i). The first of these two undesired results, it will be noted, would occur if the aggregate reimbursement to the provider were inadequate, in the sense of failing to cover all reasonable costs; the second, if that reimbursement were excessive.

Clause (ii) does not contain as exhaustive a description of its goal as clause (i); it simply requires the regulations to provide for suitable corrective adjustments where the methods of determining costs produce a reimbursement that "proves to be either inadequate or excessive." § 1395x(v)(1)(A)(ii). Reading the two clauses together, however, I think it most reasonable to take clause (ii)'s "inadequate or excessive" as shorthand for the two consequences that were just described in the same order, but more fully, in clause (i). This construction has the further virtue, of course, of support in my reading of the phrase "aggregate reimbursement produced by the methods of determining costs." For if that phrase, as I contend, refers to the amount ultimately due the provider as calculated under the Secretary's regulations (that is, according to the Secretary's "methods"), then the standard against which that amount is measured as "inadequate or excessive" must refer to some other figure (that is, a figure produced by some different method); no amount can be "inadequate or excessive" in relation to itself. Thus, in context, the phrase "inadequate or excessive" is not equivocal.

Broadening the context to all of Title XVIII only confirms the view that clause (ii) requires regulations providing for case-by-case exceptions to the methods for determining costs. Section 1395x(v)(1)(A), where clause (ii) is located, is a definitional, rather than an operative, provision; § 1395x(v) defines "[r]easonable costs." The chief operative provision to which § 1395x(v) relates is § 1395f(b), which is titled "Amount paid to provider of services"; § 1395f(b)(1) provides that under the Medicare program, providers of services are generally to be paid "the lesser of (A) the reasonable cost of such services, as determined under section 1395x(v) of this title . . . or (B) the customary charges with respect to such services." "Payments to providers of services" are covered under another section, 1395g. That section requires the Secretary "periodically [to] determine the amount which should be paid . . . to each provider of services," and requires "the provider of services [to] be paid, at such time or times as the Secretary believes appropriate (but not less often than monthly) . . . the amounts so determined, with necessary adjustments on account of previously made overpayments or underpayments." § 1395g(a). As the Court notes, ante, at ____, the petitioners argue that this section's provision for "necessary adjustments on account of previously made overpayments or underpayments" provides for the very book-balancing operation that the Secretary advances as the function of clause (ii), and thus renders clause (ii), as interpreted by the Secretary, entirely superfluous. The Court nonetheless appears to accept the Secretary's explanation that § 1395g deals with periodic adjustments to be made during the course of the fiscal year, whereas clause (ii) is directed at year-end adjustments. Ante, at ____. Two circumstances keep me from doing the same.

First, nothing in the language of § 1395g excludes "year-end adjustments" from its purview, or draws any distinction at all between periodic and year-end adjustments. All payments to providers for services to Medicare beneficiaries are made under the authority of § 1395g, since it is the only section in Title XVIII of the Social Security Act to deal with that subject; and § 1395g thus authorizes all payments to be "adjust[ed] on account of previously made overpayments or underpayments." It is doubtless this breadth which leads the Secretary to concede that had clause (ii) never been enacted, "the authority for some similar year-end mechanism might have been inferred under the Act as a whole, including 42 U.S.C. 1395g." Brief for Respondent 27, n. 16.

Second, the Secretary's proposed distinction between year-end and periodic adjustments fails to explain why Title XVIII would describe year-end, but not periodic, adjustments as "retroactive." The Secretary interprets "retroactive," as it appears in clause (ii), to mean only relating to a period for which some payment has already been made, thus rejecting the more common, stricter legal sense of the word, which implies the upsetting of some prior settled expectation or transaction. In this weak sense employed by the Secretary, however, the adjustments authorized by § 1395g are just as "retroactive" as those authorized under the Secretary's interpretation of clause (ii); they too relate to "previously made overpayments or underpayments." This leaves the Secretary with no way to explain why Congress, in passing the Social Security Amendments of 1965 (which established the Medicare program, and contained both passages, see 79 Stat. 297, 323), chose to distinguish § 1395g "adjustments" from § 1395x(v)(1)(A)(ii) "retroactive corrective adjustments."

For all of these reasons, I believe the text of the statute unambiguously requires the promulgation of regulations allowing providers (and the Secretary) to seek adjustments on the grounds that, as calculated under the methods of determining costs, the total reimbursement for a fiscal period is lower than (or higher than) the actual reasonable cost of providing services to Medicare beneficiaries. I respectfully dissent from the Court's opposite conclusion.